Repositorio Institucional de Documentos

2013 5 Edmond Pasho Tree growth responses to drought and climate variability analyzed at multiple scales Departamento Director/es Geografía y Ordenación del Territorio Vicente Serrano, Sergio Martín Camarero Martínez, Jesús Julio Director/es Tesis Doctoral Autor Repositorio de la Universidad de Zaragoza – Zaguan http://zaguan.unizar.es UNIVERSIDAD DE ZARAGOZA Departamento Director/es Edmond Pasho TREE GROWTH RESPONSES TO DROUGHT AND CLIMATE VARIABILITY ANALYZED AT MULTIPLE SCALES Director/es Geografía y Ordenación del Territorio Vicente Serrano, Sergio Martín Camarero Martínez, Jesús Julio Tesis Doctoral Autor 2013 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 Doctoral dissertation TREE GROWTH RESPONSES TO DROUGHT AND CLIMATE VARIABILITY ANALYZED AT MULTIPLE SCALES Doctoral dissertation Edmond Pasho Zaragoza 2012 Universidad de Zaragoza Departmento de Geografía y Ordenación del Territorio Tree growth responses to drought and climate variability analyzed at multiple scales Respuesta del crecimiento forestal a la sequia y variabilidad climatica analizada a diferentes escalas Edmond Pasho Tesis Doctoral 2012 Tree growth responses to drought and climate variability analyzed at multiple scales Respuesta del crecimiento forestal a la sequia y la variabilidad climatica analizada a diferentes escalas Thesis supervisors: Dr. Sergio M. Vicente Serrano Instituto Pirenaico de Ecología, Consejo Superior de Investigaciones Científicas (CSIC), Avda. Montañana 1005, Zaragoza 50080, Spain Dr. Jesús Julio Camarero Martínez Fundación “Agencia Aragonesa para la Investigación y el Desarrollo” (ARAID) - Instituto Pirenaico de Ecología (CSIC), Avda. Montañana 1005, Zaragoza 50080, Spain Departmento de Geografía y Ordenación del Territorio List of the original articles v LIST OF THE ORIGINAL ARTICLES This thesis is a summary of the following scientific articles published in international peer-review journals: I. Pasho E, Camarero JJ, De Luis M, Vicente-Serrano SM (2011). Spatial variability in largescale and regional atmospheric drivers of Pinus halepensis growth in eastern Spain. Agricultural and Forest Meteorology 151: 1106–1119. II. Pasho E, Camarero JJ, De Luis M, Vicente-Serrano SM (2011). Impacts of drought at different time scales on forest growth across a wide climatic gradient in north-eastern Spain. Agricultural and Forest Meteorology 151: 1800-1811. III. Pasho E, Camarero JJ, Vicente-Serrano SM (2012). Climatic impacts and drought control of radial growth and seasonal wood formation in Pinus halepensis. Trees, DOI: 10.1007/s00468-012-0756-x. IV. Pasho E, Camarero JJ, De Luis M, Vicente-Serrano SM (2012). Factors driving growth responses to drought in Mediterranean forests. European Journal of Forest Research, DOI: 10.1007/s10342-012-0633-6. The 2010 impact factor of the above mentioned journals based on the Journal Citation Report is as follows: 1. Agriculture and Forest Meteorology 3.228 2. European Journal of Forest Research 1.942 3. Trees 1.444 vi Table of contents vii TABLE OF CONTENTS INFORME DE LOS DIRECTORES…………………………………………………………………i ACKNOWLEDGEMENTS…………………………………………………………........................iii LIST OF THE ORIGINAL ARTICLES...…………………………………………..........................V TABLE OF CONTENTS…………………………………………………………………………..Vii LIST OF ABREVIATIONS………………………………………………………………………...iX RESUMEN (En Español)…………………………………………………………………...………Xi ABSTRACT (In English)…………………………………………………………………………XiX 1. INTRODUCCIÓN (En Español)…………………...……………………………………………..1 1.1 La dendrocronología………………………………………………………………………....3 1.2 Variabilidad climática y crecimiento forestal en la región Mediterránea: el papel de las sequías………………………………………………………………………………………..6 1.3 Justificación de la tesis, hipótesis y objetivos………………………………………………15 1. INTRODUCTION (In English)………………..…..……………………………………………..19 1.1 Dendrochronology………………….……………………………………………………….21 1.2 Climate variability and forest growth in the Mediterranean region: the role of droughts…..23 1.3 Justification of the thesis, hypothesis and objectives……………………………………….31 2. METHODOLOGY……………………………………………………………………………….35 2.1 Study sites and species………………………………………………………………………37 2.2 Dendrochronological methods..……………………………………………………………...43 2.3 Climatic data…………………………………………………………………………………45 2.3.1 Atmospheric circulation ……...……………………………………………………...45 2.3.2 Temperature and precipitation data…………………………………………………..47 2.4 Drought indexes…………...