Repositorio Institucional de Documentos
Abstract
The present thesis examined changes in the annual and seasonal distribution of daily maximum and minimum temperatures for northeastern Spain. A better understanding of the ongoing changes in the temperature means and extremes was the primary objective. Further aims involved the analysis of large-scale atmospheric circulation patterns at different geopotential levels as well as the Mean Sea Level (MSL) pressure based on climate composites analysis and canonical variates in order to quantify the driving forces beyond the observed variability. Finally, this work aimed to assess future climate projections of seasonal temperature and their spatial variations to improve the understanding and prediction of the long-term trends of temperature means and extremes simulations. To achieve all these goals, it was necessary to develop a homogenous dataset with high spatial and temporal resolution. The next few paragraphs answer the main research questions raised during this work. El Kenawy El Sayed, Ahmed Mohammed Hussain; Vicente Serrano, Sergio Martín; López Moreno, Juan Ignacio
Full text
2012 39 Ahmed Mohammed Hussain El Kenawy El Sayed Spatio-temporal variability of surface air temperature in northeastern Spain Departamento Director/es Geografía y Ordenación del Territorio Vicente Serrano, Sergio Martín López Moreno, Juan Ignacio Director/es Tesis Doctoral Autor Repositorio de la Universidad de Zaragoza – Zaguan http://zaguan.unizar.es UNIVERSIDAD DE ZARAGOZA
Departamento Director/es Ahmed Mohammed Hussain El Kenawy El Sayed SPATIO-TEMPORAL VARIABILITY OF SURFACE AIR TEMPERATURE IN NORTHEASTERN SPAIN Director/es Geografía y Ordenación del Territorio Vicente Serrano, Sergio Martín López Moreno, Juan Ignacio Tesis Doctoral Autor 2012 Repositorio de la Universidad de Zaragoza – Zaguan http://zaguan.unizar.es UNIVERSIDAD DE ZARAGOZA
Departamento Director/es Director/es Tesis Doctoral Autor Repositorio de la Universidad de Zaragoza – Zaguan http://zaguan.unizar.es UNIVERSIDAD DE ZARAGOZA
SPATIO‐TEMPORALVARIABILITYOFSURFACEAIRTEMPERATURE INNORTHEASTERNSPAIN ADissertationsubmittedfortheDegreeofDoctorinPhilosophyintheDepartmentof GeographyandLandManagement UniversityofZaragoza Presentedby AhmedMohammedHussainElKenawyElSayed MScinEnvironmentalInformatics,UniversityofLeicester,UK(2007) Supervisor Dr.SergioMartinVicente‐Serrano InstitutoPirenaicodeEcologia,ConsejoSuperiordeInvestigacionesCientificas(CSIC), Zaragoza Co‐Supervisor Dr.JuanIgnacioLópezMoreno InstitutoPirenaicodeEcologia,ConsejoSuperiordeInvestigacionesCientíficas(CSIC), Zaragoza Summer2012
gÉ My little kid Kareem
“If you cannot do great things, do small things in a great way.” Napoleon Hill (1883-1970) American writer
V ABSTRACT From the climatological point of view, northeast Spain is an area of large interest due its varied geographical, topographical and climatic characteristics. In this work, a spatially and temporarily high-resolution daily dataset of temperature was developed for northeastern Spain. Data derived from a high number of observatories (1583) spanning some period between 1900 and 2006 was tested for internal and external consistency to check data quality. To improve data completeness, a linear regression model was then utilized to infill gaps in the daily temperature series using the best correlated data from nearby sites. Discontinuities in the reconstructed series were then determined by combining the results of three homogeneity relative tests: the Standard Normal Homogeneity Test (SNHT), the Easterling and Peterson twophased regression method, and the Vincent test. The newly complied dataset seems to be more robust and reveals more coherent spatial and temporal patterns of temperature compared with the original dataset. This finding was confirmed by means of a suite of statistics (e.g., semivariance models and L-moment statistics). From the temporal and spatial perspectives, the new dataset comprises the most complete register of temperature in northeast Spain (1900-2006), with a reasonably spatial coverage. Therefore, this database provides a more reliable base for studying temperature changes and variability in the region. This thesis analyzed the surface air temperature variability and trends over the study domain for the 20th century based on the new compiled dataset. An assessment of long-term change and variability of temperature was provided using a dataset of 19 observatories from 1920 to 2006. In addition, a more detailed analysis of the spatial and temporal variability of maximum, minimum, and mean temperatures, and the
VI diurnal temperature range (DTR) was also carried out employing 128 observatories spanning the period from 1960 to 2006. In general, maximum, minimum, and mean temperatures increased significantly, mostly from contributions in the late decades of the 20st century. At the seasonal scale, the analysis revealed that the weakest trends (mostly insignificant at the 95% level [p<0.05]) were observed during autumn, while the strongest warming rates were found during summer and spring. Spatially, the observed warming was more robust in the coastal portions compared with mainland observatories. Spatial and temporal characteristics of extreme temperature events were also investigated. A total of 21 indices were used to assess changes in the cold and warm tails of the daily temperature distribution, calculated at the annual timescale. The presence of trends in temperature extremes was assessed by means of the MannKendall statistic. However, prior to assessing trends, the autocorrelation function (ACF) and bootstrap methodologies were used to account for the influence of serial correlation and cross-correlation on the trend assessment. In general, the observed changes were more prevalent in warm extremes than in cold extremes. The results indicated a significant increase in the frequency and intensity of most of warm extremes. An increase in warm nights (TN90p: 3.3 days decade-1), warm days (TX90p: 2.7 days decade-1), tropical nights (TR20: 0.6 days decade-1) and the annual high maximum temperature (TXx: 0.27ºC decade-1) was detected in the 47-year period (1960-2006). In contrast, majority of the indices related to cold extremes (e.g., cold days [TX10p], cold nights [TN10p], very cold days [TN1p], and frost days [FD0]) demonstrated a decreasing but statistically insignificant trend. Spatially, similar to temperature means, the coastal areas along the Mediterranean Sea and the Cantabrian Sea experienced stronger warming compared with mainland areas.
VII In this thesis, a procedure for classifying daily temperature extremes into homogenous regions was also presented and evaluated. This procedure employed characteristics of temperature extremes calculated for summer (JJA), including temperature frequency (e.g., warm days), intensity (e.g., warmest day), and duration (e.g., maximum length of warm spell). Following the results of the principal components analysis (PCA) and the Ward´s method of clustering, the study area was divided into four homogenous sub-regions: the Mediterranean region, the mainland and the Cantabrian region, the moderately elevated areas westward and southward, and the mountainous region. The quality of this clustering was evaluated and ensured by means of an internal cluster validation measure (Silhouette width). Overall, the delineated sub-regions were proved as homogenous in terms of both the geographic and climatic meanings. The temporal evolution of the long-term (19602006) temperature extremes was assessed for each of these sub-regions. The Mediterranean and the highly elevated regions revealed the strongest signals in both day-time and night-time extremes. This study also explored the forcing mechanisms that can explain temperature variability at seasonal timescales. The results indicated that this variability can markedly be connected to variations in the large-scale atmospheric patterns. Notably, the Eastern Atlantic (EA), the Scandinavian (SCA), and the Western Mediterranean Oscillation (WeMO) patterns exerted significant influences on temperature variations in the study domain. Temperature tended to increase during the positive (negative) modes of the EA (WeMO and SCA) patterns. Also, the possible physical processes and mechanisms favoring for the occurrence of the anomalous extreme heat events (very warm days [VWD: daily Tmax>99th percentile] and very cold nights [VCN: daily Tmin<1st percentile]) were examined. This included
VIII configurations of Seal level pressure (SLP) and 200hPa and 500hPa geopotential height fields. The occurrence of VCN was mainly associated with predominance of the meridional circulation over much area of the Western Europe, with strong advection of cool air from northern continental Europe and the North Atlantic. On the other hand, the most likely factors contributing to VWD were the north-eastward displacement of the Atlantic subtropical high and the increase in the European blockings. In spite of the small spatial extent of the study domain (approximately 160,000 km2), the results interestingly confirmed that circulation patterns had spatially variable influences on both temperature means and extremes. This work also assessed how well a set of regional climate models (RCMs) can reproduce observed changes in seasonal temperatures over northeastern Spain. The projected changes in seasonal temperatures were assessed under the A1B emission scenario of climate change for the 21st century based on comparison of the control period (1971-2000) and two future time slices: 2021-2050 and 2071-2100. The results indicated that the current substantial warming will continue during the 21st century. The largest temperature changes were projected for the second half of the 21st century, with the strongest trends being on the order of 0.7 and 0.5°C decade-1 in summer and winter, respectively. The results demonstrated a rapid increase in wintertime minimum temperature and summertime maximum temperature. Spatially, almost the whole domain will be dominated by a positive temperature anomaly during the 21st century, relative to the present climate (19712000), with the largest warming in the central Ebro valley and the least over the Pyrenees and near the Cantabrian coast. In order to assess how the lower and upper tails of temperature distributions will change over the 21st century and whether these changes will be consistent with changes in the mean in both the sign and the magnitude, an analysis of the trends in various time-varying percentiles
IX (e.g., the 10th, 25th, 75th and 90th) was performed on a seasonal basis. Linear trends of temperature percentiles showed clear warming trend over the late decades of the 21st century (2071-2100), compared with both the mid century (2021-2050) and the observed (1971-2000) changes. The largest warming was projected for the lower (upper) percentiles of the minimum (maximum) temperature distributions during summer (winter). A warming rate of 0.9ºC and 0.8ºC decade-1 was observed for the 10th and 25th percentiles of summer minimum temperature, whereas the 75th and 90th percentiles of wintertime maximum temperature distribution increased by 0.5ºC and 0.6ºC decade-1, respectively. In addition, the results indicated that, among all seasons, summer will exhibit the largest interannual variability of temperature in the future. This increase could drastically increase the probability of exhibiting more extremely warm events in the region. This study also emphasized that changes in the upper and lower tails of temperature distribution may not follow the warming rate of the mean. This suggests that much of changes in the temperature percentiles will be driven by a shift in the entire distribution of temperature rather than only changes in the central tendency. In such a complex climatic and geographic region, an assessment and attribution of regional temperature variability based on high-quality, long-term and dense network of observatories can be advantageous for extracting finer scale information, which may prove useful for the vulnerability assessments and the development of local adaptation strategies for various disciplines such as hydrology, agriculture, water resources management, ecology and human health.
X
XI RESUMEN El noreste de España es un área de gran interés por su variedad geográfica, topográfica y climática. En el marco de este trabajo se ha desarrollado una base de datos de temperatura diaria para el noreste de España de una elevada densidad espacial. Se ha utilizado un gran número de observatorios (1583) con datos desde 1900 hasta 2006. La calidad de los datos fue examinada atendiendo a su consistencia interna y externa. Los datos ausentes fueron rellenados mediante modelos lineales de regresión utilizando las series cercanas de temperatura diaria que mostraban mejor correlación. Una vez reconstruidas las series, se aplicaron tres test de homogeneidad para detectar posibles discontinuidades: El Standard Normal Homogeneity Test (SNHT), el método de regresión de dos fases Easterling and Peterson, y el test de Vincent. Diversos análisis estadísticos (e.g., semivariance modelos and L-moment statistics) han confirmado que la base de datos resultante muestra una mayor coherencia espacial y temporal que la base de datos original. La base de datos obtenida representa el conjunto de datos de temperatura más completo del noreste de España para el periodo 1900-2006, siendo una base de datos idónea para estudiar la variabilidad y el cambio de la temperatura en la región. Esta tesis analiza la variabilidad y tendencia de las temperaturas en la zona de estudio durante el siglo XX. 19 series de temperatura han sido utilizadas para analizar el periodo 1920-2006, mientras que para análisis más detallados de la variabilidad espacial y temporal de las temperaturas máximas, mínimas, medias y rango diario (DTR) se utilizaron 128 observatorios que cubren el periodo 1960-2006. Se ha detectado un incremento generalizado de las temperaturas, determinado en gran medida por la evolución que han mostrado en las últimas décadas del periodo de estudio. Estacionalmente, los análisis indican que las tendencias más moderadas
XII han sucedido durante otoño, mientras el mayor calentamiento se ha detectado en verano y primavera. Espacialmente, el calentamiento detectado resulta mucho más consistente en las zonas de costa frente al detectado en los observatorios en zonas de interior. También se ha estudiado el comportamiento espacial y temporal en la ocurrencia de eventos extremos. Un total de 21 índices fueron seleccionados para caracterizar los extremos fríos y cálidos de la distribución de frecuencias de las series de temperatura diaria. La existencia de tendencias temporales se detectaron mediante el test de Mann-Kendall. Además funciones de autocorrelación (AF) y la técnica de bootstrap sirvió para eliminar el efecto de la correlación serial y la croscorrelación en la detección de tendencias. En general, los cambios observados son más marcados en la ocurrencia de eventos fríos que cálidos. Un incremento en la ocurrencia de noches cálidas (TN90p: 3.3 días/década), días cálidos (TX90p: 2.7 días/década), noches tropicales (TR20: 0.6 días/década) and la temperatura máxima más alta anual (TXx: 0.27ºC/década) se encontró para el periodo1960-2006. La mayoría de los índices relacionados con extremos fríos (p.e., días fríos [TX10p], noches frías [TN10p], días muy fríos [TN1p], y días de helada [FD0]) mostraron un descenso que generalmente no fue estadísticamente significativo. Parecido a lo observado con las temperaturas medias, las zonas costeras Mediterráneas y Cantábricas mostraron un mayor calentamiento respecto a las zonas de interior. Se ha clasificado la zona de estudio en cuatro regiones homogéneas según la ocurrencia, intensidad y duración de eventos de temperaturas extremas aplicando Análisis de Componentes Principales (PCA) y análisis clúster (método Ward). Las regiones identificadas han sido: la Mediterránea, la Cantábrica, el sector Ibérico y
XIII las estaciones más elevadas del Pirineo y el Sistema Ibérico. La robustez de la clasificación se validó utilizando el estadístico de Silhouette. La evolución temporal de los eventos extremos en las series más largas (1960-2006) muestra un comportamiento distinto entre las diferentes regiones. La región mediterránea y los observatorios en zonas de montaña mostraron la señal más clara de calentamiento tanto en las temperaturas máximas y mínimas. Este trabajo también ha analizado los mecanismos atmosféricos que pueden explicar la variabilidad térmica. Los resultados indican que los cambios observados en las temperaturas y sus diferencias espaciales y estacionales están relacionados con patrones atmosféricos de gran escala. En particular el patrón del Atlántico Este (EA), el patrón Escandinavo (SCA) y la Oscilación del Mediterráneo Occidental (WeMO) ejercen un papel muy importante en la variabilidad térmica de la zona de estudio. La temperatura tiende a incrementarse durante las fases positivas (negativas) del patrón EA (SCA y WeMO). Los posibles mecanismos físicos que favorecen la ocurrencia de los días más cálidos (TX99) y noches frías (TN1) han sido también examinados utilizando series diarias de presión a nivel de superficie (SLP) y las alturas geopotenciales de 200hPa y 500hPa. La ocurrencia de noches muy frías (VCN) está asociada a un incremento de la circulación meridional sobre grandes áreas de Europa occidental, con una clara advección de aire frío del norte de Europa y el Atlántico norte. La situación más favorable para la ocurrencia de días muy cálidos (VWD) es un desplazamiento norte-este del anticiclón Atlántico subtropical y situaciones de bloqueo sobre Europa. Sin embargo, los resultados confirman que los patrones de circulación afectan de forma espacialmente diferenciada a las temperaturas medias y extremas.
XIV Este trabajo también ha analizado la capacidad de los modelos climáticos regionales (RCMs) de reproducir los cambios observados en las temperaturas a escala estacional en la zona de estudio. Las series de temperatura simuladas, asumiendo un escenario de emisión de gases de efecto invernadero A1B, para dos periodos del siglo XXI (2021-2050) y (2071-2100), se han comparado a los simulados para un periodo control (1971-2000). Los resultados indican que los procesos de calentamiento van a continuar durante el siglo XXI, acelerándose en sus últimas décadas. El mayor incremento corresponde a verano e invierno, con unos ratios de calentamiento de 0.7 y 0.5 ºC por década respectivamente. Espacialmente, se espera un mayor incremento térmico en las zonas del interior del valle del Ebro, y un calentamiento más suave en el área pirenaica y la costa del Cantábrico. Se ha valorado la magnitud de los cambios proyectados en distintas partes de la distribución de frecuencia de las temperaturas, y sus efectos sobre los cambios en el signo y magnitud de los valores medios. Para ello se han aplicado análisis de tendencia a las series de evolución temporal de distintos percentiles (p.e., los percentiles 10, 25, 75 y 90) tanto en el periodo control como en los dos periodos futuros (2021-2050 y 2071-2100). El mayor calentamiento ha afectado a los percentiles más bajos (altos) de la temperatura máxima (mínima) de las distribuciones de verano (invierno). Se ha estimado un ritmo de calentamiento de 0.9 y 0.8ºC por década para los centiles 10 y 25 de la temperatura mínima de verano, mientras los percentiles 75 y 90 de la temperatura máxima de invierno pueden aumentar 0.5 y 0.6ºC por década respectivamente. Los resultados indican que verano mostrará la mayor variabilidad interanual en las próximas décadas. Este incremento puede aumentar notablemente la probabilidad de ocurrencia de eventos extremos cálidos en la región. Los cambios en los extremos de la distribución de frecuencia de las series de temperatura no siguen necesariamente el ritmo de
XXI 3.6.3.4. Cloudiness………………………………………..……………..…….216 4.4.2. Influence of large‐scale atmospheric circulation on temperature extremes……………………………………………………………………………218 4.4.2.1. Influenceonmoderateextremeevents……………….218 4.4.2.1.1. SLPconfigurations……………………………..221 4.4.2.2. Influence on the anomalously severe extreme events…………………………………………………………………229 4.5. Futurechangesoftemperatureduringthe21stcentury……….245 4.5.1. Modelvalidationresults………………………………………………………………245 4.5.2. Futurechangesintemperaturemeans………………………….…......252 4.5.3. Futurechangesintemperaturestandarddeviation…………………258 4.5.4. Futurechangesinthetime‐varyingpercentiles……………………….261 4.5.4.1. Winterseason(DJF)…………………………………………261 4.5.4.2. Springseason(MAM)………………………………………265 4.5.4.3. Summerseason(JJA)……………………………………….267 4.5.4.4. Autumnseason(SON)……………………………………..269 4.5.5. Futurechangesinextremeevents…………………………………………..271 4.5.6. The dependency between changes in the mean temperature and changes in the corresponding time‐varying percentiles…………………………………………..………………………………276 5. DISCUSSION………..………………………….……………………………………….287 5.1. Observedchangesinseasonaltemperaturemeans…………....287 5.1.1. Temperaturelong‐termchanges(1920‐2006)……………..………….287 5.1.2. Temperatureseasonalandannualtrends(1960‐2006)………..….289 5.1.2.1. Maximumtemperature………………………………………289 5.1.2.2. Minimumtemperature……………………………...………..291 5.1.2.3. Meantemperature……………………………………………..291 5.1.2.4. DTR……………………………………………………...……………292
XXII 5.2. Observedchangesintemperatureextremeevents…………….…296 5.2.1. Changesinwarmextremes……….…………………………………..………..296 5.2.2. Changesincoldextremes……………………………………………………….299 5.2.3. Changesinvariabilityextremes…………………………………………....…300 5.3. Spatialregionalizationoftemperatureextremeevents…………303 5.4. Drivingforcesofobservedvariabilityandchange…………….…..307 5.4.1. Influencesonseasonalmeantemperature..………………..………....307 5.4.1.1. Influencesofteleconnections……………………………..307 5.4.1.2. Influencesofland‐atmospherecoupling……………..314 5.4.2. Influenceoflarge‐scalecirculationonextremeevents...…………318 5.4.2.1. Influenceonsummertemperatureextremes….…….318 5.4.2.2. Influence on the anomalously severe extreme events…………………………………………………………………320 5.5. Futurechangesoftemperatureduringthe21stcentury……….324 6. CONCLUSION…………………………………………………………………………...337 7. CONCLUSIONES………………….……………………………………………………351 8. OUTLOOK………………………………………………………..………………………365 9. REFERENCES……………….………………….…………..……………………..……375 10. LISTOFAPPENDICES……………………………………………………….…….407
XXIII
XXIV
XXV LISTOFTABLES Table 1.1: List of sources and description of selected datasets covering northeastern Spain………………………………………………………………………………23 Table 2.1: Summary statistics of the flagged data following the quality control procedure. The numbers refer to the fraction of the flagged data in daily maximum (Tmax) and minimum (Tmin) temperature series. The mean indicates the average of the flagged data for the whole dataset, while the lowest (highest) shows the maximum (minimum) percentage of flagged data at the station-based level………………………………………………………65 Table 2.3: Summary of the homogeneity testing results (significance level =95%)……………………………………………………………………………70 Table 2.4: Results of the cross-tabulation analysis applied to trends in annual maximum and minimum temperature series before and after homogenization. Significance is assessed at the 95% level. Numbers between brackets indicate the fraction of observatories…………………....76 Table 3.1: List of the observatories used for trend analysis from 1920 to 2006……98 Table 3.2: List of all temperature indices and their definitions………………………..101 Table 3.3: List of the indices of summer temperature extremes and their definitions...................................................................................................118 Table 3.4: Summary of RCMs simulations used in the study with their corresponding resolutions………………………………………………………………………142 Table 4.1: Seasonal and annual trends in temperature regional series for the period 1920-2006 (ºC decade-1)…………………………………………………….157 Table 4.2: Number of observatories with statistically significant trend (p<0.05) for temperature variables, classified as: + statistically significant positive; - statistically significant negative; N statistically insignificant. The trends were calculated for the period 1960-2006. Total number of observatories is 128………………………………………………………………………………158 Table 4.3: Seasonal and annual trends in temperature regional series for the period 1960-2006 (ºC decade-1)……………………………………………………..163 Table 4.4: Seasonal and annual trends in temperature regional series for the period 1920-1959 (ºC decade-1)………………………………………………………163 Table 4.5: Results of trend analysis for cold, warm and variability extremes (significance is assessed at the 95% level). Abbreviations of the indices correspond to those in Table 3.2)…………………………………………....171 Table 4.6: Cross-tabulation of the sign of the trends (statistically positive, statistically negative, and statistically insignificant) for each pair of warm and variability extreme indices……………………………………………………………….172 Table 4.7: Cross-tabulation of the sign of the trends (statistically positive, statistically negative, and statistically insignificant) for each pair of cold and variability extreme indices………………………………………………………………..178 Table 4.8: Trends of extreme temperature indices for the four sub-regions, suggested by the cluster analysis results. Only bold numbers are statistically significant at the 95% level following the Mann-Kendall results)……….............................................................................................193
XXVI Table 4.9: Pearson correlation coefficients of seasonal maximum, minimum and mean temperatures with cloud cover and soil moisture over the study domain. Bold numbers indicate statistically significant correlation at p<0.05. The correlation was calculated between regional time series calculated for the whole area for each parameter………………………………………………..214 Table 4.10: Results of the cross-tabulation analysis applied to trends in the 10th percentile time series for wintertime minima from 2071 to 2100 against the observed trends. The statistical significance is assessed at the 95% level using the Mann-Kendall Tau test. The numbers indicate the percentage of grids in the ensemble that project the same sign of changes as the observed temperature……………………………………………………………….……264
XXVII
XXVIII
XXIX LISTOFFIGURES Figure 1.1: The key GHG with positive feedback on the global energy (IPCC, 2007)…………………………………………………………………………..2 Figure 1.2: (Left) Annual anomalies of (left) the global average land-surface and (right) northern hemisphere (NH) land-surface air temperature (ºC) from 1880 to 2010. The anomalies were calculated relative to normal mean over the base period 1961-1990. Solid line indicates the smoothed curve using a 7-yrs low pass filtering. The original data were derived from CRU (Jones et al., 2001)…………………………………………..……………….7 Figure 1.3: Annual anomalies of the European land & ocean air temperature (ºC) from 1880 to 2010. The anomalies were calculated relative to the normal mean over the base period 1961-1990. Solid line indicates the smoothed curve using a 10-yrs low pass filtering. The original data were derived from CRU (Jones et al., 2001)……………………………………………….9 Figure 1.4: Surface air temperature as a key component in many hydrological, climatic, energy transfer and environmental interactions. Black (white) arrows indicate positive (negative) feedbacks……………………………11 Figure 1.5: Location of the study domain and its administrative provinces…………25 Figure 1.6: (A) Topography of the study domain, and its main physiographical units……………………………………………………………………………26 Figure 1.7: (Upper) the main hydrological divisions and, (lower) drainage network in the study domain. The percentages indicate the area represented by each basin…………………………………………….…..….……………….27 Figure 1.8: The annual cycle of mean temperature climatology in the study domain, calculated as the long-term average for the period 1960-2006. The median, 10th, 25th, 75th and 90th percentiles are shown as vertical boxes. The red line represents the mean while the dots indicate the 5th and 95th percentiles…………………………………….…………………..30 Figure 1.9: Interannual variability of mean temperature climatology at three different observatories in the study domain, calculated as the long-term average for the base period 1960-2006. The altitude is given in meters………………………………………………….……………………..31 Figure 1.10: The annual cycle of total precipitation climatology in the study domain calculated as the long-term average for the period 1960-2006. The median, 10th, 25th, 75th and 90th percentiles are shown as vertical boxes. The red line represents the mean while the dots indicate the 5th and 95th percentiles…………………………………..…………………….32 Figure 1.11: Interannual variability of precipitation climatology at three different observatories in the study domain, calculated for the base period 19602006. The altitude is g iven in meters……………..……………………....33 Figure 2.1: Location of the study area and spatial distribution of temperature observatories in the original dataset (N=1583)…….……………………..43 Fi g ure 2.2. Distribution of temperature observatories accordin g to the province….44 Figure 2.3: The cumulative distribution of the number of observatories as a function of altitude…………………………………………………………………………44
XXX Figure 2.4: The number of available daily maximum-minimum temperature observatories in the original dataset from 1900 to 2006..........................45 Figure 2.5: Scatter plot representing the relationship between the length (in years) and the percentage of missing values in the time series of the original dataset. The solid line indicates the best fitted model curve. The length is defined as the difference between the opening date of the observatory and its more recent record……………………………….……………...…46 Figure 2.6: Schematic representation of the integrated approach applied to quality control, reconstruction, and homogenization of daily temperature time series in the study area. Shaded areas indicate the main steps………48 Figure 2.7: Spatial distribution of temperature observatories from 1950 to 2006 (N=98)………………………………………………………………………….59 Figure 2.8: The percentiles of the daily summer temperature at the observatory of Pamplona (red line) and its nearby observatories. The percentiles were calculated for each day in the period from 1st June to 31st August in 1983 (total of 92 Julian days), relative to the entire daily records of each observatory…………………………………………………………………....65 Figure 2.9: The reconstructed wintertime minimum temperature series of the observatory of Calatorao Cooperative (9428E, Zaragoza) based on joining data from two nearby sites (9425I and 9432). Pearson coefficient (r) indicates the correlation between the reconstructed time series and each of the nearby observatories. The green line denotes the final time series…………………………………………………………………………..66 Figure 2.10: Box plots summarizing (a) the number of nearby observatories (b) distance threshold, (c) Pearson correlation threshold and, (d) altitude difference of the series used to create the composite reference series for break detection. Dotted line indicates the mean. The median, 10th, 25th, 75th and the 90th percentiles are plotted as vertical boxes with errors bar……………………………………………………………………………...68 Figure 2.11: The composite reference time series for (a) May maximum temperature at the observatory of Sabadell Aerodromo [Barcelona] and (b) August minimum temperature at the observatory of Reus Aeroport [Tarragona]. Pearson correlation coefficients between the original and the reference time series are provided……………………………………………………69 Figure 2.12: Frequency distribution of the correction factors (ºC) applied to a) maximum and b) minimum temperature series averaged for each season. The values of the mean (absolute mean) of the correction factors averaged for all time series are also provided……………………………72 Figure 2.13: Test results of the SNHT applied to summer maximum temperature series in Fuenterrabia aeropuerto, [Guipuzoca] (a) before and (b) after homogeneity corrections. Dashed lines indicate the 95% significant level. The test statistic (T) is plotted against the critical value…………………73 Figure 2.14. Same as Figure 2.13, but for spring minimum temperature series in Els Hostalets de Balenya [Barcelona]……………………………………........74 Figure 2.15: Spatial distribution of the trends in the annual (a) maximum and (b) minimum temperature time series before and after homogeneity corrections. Trend calculation is based on the period 1950-2006 and statistical significance is assessed at the 95% level…………………….76
XXXVII
XXXVIII
XXXIX LISTOFABBREVIATIONS A Anticyclones ACF Autocorrelation Function AEM Agencia Estatal de Meteorologia AnClim Analysis of Climate datasets (software) ANOVA One-way Analysis of Variance AR Autoregressive Process AR4 The 4th Assessment Report C Cyclones CA Cluster Analysis CDF Cumulative Distribution Function CFCs Chloroflurocarbons CH4 Methane CHE Confederación Hidrográfica del Ebro CO2 Carbon dioxide CRU Climate Research Unit CV Coefficient of Variation DJF December-January-February (winter season) DTR Diurnal Temperature Range EA East Atlantic Oscillation EAWR East Atlantic/ Western Russian Oscillation ECA European Climate Assessment ELPIS A Dataset of Local-Scale Daily Climate Scenarios for Europe EMULATE European and North Atlantic daily to MULtidecadal climATE variability E-OBS Observational Gridded Daily Dataset for Europe EV Explained Variance FD0 Frost Days GCMs Global/General Climate Models GHCN Global Historical Climatolo gy Network GHG Green House Gases GPM Geopotential meters GSL Growing Season Length ID0 Ice Days IDW Inverse Distance Weighting interpolation algorithm IPCC Intergovernmental Panel on Climate Change IVE Inter-annual Variability Error JJA June-July-August (summer season) KMO Kaiser-Meyer-Olkin Test MA Moving Average MAM March-April-May (spring season) MBE Mean Bias Error MO Mediterranean Oscillation MSL Mean Sea Level N2O Nitrous Oxide NAO North Atlantic Oscillation
XL NCDC National Climate Data Center NCEP/NCAR National Centers for Environmental PredictionNational Center for Atmospheric Research NDJFM November-December-January-February-March (Cold half of the year) NESAT North-Eastern Spain Adjusted Temperature dataset NH Northern Hemisphere NOAA National Oceanic and Atmospheric Administration O3 Ozone OLS Ordinary Least Squares PCA Principal Component Analysis PDF Probability Density Function PPM Part Per Million PPB Part Per Billion PROCLIM PROcessing CLImatological datasets (software) PRUDENCE Prediction of Regional scenarios and Uncertainties for Defining European Climate change and Effects R Pearson Correlation Coefficient RCMs Re g ional Climate Models Rho Non-parametric Spearman Test RHTs Relative Homogeneity Tests SCA Scandinavian Oscillation SDATS Spanish Daily Adjusted Temperature Series SLP Surface Level Pressure SNHT Standard Normal Homogeneity Test SON September-October-November (autumn season) SPEI Standardized Precipitation Evapotranspiration Index SPI Standardized Precipitation Index SRES Special Report on Emissions Scenarios SST Surface Sea Temperature STARDEX Statistical and Regional dynamical Downscaling of Extremes for European re g ions Stdev Standard deviation SU Summer Days TFPW Trend Free Pre-Whitening Tmax Daily maximum temperature Tmean Daily mean temperature Tmin Daily minimum Temperature VCN Very Cold Nights VWD Very Warm Days WD Warm Days WeMO Western Mediterranean Oscillation WHO World Health Organization WMO World Meteorological Organization YK Yule–Kendall Skewness measure
XLI
XLII
CHAPTER ONE INTRODUCTION
1. INTRODUCTION 1 1. INTRODUCTION The following introductory paragraphs provide an overview of the research context, the fundamental questions of this thesis, and the framework applied for this work. Finally, a summary of each main chapter is given in the thesis outline. 1.1 . Climate Change Climate change is one of the most pressing global problems, which is likely projected to become a progressively more significant threat in the next decades. Natural scientists have described it as perhaps the preeminent environmental risk confronting the world in the 21st century. Indeed, there is increasing evidence that climate is changing worldwide and most of the pronounced change is mainly observed during the past few decades (IPCC, 2007). Some potentially climate driven changes include observed changes in sea level, snow cover, ice extent, species extinctions and distributions, and extreme weather events. Changes in climate can be attributable to internal and external variability. The influence of external factors on climate system can be broadly seen in the context of radiative forcing. In its 4th Assessment Report (AR4) in 2007 the IPCC indicated that there is a very high confidence (99% probability) that recent climate change is mainly driven by anthropogenic influences rather than natural variations. Recently, Jansen et al. (2007) emphasized this notion indicating that the global climate change cannot be explained without considering the role of the human activities. This finding has also been confirmed by Hegerl et al. (2007) who indicated that the observed change in the global climate during the 20th century cannot fully be attributed to the internal variability of climate. In other words, the significant anthropogenic warming, mainly caused by burning fossil fuels, industry,
1. INTRODUCTION 2 deforestation, land-use changes that modify surface albedo and other man-made activities, has induced considerable alterations of the global energy balance leading to both positive and negative radiative forcings (Ruddiman, 2003). Overall, the key role of increasing the GHG concentrations in the atmosphere is the imbalance between incoming and outgoing radiation. A clear positive radiative forcings is the rise of greenhouse gas (GHG) concentrations such as carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), ozone (O3) and chloroflurocarbons (CFCs) (Figure 1.1). From 1970 to 2004, the atmospheric CO2 concentration has raised from 180 parts per million (ppm) to 385ppm, while the Methan has elevated from 715 parts per billion (ppb) to 1790ppb (IPCC, 2007). Figure 1.1: The key GHG with positive feedback on the global energy (IPCC, 2007)
1. INTRODUCTION 9 Figure 1.3: Annual anomalies of the European land & ocean air temperature (ºC) from 1880 to 2010. The anomalies were calculated relative to the normal mean over the base period 1961-1990. Solid line indicates the smoothed curve using a 10-yrs low pass filtering. The original data were derived from CRU (Jones et al., 2001). Similarly, In response to the global warming, much more attention has recently been paid to assess the potential impacts of future climate change on both natural environments (e.g., Thomas et al., 2004; Ryan et al., 2008) and human activities (e.g., Patz et al., 2005; Ruiz-Ramos and Minguez, 2010). These efforts are largely motivated by the fast improvement in the capability of numerical climate models, which allow not only for improving their spatial resolution but also including more physical processes (e.g., cloud formation, aerosol influences, radiation balance and atmospheric fronts). In this context, numerous European projects have designed experiments to dynamically downscale regional climate projections from GCMs to regional scale by means of
1. INTRODUCTION 10 either dynamical or statistical approaches. Numerous modeling studies have been accomplished with a selection of the high-quality simulation data provided by some of these projects, such as PRUDENCE (Christensen et al., 2007), STARDEX (Haylock et al., 2006), ENSEMBLE (Hewitt and Griggs, 2004) and MISTRA-SWECIA (Kjellstrom, 2007). With these high-resolution climate simulations, it has been possible to quantitatively assess possible future impact of global warming, to perform model intercomparison studies and to assess the possible impacts of anthropogenic climate change on natural environments and socio-economic activities. 1.3 . Temperature Changes Temperature is a climate variable of high importance from the view of various areas including, among others, hydrology, agriculture, ecology, ecosystem, health and energy. The estimation of temperature changes is critical to understanding land surface–atmosphere interactions. This is principally because it is a key variable in both the water and energy cycles in the globe. As indicated in Figure 1.4, among different feedbacks, surface air temperature determines the magnitude of fluxes of outgoing longwave, sensible, latent and ground (surface) heat. In addition, it controls, to some extent, the proportion of rainfall partitioned into runoff through evaporation process. Also, surface temperature can largely affect crop yields through changes in the length of the growing season. Similarly, the number of days that lie outside physiological tolerable limits may negatively affect the regional biodiversity (Root et al. 2003). Thus, assessing changes in surface air temperature can be of particular importance for different hydrological, environmental and societal applications.
1. INTRODUCTION 11 Figure 1.4: Surface air temperature as a key component in many hydrological, climatic, energy transfer and environmental interactions. Black (white) arrows indicate positive (negative) feedbacks.
1. INTRODUCTION 12 Due to the adverse societal, economic and environmental potential impacts, there has been increasing interest in the analysis of temperature variability and trends over the last few decades (Diaz and Murnane, 2008; Solomon et al., 2007). The spatial and temporal variability of surface air temperature have increasingly been the focus of climatic research worldwide (e.g., Jones et al., 1999a; Folland et al., 2001). Global changes in the character of temperature have been observed in the past decades (Jones et al., 1999a; Jones and Moberg, 2003). For instance, Jones and Moberg (2003) reported that the global average earth surface temperature had increased by about 0.7°C in the 20th century. The most prominently observed changes were an increase in the mean during summer and winter periods, suggesting less seasonal variations. At a more localized scale, there is a good deal of evidence on significant increases in average temperature over Europe in recent decades (Klein Tank et al., 2005). The average increase in the observed annual mean temperature across the continent was 0.8ºC. The strongest changes were observed in northwestern Russia and southern Europe, particularly the Iberian Peninsula (Beniston et al., 2007; Della-Marta et al. 2007a,b). In their study on Switzerland, Scherrer et al. (2004), for example, found the most significant changes in mean temperature during summer. Also, roughly half of the north Atlantic warming since the last ice age was noted in the last decade (NRC, 2002). While the early detection of anthropogenic change in mean temperature is of growing concern, the impacts of climate change on society are more likely to be connected to extreme events. Typically, climate change detection is more often associated with the
1. INTRODUCTION 13 analysis of changes in extreme events than with changes in the mean (Katz and Brown, 1992; Meehl and Tebaldi, 2004). Recently, there have been a number of extreme cold and heat events that have caused loss of lives and serious economic damages to private property and infrastructure. For this reason, much attention has recently been paid to learn more about the frequency and intensity of extreme temperature events, as being induced by climate change (e.g., Easterling et al., 2000; Jones et al., 2001; Frich et al., 2002; Klein Tank and Konnen, 2003; Kostopoulou and Jones, 2005; Moberg and Jones, 2005; Vincent et al., 2005; Alexander et al., 2006; Moberg et al., 2006; Brown et al., 2008). Some of impact studies gave much more concern to assess influences of extreme temperature events on different aspects of human life including: mortality, comfort, ecology, agriculture, and hydrology (Schindler, 1997; Ciais et al., 2005; Garcia-Herrera et al., 2005; Patz et al., 2005). Among them, the immediate impacts of extreme events on mortality (Karl and Knight, 1997, GarciaHerrera et al., 2005), hospital admissions (Diaz et al., 2002), forestry (Miller and Urban 1999), ecology (Chust et al., 2011), agriculture (Adams et al., 1998), water resources (Vicente-Serrano et al., 2011a), tourism (Breiling and Charamza, 1999), energy demand (Smoyer-Tomic et al., 2003) and other socio-economic sectors (Hanemann et al., 2011) have been discussed. Overall, majority of studies have analyzed temperature extremes at different spatial scales ranging from the regional to the global. In general, most of the findings revealed significant upward (downward) trend in duration and frequency of warm (cold) extremes. Examples of these extreme events are found regularly in instances around the world (e.g., Haylock and Nicholls, 2000; Klein Tank and Konnen, 2003; Barriopedro, et al., 2011). The study by Alexander et al. (2006) gave the most current assessment of changes in observed daily temperature extremes from 1951 to 2003 at the global scale. They concluded that changes in
1. INTRODUCTION 14 temperature extremes were mainly linked to daily minimum temperatures. More than 70% of the global land areas showed a significant rise (decrease) in the annual amount of warm nights (cold nights). At a narrower scale, Frich et al. (2002) similarly found positive trends in warm extremes in Europe, the USA, China, Canada and Australia. With a more focus on the European continent, one can designate the August 2003 heat wave and the January 2012 cold wave as examples for extreme temperature events causing drastic impact and damage. According to Christopher and Gerd (2003), the unusual summer heat wave of 2003 had a significant influence on numerous European communities. For the Mediterranean region, a series of studies has also put a great deal of effort into exploring the behavior of temperature extremes (e.g., Maheras et al., 1999; Meehl and Tebaldi, 2004; Della-Marta et al., 2007a,b; Kuglitsch et al., 2010). According to Meehl and Tebaldi (2004), among the European regions, the Mediterranean is more vulnerable to intense, frequent and severe extreme heat events during the late of the 20th and the early 21st century. 1.4. Temperature Changes in the Iberian Peninsula: a general context While assessment of temperature trends at a global scale is markedly important, it is also of great interest to explore temperature variability at local and regional scales, which could have wide-ranging impacts on human society and the natural environment (e.g., agriculture, hydrology, human health.... etc). Recently, numerous studies have investigated temperature behavior in the Iberian Peninsula. The spatial coverage of these works varies considerably from the whole peninsula or Spain (e.g., Oñate and Pou, 1996; Hulme and Sheard, 1999; Rodriguez-Puebla et al., 2001b; Brunet et al.,
1. INTRODUCTION 15 2001, 2005, 2006, 2007a), sub-regions (e.g., Esteban Parra et al., 1995, 2003; Morales et al., 2005; Miro et al., 2006; Martinez et al., 2010) to a specific observatory (e.g., Serra et al., 2001). According to these works, the temperature increase in the Iberian Peninsula was even higher than the global average. Hulme and Sheard (1999) noted that the annual mean temperature increased by 1.6ºC during the last century. The strongest warming was detected since the mid of the 1970s. Nonetheless, this warming did not occur everywhere and in all seasons, but since 1972 it clearly dominated during winter and summer and over land areas. Notably, the warmest years during the 20th century occurred after 1990. Similarly, Brunet et al. (2007a) analyzed long-term variability of temperature in Spain employing a new daily adjusted dataset of the longest and quality records 22 Spanish daily time series covering the period 18502005. According to this study, the mean temperature increased at a rate of 0.10ºC decade-1 over the mainland Spain. Most of this increase attributed to the warming in maximum temperature, which raised at a rate twice the minimum temperature. Among the regional studies, Quereda et al. (2000) reported a strong trend in the annual mean temperature in the Spanish Mediterranean region during the period between 1870 and 1996 (0.71ºC decade-1). This warming has also been noted by Pausas (2004) who examined temperature variability in the eastern Mediterranean Peninsula using a daily dense network of 350 observatories and Miro et al. (2006) who analyzed daily summer temperature in a Mediterranean area (Valencia) for the period 1958-2003. Over the last few decades, many studies have also put a great deal of effort into exploring the behavior of temperature extremes in Iberia (e.g., Prieto et al., 2004; Brunet et al. 2007b; Bermejo and Ancell, 2009; Rodriguez-Puebla et al., 2010). Among these studies, Brunet et al. (2007b) assessed variability of temperature extremes in
1. INTRODUCTION 16 Spain over the course of the 20th century. They reported evidence of larger changes in high temperature extremes than low temperature extremes. Based on daily data from 26 observatories coupled with a gridded dataset, Rodriguez-Puebla et al. (2010) recently investigated spatial and temporal changes in warm days and cold nights across the Iberian Peninsula during 1950-2006. At a more localized scale, Miro et al. (2006), for example, reported a significant increase in the frequency of warm and extreme temperature days in Valencia from 1958 to 2003. Lana and Burgueño (1996) examined spatial distribution of extreme minimum temperature in Catalonia (NE Spain) during the cold season (December-March). Also, Burgueño et al. (2002) analyzed daily temperature extremes at the Fabra station (Barcelona). 1.5. Knowledge gaps Anthropogenic greenhouse gas emissions are a key driver of recent changes in temperature means and extremes. However, these effects will not be homogeneous across the globe. In other words, the regional temperature change may vary markedly from the global average temperature change, and therefore assessment of temperature change and variability should be undertaken at a regional scale to explore the extent to which regional temperature variability interacts with the global warming. This kind of detailed regional analysis of temperature variations is still demanded in the Iberian Peninsula given that it is bounded by the Atlantic Ocean and the Mediterranean Sea, which makes its climate largely influenced by the complex interactions of the Mediterranean and the Atlantic configurations. As a part of the western Mediterranean, the Iberian Peninsula has been proposed by many authors as “warm spot” region in terms of the global warming (e.g., Giorgi, 2006, Coppola and Giorgi, 2010). Indeed, a considerable portion of the peninsula is mountainous with clear topographical gradient,
1. INTRODUCTION 17 which makes it more sensitive to climate change.. The complex land–sea interactions along with latitude, altitude and orography variations make the local and regional climate very variable. Topography and continentality may also produce more or less complicated patterns of temperature and therefore the impact of changes in temperature can largely depend on these local and regional conditions. For example, differences in topography can cause local variations in the angle at which the solar radiation gets to the ground surface. Also, the aspect and the slope may considerably affect the magnitude of surface heating and cooling. For instance, as a portion of the northern hemisphere, south-facing slopes of the peninsula often receive more nearly direct solar radiation relative to north-facing slopes. The impacts of these local patterns may vary from season to season. In addition, the natural environment in the peninsula is characterized by its unique flora, fauna and ecosystems which make it more vulnerable to even slight variations in temperature (Pasho et al., 2011a, b). In this context, mesoscale and regional-scale climate processes can play a key role in this diversity. Taken together, the magnitude and rate of observed, and even simulated, changes in temperature can depend on these local conditions. In Iberia, the regional studies which have investigated temperature behavior almost exclusively give much more concern to the Mediterranean region as a consequence of availability of dense station network (e.g., Quereda et al., 2000; Pausas, 2004; Miro et al., 2006, Martinez et al., 2010), but generally there is less focus on mainland and the Cantabrian regions. Also, some of these studies were restricted to very smaller number of observatories, where reliable data were available (e.g., Lana and Burgueño, 1996; Burgueño et al., 2002) for the Fabra station (Barcelona).
1. INTRODUCTION 18 From the temporal perspective, so far, there is no comprehensive view of the longterm variability of temperature at the regional and sub-regional scales in a way that can significantly contribute to more accurate and robust assessment of climate variability and change. Indeed, changes of temperature variability over the earlier decades of the 20th century have poorly been addressed. The only few exceptions (e.g., Brunet et al., 2007a) were generally based on a very limited number of observatories. In a similar way, while numerous studies were conducted to trace spatial and temporal patterns of precipitation in Iberia, as traditionally being the most important climate variable over the peninsula (e.g., Martin-Vide, 2004; BegueriaPortugues and Vicente Serrano, 2006; Costa and Soares; 2009a; López-Moreno et al., 2010), there are only very few assessments of trends in temperature means and extremes (e.g., Lana and Burgueño, 1996; Burgueño et al., 2002; Prieto et al., 2004; Brunet et al. 2007a,b; Bermejo and Ancell, 2009; Rodriguez-Puebla et al., 2010). In these regional studies, some regions (e.g., NE Spain) have received less concern. In the same context, an increasing range of studies have projected changes in precipitation in the peninsula, allowing for assessing model-to-model differences (e.g., Sumner et al., 2003; Rodriguez-Puebla and Nieto, 2010; Sanchez et al., 2011; Vicente-Serrano et al., 2011b). On the other hand, a comprehensive assessment of temperature projections at the regional and sub-regional scales is still lacking and worth investigating. Some studies have already been made to assess future temperature rise over the Peninsular Spain (e.g., Taplador et al., 2009; Errasti et al., 2011; Brands et al., 2011a). Among them, Brands et al. (2011a), for example, employed 12 different GCMs to assess their ability to simulate the Probability Density Function (PDF) of daily temperature at various pressure levels over southwestern
1. INTRODUCTION 25 main physiographic unit as it represents approximately 53.7% of the whole domain. It extends inland along a northwest-southeast axis. This depression is a semi-enclosed basin, surrounded by mountain belts including the Pyrenees (north), the Cantabrian belt (northwest), the Catalan chain (east) and the Iberian system (south and southwest) (Figure 1.6, lower panel). Figure 1.5: Location of the study domain and its administrative provinces. 1.6.3. Hydrology Hydrologically, the largest Mediterranean river of the Iberian Peninsula and the study area is the Ebro, with a basin of nearly 85,000 km2 and a channel length of 930 km (Figure 1.7). The Ebro River originates from the southern facing slopes of the Cantabrian Range and the western Pyrenees before connecting to the Mediterranean Sea through a smaller opening in the Ebro Delta at Tortosa (180 km south of
1. INTRODUCTION 26 Barcelona) (Bejarano et al., 2010). The Pyrenees contributes nearly 70% of the total water runoff (López-Moreno et al., 2006). Major tributaries include Gallego, Aragon, Cinca, Segre, Jalon, Huerva, Zadorra and Ginel. In its middle reach, the river has a floodplain of an extent of 739 km2, representing the most extensive one in the peninsula (Ollero, 2010). Figure 1.6: (Upper) Topography of the study domain and, (lower) its main physiographical units.