………………………………………………………………...49 2.5 Factors affecting growth-drought responses: climate, topography and vegetation activity…………………………………………………………………………………….51 2.6 Statistical analyses…………………………………………………………………………..53 2.6.1 Principal component analysis (PCA)…………………………………………………53 2.6.2 Correlation analysis…………………………………………………………………..53 Table of contents viii 2.6.3 Superposed epoch analysis (SEA)……………………………………………………54 2.6.4 Regression analysis…………………………………………………………………..54 3. RESULTS………………………………………………………………………………………...57 3.1. Spatial variability in large-scale and regional atmospheric drivers of Pinus halepensis growth in eastern Spain……………….………………………………................................59 3.2. Impacts of drought at different time scales on forest growth across a wide climatic gradient in north-eastern Spain ……………………………………………………….……83 3.3. Climatic impacts and drought control of radial growth and seasonal wood formation in Pinus halepensis ……………………………………………………………………….....101 3.4. Factors driving growth responses to drought in Mediterranean forests………..…………123 4. GENERAL DISCUSSION AND CONCLUSIONS ……………………………………...…....137 4.1 General discussion……………………………………………………………………..139 4.2 Conclusiones (En Español)…………………...……………………………………….145 4.2 Conclusions (In English)………………………………………………………………148 5. REFERENCES………………………………………………………………………………….151 6. ORIGINAL ARTICLES………………………………………………………………………...173 List of abbreviations ix LIST OF ABBREVIATIONS (in alphabetic order) A = Anticyclone AC1 = First-order Autocorrelation AEMET = Agencia Estatal de Meteorología AMiT = Mean minimum annual temperature AMxT = Mean maximum annual temperature C = Cyclone DBH = Diameter at Breast Height DTM = Digital Terrain Model E = Elevation EPS = Expressed Population Signal EW = Earlywood Width EVI = Enhanced Vegetation Index GIS = Geographic Information System GPS = Geographic Position System IS = Inceptisol ES=Entisol AS=aridisol JMiT = January mean minimum temperature JMxT = July mean maximum temperature LW = Latewood Width MO = Mediterranean Oscillation MODIS = Moderate Resolution Imaging Spectro-radiometer MOPREDAS = Monthly Precipitation Dataset MSx = Mean Sensitivity MW = Mean tree-ring Width NAO = North Atlantic Oscillation NDVI = Normalized Difference Vegetation Index PCA = Principal Component Analysis PDSI = Palmer Drought Severity Index PET = Potential Evapotranspiration R = Radiation S = Slope SD = Standard Deviation SE = Standard Error SEA = Superposed Epoch Analysis SPI = Standardized Precipitation Index SPEI = Standardized Precipitation Evapotranspiration Index TRW = Tree-Ring Width UTM = Universal Transverse Mercator WB = Water Balance WeMO = Western Mediterranean Oscillation x Resumen xi RESUMEN La cuenca mediterránea es una de las regiones del mundo más vulnerables a los procesos de cambio climático. La variabilidad y el cambio climático en esta región están muy afectados por los procesos de circulación atmosférica. Por lo tanto, se espera que la ocurrencia de cambios en los patrones de circulación atmosférica puedan afectar al crecimiento de los bosques. La mayoría de los estudios que analizan el impacto climático sobre el crecimiento de los bosques se han centrado, habitualmente, en las condiciones climáticas de superficie, sin tener en cuenta los patrones atmosféricos que controlan el clima en grandes regiones y que también puedan afectar los procesos de crecimiento de los árboles en el espacio y el tiempo. Debido a las proyecciones hacia un mayor calentamiento y una disminución de la precipitación en la región mediterránea, se espera un aumento del impacto de la sequía en estos bosques que afectaría a sus patrones de crecimiento. Por lo tanto, los bosques de esta región deberán adaptarse no sólo a una mayor variabilidad climática, sino también a fenómenos meteorológicos extremos, como la ocurrencia de sequías más severas y frecuentes. En la actualidad existen algunas lagunas en el conocimiento de los impactos del cambio climático y de las sequías sobre el crecimiento de los bosques mediterráneos, teniendo en cuenta la diferente vulnerabilidad de las especies forestales frente a factores de estrés y la existencia de marcados contrates espaciales en las condiciones climáticas. Por lo tanto, conocer la respuestas del crecimiento forestal a los procesos de variabilidad climática en la cuenca del Mediterráneo resulta importante para entender la sensibilidad y la capacidad de adaptación de estos bosques a las tendencias hacia una mayor aridez previstas