1. INTRODUCTION 27 Figure 1.7: (Upper) the main hydrological divisions and, (lower) drainage network in the study domain. The percentages indicate the area represented by each basin. From the hydrological perspective, the flow peak of the Ebro occurs during February, while low water levels occur during summer (particularly in June). This is mainly because the rainfall originating over the Atlantic affects the headwaters of the Ebro during winter and early spring. Historically, the Ebro has constituted a major component of the Spanish water policy. The Confederación Hidrográfica del Ebro
1. INTRODUCTION 28 (CHE) (http:// www.chebro.es), found in 1926, is the oldest basin agencies in the world. Recently, there has been intense exploitation of water resources for sectors of agriculture irrigation, electricity production, cattle breeding, domestic consumption and industrial activities. According to Bouza-Deaño et al. (2008), the water of the Ebro is usually used for agriculture and cattle breeding (89.3%), domestic supply (7.2%) and industrial activities (3.5%). 1.6.4. Climate While the study domain has a relatively small geographical space, it encompasses large variations of climates, including oceanic, Mediterranean, continental and mountainous. According to the Koeppen classification (1936), much of the study domain is defined as a semiarid Mediterranean climate (BWh); a classification that has also been confirmed by Thornthwaite (1948) (see Peel et al., 2007). In the study domain, there are clear climate contrasts between the continental portions and the closing coasts. Also, there is a strong gradient from central flatland areas to adjacent mountainous belts. More specifically, the climate is highly varied, from semi-arid conditions in central portions (i.e., the Ebro valley), moderate conditions along the Mediterranean and the Cantabrian regions to mountainous climate. This diversity in climate comes mostly from weather systems interactions with a complex terrain that includes from sea-land interfaces to mountainous ranges. The interaction of the atmospheric circulation, the latitude, the altitude, the terrain topography, vegetation cover, the land-sea interactions and the atmospheric circulation are the main factors contributing to climate variations in the region (Rodriguez-Puebla et al., 1998; LópezBustins et al., 2008; Vicente-Serrano et al., 2009). From a global perspective, this region encompasses a climatic gradient between mid-latitude and subtropical regimes.
1. INTRODUCTION 29 Also, it is situated in a transitional zone where the Mediterranean configurations and the Atlantic influences at the mid-latitudes are the main driving forces of the regional climate. These closing seas are the most important source of moisture for surrounding land areas. The climate is therefore influenced by the large-scale configurations, originating from the north Atlantic and the Mediterranean. The various modes of the large-scale atmospheric circulation include, among others, the north Atlantic subtropical High and Sea Surface Temperature (SST). Generally, winters are characterized by cyclonic disturbances and negative pressure anomaly whereas subtropical high pressure system is prevalent during summers. Figure 1.8 indicates the annual cycle of mean temperature in the region. As indicated, January is the coldest month with an average surface air temperature of 5.2ºC, meanwhile July is the warmest month with an average of 21.4ºC. The mean annual temperature fluctuations are about 16.2ºC in the whole region. The annual temperatures range from about 17.9ºC in the lowest sites in the Ebro valley to 5ºC in the mountain areas. However, due to the complex topography of the region, the annual diurnal range, defined as the difference between the highest maximum temperature and the lowest minimum temperature in the year, is apparently high suggesting more continental influences in the region. For example, the mean maximum temperature during summer may reach 45ºC in the Ebro valley; meanwhile winter mean minimum temperatures may fall below -15ºC in very elevated sites in the Iberian system and the Pyrenees. This also suggests great interannual contrasts. This contrast is also evident over space, as revealed in Figure 1.9, in which interannual variations of mean temperature at three different observatories with varied elevation are plotted. It can be noted that temperature variations are elevation dependent. The highest temperatures
1. INTRODUCTION 30 often occur in the flattened areas in the Ebro valley (e.g., Zaragoza airport, Zaragoza) and in areas close to the Mediterranean Sea (e.g., Sort Piraguisme, Lleida). On the other hand, mountainous areas (e.g., Pantano de Compuerto, Palencia) and areas close to the Cantabrian Sea are relatively cooler. However, it is noteworthy indicating that this temperature-elevation dependency may vary from one season to another as a consequence of influences of other local variables such as the slope, aspect, land cover, vegetation canopy, and maritime influences. Figure 1.8: The annual cycle of mean temperature climatology in the study domain, calculated as the long-term average for the period 1960-2006. The median, 10th, 25th, 75th and 90th percentiles are shown as vertical boxes. The red line represents the mean while the dots indicate the 5th and 95th percentiles. The study region also includes diverse pluviometric regions, varying from arid, semiarid to humid. According to Martin-Vide and Olcina (2001), the precipitation isoline between 600 and 800 mm/year distinguishes between the humid and arid areas,
1. INTRODUCTION 31 meanwhile isoline between 300 and 350 mm/year differentiates the arid from the semi-arid regions. The humid area (above 1000 mm/year) is mainly located to the north close to the Cantabrian Sea and over the mountain ranges (e.g., the Iberian, Cantabrian and Catalan systems and the Pyrenees). Dry conditions occur in central portions, particularly over the Ebro valley and along the Mediterranean littoral area to the east. Figure 1.9: Interannual variability of mean temperature climatology at three different observatories in the study domain, calculated as the long-term average for the base period 1960-2006. The altitude is given in meters. As illustrated in Figure 1.10, precipitation is mainly concentrated in spring and autumn. The dry conditions typically prevail in the period from June to September, The average of total precipitation varies from 25.6mm (July) to 64.1mm (October). There are also large spatial variations, with a remarkable gradient from south to north and from east to west (De-Castro et al., 2005; Gonzalez-Hidalgo et al., 2010). These spatial contrasts are presented in Figure 1.11, which compares the interannual variations of
1. INTRODUCTION 32 total precipitation at three different localities in the region. As illustrated, the precipitation peak is located in October, with values ranging from 22.4 mm at Artieda (Navarra) to 71.2 mm at Segorbe (Castellon). On the other hand, total precipitation in the drier month varied from 7.9 mm in August at Artieda (Navarra) to 24.2 mm in July at Huesca Monflorite (Huesca). Most annual precipitation falls as snow in mountainous regions (e.g., the Pyrenees). A more detailed summary of the climate of the study area can be found in Capel-Molina (1981) and Font-Tullot (1983). Figure 1.10: The annual cycle of total precipitation climatology in the study domain, calculated as long-term average for the period 1960-2006. The median, 10th, 25th, 75th and 90th percentiles are shown as vertical boxes. The red line represents the mean while the dots indicate the 5th and 95th percentiles.
1. INTRODUCTION 33 Figure 1.11: Interannual variability of precipitation climatology in three different there observatories in the study domain, calculated for the base period 1960-2006. The altitude is given in meters. 1.7. Aims of the thesis The main focus of this thesis is to improve the understanding of the regional variability of temperature in northeast Spain using daily temperature dataset. The research line examines the observed space-time variability of temperature means and extremes from 1960 to 2006. The primary objectives of this study were: - To develop a complete, reliable and homogenous daily temperature dataset for northeastern Spain, with the aim of improving the spatial and temporal coverage of temperature time series in the study domain. This new compiled dataset was developed using a very dense network of 1583 daily raw time series distributing across the region. Herein, this research also aspires to assess the methodology used to quality control, reconstruct and adjust breaks
1. INTRODUCTION 34 (inhomogeneities) in the series on trends, spatial and temporal coherence, and statistical properties of the final series. In order to draw a comparison between the series before and after adjustments, a suite of different statistical methods (e.g., semivariance models, L-moment statistics, and Pearson order moment correlation) was used to accomplish this task. There are emphasizing needs to develop temperature dataset with high spatial and temporal resolution for the study area. - To examine spatial and temporal variability of surface air maximum, minimum and mean temperatures and daily temperature range (DTR) at both seasonal and annual timescales during the period from 1960 to 2006. This aim extends further to analyze the possible physical causes behind the observed regional variability. The purpose was to quantify possible large-scale atmospheric configurations that control the observed local trends. - To analyze extreme heat events based on (i) defining extreme events by means of both arbitrary and percentile-based thresholds, (ii) calculating their linear trends, and (iii) tracing spatial variability of the observed trends. To meet this goal, trends in daily maximum and minimum temperatures were assessed using 21 extreme temperature indices. - To divide the study domain into regions as homogenous as possible based on the information derived from a set of extreme temperature indices using the Principal Component Analysis (PCA) and Cluster Analysis (CA), and to assess the connections between spatial and temporal variability of temperature
2.DATASETDEVELOPMENTANDDESCRIPTION 41 2. DATASET DEVELOPMENT AND DESCRIPTION 2.1. Introduction In recent years, there is ongoing global increase about climate change and its natural and socioeconomic consequences (e.g., water resources, hydrology, forestry, agriculture, energy...etc). Assessing these impacts requires climate datasets of high spatial and temporal resolution. Beside data availability, the quality and homogeneity of climate series are prerequisites for detailed and trustworthy climate studies. Complete, reliable and spatially dense climatic datasets are mandatory for different types of climatic analyses (Eischeid et al., 2000). For example, an appropriate analysis of climate variability and trends necessitates climatic datasets of fine spatial and temporal resolution. If the inhomogeneities are not accounted for properly, the results of climate analyses based on the non-adjusted data can be erroneous and misleading (Peterson et al., 1998). The study domain is characterized by its complex topography, moderately high latitude, and transitional location between the Atlantic and the Mediterranean configurations. In addition, this domain encompasses diverse climate regimes, varying from semi-arid to humid, continental to maritime and lowlands to mountainous. Therefore, assessing temperature changes in a region of these high geographical and climatic contrasts necessitates a dataset of high spatial and temporal resolution. Unfortunately, the insufficient number of temperature observatories and its uneven spatial distribution has been the main features of recent studies focusing on air temperature change and variability in the region. In addition, most long-term time series covering the region are generally affected by some inhomogeneities caused by changes in instruments, observers, site displacements, observing practices, and the
2.DATASETDEVELOPMENTANDDESCRIPTION 42 surrounding environment (e.g., urbanization, land use and vegetation canopy). These inhomogeneities in temperature data do not make possible to consider temporal variations as induced only by climate processes. In this chapter, a detailed description of the different techniques conducted to reconstruct, quality control and test homogeneity of the available time series is provided. In general, the rationale behind this multistep procedure was to improve quality, temporal extension and spatial coverage of temperature time series in the study region. This chapter is structured in four main parts. A description of the original dataset is given in section 2.2. The methodology employed to check data quality, infill gaps, and test homogeneity of daily temperature time series are outlined in sections from 2.3 to 2.7. An assessment of the possible impact of data adjustment (correction) on spatial and temporal characteristics of the final temperature time series is included in section 2.8. Finally, a description of the final adjusted dataset is presented in section 2.9. 2.2. Raw dataset description This study employed a database of 1583 daily maximum and minimum temperature time series spanning some period between 1900 and 2006. This dataset was provided by the Spanish Meteorological Agency (Agencia Estatal de Meteorologia, AMET). The spatial distribution of temperature observatories is illustrated in Figure 2.1. A quick inspection of the network of stations clearly reveals that the spatial density of stations is inhomogeneous across the administrative 18 provinces. As depicted in Figure 2.2, the most data-rich provinces are Barcelona, Guipuzcoa and Cantabria with a density of 1 station for each 481.9, 493.6 and 515.5 km2, respectively. In contrast, the coarser
2.DATASETDEVELOPMENTANDDESCRIPTION 43 coverage is found in Soria, Guadalajara and Burgos, with a density of 1 station for an extent of 3430.4, 2711.8 and 2458.9 km2, respectively. The average of inter-station distance in the whole dataset is nearly 23.7 km. In terms of topography, majority of observatories are located in lowlands and moderately elevated areas. Figure 2.3 informs that 34, 52.7, 73.6 and 86 % of observatories are situated in sites below than 300, 500, 800 and 1000 m, respectively. Areas above 1500 m are only represented by 2.6% of the network observatories. Temporarily, the stations density markedly differs with a sharp increase from 1970 onwards (Figure 2.4). About 21.3 % of stations extend back to 1950 or earlier, while 98.7 % of the observatories have their most recent records until 2002. Figure 2.1: Location of the study area and spatial distribution of temperature observatories in the original dataset (N=1583).
2.DATASETDEVELOPMENTANDDESCRIPTION 44 Figure 2.2: Distribution of temperature observatories according to the province. Figure 2.3: The cumulative distribution of the number of observatories as a function of altitude. Altitude is given in meters.
2.DATASETDEVELOPMENTANDDESCRIPTION 45 Figure 2.4: The number of available daily maximum-minimum temperature observatories in the original dataset from 1900 to 2006. As illustrated in Figure 2.5, the raw dataset, particularly those series with long-term temperature records (e.g.,> 50 yrs), suffers from presence of missing values and temporal gaps. Changes in stations opening, closing and locations are being the reason. More importantly, Figure 2.5 indicates that parts of the available observatories have archives of instrumental climate records which, in some cases, date back to the first decades of the 20th century. Also, the region provides promising amount of data, as revealed by the dense network of observatories, which can allow for potential reconstruction of very useful long-term temperature time series making a good use of data from short interval time series (e.g., < 30yrs).
2.DATASETDEVELOPMENTANDDESCRIPTION 46 Figure 2.5: Scatter plot representing the relationship between the length (in years) and the percentage of missing values in the time series of the original dataset. The solid line indicates the best fitted model curve. The length is defined as the difference between the opening date of the observatory and its more recent record. In the following section, a full description of the multistep approach used for quality control, reconstruction and homogenization of the daily temperature series is detailed. A scheme that represents the different steps of this procedure is summarized in Figure 2.6.
2.DATASETDEVELOPMENTANDDESCRIPTION 47 2.3. Quality control of daily time series Quality control is a fundamental task to remove incorrect data and to check for data consistency and reliability (Feng et al., 2004). In this work, the original series were subjected to several quality checks. First, typical tests were performed to identify systematic errors, which resulted from different sources (e.g., archiving, transcription, and digitization). This can include non-existent dates, Tmin ≥ Tmax, Tmax > 50ºC, Tmin < - 50ºC and runs of at least 7 consecutive days with identical records. Next, the data were screened for internal consistency by comparing the value in question against other values in the same time series following Reek et al. (1992). Finally, the data were checked for external consistency by comparing each time series with nearby sites. The rationale behind this procedure was to trim outliers that markedly differ from the majority of neighbors while keeping the valuable extreme information. To accomplish this task, daily data of each testable observatory were compared with a minimum of 3 nearby observatories located within a maximum distance of 30 km. More specifically, the rank of each value in a testable time series was obtained as a quantile of all values in this series. Then, these quantiles were compared with the quantiles of nearby stations for each particular day. Values which exceed a user-defined difference threshold in this between-station comparison was flagged and set to missing values. In order to screen for appropriate interquantiles range threshold, this research tested different thresholds (e.g., 0.1, 0.2, 0.3, etc) using a random subset representing 1 % of all time series.
2.DATASETDEVELOPMENTANDDESCRIPTION 48 Figure 2.6: Schematic representation of the integrated approach applied to quality control, reconstruction, and homogenization of daily temperature time series in the study area. Shaded areas indicate the main steps.
2.DATASETDEVELOPMENTANDDESCRIPTION 49 The results suggested a threshold of 0.5 as a more suitable compromise for all seasons apart from winter, whereby a more restricted threshold (0.3) was set. This was principally done to account for the thermal and dynamical conditions originating from topography influences in winter. As the study domain is topographically complicated, the topography can significantly influence different atmospheric and climate processes through modification or blocking atmospheric flows and air masses. In areas of strong topography gradient, meso-scale weather systems can be found as a result of thermodynamic processes associated with the vertical (ascending and descending) movement of air (Barry and Chorley, 1987; Whiteman, 2000). Accordingly, the spatial dependency among observatories can markedly degrade over short distances. A similar approach was recently adopted by Stepanek et al. (2009) and Vicente-Serrano et al. (2010) to detect erroronous records in climate data. 2.4. Reconstruction of daily time series Short-term and fragmented time series may introduce noise to estimates of long-term climate changes. This is the typical case in many climatic datasets worldwide, where many gaps are introduced in the series. Peterson and Vose (1997) indicated that short or fragmented series may significantly alter the magnitude and sign of climate variability trends, particularly at regional-scale resolution. For this reason, many techniques have recently been applied for serial reconstruction in climatology including, for example, the neural artificial network (Rigol et al., 2001), interpolation algorithms (Vicente-Serrano et al., 2010), and the singular spectrum analysis (Ghil et al., 2002). The basic idea behind these methods is to obtain long-term complete dataset using all available data within a region. In general, the reconstruction can be
2.DATASETDEVELOPMENTANDDESCRIPTION 50 carried out as a function of the weighted distance (Shen et al., 2001) and/or correlation (Young, 1992). Nevertheless, these approaches differ in their applicability according to terrain complexity and spatial density of stations (Allen and DeGaetano, 2001; Hofstra et al., 2010). For instance, many interpolation techniques (e.g., optimal interpolation, Kriging, and splines-surface fitting) can give good results in regions with no gradients (Jarvis and Stuart, 2001a, b). Contrarily, their performance in terms of both bias and amount of variance is predominantly poor in areas of complex terrain or uneven density of observatories (Romero et al., 1998; Eischeid et al., 2000; Hubbard and You, 2005). In such environments, the regression-based methods can be a superior solution. These methods take into account data from adjoining stations with the best temporal correlation and the highest spatial dependence. The primary advantage of these methods is that they show less sensitivity to outliers (Romero et al., 1998), and can therefore better represent systematic temperature differences associated with topographic influences in heterogeneous regions (Hubbard and You, 2005). Compared to other meteorological variables, such as precipitation, temperature often has a statistically normal distribution. Accordingly, the regression-based approaches seem to be an adequate choice for this research. In literature, numerous previous studies proved the advantages of the regression methods to reconstructing temperature time series relative to other methods (e.g., Eischeid et al. 1995; Allen and DeGaetano, 2001; Hubbard and You, 2005; You et al., 2008). In this work, the standard linear regression model was simply applied to estimate missing value at the target observatory on a given day (i), as follows: ii bxaY += (2.1)
2.DATASETDEVELOPMENTANDDESCRIPTION 57 and/or the end of the series, particularly if they are not fully explained by metadata (e.g., Gonzalez-Rouca et al., 2000, Gokturk, et al., 2008). The homogeneity adjustments make it possible to consider temporal variations of temperature time series as caused only by climate variations. Therefore, an adjustment (correction) model was applied to account for statistically significant abrupt changes. The correction factor was computed for each month individually based on a comparison of percentiles of the differences between the candidate and the reference series before and after any detectable break. To obtain daily corrections, the predefined monthly corrections were interpolated into the daily time series following the approach described by Sheng and Zwiers (1998) and recommended by Vincent et al., (2002). This procedure is advantageous in various ways. First, it reduces discontinuities on the first and last days of each month (Vincent et al., 2002). Second, it maintains characteristics of daily extreme events, such as frequency and intensity. Lastly, it preserves temperature trends and variability presented in monthly temperatures. Altogether, it can be expected that the homogenized daily temperature series will be compatible with those of monthly resolution. Finally, it is worthwhile to indicate that testing homogeneity was applied iteratively as all time series in the dataset were considered repeatedly as candidate and reference series. However, more restricted criteria were considered in this iterative procedure. First, multiple reference series were used as alternative to the composite reference series to test homogeneity. Accordingly, each candidate time series was tested independently and iteratively against each of the best correlated 5 neighboring stations. Although this procedure involved intensive use of the data, it reduced the possible effects of temperature spatial variation originating from terrain complexity in
2.DATASETDEVELOPMENTANDDESCRIPTION 58 the study domain. In the following iterations homogeneity was tested again using the composite reference series, but applying more restricted criteria (e.g., r= > 0.9, number of observatories<= 10). This procedure mainly aimed to verify the regional consistency among nearby sites. At this stage, combining both multiple and composite reference series and testing all the time series several times was important to ensure that the final dataset is relatively free from any significant breaks. 2.6. Impacts of the adjustment protocol on trends and statistical properties of the series Evaluation of the impact of series adjustment is of great importance to determining the reliability of the final time series for different climate analyses such as trends and climate extremes. Given that homogenization procedures are more subjective and therefore can have adverse and complex influences on time series, many authors (e.g., Peterson et al., 1998; Vincent et al., 2002) highlighted the importance of identifying the possible impacts of inhomogeneity adjustment on climatic data. In this research, a set of techniques was used to account for such possible impacts. Herein, the main hypothesis was that the degree to which a station’s data reveals spatial and temporal coherence with their immediate surroundings can be a good indication of the reliability of the applied methodology. Therefore, the validity of the final dataset can be evaluated by examining spatial and temporal characteristics of the new dataset, as compared to the original dataset. All the assessment tests were applied to a set of 98 observatories covering the period from 1950 to 2006. The spatial distribution of these observatories is given in Figure 2.7.
2.DATASETDEVELOPMENTANDDESCRIPTION 59 Firstly, linear trends of maximum and minimum temperatures were calculated at seasonal and annual timescales before and after eliminating inhomogeneities. The main purpose was to assess the relative influence that the homogeneity procedure exerted on the trend assessment in terms of both the magnitude and sign (direction) of the trend. The significance of the trends was assessed using the nonparametric Spearman (Rho) test at a confidence level of 95% (p value < 0.05). The Spearman Rho statistic was computed, as follows: () )1( 6 12 1 2 − − −= ∑ = nn iR Rho n i i (2.3) where i R is the rank of ith observation in the time series and n is the length of the time series. Figure 2.7: Spatial distribution of temperature observatories from 1950 to 2006 (N=98).
2.DATASETDEVELOPMENTANDDESCRIPTION 60 This test is robust to outliers and does not assume prior probability distribution of the residuals. The slope was estimated using the ordinary least squares (OLS) fitting and expressed in ºC per decade. Seasonal averages were obtained from monthly data for each year and defined as winter (December-February; DJF), spring (March-May; MAM), summer (June-August; JJA), and autumn (September-November; SON). A comparison between the trends sign (direction) before and after corrections was conducted by means of the cross-tabulation analysis, which illustrated pairwise relationships between the categories of the trend signs (i.e., significant positive [+], significant negative [-], and statistically insignificant [N]). In this context, the pivot tables were constructed to represent the cross-categorized frequency data in a matrix format following the results of the trend assessment. To assess the degree of spatial dependence between the seasonal and annual trends before and after homogeneity correction, the semivariance models were computed for the magnitude of change. A detailed explanation of semivariance theory can be found in Webster and Oliver (2001). Semivariance analysis has widely been applied in ecology (e.g., Urban et al., 2000) and climatology (e.g., Vicente-Serrano et al., 2010) to detect scales of variability in spatial data. The semivariance describes the spatial variance as a function of distance between the observatories. The value of the semivariance decreases as the pair of data points is separated by a short distance, meanwhile it increases when the distance increases. In this study this geostatistic was employed to analyze variations in spatial structure of the trends before and after homogeneity correction. It can be expected that the semivariance will be lower after correcting inhomogeneities as a consequence of introducing a relatively free-of-error and homogenous series. Spatial semivariance was obtained, as:
2.DATASETDEVELOPMENTANDDESCRIPTION 61 (2.4) where y(h) is semivariance at lag distance h, N (h) is the number of station pairs that are separated by h distance, Z (si) and Z (si+h) are the values of Z variable in stations si and si + h. Secondly, adjustment of the series may alter the probability distribution of extreme temperature (e.g., frequency and intensity). Therefore, this study used two extreme temperature indices to assess possible impacts of adjustment on occurrence of extreme values. In practice, the annual count of warm and cold days before and after breaks adjustment was computed for each daily time series from 1950 to 2006. The warm (cold) day was simply defined as the day higher (lower) than the 90th (10th) percentile of the average maximum (minimum) temperature. The definition of extreme values based on the percentiles is objective, site-independent and facilitates direct comparisons between regions of diverse climates (Jones et al., 1999). However, it is worthwhile to indicate that the key question behind this analysis was to assess the way in which the adjustment procedure affected spatial continuity of the trends rather than the trends themselves. Therefore, the spatial continuity of the extreme temperature trends before and after homogenization was compared using semivariance of the magnitude of change. Thirdly, to determine how the adjustment procedure affected the statistical distribution of daily temperature values in the series, a suite of statistical indicators were computed for the time series before and after correction. In this regard, L-moment statistics were computed for each independent daily maximum and minimum time series before and [] ∑ = −−= )( 1 2 )()( )(2 1 )( hN i ii hsZsZ hN h γ
2.DATASETDEVELOPMENTANDDESCRIPTION 62 after homogeneity correction. These statistics provide information on the scale (Lcoefficient of variance [t2]), shape (L-coefficient of skewness [t3]), and peakedness of the series (L-coefficient of kurtosis [t4]). L-moment statistics are independent of the sampling size and also resistant to outliers. Moreover, they show less bias in comparison with other conventional product moments, such as standard deviation, skewness and kurtosis. These statistics have therefore been used in climatological and hydrological studies (e.g., Guttman, 1993; Guttman et al. 1993; Vicente Serrano et al., 2010). More details on the L-moment theory are given in Asquith (2003). In short, the L-moment coefficients are defined according to Hosking and Wallis (1997), as follows: 1 2 2 λ λ τ = (2.5) (2.6) 2 4 4 λ λ τ = (2.7) where ( 1 λ ,2 λ ,3 λ ,4 λ ) are the probability-weighted moments (PWMs) defined by Greenwood et al. (1979), as: 01 β λ = (2.8) 012 2 β β λ −= (2.9) 2 3 3 λ λ τ =
2.DATASETDEVELOPMENTANDDESCRIPTION 63 0123 66 β β β λ +−= (2.10) 01234 123020 β β β β λ − + −= (2.11) The PWMs of order r is given by: () i r N i ir xF N∑ = −= 1 1 1 β (2.12) where i F is the cumulative distribution of a given ordered sample ( 1 x<2 x<3 x<i x) calculated, as: N i Fi 35.0− = (2.13) Also, to get a closer picture of the spatial structure of L-moment coefficients, values of these statistics were also plotted against distance by means of the semivariance models before and after inhomogeneities correction. Lastly, while the statistical methodologies described above can provide a guidance to assessing the reliability of the final dataset, perhaps the most outstanding feature to assess the effect of the homogeneity procedure is to account for the correlation between the time series before and after correcting breaks. In this research an assessment of this effect was achieved by deriving the Pearson correlation matrix between temperature series at different distance orders. The correlation matrix was computed for both the original series (i.e. raw and corrected) and the series of first differences (Xt+1-Xt) for the raw and corrected data. Considering the correlation between the first difference series is of special interest to give insights into the strength of the temporal dependency among the series after isolating the influence of
2.DATASETDEVELOPMENTANDDESCRIPTION 64 inhomogeneities which may be introduced in the independent series. Overall, introducing inhomogeneities in the series is assumed to degrade correlation among nearby sites and, in this study, the inter-station correlation could act as a kind of filtering for non-adjusted series. 2.7. Homogeneity testing results Table 2.1 summarizes the percentage of flagged data following the quality control procedure. On average, the percentage of erroneous data was higher for minimum (0.11%) than for maximum temperature (0.09%). The highest errors for maximum and minimum temperatures occurred during spring and summer, respectively. Spatially, the fraction of flagged data was greater in station-dense areas (e.g., Catalonia and Cantabria), compared with relatively sparse areas (e.g., Burgos and La Rioja). One representative example for the erroronous data was the daily minimum temperature at the observatory of San Mateo (Castellon) which recorded -88.8ºC on the second day of January in 1960. Figure 2.8 also illustrates an example of one outlier detected in daily maximum temperature at the observatory of Pamplona on 23rd June 1983, in which the percentile of this daily value was 0.37, which differed largely from other neighboring observatories (0.75±0.82). This value was flagged and set as missing value. After quality control checks, a linear regression model was used to comprise nearby and best correlated series of short-term span in order to rebuild long-term time series. Figure 2.9 illustrates an example of the reconstruction procedure. As presented in Figure 2.9.A, the reconstructed wintertime daily minimum temperature time series at
2.DATASETDEVELOPMENTANDDESCRIPTION 65 the observatory of Calatorao Cooperative (9428E, Zaragoza) has records from 1940 to 2006 as a consequence of joining data from two nearby sites (9425I and 9432). Table 2.1: Summary statistics of the flagged data following the quality control procedure. The numbers refer to the fraction of the flagged data in daily maximum (Tmax) and minimum (Tmin) temperature series. The mean indicates the average of the flagged data for the whole dataset, while the lowest (highest) shows the maximum (minimum) percentage of flagged data at the station-based level. Winter Spring Summer Autumn Annual Tmax Lowest 0 0 0 0 0 Highest 0.47 0.97 0.55 0.6 1.99 Mean 0.01 0.04 0.02 0.02 0.09 Tmin Lowest 0 0 0 0 0 Highest 0.52 1.41 1.35 0.99 3.85 Mean 0.02 0.02 0.04 0.03 0.11 Figure 2.8: The percentiles of the daily summer temperature at the observatory of Pamplona (red line) and its nearby observatories. The percentiles were calculated for each day in the period from 1st June to 31st August in 1983 (total of 92 Julian days), relative to the entire daily records of each observatory.
2.DATASETDEVELOPMENTANDDESCRIPTION 66 Figure 2.9: The reconstructed wintertime minimum temperature series at the observatory of Calatorao Cooperative (9428E, Zaragoza) based on joining data from two nearby sites (9425I and 9432). Pearson correlation coefficient (r) indicates the correlation between the reconstructed time series and each of the nearby observatories. The green line denotes the final time series.
2.DATASETDEVELOPMENTANDDESCRIPTION 73 As shown, T statistic reached its critical value (95% significance level) around the year 1968. Nonetheless, the displacement from the mean disappeared after eliminating the detected inhomogeneity. Another example is presented in Figure 2.14, which corresponds to spring minimum temperature at Els Hostalets de Balenya (Barcelona). As illustrated, a statistically significant break was presented close to the year 1956. The differences between the candidate and the reference series and their t-test results indicate that T-value for the adjusted data was below the 95% confidence limit. Figure 2.13: Test results of the SNHT applied to summer maximum temperature series at Fuenterrabia aeropuerto, [Guipuzoca] (a) before and (b) after homogeneity corrections. Dashed lines indicate the 95% significance level. The test statistic (T) is plotted against the critical value.
2.DATASETDEVELOPMENTANDDESCRIPTION 74 Figure 2.14: Same as Figure 2.13, but for spring minimum temperature series at Els Hostalets de Balenya [Barcelona]. 2.8. Indicators on the reliability of the adjusted dataset In the following section, an evaluation of the reliability of the newly developed daily adjusted dataset is outlined. The aim was to assess the sensitivity of temperature series to the adjustment procedure in terms of any significant change in their statistical properities (e.g., the mean, variance, extremes and statistical distribution). 2.8.1. Impact of the homogeneity protocol on trends Table 2.4 presents the results of the cross-tabulation analysis applied to trends in the annual maximum and minimum temperature for each pair of stations (i.e. before and after homogeneity correction). In this section, construction of the pivot tables was only
2.DATASETDEVELOPMENTANDDESCRIPTION 75 restricted to the annual trends to give an indicative example of the cross-categorized frequency of the trends. In general, the results did not largely reflect considerable differences in the sign (direction) of the trends. As shown in Table 2.4, the cells lie off the diagonal of the pivot table (grey shading) implies that the corrections did not affect trends direction The oppositely directed trends were only evident in 25 (25.5 %) and 44 (44.9 %) of maximum and minimum temperature series, respectively. This clearly implies that in most cases the adjustment had no discernible effect on the direction of the gained trends. Perhaps this arises from the small correction factors applied to the majority of observatories at the monthly scale, which in turn disappear when aggregated to the annual timescale (refer to Figure 2.12). Another possible explanation can be linked to the fact that most of the detectable inhomogeneities were found in the 1970s and 1980s, a period which exhibited a remarkable warming trend worldwide (Jones and Moberg, 2003). Under this warming, the low magnitudes of correction factors failed to alter the direction of the observed variability in the series. Spatial differences between the trends in the raw and corrected annual temperature series are illustrated in Figure 2.15. It seems that the adjustments had a very local impact on trends, whereby observatories located along the Mediterranean coast experienced more warming after homogeneity correction. This is more apparent for minimum temperature. The picture is almost uniform since the trends of the series before and after adjustment showed coherent spatial structure of the signs. A quick inspection of the trend assessment results indicates that the most convincing impact of adjustments on trends was mainly related to changes in the magnitude rather than the sign (direction) of the trend.
2.DATASETDEVELOPMENTANDDESCRIPTION 76 Table 2.4: Results of the cross-tabulation analysis applied to trends in annual maximum and minimum temperature series from 1950 to 2006 before and after homogenization. Significance is assessed at the 95% level. Numbers between brackets indicate the fraction of observatories. After homogenization Before homogenization Maximum temperature Minimum temperature Positive Negative Insignificant Positive Negative Insignificant Positive 63 (64.3 %) 2 (2 %) 18 (18.4 %) 49 (50 %) 5 (5.1 %) 36 (36.8 %) Negative 0 (0 %) 0 (0 %) 0 (0 %) 0 (0 %) 0 (0 %) 0 (0 %) Insignificant 5 (5.1 %) 0 (0 %) 10 (10.2%) 1 (1 %) 2 (2 %) 5 (5.1 %) Total 98 98 Figure 2.15: Spatial distribution of the trends in the annual (a) maximum and (b) minimum temperature time series before and after homogeneity corrections. Trend calculation is based on the period 1950-2006 and statistical significance is assessed at the 95% level.
2.DATASETDEVELOPMENTANDDESCRIPTION 77 Figure 2.16 depicts scatter plots showing the differences in the magnitude of the linear trends (ºC decade-1) between the original and adjusted time series. For most seasons, the trends after correction were not linearly consistent with those before correction. The correction was generally in the range of -0.23 (summer minima) to 0.17 (autumn maxima). This suggests a considerable impact of the adjustment procedure on the slope of both seasonal and annual time series. One typical example corresponding to the observatory of Yebra de Basa (Huesca) is shown in Figure 2.17. The linear fit of the raw and homogenized series indicates that both series showed uptrend over the period from 1950 to 2006. Nevertheless, the tendency toward warmer conditions was weaker after adjusting the series (0.1ºC decade-1), compared with the series prior to correction (0.3ºC decade-1). Another example corresponding to the annual minimum temperature at the observatory of San Sebastian “Igueldo” [Guipuzcoa] confirmed the same finding as minimum temperature showed slightly more warming (0.22ºC decade-1) after adjusting the series (Figure 2.18). In the same sense, Figure 2.19 strongly suggests that the temporal behavior of the adjusted series as predicted by the semivariance models was spatially more dependent compared with the raw series. This higher spatial continuity was markedly apparent in all seasons for both maximum and minimum temperatures. Given that eliminating inhomogenities likely reduces signal-to-noise ratio in the time series, it can thus be expected that the homogeneity adjustment accounted for the improvement in the spatial continuity of the trends after correction.
2.DATASETDEVELOPMENTANDDESCRIPTION 78 Figure 2.16: Scatter plots of the magnitude of the trends (ºC decade-1), as derived from the seasonal and annual trend analysis for (a) maximum and (b) minimum temperature before and after homogeneity corrections. The trends were assessed for the period from 1950 to 2006.
2.DATASETDEVELOPMENTANDDESCRIPTION 79 Figure 2.17: Trends in wintertime minimum temperature at the observatory of Yebra de Basa (Huesca) (1950-2006) (a) before and (b) after homogeneity corrections. Gray line corresponds to a low-pass filter of 9-years. Figure 2.18: Same as Figure 2.17, but for annual-average minimum temperature at the observatory of San Sebastian “Igueldo” [Guipuzcoa].
2.DATASETDEVELOPMENTANDDESCRIPTION 80 2.8.2. Impact of the homogeneity protocol on extreme events The homogeneity procedure applied in this work may affect extreme temperatures in different ways including: frequency, intensity and persistence. Figure 2.20 compares the semivariance models of the trends in warm and cold days indices before and after homogeneity correction. A quick visual inspection of the semivariance models indicates that the spatial heterogeneity of the series markedly decreased after applying the homogeneity correction. Trends in extreme events did show high level of spatial consistency for the adjusted series, as neighboring observatories tended to have more identical patterns. In contrast, the raw series exhibited certain abrupt jumps over small distances. This implies that the variance of the raw data was higher, suggesting a weaker spatial component. By contrast, the high spatial continuity of the corrected data simply indicates that the spatial coherence among observatories was predominantly attributed to similar temporal evolution at short distances. It is worthwhile to indicate that the influence of adjustments on spatial dependency of cold days climatology was slightly less apparent compared with warm days (Figure 2.19). This can primarily be linked to the fact that minimum temperature during wintertime shows higher spatial variability compared with other seasons, as a consequence of the joint effect of strong circulation influences and high topographic-induced thermal contrasts. Overall, the high degree of spatial coherence in temperature extremes after correcting inhomogeneities provides a strong guidance on the reliability of the adjusted dataset.
2.DATASETDEVELOPMENTANDDESCRIPTION 81 Figure 2.19: Semivariance of the magnitude of the annual and seasonal trends for (a) maximum and (b) minimum temperature before and after homogeneity corrections.
2.DATASETDEVELOPMENTANDDESCRIPTION 82 Figure 2.20: Semivariance of the magnitude of trends in the annual count of (a) warm and (b) cold days calculated for the 57-year time series (1950-2006) before and after homogeneity corrections. 2.8.3. Impact of the homogeneity protocol on statistical properties of the series Figure 2.21 shows the relationships between L-moment statistics (i.e. variance, skewness and kurtosis) before and after homogenization. A comparison of the parent distribution of L-moments permits the assumption that the frequency distribution of temperature series before and after adjustment was generally coincided with a clear linear well-fit. This was evident for both maximum and minimum temperature time
2.DATASETDEVELOPMENTANDDESCRIPTION 89 distribution is expected given that higher elevations were not adequately represented by the data in the original dataset. Figure 2.26: Distribution of the final dataset according to the province from 1970 to 2006. Figure 2.27: The cumulative distribution of the number of observatories in the final dataset as a function of altitude from 1970 to 2006.
2.DATASETDEVELOPMENTANDDESCRIPTION 90 2.10. Summary In this work the procedure used to create the daily adjusted dataset for northeast Spain can be seen with confidence. This dataset passed through multiple steps to assure its quality, completeness and homogeneity. The original dataset was first quality controlled to remove anomalous and suspious data. This procedure was proved to be sensitive to trim outliers and keep only valuable extremes. A linear regression model was then undertaken to infill gaps in daily temperature series using information captured from the surrounding observatories. This regression model is simple, straightforward, applicable and robust when dealing with extreme values. Furthermore, it accounted for the dependency between temperature and topography gradients, particularly in areas of complex terrain. To account for possible breakpoints in the reconstructed time series, three well-established RHTs were used to test homogeneity of the series at monthly, seasonal and annual timescales. Afterwards, a monthly correction factor was calculated and interpolated to daily values to eliminate inhomogeneities from the daily series. A combination of the results of three homogeneity tests was advantageous because these tests had different sensitivities to defining discontinuities in the time series. Moreover, this approach helped determining not only strong breakpoints in temperature time series, but also small shifts. In this thesis, a great deal of effort being put into developing techniques to assess how the homogeneity methodology affected spatial and temporal characteristics of the final time series. For this reason, this work provided a suite of statistical tests for screening of different aspects in which the break correction can affect temperature series. By means of the spatial semivariance statistic and L-moment statistics, it was possible to look at a broad array of time series characteristics (e.g., means, extreme values and
2.DATASETDEVELOPMENTANDDESCRIPTION 91 frequency distribution). These methodologies were applied comparatively to the series before and after adjustment to realistically evaluate impacts of the break correction on the final dataset. In practice, the semivariance models were proven as efficient in describing and analyzing spatial structure of temperature means and extremes. Lmoment statistics also provided a consistent tool to assess changes in the statistical properities of the series (i.e., variance, kurtosis, and skewness). Following these techniques, it was clearly evident that the homogeneity adjustment significantly improved the spatial and temporal structure of both temperature means and extremes. Given that evaluation of the impact of homogeneity routines on final climatological products has not obtained much attention in the literature, these methodologies can offer a useful approach for the assessment of homogeneity impacts on climate time series. In general, these methodologies are objective, flexible, reproducible, and can therefore be applied in similar environments. To conclude, this newly compiled database comprises the most complete and homogenous time series of maximum and minimum temperature for northeastern Spain. From the spatial and temporal perspectives, this dataset is unique and represents an advance compared with previous datasets available for surface air temperature in the Iberian Peninsula. Considering the high spatial and temporal resolution of the final dataset, it can contribute to better understanding of space-time variability of temperature and its driving causes at both local and regional scales. Moreover, this dataset can be useful to understanding the causes and impacts of local and regional climate changes on hydrological systems, ecosystems, natural resources and human activities. In the study domain, this feature is of high importance due to its complex topography and diverse climates. Moreover, this dataset considerably
2.DATASETDEVELOPMENTANDDESCRIPTION 92 overlaps with some available precipitation datasets over the region (e.g., VicenteSerrano et al., 2009). For instance, from 1960 to 2006, 82.8% of temperature observatories have adjacent precipitation data. This sounds important as it permits a detailed regional analysis of the combined effect of temperature and precipitation on different disciplines in the region. Finally, this developed climatology can enhance the grid resolution of any climatic study in future with more potential to validate climate simulations from Regional Climate Models (RCMs).
CHAPTER THREE METHODOLOGICAL FRAMEWORK
3. METHODOLOGICAL FRAMEWORK 95 3. METHODOLOGICAL FRAMEWORK 3.1. Introduction The primary objective of this thesis was to determine quantitatively the observed and projected changes of surface air temperature in northeastern Spain and to assess the main drivers behind this variability. The following paragraphs will elaborate the methods used to achieve this goal in more detail. Unless mentioned, the analyses were based on daily maximum and minimum temperatures of a dataset of 128 observatories spanning the period from 1960 to 2006. The prerequisites for selecting these observatories for subsequent analyses include length, data completeness (i.e., no missing values), and quality (i.e., consistence and homogenous data). Also, as previously detailed, this dataset is free from the effects of non-climatic factors (e.g., urbanization, industrialization and changes in surrounding environments). This is critical to ensure that the observed changes in temperature across the study domain reflect natural variability and/or global warming rather than non-climatic processes. The (1960-2006) time interval was selected as a base period in this thesis due to the observed rise in the global temperature since the 1960s, as reported in many previous studies (e.g., Jones et al., 1999). Additionally, the dataset offered a reasonable spatial density of temperature observatories from 1960 onwards, which can adequately give insights into the main climate regimes across much of the study domain. Locations of the weather observatories belonging to the 1960-2006 period are illustrated in Figure 3.1. The analyses were conducted at both seasonal and annual timescales. Seasons were defined as: winter (December-February; DJF), spring (March-May; MAM), summer (June-August; JJA) and autumn (September-November; SON).
3. METHODOLOGICAL FRAMEWORK 96 3.2. Observed changes in seasonal and annual mean temperatures Seasonal and annual changes in maximum, minimum, and mean temperatures and DTR were assessed using the ordinary least squares (OLS) method, while the significance of the observed trends was tested at a confidence interval of 95% using the non-parametric Spearman Rho statistic. In case of normally distributed and serially independent climate variables (e.g., temperature), the OLS method often yields the maximum likelihood estimator of the regression coefficients. In the same context, the Spearman Rho is advantageous because it is robust to outliers and does not assume prior probability distribution of the residuals (Sneyers, 1990). The slope of the least squares regression was used to assess the magnitude of change and was expressed in units of ºC per decade. In this analysis, the trend was firstly calculated for a subset of 19 observatories that has data from 1920 to 2006. The availability of a reasonable number of temperature time series, which dates back to the earlier decades of the 20th century, motivates a detailed assessment of such long-term changes in temperature across the study domain. This kind of assessment has poorly been addressed in previous studies in the region in particular and over Iberia in general. The spatial distribution of temperature observatories covering the period from 1920 to 2006 is presented in Figure 3.2 and their main characteristics are provided in Table 3.1. The analysis was then extended to the denser network from 1960 to 2006 in order to obtain a more detailed regional assessment of temperature variability. In this work, 3000 Monte Carlo simulations were used following the procedure described by Kundzewick and Robson (2000). This procedure was applied to each season and annually. Numerous previous studies applied this technique to resampling climate data at random. This procedure aimed to ensure that the observed variability in climate data is
3. METHODOLOGICAL FRAMEWORK 97 real and not a statistical artifact originating from suspious variability (e.g., Lettenmaier at al., 1994; Adamowski and Bougadis, 2003). Figure 3.1: Spatial distribution of temperature observatories from 1960 to 2006. Figure 3.2: Spatial distribution of temperature observatories from 1920 to 2006.
3. METHODOLOGICAL FRAMEWORK 98 A low-path Gaussian filter was then applied to the original time series to smooth out the interannual variability and provide more robust trends by removing high-frequency fluctuations from the data (Sneyers, 1992). Table 3.1: List of the observatories used for trend analysis from 1920 to 2006. WMO CODE OBSERVATORY PROVINCE 0120 MOIA BARCELONA 0213 CARDEDEU BARCELONA 1006 SANTESTEBAN NAVARRA 1024E SAN SEBASTIAN 'IGUELDO' GUIPUZCOA 1035U AYA-LAURGAIN GUIPUZCOA 9174 SARTAGUDA NAVARRA 9198 CANFRANC LOS ARA/ONES HUESCA 9246 CARCASTILLO LA OLIVA NAVARRA 9269 ALSASUA NAVARRA 9281 FALCES NAVARRA 9283 CADREITA NAVARRA 9390 DAROCA OBSERVATORIO ZARAGOZA 9434 ZARAGOZA AEROPUERTO ZARAGOZA 9474 LA PEðA 'EMBALSE' HUESCA 9547 LA PUEBLA DE HIJAR TERUEL 9562 MORELLA CASTELLON 9657 ESTERRI D'ANEU LLEIDA 9900 NUENO HUESCA 9910 PALLARUELO DE MONEGROS HUESCA To reveal the overall picture of temperature variations in the study area as a whole, regional series of maximum, minimum, and mean temperatures and DTR were constructed on an annual basis and for all seasons. The regional series were built using two different manners. Over the period 1920-2006, a regional series based on the arithmetic average of the daily records from the 19 observatories belonging to this time interval was composed. This approach is simple, straightforward, and standard when dealing with a limited and a sparse station network. Recalling that the regional
4. RESULTS 201 Figure 4.25: Scatter plots of the Pearson correlation coefficients (r) calculated between the detrended anomalies of seasonal and annual maximum, minimum, and mean temperatures and DTR and the main large-scale atmospheric circulation. Dotted lines show the upper and lower limits of the 95% significance level. The median, 10th, 25th, 75th and the 90th percentiles as vertical boxes are plotted with errors bar.
4. RESULTS 202 However, given that most thermal variables (i.e., maximum, minimum and mean) showed more or less similar connections with atmospheric modes in terms of the correlation sign (direction), as revealed in Figure 4.25, the intention was only confined to explore the dependency between the leading atmospheric circulation modes in the region and mean temperature. The spatial structure of the relationship between seasonal mean temperature and atmospheric patterns is presented in Figures from 5.26 to 5.29. Figure 4.26 depicts the spatial patterns summarizing the response of winter mean temperature to atmospheric circulation. As illustrated, the connection between the EA pattern and mean temperature showed a broad and consistent spatial pattern, suggesting a large-scale signature of this pattern over the study domain. Nonetheless, this relationship was apparently stronger close to the Cantabrian Sea and in the northeastern portions of the study domain. By contrast, the influence of the MO, NAO, and SCA patterns on wintertime temperature means reached the significance threshold only at highly elevated areas in the Pyrenees and west of the Iberian system. As depicted, the EAWR and WeMO influences on mean temperature were markedly weaker in complex terrain areas during winter months. Also, the dependency between mean temperature and the EAWR and WeMO modes showed a NW-SE gradient. This relationship was markedly weaker to the west, particularly along the Cantabrian coastland, and to the northeast (Catalonia). On the other hand, the NAO influence on mean temperature variability seemed to have a clear south-north gradient.
4. RESULTS 203 Figure 4.26: Spatial distribution of Pearson correlation coefficients between standardized anomalies of winter mean temperature and the general atmospheric circulation. In some cases, the scale of the maps was set to improve their readability. However, the scale of the Pearson coefficient remained constant for all maps. Correlation coefficients above 0.28 were statistically significant at the 95% level.
4. RESULTS 204 Figure 4.27 indicates the spatial distribution of Pearson correlation coefficient between spring mean temperature and atmospheric patterns. The EA, SCA and WeMO modes were also the main drivers of mean temperature variability. Although the EA pattern had a broad spatial pattern across the study domain, the WeMO and SCA modes experienced a south-north gradient. In particular, the influence of these modes on springtime mean temperature was more pronounced northward. Also, areas close to the Mediterranean Sea indicated higher correlation with the SCA and MO, while they showed less and statistically insignificant correlation with the WeMO. As illustrated in Figure 4.28, the influence of the EA pattern on summer mean temperature was higher close to the Mediterranean Sea, particularly in Catalonia. Similarly, the Pearson correlation between summer mean temperature and the SCA mode was less evident close to the Cantabrian Sea, while it increased eastward and southward. The picture summarizing the role of the EAWR and MO modes only showed significant correlation in the Pyrenees and along western parts of the Iberian system. For the NAO, the spatial patterns were almost uniform with less spatial differences. Similar to spring, the WeMO experienced a south-north gradient, with more intense association in areas close to the Cantabrian Sea northwest. In comparison to other seasons, the spatial contrasts in the relationships between autumn mean temperature and the general circulation patterns were less apparent (Figure 4.29). Overall, the EAWR, MO, and NAO patterns showed a uniform pattern, with no marked differences.
4. RESULTS 205 Figure 4.27: Same as Figure 4.26, but for spring.
4. RESULTS 206 Figure 4.28: Same as Figure 4.26, but for summer.
4. RESULTS 207 Conversely, the relationship between the EA pattern and mean temperature was higher close to the Mediterranean Sea and the Cantabrian Sea, whereas it became weaker in mainland areas. Figure 4.29: Same as Figure 4.26, but for autumn.
4. RESULTS 208 Interestingly, the connection of autumn mean temperature with the SCA showed a clear dependency with elevation. In particular, the correlation was stronger at highly elevated localities in the north (the Pyrenees) and to the south and the west (the Iberian system); meanwhile it was less apparent in coastal and inland areas of low elevation. To explain in detail the SLP configurations beyond the observed spatial structure of the dependency between temperature and teleconnection indices, features of anomalous seal level pressure (SLP) corresponding to the key atmospheric circulation (herein, the EA+, WeMOand SCA-) were presented by means of the climate composite maps. These physical processes can be seen as a good indicator of air advection through land-sea interactions. The results on the configuration of mean SLP anomalies associated with each dominant pattern are presented in Figure 4.30. The major centers of atmospheric action and their seasonal variations in terms of both strength and position are also described. In winter, the interannual variability of temperature was mainly dependent on the positive EA. As illustrated in Figure 4.30, the positive EA is primarily linked to a north (low)-south (high) dipole of pressure anomalies, with a vast anticyclonic center over North Africa and the Mediterranean and a negative anomaly across northwestern Europe and west of the British Isles. This configuration would tend to bring anomalous S and SW warm air to the Iberian Peninsula, causing higher temperatures. As the study area is located downstream from the Atlantic Ocean, the Atlantic air advections bring relatively warmer air over the study domain.
4. RESULTS 209 Figure 4.30: (a) Average sea level pressure (SLP) anomalies (hPa) over the period 1960-2006 corresponding to the most significant atmospheric patterns. The anomalies were derived from the daily NCEP/NCAR reanalysis data provided by the NOAA/OAR/ESRL PSD (http://www.nws.noaa.gov) for the base period 1960-2006.
4. RESULTS 210 While the positive phase of the NAO significantly correlated with maximum temperature during winter, it merely correlated with mean temperature in terrain complex areas in the north and the west (see Figure 4.26). Overall, as depicted in Figure 4.30, the positive phase of the NAO is largely associated with an increase in zonal circulation over Western Europe. This feature was clearly evident where the isobars had zonal direction. Figure 4.30 also indicates that the positive EA during springtime has two main anomaly centers, located over the Sahara and NW Europe. Pressure anomalies during this mode reveal that the study area is particularly influenced by dry and warm air flows originating from the positive pressure anomaly over North Africa and moving toward the Iberian Peninsula. On the other hand, the anomalous pressure associated with the negative SCA during spring is mainly characterized by the zonal circulation, where the Atlantic blocks move eastward to cover broad areas of the Iberian Peninsula and Western Europe. This configuration often corresponds to very weak Icelandic lows, enhancing the advection of the westerlies from the Atlantic Ocean. The westerly and southwesterly flows affect much of the western areas of the study domain, explaining the stronger correlation found in these areas. In the years of negative WeMO, a distinct dipole in SLP anomaly is observable, with an anticyclonic anomaly placed over northern Europe and a negative anomaly located throughout North Africa and the Azores. Throughout this configuration, the Azores High moves northward covering broad areas of the British Isles and northwestern Europe. In short, spring is mainly characterized by the northerly displacement of the subtropical high corresponding to the positive EA. This positive anomaly is located over vast areas of the
4. RESULTS 217 temperature variations during spring and summer than in winter and autumn. These feedbacks were generally negative assuming that an increase (decrease) in temperature anomalies often corresponds to a negative (positive) anomaly of soil moisture and cloud cover. Winter was the only season in which both cloudiness and soil moisture experienced positive feedback with temperature. Second, the influence of the examined land surface-atmosphere forces had a diurnal cycle, with more influence during daytime than nighttime. These diurnal differences were more apparent during spring with regard to other seasons. Figure 4.32: The positive feedback between cloud cover and wintertime minimum temperature in the study domain from 1960 to 2006. Standardized anomalies of both fields are provided from 1960 to 2006.
4. RESULTS 218 4.4.2. Influence of large-scale atmospheric circulation on temperature extremes 4.4.2.1. Influence on moderate extreme events In this section, the influence of the atmospheric circulation patterns on temperature extremes was assessed based on the spatial regionalization of summertime extreme events defined in section 5.3.1 and plotted in Figure 4.20 (CL1 to CL4). Figure 4.33 depicts Pearson correlation values between the general atmospheric circulation and the regional time series of extreme temperature indices in the period 1960-2006. This association was calculated for the regionally weighted time series computed for the delineated four clusters. Similar to mean temperature conditions, the connection between the circulation patterns and temperature extremes was found statistically significant (p<0.05) only for the EA+, SCAand WeMOpatterns. Among them, the WeMO represents an east-west dipole; meanwhile the EA and SCA are north-south dipoles. On the other hand, the NAO was found a weak predictor of temperature extremes during summer season. Similarly, the role of the MO pattern was irrelevant in the region. For the established homogenous sub-regions, it seems that the SCA pattern is a key controller of temperature extremes, with correlation coefficients ranging between -0.20 and -0.64. However, it can be noted that the influence of the SCA on day-time extremes (e.g., max_summer, WD and TX90p) was much stronger than its influence on night-time extremes (e.g., min_summer, TXn, TNn and TN90p). In particular, correlation coefficients with day-time extremes were in the
4. RESULTS 219 range of -0.41 and -0.54, whereas they varied from -0.29 to -0.49 for night-time extremes. Figure 4.33: Pearson correlation coefficients (r) between detrended time series of summer extremes and the main modes of atmospheric circulation over the period 1960-2006. Dotted lines show the upper and lower limits of the 95% level of significance.
4. RESULTS 220 A quick view of the association between the SCA circulation pattern and temperature extremes at sub-regional scale reveals some significant spatial differences (Figure 4.33). The influence of the negative SCA on most of summer temperature extremes was generally less marked in the Mediterranean region (CL1) and, in contrast, much stronger in the highly elevated regions (CL3 and CL4). A clear example corresponding to DTR showed a significant correlation with the negative SCA in all the defined sub-regions, but with higher values in CL3 (r = -0.60) and CL4 (r = -0.56), compared with CL1 (r = -0.43) and CL2 (r = - 0.38). This spatial pattern likely resembles that of the Spell index, which exhibited the strongest relationship with the SCA negative phase in CL4 (r = -0.64) and CL3 (r = -0.56), relative to CL1(r = -0.41) and CL2(r = -0.52). It can thus be inferred that the influence of SCA on extreme temperature in the region is clearly elevation dependent, with more (less) influence at high (low) elevation sites. The correlation of summer temperature extremes with the positive EA was generally positive although it did not necessarily reach the statistical significance threshold, as being the case with the WD, INTR, SU25 and Spell indices. In contrast to the SCA pattern, the strongest association between the EA positive phase and temperature extremes was markedly apparent along the Mediterranean coast (CL1) for the majority of the indices. Contrarily, this influence declined in mainland and over complex terrain sub-regions. Similar to the SCA pattern, the WeMO correlated better with maximum temperature indices than with minimum temperature indices. Spatially, it can be noted that the impacts of the WeMO pattern on temperature extremes were more pronounced in continental and low elevated areas than in coastal and highly elevated regions.
4. RESULTS 221 In summary, among all circulation patterns, the most significant during summer season were the EA+, WeMOand SCA-. Figure 4.34 indicates the temporal evolution of the atmospheric circulation patterns from 1960 to 2006. As depicted, a significant positive trend was exclusively exhibited for the EA index, meanwhile a statistically downward trend with remarkable decadal variability was found for the EAWR, MOI and SCA patterns. In particular, the EA experienced a negative phase until the end of the 1970s, followed by a noticeable rise since the mid of the 1990s. By contrast, a decline in the MOI, WeMO and SCA patterns was observed since 1980. This figure also indicates that the NAO showed a negative trend from the mid of the 1980s, which comes in direct contrast with the strong signal of temperature during this period. 4.4.2.1.1. SLP configurations In order to better understand the mechanisms of atmospheric flows and land-sea interactions during summer season, the SLP configurations corresponding to the significant circulation patterns in the region (i.e., EA+, WeMOand SCA-) were summarized by means of the composite climate analysis and the canonical correlation functions. For each dominant circulation mode, the composite maps were plotted based on SLP anomalies corresponding to summers of positive or negative values from 1960 to 2006, depending on the phase under investigation (i.e., positive or negative). This dependency is depicted in Figures from 5.35 to 5.37. SLP canonical variates are plotted in left panels, whereas their corresponding temperature variates are illustrated in right panels. The sign of canonical correlation coefficients for the SLP and temperature anomaly indicates the direction of the co-variability (i.e., correlated (+) or anticorrelated (-)).
4. RESULTS 222 Figure 4.35 (panel A) shows the averaged anomalies of SLP over western and southwestern Europe during summers with positive EA values. As illustrated, the positive mode of the EA pattern is mainly associated with a strong dipole over the North Atlantic (approximately 50ºN, 30ºW) and dominance of anticyclones over the Mediterranean and central and Eastern Europe. This situation also corresponds to an increase in the anticyclonic activity over the Iberian Peninsula, while the Azores High extends northward. Figure 4.34: Temporal evolution of the large-scale atmospheric circulation patterns during summers of the period 1960-2006. Bold lines refer to a low-pass Gaussian filter of 9 years.
4. RESULTS 223 Under this SLP configuration, the advection of the westerly and southwesterly winds over the Iberian Peninsula is enhanced, while the easterly flows transport moisture from the Mediterranean along a SE-NW gradient. Correspondingly, there is a weak moisture transport from the north Atlantic during this configuration, which induces a decrease in precipitation over the study domain. The anomalous low precipitation demonstrates positive feedback with soil moisture anomalies, which in turn has a positive feedback with temperature anomaly. The impacts of the westerly, southwesterly and easterly flows on the study domain are largely constrained by local terrain. For example, the central portions (e.g., the Ebro valley) are weakly affected by these flows as a consequence of the surrounded orographic lifting (i.e. the Iberian and Cantabrian systems). Contrarily, the EA exerts more important influence along the Mediterranean coast, particularly with the increase in the warm and moist flows from the Mediterranean and the mid of the Atlantic. Owing to the weak contrast between SLP over the Mediterranean Sea and closing land areas during the positive EA as revealed by SLP isopleths, the influences of these inflows can not extend to mainland, particularly with prevalence of high pressure anomalies in most of the peninsula as a consequence of local heating effects (see Figure 4.35, panel A). In short, the EA positive phase exerted a remarkable influence on the increase in frequency and intensity of summer temperature extremes along the Mediterranean coast and high elevation sites. Figure 4.35 (panels B and C) illustrates respectively the first and second canonical functions that explain the large proportion of variability in both SLP and temperature during the positive EA mode.
4. RESULTS 224 Figure 4.35: (A) Composite of sea level pressure (SLP) anomalies (hPa) corresponding to summers with positive EA values from 1960 to 2006. SLP anomalies at each grid were computed using the monthly mean and standard deviation for the long-term period (1960–2006). (B) The first leading canonical function of SLP and temperature anomalies co-variability during summers of the positive EA mode, (C) the same as (B) but for the second leading function. Isopleths show the Pearson correlation coefficient. Coefficients above 0.23 are statistically significant at the 95% level.
4. RESULTS 225 As shown, the canonical correlation coefficients of the first and second functions were 0.78 and 0.60, respectively. Spatially, it was found that temperature over central and western areas of the study domain are mainly controlled by SLP anomaly over the Mediterranean region, particularly the western basin (r = -0.7), whereas it is anticorrelated with SLP anomalies over the northern Atlantic. The second function suggests that temperature variations over the eastern portions of the study area were significantly influenced by SLP anomaly over central Europe and northern Africa, while they correlated negatively with SLP over the eastern Atlantic (panel C). Taken together, it can be noted that above-normal temperatures during summers of the positive EA values are mainly associated with a positive SLP anomaly over the Mediterranean, central Europe and North Africa on the one hand and a negative SLP anomaly placed over the east Atlantic on the other hand. Figure 4.36 (panel A) illustrates the composite of SLP anomaly during summers of negative SCA values from 1960 to 2006. As depicted, the negative phase of the SCA pattern produced similar configuration to the positive EA, with clear high pressure anomaly over central Europe. This high pressure anomaly extends westward to cover the Iberian Peninsula and southward to include parts of the Mediterranean region and North Africa. On the other hand, a low pressure system predominates over the north Atlantic and vast areas of northern Europe and Scandinavia. This atmospheric circulation indicates that the advection of northern flows over the peninsula is largely restricted during the negative mode of SCA, as a consequence of dominance of
4. RESULTS 226 anticyclonic conditions over the European mainland, giving rise to increased above-normal temperature. This situation comes clearly in contrast with SLP features during summers of positive SCA values, where the anticyclonic anomalies predominate over the Scandinavian region favoring for the advection of northern cold-air flows toward the Iberian Peninsula. Figure 4.36: Same as Figure 4.35, but for summers with negative SCA values.
4. RESULTS 233 Figure 4.39: Same as Figure 4.38, but for VCN.
4. RESULTS 234 Correspondingly, the positive anomaly showed a strong gradient over the Northern Atlantic and Europe (Figure 4.39, panel B). This configuration was also visible at the surface level and low and mid troposphere, enforcing the meridional circulation over the study domain with strong advection of northerly and northeasterly cooler air flows. To explore air advections controlling very heat temperature variations in the region, the connection between the spatial variability of VWD/VCN and the main centers of large-scale atmospheric circulation patterns was analyzed by means of the canonical correlation. For each canonical pair (variate), the results are presented in two maps: one for the dependent variable (i.e., anomaly of atmospheric fields) and the latter for independent variables (i.e., temperature anomaly during these VWD/VCN). This pair of maps provides an explanation of the physical reasoning of VWD/VCN canonical variate by identifying the main centers of action of circulation patterns that significantly correlate with the spatial modes of VCN/VWD over the study domain. 4.4.2.2.1. Mean Sea Level (MSL) Figure 4.40 depicts three surface synoptic patterns that explained regional variability of VWD over northeast Spain. The first leading function (canonical r = 0.7) revealed strong and statistically significant positive correlation between anomalously higher temperature over the western areas of the study domain and SLP regime over the Iberian Peninsula near to the Cantabrian Sea. The second and third functions suggested that the behavior of VWD temperature along the coastal regions was mainly driven by SLP anomaly over the mid latitudes of the
4. RESULTS 235 Atlantic Ocean and the Western Mediterranean. As illustrated, it can be seen that positive temperature anomalies along the Cantabrian Sea were well correlated with pressure anomaly west of the Mediterranean basin (the second function), while the intensity and frequency of VWD in the Mediterranean observatories were strongly driven by the positive SLP anomalies over North Africa and Central and Northern Europe (the third function). An interesting note is that the influence of increasing convection as a consequence of local heating effects, as revealed in the first function, was broadly consistent across the entire domain. This pattern was monopole with positive correlation with temperature anomalies in all sites although the highest significance occurred close to the Cantabrian system in the west and the lowest were found near to the Mediterranean Sea. On the other hand, the second and third functions suggested more spatial contrasts, indicating that more frequent VWD on the Cantabrian Sea always correspond to relatively fewer number of VWD on the Mediterranean Sea and vice versa. Figure 4.41 represents three main synoptic conditions that explained subregional variability of VCN at sea level. All these patterns imply that the anomaly pattern over the North Atlantic region was the main controller of the occurrence of VCN over northeast Spain. The first function (canonical r = 0.65) indicated a generally negative correlation between SLP north of the British Isles and magnitude of temperature during VCN over the study area. In particular, high (low) pressure anomaly north of Ireland implies more (less) severe cold nights in the whole domain, with few spatial differences (Figure 4.41, panel A).
4. RESULTS 236 Importantly, the regional response of VCN to SLP anomaly over the Atlantic waters was largely dependent on the location of the Atlantic SLP center action. The first function revealed the strong influence of the Arctic flows when the highs are located over the British Isles. Figure 4.40: Pearson correlation coefficient between anomalies of SLP (left) and temperature anomalies (right) during the VWD across the study domain (coefficients above 0.15 are statistically significant at the 95% confidence level).
4. RESULTS 237 Figure 4.41: Same as Figure 4.40, but for VCN. The second function suggested less (more) intense VCN over the elevated areas to the west and the Pyrenean mountains northward (r > 0.4) when high (low) anomaly was placed near to the Iberian Peninsula. The third function suggested the possible influence of the Azores high that can extend further to the Eastern Mediterranean during winter months. The displacement of the Azores assumes that temperature intensity in the whole domain, apart from Catalonia in northeast,
4. RESULTS 238 was negatively influenced demonstrating warmer conditions and less frequent VCN. 4.4.2.2.2. 200hPa Level Figure 4.42 suggests three canonical functions, with correlation coefficients varying from 0.14 to 0.64. The first function demonstrated that the positive 200hPa anomaly over the Iberian Peninsula was highly correlated with VWD temperature in northeast Spain. High anomaly at this geopotential induces stability at this upper level causing above-normal temperature and in turn clear skies. This synoptic pattern fitted well with the first function detected at the surface level (Figure 4.40). The very high correlation (highest values above - 0.9) ensured that, during summer months, very high temperature over the western parts of the study area significantly relate to high geopotential anomaly over the Cantabrian Sea. For temperature variate, the highest values of correlation coefficient (larger than 0.8) occurred over the western portions. Contrarily, low pressure anomalies over the peninsula imply less frequent and intense warm days in the western part of the study domain. The second function indicated that the most extreme warm days along the Mediterranean were mainly connected to the pressure anomaly over the British Isles (r > 0.6, p<0.05) and the Western Mediterranean (r > 0.4, p<0.05) although the direction was reversed. In particular, high anomaly over the western Mediterranean basin was positively correlated with more intense over the Mediterranean observatories in the region. Also, a negative (positive) anomaly across northern Europe and the British isles assumed more (less) intensity of VWD close to the Mediterranean coast. These anomalies bring warm and moist air from the high anomaly over the eastern
4. RESULTS 239 Mediterranean and North Africa to the peninsula, causing above-normal temperature in coastal observatories. This dependency is reversed in mainland areas (i.e., the Ebro valley) as the correlation became negative, but statistically insignificant. The third function characterized the possible influences of the warm masses from the Sahara and North Africa. The intensive warm days over the study domain were associated, to large extent, with the strong advection of warm and very dry air originating over North Africa. The correlation coefficients on temperature variate suggested a markedly SE-NW gradient, with more (less) intense warm days in south and east (north and west). Figure 4.43 illustrates the main leading 200hPa synoptic patterns that explain VCN spatial variations during the winter season. Overall, the first function (canonical r = 0.73) indicated that VCN minimum temperature anomaly in vast areas of the study domain was positively correlated with 200hPa anomaly over the peninsula. Low anomaly assumes more frequent and intense cold nights. This situation also highlights the strong influence of the Icelandic High as they cause anomalous advection of cold winds from the anticyclones near to the British Islands and Scandinavia to the deep cyclones over the Iberian Peninsula. As shown in Figure 4.43, while the first and second canonical functions highlight the possible influence of meridional circulation as the cooler air flows moved from the north (Islands High) to southern Europe and the Mediterranean, the third function mainly featured the influence of the zonal circulation. The anomalous high (low) over the north Atlantic around (55ºN, 25ºW) and the anomalous low (high) in the Central Europe suggested more influence of the zonal circulation with western and northwestern (eastern and northeastern) advections over
4. RESULTS 240 southern and southwestern Europe. The enhanced (weaker) westerlies cause less (more) frequent VCN in northeast Spain. Figure 4.42: Pearson correlation coefficient between anomalies of 200hPa geopotential height field (left) and temperature anomalies (right) during the VWD across the study domain (coefficients above 0.15 are statistically significant at the 95% confidence level).
4. RESULTS 241 Figure 4.43: same as Figure 4.42, but for VCN. 4.4.2.2.3. 500hPa Level Figure 4.44 reveals three statistically significant canonical functions (at 95% confidence level). The canonical correlation coefficients ranged from 0.31 to 0.72. Similar to SLP and 200hPa patterns, the first function (r = 0.77) indicates a statistically significant positive correlation between 500hPa geopotential anomaly north to the peninsula and high temperature anomalies across the whole domain. This suggested that VWD were almost entirely controlled by the activity of
4. RESULTS 242 anticyclones formed over the Peninsula in general and the Cantabrian Sea in particular. Also, the severity of warm temperature extremes increased westward over the study domain. The intensity of temperature on this function was suggested to be enhanced by the Azores anticyclones, which induce a strong heat and humid advection from the Atlantic. The second function (r = 0.45) highlighted the influence of high anomaly of 500hPa over the western Mediterranean and North Africa on intense temperature in the Mediterranean portions of the study domain. The correlation coefficient reached its highest values (r = - 0.6) in the central Ebro valley. Contrarily, the 500hPa positive anomaly in the western Mediterranean was anticorrelated with intensity of VWD close to the Cantabrian Sea. In accordance of the first function explaining 200hPa-VCN co-variability, the first function, as illustrated in Figure 4.45, indicates that a deep depression placed over the Iberian Peninsula is favoring for more intense cold nights as it encourages strong advection of northerly cooler flows from the north (Figure 4.45). The results also suggested a statistically significant positive relationship between 500hPa anomaly over the Atlantic Ocean west to the Iberian Peninsula and cold extreme days in the western windward areas of the study domain. Figure 4.45 obviously reveals that cooler air masses over the study domain were strongly linked to low anomaly over the mid latitudes of the Atlantic Ocean, suggesting advections from cooler continental areas in Central Europe. Conversely, this function also implies that zonal advection from the Atlantic was dominated by high humidity advection enhanced by strong anomaly of the westerlies, which was favoring for less intense cold nights.
4. RESULTS 249 ensembles were also reasonably good at reproducing the observed symmetry of seasonal minimum temperature series, as revealed by the YK statistic. Figure 4.47: Same as Figure 4.46, but for minimum temperature.
4. RESULTS 250 Through the PCA, the ICTP, KNMI and METO ensembles also simulated well the spatial patterns of seasonal trends, implying that they can provide a reasonable skill in analyzing changes and variability of the simulated data in the future. Overall, according to the validation results for simulated minimum temperature data, the ICTP, KNMI and METO ensembles correctly reproduced the main characteristics of the observed climate (e.g., the mean, variance, skewness and symmetry). Accordingly, the models with the best agreement with the observed station data were selected. Herein, the decision was made to select the ICTP, KNMI and METO, as being the models with the overall best performance, to simulate future changes in minimum temperature. As presented in Table 3.4, this group of models is driven by three different GCMs (Arpège, ECHAM5 and HadCM3Q16). Herein, inclusion of more than a single model is advantageous to estimate a wider range of possible responses of the climatic system to elevated GHG emissions. In the attempt to assess changes in extreme events in the future, an additional pair of statistics was also used to assess the sensitivity of the models to capture the anomalous extremes events (i.e., VCN and VWD). In this context, the coefficient of variance (CV) and changes in the magnitude of the 1st (99th) percentiles for the modelled and observed data were computed for the 9 different models. This assessment was important to test the ability of the models to produce the values of the anomalous temperature as well as the variance of the series. It is noteworthy indicating that these measures were only calculated for summer (MJJA) and winter (NDJF) seasons. The results are presented in Figure 4.48.
4. RESULTS 251 For summer (left panel), the results reveal a good performance of the MPI and KNMI models. The 99th percentile was well modeled by these ensembles although they were a bit underestimated by -1.19ºC and -1.16ºC, respectively. Similarly, the ratio of coefficient of variance (CV), calculated for the modelled and observed data gave good results for the MPI and KNMI simulations with values generally close to 1. Figure 4.48: Comparison between (upper) differences in the magnitude of the 99th and 1st percentiles, and (lower) changes in the coefficient of variance. Red line shows the average, while the black line indicates the median. Left (right) panels belong to summer (winter). For winter (right panel), the models with the best performance in terms of the ratio of the coefficient of variance were the METO, ICTP and KNMI. These models were also able to capture the 1st percentile magnitude of wintertime
4. RESULTS 252 minimum temperature, with differences of 0.2, -0.7 and 0.1ºC, respectively. Interestingly, these biases were markedly smaller compared with differences in the magnitude of the 99th percentile for daily maximum temperature during summer season. Interestingly, this was also the case for extreme maximum temperatures in highly elevated areas with cooler temperatures. Following the validation results, two different RCMs were used to simulate VWD (MPI and KNMI), while three models were used to project changes in VCN (METO, ICTP and KNMI). 4.5.2. Future changes in temperature means After having confidence in model performance for the control period (1971-2000), the projected future changes in the seasonal maximum and minimum temperatures for the time slices 2021-2050 and 2071–2100 were assessed, relative to the control period. Figure 4.49 shows the seasonal maximum and minimum temperature anomalies for the 20th and 21st century over the whole region. For each individual simulation, the anomalies were first computed relative to the 30-year observed (1971-2000) long-term mean for the region grid points. In order to detect temperature change signal for each season under the A1B emission scenario, an inter-model anomaly was then calculated as the average of the anomalies of the selected simulations. Figure 4.49 informs that both mean maximum and minimum temperatures exhibited positive anomaly with regard to the 1971-2000 period. This anomaly was much stronger during the last decades of the 21st century. However, this warming had some seasonal differences.
4. RESULTS 253 Figure 4.49: The observed and simulated time series of areally averaged seasonal temperature from 1971 to 2100, presented as bars. A 7-yrs low pass filter is calculated and mapped as solid lines. The anomalies are calculated relative to the control period (1971-2000) for each season.
4. RESULTS 254 From 2021 to 2050, the predicted warming during the summer and winter periods, under the A1B emission scenario, were larger than the transition seasons (i.e., spring and autumn). The projected increases were in the order of 2, 1.1, 0.9 and 0.6ºC for summer, winter, autumn and spring maximum temperatures, respectively. Similarly, minimum temperature became warmer, but at rates faster than those of maximum temperature. In particular, the average minimum temperatures were warmer than observations by approximately 0.5– 2.3°C, depending on the season. The highest increase occurred during summer (2.3ºC) and winter (2.1ºC), relative to autumn (1.1ºC) and spring (0.5ºC). Similarly, the projected changes of maximum and minimum temperatures over the period from 2071 to 2100 showed rapid shift toward more warming conditions relative to both the baseline (1971–2000) and near future (2021–2050) simulations. This trend was evident for all seasons, with rather similar rates of increase for maximum and minimum temperatures. As presented, there was a high likelihood of increase in maximum temperature by 5.1, 2.8, 2.6 and 2.1ºC during summer, autumn, spring and winter, compared with a simulated warming of about 5, 2.9, 2.2 and 3.9ºC for minimum temperature, respectively. The spatial structure associated with changes in the maximum and minimum temperature anomalies are given in Figure 4.50. These patterns are presented for the 2021-2050 and 2071-2100 time slices, based on computing an intermodel average for the best-validated simulations for each season ( as defined previously in section 5.5.1). As depicted in Figure 4.50 (panel A), the main spatial pattern corresponding to maximum temperature was the high positive anomaly over the central Ebro valley, with higher increases in winter and summer. In the
4. RESULTS 255 Ebro valley, the projection for the A1B emission scenario over the 21st century was that maximum air temperature may increase in the range of 3-4ºC and 1-2ºC during summer and winter respectively compared to the present climate. From 2021 to 2050, all solutions also produced a slight negative anomaly over parts of the Pyrenees, which had an opposite sign to the entire region in all seasons, with a maximum reduction in cold seasons (winter [2-3°C] and spring [1-2ºC]). These changes were less pronounced during warmer seasons (i.e., summer and autumn). However, a projected increase in the temperature maxima across the Pyrenees is expected by the end of the century, particularly during summer. In particular, the RCMs predicted an increase in the summer maximum temperature by 3.5ºC to 5ºC over the region by the end of the century. Another notable feature was that the warming pattern exhibited a meridional gradient with somewhat higher values in the south than in the north. In areas close to the Cantabrian Sea, maximum temperature was generally lower relative to the continental and southern portions by roughly 2ºC in winter and 3ºC in summer. Figure 4.50 (panel B) illustrates the spatial distribution of minimum temperature changes during the 21st century with regard to the present climate. As depicted, a positive anomaly prevailed during winter and summer from 2021 to 2050, particularly over continental grids. Correspondingly, a slight increase was experienced during spring and autumn. Similar to maximum temperature, the Pyrenees was expected to exhibit a slightly negative anomaly during spring and autumn. From 2071 to 2100, temperature anomalies showed a steady increase on an east-west gradient. While the Cantabrian system to the west and southwest exhibited the strongest positive anomaly (above 4.5 and 6ºC in winter
4. RESULTS 256 and summer, respectively), a relatively less warming occurred close to the Mediterranean and the Cantabrian Sea. Figure 4.50: Spatial distribution of seasonal maximum and minimum temperature anomalies (ºC) over the study domain. The anomaly was defined as departures from the observed long-term mean (1971-2000). Figure 4.51 reveals the Gaussian probability distributions of the winter and summer maximum temperatures calculated from the Pyrenean grid cells with altitude above 1000 m. The probability distribution had a near normal distribution during the period from 1971 to 2000, with a mean value of 6.8ºC for winter and
4. RESULTS 257 22.2ºC for summer. On the other hand, those of the future periods were quite diverse but generally very elongated, with significant shifts toward the positive tail. The only exception corresponds to the winter temperature from 2021 to 2050, which exhibited a slight decrease in the mean by 0.3ºC relative to the observed climate. Figure 4.51: Gaussian distribution for (upper) wintertime and (lower) summertime maximum temperatures for the 1971-2000, 2021-2050 and 2071-2100 periods for the Pyrenees region. The inter-model average was calculated as an average of temperature from the grids with altitude above 1000 m in this region. Numbers between brackets indicate the mean value corresponding to each bell curve.
4. RESULTS 258 Conversely, there was a general consistency about high likelihood of increase in winter and summer maximum temperatures by the end of the century. This warming was more pronounced during summer with an increase rate of 4.7ºC, compared with the present climate. In summer, the temperature distribution was flatter, suggesting that temperature values are more disperse from the mean, a situation which favors for more warm events. This feature was enhanced by the increase in the standard deviation during summer, which came in contrast with maximum temperature during winter. 4.5.3. Future changes in temperature standard deviation Figure 4.52 shows the spatial distribution of changes in the standard deviation of the seasonal mean maximum and minimum temperatures across the region over the 21st century. For each grid box, changes in the standard deviation were simply calculated as the difference between the standard deviation of the future simulations (2021-2050 and 2071-2100) and the reference period (1971–2000). A quick inspection of Figure 4.52 clearly reveals that, in accordance with changes in the mean, standard deviation presented an overall increase during the 21st century, except for the winter season. Also, this warming was much higher at the end of the century. The projected changes were larger in summer and spring than in the winter and autumn. The only exception corresponded to the autumn maximum temperature as the standard deviation during the first half (2021-2050) of the century was larger than that of the latter half (2071-2100). Averaged over the whole region, the models simulated an increase in the standard deviation of summer maximum temperatures from 2071 to 2100 by 0.5ºC (proportional to
4. RESULTS 265 the end of the century corresponded to a decrease in the standard deviation of wintertime temperature. 4.5.4.2. Spring season (MAM) Results on changes in the time-varying percentiles during spring season (MAM) from 1971 to 2100 are summarized in Figure 4.54. A visual inspection reveals two main findings. First, the percentiles steadily showed an increase in their magnitude, indicating that the current warming will continue during the 21st century. In contrast to the winter season, rates of warming were similar for the two comparable periods: 2021-2050 and 2071-2100 for both maximum and minimum temperatures, with negligible differences. In particular, changes in maximum temperature percentiles varied from 0.66 to 3.5ºC during the course of the 21st century, which corresponded to an upward of 1.1 to 3.4ºC for minimum temperature percentiles. One interesting note is that the warming rates of the time-varying percentiles during spring were markedly much higher than those of the mean. A representative example is the 75th and 90th percentiles from 2021 to 2050 which moved towards more extremes, with increases of 1.8 and 1.9ºC, respectively, while mean maximum temperature roughly warmed by 0.6ºC, relative to the observed climate. This finding, together with the strong increase of the standard deviation during spring, implies that high-intensity temperature will get warmer at rates exceeding that of the mean. Second, there were remarkable differences between the interannual variability of the percentiles during the observed and the simulated climate.
4. RESULTS 266 While the percentiles of spring season showed strong warming trend from 1971 to 2000, they exhibited less variability during the 21st century, especially for the cold percentiles. Interestingly, the upper percentiles (75th and 90th) of maximum temperature exhibited a downward tendency from 2021 to 2050, at linear rates of -0.23 and -0.32ºC decade-1, respectively. Accordingly, it can again be noted that, in the future, the increase in the magnitude of temperature percentile does not necessarily correspond to more interannual variations. Figure 4.54: Same as Figure 4.53, but for spring.
4. RESULTS 267 4.5.4.3. Summer season (JJA) Changes in the magnitude, standard deviation and interannual variability of the time-varying percentiles during summer season are given in Figure 4.55. Similar to the wintertime, the magnitude of the percentiles exhibited a “systematic” future increase for both maximum and minimum temperatures. Also, changes in the percentiles corresponding to minimum temperature were faster than those of maximum temperature. With regard to the current climate, the magnitudes of the 10th, 25th, 75th and 90th percentiles increased by 1.4, 1.3, 1.8 and 1.8ºC for maximum temperature from 2021 to 2050, while their changes had the order of 2.4, 2, 2.3 and 2.5ºC for minimum temperature, respectively. This picture was markedly reversed at the latter half of the 21st century, as the percentiles of maximum temperature increased faster than those of minimum temperature. In contrast to wintertime temperature, Figure 4.55 (panel A) informs that changes in the upper percentiles exceeding those of the lower percentiles. One clear example is the change in the 90th percentile from 2071 to 2100, which corresponded to a rise of 5.9 and 5.3ºC for maximum and minimum temperatures, respectively, compared with 4.3 and 4.8ºC for the 10th percentile. Similarly, the time-varying percentiles of summer experienced a clear increase in their standard deviation, relative to the control period, which also came in contrast to the winter season. This has been the case for both maximum and minimum temperatures although the increasing rates were much stronger for the cold tail 10th and 25th percentiles.
4. RESULTS 268 Figure 4.55: Same as Figure 4.53, but for summer.
4. RESULTS 269 For the period 2071-2100, the results suggested a rapid increase in the standard deviation of the 10th (0.7ºC) and 25th (0.6ºC) percentiles. In response to the projected increase in the magnitude and standard deviation of the time-varying percentiles, the summer season was expected to exhibit a remarkable increase in the temperature interannual variability over the 21st century. This feature was clearly observed at the late of the century, particularly for the cold tail percentiles (i.e., the 10th and 25th). All the trends were positive and statistically significant (at the 99% level), with magnitudes of change in the range of 0.5-1.7ºC decade-1 for the 10th percentile and 0.5-1.4ºC decade-1 for the 25th percentile. 4.5.4.4. Autumn season (SON) Figure 4.56 depicts changes in the magnitude, standard deviation and trends of the time-varying percentiles calculated for autumn maximum and minimum temperatures during the 20th and 21st century. In accordance with other seasons, the results also suggest increase in the magnitude of all percentiles during the 21st century, being more highlighted at the end of the century. Similar to spring season and also in contrast to summer and winter, there were no significant differences between changes in the upper and lower percentiles either for maximum or minimum temperature. On the other hand, while standard deviation of maximum temperature time-varying percentiles showed a markedly rapid warming with respect to the 1971-2000 period, there were no changes in the standard deviation of minimum temperature percentiles. The main modes of variability characterizing autumn time-varying percentiles were the markedly increase (decrease) in the interannual variations of upper (lower) percentiles of maximum (minimum) temperature.
4. RESULTS 270 Figure 4.56: Same as Figure 4.53, but for autumn.
4. RESULTS 271 4.5.5. Future changes in extreme events Figure 4.57 shows changes in the 99th percentile magnitude over the periods (2021-2050) and (2071-2100) as compared to the control period (1971-2000). Under the A1B emission scenario, all models agree that the 99th percentile magnitude will be reduced by an average of -1.6ºC and -1ºC following the KNMI and MPI simulations during the period 2021-2050 respectively. Nonetheless, Figure 4.57 also depicts clear spatial differences as grid points along the Mediterranean coast and across the Ebro valley exhibited positive differences in the 99th percentile values (nearly by 2-3ºC), suggesting warmer conditions than the observed climate. Interestingly, other regions demonstrated a decrease in the magnitude of the 99th percentile. Interestingly, this spatial structure was reflected by both the MPI and KNMI models. Although these extremely severe warm events are characterized by high natural variability, the similarity between the models in terms of the magnitude, sign and spatial variability of simulated changes confirms the skill of these models in responding a similar parameterization scheme. During the period 2071-2100, this magnitude will markedly be more intensified, with an average increase of 3 and 4.2ºC for the KNMI and MPI ensembles, respectively. These substantial projected increases in the 99th percentile temperature magnitudes strongly indicate that more heat stress could dramatically be intensified across the region during the last decades of this century, which could have considerable impacts on different disciplines in the study domain, such as hydrology, agriculture, water resources management, forestry, and human health in .
4. RESULTS 272 From the spatial perspective, the regional differences among the models from 2071 to 2100 showed similar patterns. The distribution of the 99th percentile changes seemed to be elevation dependent, with rapid increase at lowest elevations (e.g., the Mediterranean coasts and the Ebro valley) and weak warming at high elevation sites to the west (i.e., the Cantabrian system) and southwest (i.e., the Iberian system). Figure 4.57: Change (prediction minus current climate) in the magnitude of the 99th percentile daily maximum air temperature during summer. Figure 4.58 depicts changes in summer frequency of VWD during the periods 2021-2050 and 2071-2100, as compared with the control integration. As illustrated in the upper panel, a generalized decreasing in the number of VWD is
4. RESULTS 273 detected for the near future (2021-2050). This cooling of VWD was more pronounced along the Cantabrian, the Mediterranean and at elevated sites in the west, whereas it was less evident in lowlands and inland. A similar pattern was also observable during the period 2071-2100 (lower panel), but with rapid increase in the frequency of VWD near the coast, while there was a relatively weaker warming in continental areas. Simulated changes in the magnitude of the 1st percentile of wintertime minimum temperature, calculated as difference between this magnitude for the 30 yr minimum temperatures in baseline (1971-2000) and future scenario periods (2021-2050) and (2071-2100) are presented in Figure 4.59. Analogous to the expected warmer climate in future, changes in the 1st percentile magnitude tended to increase gradually over the two time slices: 2021-2050 and 2071-2100 under the A1B emission scenario. This result suggests that the VCN will become substantially less severe in future as the cold tail of daily minimum temperature shifts toward warmer conditions. However, this increase slightly occurred over 2021-2050 with an average of 0.16, 0.29 and 0.43ºC for the ICTP, KNMI and METO ensembles, respectively. Contrarily, the projected warming of the 1st percentile will sharply be larger over the last three decades of the 21th century, with an average of 1.94, 2.4 and 3.3ºC for the ICTP, KNMI and METO ensembles, respectively. For the two time slices under investigation, spatial variations in the magnitude of the 1st percentile appeared robust and similar between models, with more warming in the west and close to the Cantabrian Sea and less increase and, even a decrease, in the eastern parts and over the northern Pyrenees.
4. RESULTS 274 Figure 4.58: Same as Figure 4.57, but for changes in the number of days per each summer in which the maximum temperature exceeded the 99th percentile of summertime daily distributions. Changes in the frequency of VCN are given in Figure 4.60. The observed changes fitted well with spatial changes in the magnitude of the 1st percentile. Given that the last period of this century will be warmer than the near future decades, a less frequent and intense VCN will be expected by the end of the century. The results suggest that the most elevated sites to the west will exhibit less severe cold conditions in the two simulated periods than eastern portions, particularly those close to the Mediterranean Sea.
4. RESULTS 281 For winter, Figure 4.62 suggests that the interannual variability of the 75th and 90th time series of minimum temperature from 2021 to 2050 can be expected to decrease as the mean winter gets warmer. This clearly implies that an increase in wintertime minimum temperature corresponds to a decrease in the variability of the 75th and 90th time over the period 2021-2050. For summer season, a comparison of Figures 5.61 and 5.62 indicates that the 10th and 25th percentile correlated weaker for minimum temperature than for maximum temperature, while the correlation of the 75th and 90th increased more with minimum temperature than with maximum temperature. This gives indication that extreme events corresponding to warmer minimum temperature would increase more rapid than those of maximum temperatures. Contrarily, changes in the 10th percentile of winter minimum temperature were more consistent with changes in the mean, while it poorly responded to changes in the mean for the 10th percentile of maximum temperature. Moving to spring season, it can be seen that changes in upper tail of the minimum temperature distribution tended to be significantly driven by changes in the mean, with more shift toward warmer values by the end of the century. This feature remarkably agrees well with the pattern detected for the observed temperature. In contrast to other seasons, the warm percentiles of minimum temperature in autumn show a slight negative correlation with the mean trends over the study domain from 2021 to 2050. This association was statistically insignificant (p<0.05), with r values of -0.03 and -0.05 for the 75th and 90th percentiles, respectively. Also, it seems that the relationship between changes in autumn
4. RESULTS 282 minimum temperature and changes in the warm tail of daily minimum temperature was markedly much stronger than changes in the cold tail. In general, this relationship was found consistent over the 20th and 21st century. 4.5.7. The dependency between changes in the mean temperature and the frequency of extreme events This research also looked at changes in the relationship between the mean (hereafter: M) temperature and standard deviation (hereafter: S) in the one hand and the frequency (hereafter: F) of extreme events on the other hand. This was mainly to explore whether changes in the frequency of anomalously severe heat events (i.e., VCN and VWD) over the study domain are more linked to changes in the occurrence of these days (as revealed by changes in the standard deviation) and/or to changes in the intensity (as revealed by changes in the mean) of temperature during these days. For each grid, the association between the frequency of these days (F+/F-) and changes in (i) the mean (M+/M-) (ii) standard deviation (S+/S-) and (iii) combined effects of changes in the mean and standard deviation (e.g., M+/S+, M+/S-, M-/S- , M+/S-) were examined for each ensemble. For instance, this assessment helps exploring whether an increase in the mean (M+) and/or the standard deviation (S+) could induce an increase in the frequency (F+) of these extreme events. The results are summarized in Figure 4.63. The findings suggested that intense severe warm events during the period 2071-2100 can be explained by the combined effect of positive changes in both the standard deviation and the mean of summer maximum temperature.
4. RESULTS 283 Figure 4.63: Relationships between simulated changes in the frequency of extreme events (labeled as F+ for increase in the occurrence and Ffor the decrease) and changes (i.e., increase [+] vs. a decrease [-]) in the mean (M) and/or the standard deviation (S).
4. RESULTS 284 The positive change in the mean (M+) was spatially consistent with the increase in the occurrence (frequency) of warm events (F+) in more than 80% of the grids across the study area. The same finding was observed for S+ with F+ and also for M+S+ with F+. However, this effect was markedly minimized during the 20212050 period, with less frequent VWD, due to the weak change in the standard deviation of summer maximum temperature. The results also suggest that the decrease in the standard deviation had more influence on VWD when compared with the impact of the decrease in the mean. For VCN, on the other hand, this dependency was more stable with a steady decrease in the standard deviation in both time slices (2021-2050) and (20712100) and an increase in the minimum mean temperature in spite of its slow rates in the first decades of the 21st century. Accordingly, it can be demonstrated that the shift toward higher minimum temperatures in the next decades cannot be the solely factor responsible for the observed decrease in the frequency of very cold nights across the study domain.
CHAPTERFIVE DISCUSSION
5. DISCUSSION 287 5. DISCUSSION 5.1. Observed changes in seasonal temperature means 5.1.1. Temperature long-term trends (1920-2006) The results concerning the observed trend in the mean temperature from 1920 to 2006 reveal that the study domain exhibited an increase during the 20th century. This warming was much more pronounced during the last few decades, with the largest increase occurred during summer and spring periods. The observed trend in temperature from 1920 to 2006 has notably been consistent with earlier findings in the Iberian Peninsula (e.g., Quereda et al., 2000), the Mediterranean (e.g., Brunetti et al., 2006) and Europe (e.g., Parry, 2000). For instance, according to the present work, the annual mean temperature increased at a rate of 0.96ºC from 1920 to 2006. In the Mediterranean region, Parry (2000), for example, reported that the annual mean temperature over Europe has risen at a rate of 0.80ºC over the 20th century. Similarly, Brunetti et al. (2006) recorded an increase of 1ºC in the Italian mean temperature during the last century. Over the Iberian Peninsula, Quereda et al. (2000) noted that the annual mean temperature in the Spanish Mediterranean region has increased at a rate of 0.71ºC in the last century. Similarly, over the period from 1920 to 2006, the study domain experienced an increasing trend of 0.7ºC in the mean maximum temperature, which has also been analogous to the finding of Karl et al. (1993) and Easterling et al. (1997) for the globe (0.82ºC/century). Nonetheless, although the annual minimum temperature showed stronger rise (1.22ºC) from 1920 to 2006, compared with both maximum and mean temperature, this trend is still lower than the global trend reported by Easterling et al. (1997) (1.79ºC/century).
5. DISCUSSION 288 This work also indicates that the temporal evolution of temperature over the study domain showed a remarkable interdecadal variability for all analyzed thermal variables (i.e., maximum, minimum and mean temperatures), with main departure toward cooling during the 1920s, 1930s and 1960s. Contrarily, temperature exhibited a general positive anomaly during the 1940s, 1970s, 1980s, 1990s and 2000s. This interdecadal variations of seasonal temperature generally fit well with other regional and global studies (e.g., Tett et al., 1999; Rodriguez-Puebla et al., 2001a). Notably, there has been a clear tendency toward warming at both the annual and seasonal timescales since the mid of the 1970s. This general behavior also agrees well with numerous previous regional (e.g., Labajo et al., 1998; Brunet et al., 2002; EstebanParra et al., 2003; Morales et al., 2005; Brunet et al., 2007a) and global studies (e.g., Parker et al., 1994; Tett et al., 1999; Jones and Moberg, 2003; Luterbacher et al., 2004). According to Brunet et al. (2002), a clear warming has been mainly registered over the peninsular Spain in two periods: during the 1940s and from the mid-1970s onwards. Other regional studies (e.g., Labajo et al., 1998; Morales et al., 2005) also reported a remarkable change in the behavior of the annual maximum, minimum, and mean temperatures after the year 1972. In accordance with the IPCC (2007), the unusual warmest years over the whole period (1920-2006) were restricted to the past two decades. Over this period, the study area also witnessed anomalous warming temperatures in 1990, 1997, 1998, 2003, and 2006. Herein, it is also worthwhile to indicate that the network covering the period 1920-2006 was not dense enough, as being represented only by a sample of 19 observatories. Nonetheless, the final conclusion on the observed trends can still be seen with high level of confidence. The derived results seem to clearly delineate similar temporal
5. DISCUSSION 289 patterns to those observed for Iberia, Europe and the globe. In addition, given that the observatories employed in this study are not only located in rural sites and small towns, but also include urban cities, the “global” warming observed in much of the study domain implies that this trend is a consequence of natural variability reflecting global warming conditions. This sounds interesting since it confirms the assumption that the observed temperature variations in the region are only due to climatic processes rather than other non-climatic factors such as urbanization or industrialization. Also, given that the long-term changes of temperature are not well captured in many studies over Europe in general and in the Iberian Peninsula in particular due to lack of reliable data in the early decades of the past century, these results can considerably contribute to improving the current knowledge on long-term temperature variability and change within a larger spatial context, including the Iberian Peninsula, the Mediterranean and the Western Europe. 5.1.2. Observed seasonal and annual temperature trends (1960-2006) 5.1.2.1. Maximum temperature The linear trend analysis of maximum temperature suggests a warming trend in all seasons. The strongest signal occurred during summer and spring. The only exception is seen in autumn since 43% of observatories exhibited a negative tendency at the 95% level. This finding is of special interest given that earlier studies (e.g., Houghton et al., 2001; Klein Tank et al. 2005; Martinez et al., 2010) also reported a downward trend in autumn temperature in Spain and wide regions across Europe, being the sole exception among all seasons. For example, Martinez et al. (2010) reported a generalized decreasing in autumn maximum temperature in Catalonia (NE Spain).
5. DISCUSSION 290 In general, there is a conclusive evidence on the increase in maximum temperature in NE Spain, which is comparable to previous works in the Iberian Peninsula (e.g., Quereda et al., 2000; Esteban-Parra et al., 2003; Morales et al., 2005; Miro et al., 2006; Rodriguez-Puebla et al., 2010; Martinez et al. 2010), the Mediterranean (e.g., Maheras and Kutiel, 1999; Brunetti et al., 2000, 2006) and Europe (e.g., Jones et al., 1999a). For instance, Martinez et al. (2010) indicated a general warming trend in the annual maximum temperature in Catalonia (NE Spain) with a value of 0.5ºC decade-1, the greatest increase occurred during summer and spring. Spatially, it is noted that the largest warming occurred in coastal areas along the Mediterranean and the Cantabrian Seas, whereas mainland observatories experienced less warming. However, this general picture still has some seasonal differences. An interesting aspect is the dominance of a statistically insignificant pattern of trends in the Ebro basin during autumn. A possible explanation of this pattern is probably linked to the location of this semi-closed basin between the Pyrenees northward, the Catalan system to the east and the Iberian system to the south and southwest. This feature is favorable for below-normal temperature as a consequence of the frequent occurrence of fog, particularly during wind-calm days corresponding to anticyclonic conditions. Also, during summer and spring seasons, there is a noticeable south-north gradient of warming, with the strongest warming in southern part of the study domain. This feature can largely be attributed to the strong advection of warm and dry air masses from the Sahara during these periods of the year. Another important spatial feature is that the spatial distribution of the trends in annual temperature seems to be spatially more consistent with the distribution of trends in summer and spring than in winter and
5. DISCUSSION 297 studies also assessed the impact of climate change on temperature extremes (e.g., Klein Tank and Können, 2003; Kostopoulou and Jones; 2005; Diffenbaugh et al., 2005; Hertig et al., 2010). Among them, Klein Tank and Können (2003) found a significant positive trend in the warm tails of the European daily temperature over the second half of the 20th century. The same finding has recently been confirmed by Kostopoulou and Jones (2005) for the eastern Mediterranean and by Politano (2008) and Hertig et al. (2010) for the Western Mediterranean. For the Iberian Peninsula there has been few limited number of studies that examined the behavior of warm extremes (e.g., Burgueño et al., 2002; Miro et al., 2006; Brunet et al., 2007b; Della-Marta et al., 2007a,b). Among them, Brunet et al. (2007b) gave evidence on larger changes in high temperature extremes in Spain from 1955 to 2006. Similarly, Della-Marta et al. (2007b) pointed out that temperature of the Western Europe, including Iberia, has become more extreme, with the increase being more confined to summer. At the regional scale, Miro et al. (2006) reported a significant increase in the frequency of warm and extreme temperature days in Valencia from 1958 to 2003. The increase in the frequency and severity of warm temperature extremes can mainly be linked to the rapid warming in maximum temperature compared with minimum temperature. In their study covering the whole Spain, Brunet et al. (2007b) found higher rates of change in maximum temperature rather than in minimum temperature over the period 1850-2005 (0.11º versus 0.08ºC decade-1). In the study domain, this finding has been confirmed, suggesting that the most remarkable warming in mean temperature from 1960 to 2006 occurred during spring (0.66ºC decade-1) and summer (0.41ºC decade-1). Warm extremes are more likely to occur during these periods of the year.
5. DISCUSSION 298 Spatially, the trends in warm extremes were more pronounced in the coastal areas along the Mediterranean Sea and the Cantabrian Sea, whereas less warming was evident in mainland areas. The roughly contrasting coastal-continental pattern of warm extremes can be due to the prolonged strength of sea surface temperature (SST). Recently, many studies reported strong relationship between temperature extremes and SST. For example, Vincent et al. (2011) found strong influence of SST on variability of warm extremes in the countries of the Western Indian Ocean, with correlation coefficient ranging between 0.6 and 0.87. Black and Sutton (2006) also linked the 2003 European heat wave with variations in SST anomalies of both the Mediterranean Sea and the Indian Ocean. However, this warming can also be attributed to the strong influence of the general large-scale atmospheric circulation. Rodriguez-Puebla et al. (2010) attributed much of the warming in warm days (WD) across the Iberian Peninsula to changes in the Scandinavian pattern and the 500 hPa geopotential height field over the North Atlantic. Similarly, Politano (2008) linked the 1998 unprecedented warm winter in the Mediterranean region with the anomalous geopotential height at 500hPa and 200hPa levels. Fischer et al. (2007) also suggested the anticyclonic activity as a key driver of warm events in Europe. Cassou and Philips (2005) found strong association between the 1994 warm summer over France and the Atlantic blockings. Altogether, these results strongly suggest the possible linkage between warm extremes during summertime over the domain and large-scale atmospheric circulation and geopotential heights. Given the notable increase in the frequency and severity of warm temperature events, which could have significant implications for hydrology,
5. DISCUSSION 299 ecology and agriculture, assessing this kind of research could provide a concrete base to better understanding of the spatial and temporal variability of warm extremes at this sub-regional scale. In the same context, the results of this work show more warming in warm extremes close to the coasts. Thus, it is also important to assess the impact of closing water bodies on warm extremes by means of simulations from different atmosphere-ocean climate models. 5.2.2. Changes in cold extremes The Mann-Kendall results confirm that most of the cold temperature extremes related to the frequency and intensity showed a decreasing but statistically insignificant trend (p<0.05). The only exception corresponds to the annual high minimum temperature (TNx), which showed upward trend of 0.3ºC decade-1 in the whole domain. This uptrend can be linked to the increase in the mean minimum temperature during the summer season, which has been confirmed in other regional (e.g., Zhang et al., 2005) and global studies (e.g., Frich et al., 2002). In the study domain maximum temperature has increased faster than minimum temperature over the whole period 1960-2006, which suggests lower variance in the cold tail of temperature distribution. A number of studies worldwide confirmed this little change in cold extremes. For example, New et al. (2006) pointed out that indices related to the minimum temperature for southern Africa showed less warming compared with those of maximum temperature. In central and south Asia, Klein Tank et al. (2006) reported that most cold temperature indices exhibited significant warming in the period from 1961 to 2000. Also, Moberg et al. (2006) gave consistent conclusions in their study of trends in daily temperature extremes in Europe. In Iberia, Brunet et al. (2007b) reported a decrease in cold temperature extremes across much of Spain during the last few decades. In this
5. DISCUSSION 300 context, it is worthwhile to indicate that the occurrence of cold temperature extremes showed a clear decadal variation in the study area. Cold events were more frequent during the 1960s and 1970s. This situation has been reversed from the late 1970s onwards, due to the rapid warming in minimum temperature in the last two decades. This observation agrees well with the results of Bermejo and Ancell (2009), who reported a significant increase in minimum temperature across Spain after 1980. Accordingly, it is important to seek out the driving forces that can describe the interdecadal variability of cold extremes. Physical factors such as atmospheric circulation and sea surface temperature could be among other forces. 5.2.3. Changes in variability extremes Similar to cold extremes, trends in variability extremes were less prevalent since insignificant trends were noted for majority of indices, with the exception of temperature sums (Tsums). For instance, the growing season length (GSL) significantly increased in only 14.8% of observatories. This finding is in line with Alexander et al. (2006) who indicated that around 16.8% of land observatories worldwide have experienced significant positive trends in the growing season length (GSL). In the study area, the less temporal variability of most of the indices can largely be explained by the inconsistence changes in maximum and minimum temperature during the past few decades. The evolution of the diurnal temperature range (DTR) is a clear example that summarizes the asymmetric evolution of both maximum and minimum temperature. As noted by many studies (e.g., Dai et al., 1997; Easterling et al., 2000), the globe has experienced a general negative trend in diurnal temperature range (DTR) that is largely a consequence of the rapid increase in minimum temperatures rather than maximum temperatures. In this work, the trends in DTR were
5. DISCUSSION 301 mixed between positive (49.2%) and negative (50.8%), being statistically insignificant in 71.9% of all observatories. This behavior is mostly due to the rapid warming of maximum temperature in recent decades, which is inconsistent with changes in minimum temperature over the same period. An inspection of DTR variability from 1960 to 2006 clearly revealed that DTR decreased during the 1960s, which was then reversed from the 1970s, suggesting less marked changes over the whole period (1960-2006). The evolution of the intra-annual extreme temperature range (Intr) index can also be explained by the asymmetric evolution of very warm days (TX99p) and very cold days (TN1p) indices. Although these two indices summarize the evolution of the most extreme events in terms of their frequency, they can, to some extent, provide a good indication on the variability in Intra-annual extreme temperature range (Intr). The relationship between trends in both TX99p and TN1p is given in Figure 5.2, which are not linearly well-fitted (r = 0.01). This weak dependence between the two indices can be seen in the context that the frequency of very warm days (TX99P) showed a strong uptrend; more highlighted during the last few decades (refer to Figure 5.11). In contrast, the decrease in the frequency of very cold nights (TN1p) did not occur at the same rate (see Figure 5.13). Overall, a comparison of the trends in warm and cold temperature extremes indicates that the impact of climate change on temperature is mainly accompanied by higher shift in the warm tail rather than in the cold tail. Trends in warm extremes are of greater magnitudes than cold extremes. These findings are compatible with previous works (e.g., Klein Tank and Können 2003; Kostopoulou and Jones 2005; Alexander et
5. DISCUSSION 302 al., 2006; Moberg et al., 2006). In the study domain the major changes in temperature extremes are focused on frequency and intensity of warm extremes. This finding fits well with the results of Klein Tank and Können (2003) for Europe, Politano (2008) for the Mediterranean, Beniston (2009) for Switzerland and Brunet et al. (2007b) for the peninsular Spain. The results demonstrate that changes in cold extremes were largely related to changes in the magnitude (e.g., CN, TNx, and TNn) rather than changes in the frequency of cold events (e.g., TX10p, TN10p, ID0, and FD0). By contrast, changes in warm extremes were due to the combined change of frequency and magnitude. Taken together, it can be inferred that the study area seems to be more sensitive to global warming during the warmer periods of the year, while it shows less sensitivity to this warming during the cold periods. Figure 5.2: Scatter plot of the relationship between the observed trends in TX99p and TN1p.
5. DISCUSSION 303 5.3. Spatial regionalization of temperature extreme events In this thesis, an attempt was made to divide the study domain into sub-regions as homogenous as possible based on characteristics of summer extreme events (e.g., frequency, intensity and persistence). The motivation for this classification was the high spatial variability of climate in the region due to its complex topography. The summer season was selected for this regionalization because the present study reported that the mean summer surface air temperature in northeastern Spain has increased by about 1.9ºC since 1960, with a warming rate of about 0.41°C decade-1, which represents the strongest signal among all seasons. This can be clearly seen, at a broader scale, in the numerous summers with the anomalously record-breaking warm events in the recent decades over the Mediterranean and Europe (e.g., 1998, 2003, 2005 and 2010). These unusual events caused various drastic impacts on both the physical (e.g., agriculture, ecology, forest fire, and hydrology) and human environments (e.g., mortality and energy demand). In this research, the multivariate statistics, including the principal component analysis (PCA) and cluster analysis (CA), were employed to divide the study domain into relatively homogenous climate regions. This kind or regionalization helps assessing the driving forces beyond the detectable spatial modes. Although the procedure followed to obtain homogenous regions can be seen as arbitrary and user defined given that the selection of clustering algorithm and the number of clusters was subjectively defined, the obtained results can be seen as satisfactory for many reasons. First, this study followed the standard procedure applied by many previous studies (e.g., Baeryswil and Rebetez, 1997; Romero et al., 1999; Papadimas et al., 2011) to obtain climatic homogenous regions. In particular, the climatic data were first
5. DISCUSSION 304 summarized by means of factor analysis to reduce data dimensions. Second, the scores of the retained factors were examined by cluster analysis to delineate the final homogenous regions. Also, given that the clustering procedure is unsupervised as the number of clusters is defined objectively, it was important to verify its goodness of fit. For this reason, the obtained classification was validated by means of the Silhouette width index to ensure its stability and goodness. Through this two-step statistical procedure, this work intended to use multiple statistics (i) to define the number of retained factors, (ii) to select the appropriate clustering algorithm, (iii) to detect the number of retained clusters, and (iv) to validate the clustering outputs. Combining the results from different statistics is advantageous to check for consistency between various statistics and in turn ensure the reliability of the findings. According to this scheme, four sub-regions were indentified: the Mediterranean region, the Cantabrian region and the inland region, the moderately elevated areas, and the highly elevated areas. These sub-regions have clear climatic and geographical meanings, with relatively clear physiographic boundaries. According to previous knowledge, these defined sub-regions match well with the dominant climate regimes over the study domain. Given that the regionalization of extremes is challengeable due to very rarity of these events, relative to regionalization of the mean values, the obtained sub-regions are reasonably found homogenous and well separated. Interestingly, although the network density in high-elevation sites is generally irregular compared with other data-rich regions (only 14.4 % of the observatories are located above 1000 m), the clustering procedure, as being distinctly identified in CL4, skillfully captured the variability of temperature extremes at these highly elevated and scattered localities. This can probably be explained by the free-air advection at the mountains
5. DISCUSSION 305 summits and along the free-drainage slopes. Pepin and Lundquist (2008) confirmed this finding for elevated sites with annual 0º isotherm across the globe, suggesting that inter-site variance of temperature at those sites is expected to be lower than moderate and low elevation sites, as a consequence of the weak influence of local factors near surface such as land use changes. A detailed assessment of the linear trends in extreme temperature indices in the established sub-regions indicates some important findings. Overall, there is a general tendency toward warming in temperature extremes for all sub-regions. However, this warming has a spatial component. For summer cold and warm temperature indices, the strongest signals were found in the most elevated areas (CL4) and along the Mediterranean (CL1). This probably suggests that orography and distance to the Mediterranean Sea play a key role in the temporal evolution of summer extremes in NE Spain. Accordingly, it can be assumed that changes in the summer extreme temperature are relatively complex and thus behave disproportionately over space as the climate warms. In this small area of complex orography, this finding implies that changes in these extremes are governed by different physical and dynamical considerations within the climate system. The temporal evolution of summer temperature extremes close to the Mediterranean Sea (CL1) coincides with the observed changes found across many Mediterranean areas (e.g., Frich et al., 2002; Klein Tank and Können, 2003; Kostopoulou and Jones, 2005; Brunet et al., 2005; Hertig et al., 2010). For example, Klein Tank and Können (2003) found upward trend in warm temperature extremes over the Mediterranean region from 1976 to 1999. Also, Kostopoulou and Jones (2005) found that summer
5. DISCUSSION 306 was the season of the most significant increase in maximum temperature extremes in the eastern Mediterranean. For the western Mediterranean, Hertig et al. (2010) recently observed a strong warming trend in summer maximum temperature, more intense over the Iberian Peninsula. Brunet et al. (2006) also reported rapid increases in warm days over Spain since 1973, more apparent close to the coasts. The same finding has also been previously confirmed by Brunet et al. (2007b) who assessed variability of extreme temperatures in Spain, providing evidence on larger changes in warm temperature extremes during the 20th century, as compared with cool extremes. Given that the Mediterranean is a close basin, temperature variability at coastal sites seems to be closely linked to Sea Surface Temperature (SST) variations. Santoleri et al. (1994) found an increase of 1.5ºC in mean SST across the western Mediterranean, mostly faster during summer and winter compared with spring and autumn. Xoplaki et al. (2003a) also noted a significant warming trend of the Mediterranean Sea Surface Temperature (SST) west of 20°E over the period 1950-1999, while the eastern Mediterranean basin experienced cooling. More recently, Salat and Pascual (2007) showed a similar upwarding trend in SST along the Catalan coast (NW the Mediterranean). The results also confirm that high mountain areas are more vulnerable to the global warming, compared with lowlands. These mountain environments are more likely to be affected by climate change and therefore they can be an early indicator of climate variability and change for the nearby low elevated areas. Accordingly, these results can be of particular importance in the context of the possible impacts of the global climatic change on behavior of temperature extremes in areas of complex topography. At the global scale, few studies also have been undertaken to assess elevation
5. DISCUSSION 313 Figure 5.4: (A) Seasonal and annual temporal variations of mean SLP (hPa) over the study domain and, (B) differences in SLP in the two sub-periods 1960-1980 and 19812006, as indicator of blocking behavior. All calculations were made based on the averaged values of the 2.5º by 2.5º grid resolution covering he study domain. In right panels, changes in the mean values are represented by the vertical lines.
5. DISCUSSION 314 In addition, the persistence of anticyclones enhances the advection of air flows from one direction for consecutive days, which can be responsible for above normal temperature. For example, the EA+ configuration during winter and summer is mainly associated with predominance of anticyclonic conditions over the Mediterranean and North Africa, which would tend to bring anomalous S and SW warm air to the Iberian Peninsula, causing higher temperatures over the region. This configuration has been noted in earlier works (e.g., Sumner et al., 2001; Muñoz-Diaz and Rodrigo, 2004). Similarly, the SCAduring summer is characterized by prevalence of anticyclonic conditions at the surface level over Iberia, Central Europe, and North Africa. This configuration reinforces easterly and southerly flows from the Mediterranean, causing above-normal temperature over the study region. 5.4.1.2. Influence of land-atmosphere coupling While numerous studies explained interannual variability of temperature as driven by large-scale atmospheric circulation (e.g., Sáenz et al., 2001a, b; Brunet et al., 2007b; Rodriguez-Puebla et al. 2010), the impact of these circulations is season dependent, with the stronger (weaker) effect during winter (summer). In particular, summer is generally characterized by a quasi-stationary circulation anomaly. Therefore, in response to the localized diabatic heating at the surface level, distinctive cyclonic (anticyclonic) circulation predominates in the lower (upper) layers of the troposphere during summertime (Chen, 2001). This thermally forced circulation is also coupled with other heat sources during this season, including heat radiation, maximum insolation, clear skies and light wind. Warmer temperature during summer can therefore be partially driven by this large stability and strong persistence in atmospheric circulation.
5. DISCUSSION 315 Over Iberia, this configuration can act as blockings to the passage of the Icelandic cyclones and Atlantic fronts, which bring cold air from Scandinavia and Eurasia. Trigo et al. (1999) defined a major cyclonic center over Iberia during summer season (36º42ºN, 10ºW-0º). It is mainly originated from the thermal effect of warm land, besides the influence of land-sea interaction. Thomas et al. (2010) indicated that the dominance of cyclonic conditions during summer is more highlighted in the mainland peninsula and over mountainous regions, leading to more stability. While the large-scale atmospheric modes can largely be responsible for temperature variations in some seasons (e.g., winter), land-atmosphere coupling processes (e.g., cloudiness and soil moisture) can explain large proportion of temperature variability in the summer periods. Based on simulated data from four different RCMs, Seneviratne et al. (2006), for example, noted that the projected changes in the interannual variability of climate in Europe would largely be driven by land-surface-atmosphere coupling. Among the land surface-atmosphere forces, cloudiness and soil moisture might be considered the two most important parts with significant feedbacks with temperature. Both are key drivers of mass and energy transfer in the globe. Previous studies linked changes in land surface, including soil moisture, with climate variations (e.g., Huang et al. 1996; Douville 2003; Koster and Suarez 2003; Koster et al., 2004, Seneviratne et al. 2006; Fischer et al., 2007). Soil moisture plays a critical role in influencing surface energy and water balance components mainly through its effects on evapotranspiration or latent heat flux. Given that the study area is located in the mid-latitudes between dry and wet conditions; its climate is more likely to be influenced by changes in soil moisture. In this semi-arid region, soil moisture is a
5. DISCUSSION 316 limiting factor for evapotranspiration and thus exerts strong impact on the land energy balance. In this context, an attempt was made to quantify the relationship between the variability of surface temperature and soil moisture availability. It was revealed that changes in soil-atmosphere interactions can partially contribute to the observed trends in temperature evolution, particularly during spring and summer. This feedback was found negative and statistically significant at the 99% level as low soil moisture at the surface level during summer and spring seasons always causes a decrease in latent cooling and in turn above-normal temperature. A recent study by Zhang and Dong (2010) found that soil moisture feedbacks accounted for 5-20% of temperature variability in the transitional zones of eastern Asia. Similarly, Seneviratne et al. (2006) attributed much of variation (60%) in summer temperature variability in the transitional zones over Europe to soil moisture feedbacks. In summer and spring, soil moisture deficit can damp evapotranspiration and consequently more energy is partitioned into sensible heat, enhancing surface air temperature. Moreover, soil moisture can also modify surface air temperature through altering other components of surface energy balance (e.g., surface albedo, atmospheric water, clouds, and thermal properties of soil). Recalling that surface evapotranspiration is likely to inhibit the rising of daytime temperature through evaporative cooling, it can be expected that the dependency of soil moisture will be stronger during daytime (i.e., maximum temperature), while it decreases with nighttime temperature. This finding agrees well with previous works, which found strong feedback between soil depletion and maximum temperature than with minimum temperature (e.g., Dai et al., 1999; Durre et al., 2000; Diffenbaugh et al., 2005; Alfaro et al., 2006; Fischer et al., 2007; Zhang et al., 2009; Zhang and Dong, 2010; Lorenz et al., 2010).
5. DISCUSSION 317 Spatially, soil moisture feedback is likely to be greater on dry slopes, where soil moisture is determined by recent precipitation and time since snowmelt, and least near streams, where soil moisture persists at higher levels. In the study domain, a strong coupling of soil moisture with precipitation is expected during summertime. This is particularly because the study domain is located in a transition zone between dry and wet climates. Also, there is a high spatial and temporal variability of precipitation in the region. The influence of reduced soil moisture on summer temperature variations is also expected to be more highlighted along the coastal and mountainous regions compared with lowlands. This is basically because these regions often receive higher amounts of precipitation compared with areas of low altitude. Thus, the persistence of negative soil moisture anomalies is expected to be higher when there is a decrease in the amount of precipitation. This feature may partially explain the higher temperature warming observed in coastal areas and at highly elevated sites with respect to continental and lowlands. For example, Vautard et al (2007) indicated that a shortage in winter precipitation over the Mediterranean region often causes above-normal temperature in the following summer season. Fink et al. (2004) also demonstrated that the 2003 anomalous warm summer in Central and Western Europe was accompanied by a remarkable deficit in precipitation during the preceded winter. Recently, Seneviratne et al. (2006) showed observational evidence on strong impact of deficit in soil moisture on warm extremes in southeastern Europe. Another important factor affecting temperature variations is cloudiness through radiation feedbacks. Cloudiness directly affects the global climate system by transferring energy in the atmosphere. The decrease in cloudiness often affects energy and heat transfer throughout insolation, suggesting above-normal temperature
5. DISCUSSION 318 during the daytime. Conversely, cloudy days are mostly linked to decrease in sunshine and in turn more evaporation and cooling. However, this study strongly suggests that the relationship between cloudiness and temperature is highly dependent on season, with positive (negative) feedbacks during winter (spring and summer). In contrast to summer, increasing cloudiness causes warmer temperature during winter. Dai et al. (1997) explained the association between surface air temperature and cloudiness in the context of radiation fluxes. In short, the cloudiness effect has a negative feedback during summertime as, under decreasing cloudiness, the incoming shortwave radiation is lower than the outgoing longwave radiation. This feature would be reversed in cloudy days. In winter, this association shows highly significant inverse relationship, indicating warmer minimum temperature during cloudy weather. In their study on the USA, Plantico et al. (1990) indicated that assessing interrelationships between trends or anomalies of temperature and cloud cover could significantly contribute to the understanding of climate change processes. They found a correlation coefficient in the order of -0.44 and -0.13 between cloud cover and maximum and minimum temperature, respectively, during summer (JJA) season. Herein, it is also worthwhile to indicate that this correlation is expected to be higher during years of positive anomalies of sea surface temperature (SST) as warmer surface water leads to less cloud amount. 5.4.2. Influence of large-scale circulation on extreme events 5.4.2.1. Influence on summer temperature extremes Pearson correlation values between the general atmospheric circulation and the regional time series of extreme temperature during summers (JJA) of the period 1960-
5. DISCUSSION 319 2006 were computed for the established sub-regions following principal component ad cluster analyses. In general, this relationship was found statistically significant (p<0.05) for the EA+, SCAand WeMOpatterns. Though the relationship between the NAO and winter temperature over large portions of the Mediterranean and Europe is confirmed (Hurrell, et al., 2003), the NAO seems to be a weak predictor for temperature extremes during summer season. This finding also agrees well with Trigo and Palutikof (2001) who found that the NAO poorly explained variability in atmospheric circulation during summer months, as compared with other seasons. In their study on the entire Europe, Beranova and Huth (2008) found the strongest connections between the EA mode and temperature in southern France and NE Spain. The obtained results suggest that the behavior of temperature extremes during summer is mainly driven by atmospheric circulation during the positive EA, and the negative modes of the SCA and WeMO. This clearly implies that the dependency between temperature extremes and teleconnection indices resembles what was previously obtained for the mean temperature conditions. Similarly, the impacts of these configurations seem to have a spatial structure, with clear regional contrasts among the defined sub-regions. Accordingly, the co-variability between SLP as independent variable and summertime (JJA) temperature as dependent variable was explored for the period from 1960 to 2006. This co-variability was explored for the leading circulation modes by means of the composite climate analysis and the canonical correlation. In summary, it can be concluded that the variability of summer temperature extremes in NE Spain is particularly related to the circulation modes that produce high pressure
5. DISCUSSION 320 anomalies over much of Europe and the Mediterranean Sea. A series of studies found a statistically positive trend in SLP over the whole Mediterranean and most of the continental Europe during warm summers of recent decades (e.g., Reddaway and Bigg, 1996; Xoplaki, 2002). For example, Xoplaki (2002) noted an upward trend in both surface pressure and different geopotential heights over the eastern Atlantic and most of continental Europe west of 30ºE. However, the canonical functions obtained in this work suggest that the spatial variability of temperature anomalies over northeastern Spain varies considerably according to SLP anomaly variations, which markedly differs its position, strength and influence domain from one prominent mode to another 5.4.2.2. Influence on the anomalously severe temperature extremes With a focus on anomalous and very extreme temperature events, the PCA results denote that that changes in these events do not scale proportionately over the study domain. Changes are not found uniformly in all areas of the domain. The large-scale atmospheric circulation at SLP, 200hPa and 500hPa levels was proven to be able to explain spatial variability in very extreme temperature events (i.e., VCN and VWD). The results derived from both the composite maps and the canonical correlation analysis pointed out that the patterns of the 200hPa and 500hPa anomaly fields resemble that of the SLP modes. The spatial patterns of SLP and upper air (i.e., 200hPa and 500hPa) anomalies remain more or less consistent during both VCN and VWD. This more or less similarity implies that cyclones/anticyclones are developed simultaneously at these different levels. This indicates a stationary vertical structure of pressure, suggesting that extreme temperature variability at the seasonal scale (i.e., winter vs. summer) is forced by similar modes of pressure at mid and shallow
5. DISCUSSION 321 troposphere; a result that has been confirmed in earlier works over Western and southwestern Europe (Ulbrich et al., 1999). The results also indicate that changes in VCN can be attributed to two main synoptic conditions. First, they are related to low surface and geopotential anomalies at the 200hPa and 500hPa geopotential heights over the Peninsula, which encourage strong advection of cold polar air masses originating from the anticyclones over Scandinavia. This finding has been reported in recent works (e.g., Klein Tank and Können, 2003; Prieto et al., 2004). For example, Klein Tank and Können (2003) attributed much of variation in cold extremes over the Western Europe and the Mediterranean winters to cold airflows from the snow-covered European continent and the northern Atlantic Ocean. Similarly, Prieto et al. (2004) identified the Arctic synoptic pattern as responsible for much of variability in cold days over the Iberian Peninsula. According to this explanation, the decrease in the frequency and intensity of these cold events can mainly be linked to a decrease (increase) in the meridional (zonal) circulation over the Western Europe in recent years. Werner et al. (2000) reported an increase in the mean residence time of zonal circulation during wintertime in the North Atlantic/European sector since the 1970s. This also agrees well with the previous finding which indicates that warmer winters in northeast Spain are mainly linked to the increase in zonal circulation, as revealed by more warm advections from west and southwest. This situation corresponds to weaker flows from cooler continental areas in north Europe as a consequence of the increase in the frequency of the Atlantic blockings. The second dominant pattern corresponding to the occurrence of VCN is linked to the presence of deep anticyclones over the study area for several days, which help local factors (e.g., fog) to generate VCN. Prieto et al. (2004) demonstrated
5. DISCUSSION 322 that this situation is responsible for the occurrence of cooler temperatures in the Iberian Peninsula when there is a general absence of any significant pressure gradient. The persistence of this circulation for uninterrupted days can likely enhance the occurrence of VCN conditions. The findings on VWD assume that the warmest days during summer corresponded to strong anomalies at different levels (i.e., SLP, 200hPa and 500hPa) over Central Europe, while a negative anomaly is located over the Atlantic Ocean near to the Iberian Peninsula. Numerous studies reported a significant increase in pressure at MSL and different geopotential levels (e.g., 200hPa, 500hPa, 850hPa and 1000hPa) over the Mediterranean and Western Europe during summers of recent decades (e.g., Maheras et al., 1998, 1999; Wanner et al., 1997; Schonwiese et al., 1998; Xoplaki, 2002). For instance, Maheras et al. (1998) indicated a statistically significant increase for the 500 hPa level over the western Mediterranean. This finding has also been confirmed by Wanner et al. (1997), but for a broad geographical domain including the eastern Atlantic sector and much of the continental Europe. More recently, Xoplaki (2002) found statistically significant uptrend in both the 500hPa and 1000hPa geopotential heights from the tropics to the midlatitudes during summer months, the European regions west of 30ºE being the areas with the most significant trend. This configuration enhances advance of warm-dry air flows from the enhanced ridges from overheated European plains to the west and southwest. This situation comes in agreement with previous regional (e.g., Trigo and DaCamara, 2000; Lorenzo et al., 2008) and continental studies (e.g., Post et al., 2002). For example, Lorenzo et al. (2008) showed that the northeastern flows showed their high frequency over northern Spain during warmer summers, while southwestern advections had their least
5. DISCUSSION 329 precipitation, soil moisture, snow-albedo feedbacks, surface fluxes and convective parameters. Another possible reason might be related to the uncertainty related to the use of the Inverse Distance Weighting (IDW) interpolation algorithm to obtain a gridded dataset based on the observational data (1971-2000). This Pyrenees has a relatively uneven network of observatories. Accordingly, the observed temperature might be overestimated as a consequence of the combined effect of topography gradient and the edge impact as access to data from the French Pyrenees was unavailable to limit this effect during interpolation. Overall, according to this finding, a late snow cover melting could be expected in the region in the next few decades, which could have potential implications in the area of water resources management. This picture has completely been reversed at the end of the 21st century, fitting the result of LópezMoreno et al. (2008b) that projected a 1.7–3.1°C rise in the wintertime temperature across the Pyrenees by the end of the current century under the A2 emission scenario using a set of RCMs developed under the PRUDENCE project. The results also indicate that changes in the mean temperature are accompanied by changes in the corresponding time-varying percentiles. This denotes that changes in temperature extremes are mainly due to the shift of the whole distribution rather than changes only in the mean. This also suggests that changes in extremes of daily temperatures in the study domain were due to changes in both the mean and the variance. The overall warming is more highlighted during summer, implying that there is consensus among models and the ensemble mean about high likelihood of increase in extreme warmer minimum and maximum temperature in the future. This range comes in agreement with Hertig et al. (2010) who projected an increase in the
5. DISCUSSION 330 magnitude of the upper percentiles of summer maximum temperature over the western Mediterranean during the 21st century. A comparison between the role of changes in the mean temperature on trends of the maximum and minimum temperature time-varying percentiles during the 21st century reveals that the projected increase in both the mean and magnitudes of the timevarying percentiles generally corresponds to lower interannual variability in the future. The results gave little evidence on a statistically significant change in inter-annual temperature variations. The only exception was found during summer, which exhibited a future increase in the interannual variability of temperature. This implies a more likelihood to exhibit severe extreme warm events in the future. This finding was confirmed by Schär et al. (2004) for the whole Europe. This study highlights the finding that the large absolute change in the mean during the first and the late halves of the 21st century does not necessary correspond to an increase in the temperature variability in the future. Therefore, future temperature changes over the study domain cannot only be inferred from changes in the mean, especially with respect to temperature changes at the lower and upper ends of the maximum and minimum temperature distributions. These results imply that we cannot only rely on changes in the mean to infer changes in the corresponding time-varying percentiles. Hertig et al. (2010) confirmed the same finding for the whole Mediterranean region, indicating that changes in the 90th and 5th percentiles of summer maximum and winter minimum temperatures, respectively, do not follows the same rates of changes of the mean values.
5. DISCUSSION 331 Based on the performance of 9 RCMs over the study domain, future changes in VWD and VCN were also assessed for the 2021-2050 and 2071-2100 periods. It can be seen that changes in the VCN over the study area are more linked to changes in the intensity of temperature during these days than to changes in the VCN frequency. This can be seen when comparing rates of change in the 1st percentile with changes in the frequency of VCN. This comparison may suggest that changes in the mean of minimum temperature during winter season are significant, while changes in the standard deviation are less evident. This situation leads to few changes in the frequency of these events while there is a significant change in the magnitude of temperature recorded during these days. This finding has been confirmed by Prieto et al. (2004) for cold days across the Iberian Peninsula, suggesting no significant changes in the standard deviations of daily minimum temperature distribution. Similarly, the magnitude of the 99th percentile of summer maximum temperature increased at rates greater than those of the 1st percentile of winter minimum temperature during the second half of the 21th century. This indicates a rapid increase in the warm tail of daily temperature distribution than in the cold tail. This picture has been reversed in the earlier decades (2021-2050) as the 1st percentile of daily minimum temperature has risen at rates higher than the 99th percentile of daily maximum temperature, suggesting rapid warming of minimum temperature than maximum temperature in earlier decades of the 21st century. The results on changes in both the mean and the time-varying percentiles can be valuable to assess the potential impacts of climate change on hydrology, human activities, agriculture and economy and can also be useful to monitor these influences at the regional scale. Recalling that warm extreme events are mostly expected to
5. DISCUSSION 332 occur during summer season (JJA), the enhanced summer temperature increase for the study domain, particularly, at the end of the current century, could have various hydrological and environmental consequences. These impacts could include, for example, the increase in drought severity, forest fires and energy consumption and the degradation of agricultural production. Moreover, it can lead to higher evaporation rates that transport larger amounts of water vapor into atmosphere, inducing accelerated changes in the hydrological cycle over the region. In addition, shallow snowpacks as caused by the projected warming can lead to drier soils, longer growing season, and higher moisture stress for plants. In other words, it can be expected that the growing season will begin earlier due to melting of snowpacks, while soil moisture will become depleted sooner as a consequence of the rapid snow melting. Also, this warming can significantly influence the timing of streamflow in the study domain, which may increase water demands and stress in the region. Following Rodriguez et al. (2005), an increase of 1ºC in the Spanish mean annual temperature in the future can be responsible for a reduction of nearly 5-14% in water yields, while a more intense increase of 4ºC could reduce water yields by 22%. The projected rapid warming at high elevation regions by the end of the century (e.g., the Pyrenees and the Iberian systems) could have a wide range of impacts with potentially severe consequences for the biodiversity in these vulnerable environments. For example, changes in the timing of snowmelt could affect plant phenology and induce early season temperature regimes (Inouye et al. 2002; Dunne et al. 2003). This warming may also alter conditions that determine the distribution of habitats and phenologies of the plants. Thuiller et al. (2005) indicated that a projected global warming of 3.6°C could induce a loss of more than 50% of plant species in the
5. DISCUSSION 333 Mediterranean mountain region. This loss is likely to be maximized over northern Spain, reaching 80% of species. In the same context, the model results projected a rapid increase in maximum temperature over the Ebro valley. This suggests that the Ebro valley may become warmer and drier in the future. This projected warming may intensify the hydrological cycle resulting in higher evaporation losses, higher irrigation water requirements, and an overall increase in water resource demand for domestic, agricultural, and industrial use. Such changes in hydrological system and water resources could have a direct effect on society, environment and economy. Taken together, the results of this work on the projected temperature future changes should be taken with much consideration by local decision makers, highlighting the need to adopt their future policy and development plans to meet future demands.
CHAPTERSIX CONCLUSION
6.CONCLUSION 337 6. CONCLUSION The present thesis examined changes in the annual and seasonal distribution of daily maximum and minimum temperatures for northeastern Spain. A better understanding of the ongoing changes in the temperature means and extremes was the primary objective. Further aims involved the analysis of large-scale atmospheric circulation patterns at different geopotential levels as well as the Mean Sea Level (MSL) pressure based on climate composites analysis and canonical variates in order to quantify the driving forces beyond the observed variability. Finally, this work aimed to assess future climate projections of seasonal temperature and their spatial variations to improve the understanding and prediction of the long-term trends of temperature means and extremes simulations. To achieve all these goals, it was necessary to develop a homogenous dataset with high spatial and temporal resolution. The next few paragraphs answer the main research questions raised during this work. (1) To what extent the daily temperature dataset can be trustworthy to examine temperature changes in the study domain? In addition to data availability, the quality and homogeneity of temperature time series are prerequisites for detailed, reliable and trustworthy assessment and attribution of temperature changes. In this work a dense daily temperature database spanning the period between 1900 and 2006 has been developed for the study domain. The raw data provided by the Spanish Meteorological Agency (AMET) were subjected to a vigorous quality control procedure to eliminate any spurious values. Then, a reconstruction scheme was performed to fill in missing values by linear regression.
6.CONCLUSION 338 Potential discontinuities in the time series, as caused by any of the non-climatic factors (e.g., changes in locations, instruments, observers, observing practices, and surrounding environments) were also evaluated. When a statistically significant breakpoint was identified, a correction model was applied to adjust the detected breaks. A monthly correction factor based on the combined results of all homogeneity tests was then computed and interpolated to daily data. This dataset comprises the most long, complete, reliable and spatially dense time series over northeastern Spain, encompassing its major climate regimes (i.e., Mediterranean, oceanic, continental and mountainous). (2) How are seasonal temperature variations distributed in space and time in the study domain? The evolution of seasonal and annual temperatures has been investigated over the period 1920-2006 and the sub-period 1960-2006. By means of the non-parametric Spearman Rho statistic, it was possible to assess presence of trends in the temperature series. Overall, there is strong evidence on an increasing trend in temperature at both seasonal and annual timescales. This finding implies that the regional trends of temperature on either yearly or seasonal scales are closely related to changes in the global climate system. The largest warming occurred during the last few decades, especially from the mid of the 1970s. This warming was faster during spring and summer than in winter and autumn. Spatially, the coastal areas warmed at higher rates than in the mainland areas, demonstrating that there is a distinct coastalcontinental gradient in temperature variations across the region.
6.CONCLUSION 345 mean during the first and the late halves of the 21st century does not necessary correspond to an increase in the temperature variability. Therefore, future temperature changes over the study domain cannot only be inferred from changes in the mean, especially with respect to temperature changes at the lower and upper ends of the maximum and minimum temperature distributions. Changes in the variance of the temperature distribution could have larger impacts on extreme events like change in the mean. These seasonal differences also imply that we cannot only rely on changes in the mean to infer changes in the corresponding time-varying percentiles. (7) What are the potentials of the results obtained in this work? The results derived from this study can contribute to understanding the climate change signal associated with the seasonal temperature variability in northeast Spain. This work represents one of the first attempts to explore spatial and temporal characteristics of temperature variations at sub-regional scale in the Iberian Peninsula. Given the high spatial and temporal scales of the dataset used in this work, the present study can contribute to the limited number of studies focusing on temperature variability at the regional scale in Iberia. Moreover, this work provides an insight into the possible mechanisms and physical processes that may relate spatial and temporal variability of regional temperature with the large-scale atmospheric circulation patterns. Given that the study domain is characterized by complex topography and geography, which play a considerable role in determining the climate and weather at regional and local scales, the results on long-term variability of temperature in this Atlantic/ Mediterranean region can be placed in a larger climate context, providing insights into temperature variability in the Mediterranean and southwestern Europe. In addition,
6.CONCLUSION 346 information on the behavior of temperature on this fine scale can be essential for different impact assessment applications in the region. The results obtained in this work could therefore be meaningful for various applications related to hydrological modeling, agroclimatology, water resources management and drought monitoring. Northeastern Spain has several characteristics that make it interesting for the study of temperature trends. These features include the altitude, latitudinal location, complex topography, and the high spatial and temporal variability of climate. Accordingly, there are great disparities in the variability of temperature at various spatial and temporal scales. Owing to its diversity, the study domain provides an interesting test ground to assess the ability of different RCMs to reproduce temperature future projections. Unfortunately, in contrast to the global scale, there is generally less confidence in estimates of how the climate will change to global warming at regional scale. Only few projections have been studied in the region due to lack of high quality data on a daily basis. The projected change in temperature, as simulated by RCMs forced by different GCM experiments, was within the range of changes simulated by other works in the Mediterranean and Western Europe. The results of this work can thus represent an important milestone step in the prediction of future temperature, as caused by the global warming, under different climate scenarios in the Iberian Peninsula. Therefore, the results intend to complement the series of studies on climate future changes in Iberia by evaluating projected changes in temperature means and extremes over the study domain during the 21st century. The results on future simulations can also contribute to enhancement of climate change adaptation and disaster management as the observed changes could have considerable implications in various areas, such as hydrology, ecology, mortality rates and energy demand and consumption. For
6.CONCLUSION 347 instance, the enhanced summer temperature increase for the study domain, particularly at the end of the current century, could have various hydrological and environmental consequences. These impacts could include, among others, the increase in drought severity, forest fires and energy consumption and the degradation of agricultural production. Moreover, it can lead to higher evaporation rates that transport larger amounts of water vapor into atmosphere, inducing accelerated changes in the hydrological cycle over the region. Accordingly, there is a need to understand the hydrological processes possibly altered by climate change, such as evaporation, surface runoff, drought conditions and water availability. To conclude, this study can be advantageous compared with earlier studies examining temperature changes and variability in the study domain in several ways. First, it depends on a dataset of long, complete, reliable and spatially well-covered time series, which encompasses the main climate regimes in northeastern Spain (i.e. oceanic, Mediterranean, and continental). Previous studies were only restricted to very smaller number of observatories, where reliable data were available. Second, this study can provide a more comprehensive view of long-term variability on seasonal and annual timescales in a way that can significantly contribute to more accurate and robust climate projections. Third, the high spatial and temporal resolution of this dataset suggest that the projected future changes must be taken with much consideration by local decision makers, highlighting the need to adopt their future policy and development plans on a more local scale to meet future demands.
CHAPTERSEVEN CONCLUSIONES
7. CONCLUSIONES 351 7. CONCLUSIONES En esta tesis doctoral se han analizado los cambios en la distribución anual y estacional de las temperaturas máximas y mínimas diarias en el noreste de España. Uno de los principales objetivos fue obtener una mejor comprensión de los cambios en los valores medios y en los extremos térmicos. Otros objetivos incluyeron el análisis de la influencia de los patrones generales de circulación atmosférica a diferentes niveles de geopotencial, así como a nivel del mar (SLP); todo ello basado en clasificaciones y variables canónicas para cuantificar los factores atmosféricos que controlan la variabilidad de la temperatura. Finalmente, este trabajo también ha tenido como objetivo evaluar las proyecciones futuras de la temperatura estacional y sus variaciones espaciales, con el fin de mejorar la comprensión y la predicción de las tendencias a largo plazo, tanto en los valores medios como en los extremos. (1) ¿En qué medida la base de datos generada a escala diaria resulta adecuada para identificar los cambios térmicos en el área de estudio? Además de la disponibilidad de datos, la calidad y homogeneidad de las series de temperatura son un requisito esencial para llevar a cabo una evaluación detallada y fiable sobre los cambios de temperatura en la región. En este trabajo se ha desarrollado una base de datos diaria de temperatura con una elevada densidad espacial. Los datos brutos proporcionados por la Agencia Estatal de Meteorología (AEMET) fueron sometidos a un cuidadoso procedimiento de control de calidad para eliminar los valores falsos de las series. A continuación, se llevó a cabo un proceso de reconstrucción para unir diferentes observatorios y rellenar determinados valores. Se evaluaron potenciales discontinuidades en las series temporales no causadas por
7. CONCLUSIONES 352 factores climáticos (p. ej., cambios en la ubicación de la estación, instrumental, los observadores, prácticas de observación y entorno). Cuando se identificó un punto de ruptura estadísticamente significativo, se aplicó un modelo de corrección para ajustar los problemas detectados. Se obtuvo un factor de corrección mensual, basado en el resultado combinado de diferentes pruebas de homogeneidad, que se interpoló a los diferentes datos diarios. La base de datos desarrollada constituye la serie temporal a escala diaria más completa, fiable y densa espacialmente sobre el noreste de España. (2) ¿Cómo son las variaciones estacionales de temperatura en el área de estudio? La evolución de las temperaturas estacionales y anuales se ha caracterizado para los periodos 1920-2006 y 1960-2006. Las tendencias se evaluaron mediante un test estadístico no paramétrico. En general, se registra una clara tendencia hacia el aumento de la temperatura a ambas escalas: estacional y anual. Este resultado implica que las tendencias regionales identificadas en la región, tanto anual como estacionalmente, están muy relacionadas con los cambios observados en el sistema climático mundial. Las principales tasas de calentamiento se han registrado en las últimas décadas, especialmente desde mediados de la década de 1970. El calentamiento es más rápido en primavera y verano que en invierno y otoño. Espacialmente, las zonas costeras muestran mayores tasas de calentamiento que las zonas continentales. (3) ¿Cómo varían los índices de temperatura extrema en el espacio y el tiempo en el área de estudio?
7. CONCLUSIONES 353 Se analizó la variabilidad espacial y temporal de las temperaturas extremas durante el período 1960-2006 mediante datos diarios de temperatura máxima y mínima de 47 en 128 observatorios meteorológicos. La tendencia se evaluó por medio del test de Mann-Kendall después de eliminar la correlación serial mediante un procedimiento de prewhitening. El análisis de tendencias de los eventos extremos indica un aumento en la frecuencia e intensidad de los extremos cálidos (TX90p, TN90p, TR20 y Txx) frente a la evolución de los extremos fríos (TN10P, FD0, ID0 y TNn). La frecuencia de tendencias significativas fue menor para los eventos fríos que para los cálidos, lo que sugiere un cambio hacia condiciones con extremos más cálidos. Esta tendencia al alza en los extremos cálidos ha sido más pronunciada en las dos últimas décadas, correspondiéndose con un rápido aumento en las temperaturas máximas. La variabilidad de la temperatura máxima extrema parece tener un componente espacial. Las mayores tendencias positivas registradas a lo largo de las costas mediterránea y cantábrica sugieren posibles efectos diferenciadores de los patrones de circulación atmosférica y de la temperatura del mar. Se ha descrito un completo procedimiento para clasificar los valores diarios extremos de temperatura durante el verano. El principal objetivo consistió en delimitar espacialmente regiones coherentes a partir de 14 indicadores basados en las temperaturas máximas entre 1960 y 2006. Se utilizó estadística multivariante (análisis de componentes principales y análisis clúster) para llevar a cabo una clasificación espacial coherente con información sobre la variabilidad espacial de las temperaturas extremas de verano. El grado de bondad de la clasificación se evaluó por medio del índice de Silhouette. Se identificaron cuatro sub-regiones: i) la región del Mediterráneo, ii) la región cantábrica y zonas del interior, iii) las áreas más
7. CONCLUSIONES 354 occidentales y meridionales y iv) las zonas más elevadas. Se examinó la evolución temporal de las temperaturas extremas de verano para las diferentes regiones establecidas, destacando que, si bien el dominio espacial del estudio queda muy limitado espacialmente (≈ 160.000 km2), existe un elevado grado de variabilidad espacial en las características de las temperaturas extremas (frecuencia, intensidad y persistencia). En general, la tendencia hacia un mayor calentamiento se observó en los observatorios ubicados a mayor altitud y a lo largo del litoral mediterráneo. (4) ¿Hasta qué punto los cambios en los valores medios y en los eventos extremos pueden explicarse por patrones de circulación atmosférica a gran escala? En este trabajo también se ha analizado la conexión de la temperatura en la región con la variabilidad en la circulación atmosférica para el período 1960-2006. El incremento térmico observado en las temperaturas máximas, mínimas y medias puede estar relacionado con cambios en algunos patrones de circulación: i) en primer lugar, con el aumento de la circulación zonal con relación a la circulación meridional, lo que implica un aumento de los flujos del oeste respecto a los flujos del norte o del sur y ii) en segundo lugar, un mayor dominio de los anticiclones atlánticos en Europa Occidental, especialmente coincidiendo con los modos negativos de la SCA y WeMO. La correlación de la circulación atmosférica a gran escala con las temperaturas extremas también mostró que las temperaturas extremas en el NE de España se explican principalmente por las configuraciones de tres patrones generales de circulación (SCA, WeMO y EA). El modo negativo del patrón SCA explica la mayor parte de la variabilidad de las temperaturas extremas de verano a escala subregional, con una mayor influencia en los observatorios de mayor altitud. En contraste, el