para finales del siglo XXI. Esto resulta de vital importancia para mejorar nuestra capacidad de predicción de las consecuencias del cambio climático sobre el crecimiento de los árboles y para desarrollar estrategias de silvicultura adecuadas para mitigar los impactos en los ecosistemas forestales mediterráneos. Teniendo en cuenta las cuestiones antes mencionadas, la investigación que presenta esta tesis analizó en primer lugar la influencia de los patrones de circulación atmosférica general y regional, Resumen xii resumidos a partir de diferentes índices atmosféricos y de series de frecuencias de tipos de tiempo, y de las temperaturas y precipitaciones sobre la variabilidad espacio-temporal en la formación de madera temprana (earlywood, EW) y madera tardía (latewood, LW) en bosques de Pinus halepensis del este de España. Además, la investigación también analizó el impacto de la sequía cuantificada a diferentes escalas temporales sobre el crecimiento radial de ocho especies arbóreas (Abies alba, Pinus halepensis, Quercus faginea, Pinus sylvestris, Quercus ilex, Pinus pinea, Pinus nigra y Juniperus thurifera) a lo largo de un amplio gradiente climático en el noreste de España. Los principales factores geográficos y ambientales que controlan la respuesta del crecimiento forestal a la sequía también fueron investigados e identificados. Se utilizaron técnicas dendrocronológicas para cuantificar la variabilidad del crecimiento radial de todas las especies consideradas y analizar la influencia de los diferentes parámetros climáticos y de la sequía en el crecimiento de los árboles. En particular, la respuesta del crecimiento de los árboles a la variabilidad climática, incluyendo procesos de circulación atmosférica, se evaluó únicamente en bosques de P. halepensis, mientras que la respuesta espacio-temporal en el crecimiento de los árboles a la sequía y los factores que condicionan esa misma respuesta se analizaron teniendo en cuenta todas las especies antes mencionadas. La investigación se ha centrado principalmente en los bosques de P. halepensis porque se trata de la especie dominante en las zonas más secas del área de estudio y por estar esta especie bien representada en buena parte de la región. Los sitios de muestreo se seleccionaron para capturar la mayor parte de la variabilidad climática de la región. En cada sitio de muestreo se seleccionaron al azar y muestrearon entre 10 y 35 árboles. Se obtuvieron muestras a una altura aproximada de 1,3 m del suelo mediante una barrena de tipo Pressler. Las diferentes muestras de madera se procesaron utilizando métodos dendrocronológicos estándar con la finalidad de obtener información de la variabilidad de EW, LW y de la anchura del anillo de crecimiento anual. Con respecto a la circulación general atmosférica, se ha analizado la influencia de los tres principales patrones de circulación atmosférica que afectan a la región mediterránea occidental: la Resumen xiii Oscilación del Atlántico Norte (North Atlantic Oscillation, NAO), la Oscilación del Mediterráneo Occidental (Western Mediterranean Oscillation, WeMO) y la Oscilación del Mediterráneo (Mediterranean Oscillation, MO). Se obtuvieron series estacionales para otoño (septiembre a noviembre), primavera (abril-mayo), verano (junio a agosto) e invierno (diciembre a marzo) de los índices mensuales de los tres patrones de circulación mencionados obtenidos a partir de series de presiones a nivel del mar (Sea level pressure, SLP). Además, se obtuvieron series de tipos de tiempo a partir de una rejilla de series de SLP del conjunto de la Península Ibérica para comprobar su influencia en el crecimiento radial de los bosques de P. halepensis. También se trabajó con series mensuales de precipitación total y temperatura media obtenidas a partir de dos bases de datos climáticos homogéneos y con una elevada densidad espacial de observatorios. Con ello se determinaron los mecanismos que condicionan la influencia de los procesos de circulación atmosférica en el crecimiento de los bosques de P. halepensis en la región y el impacto directo de las condiciones climáticas de superficie sobre el crecimiento radial de esta especie. Para evaluar el impacto de la sequía sobre el crecimiento de diferentes especies arbóreas en el noreste de España, se utilizó el Índice de Precipitación Estandarizada (Standardized Precipitation Index, SPI), calculado a diferentes escalas temporales (de 1 a 48 meses). Además, se consideró un conjunto de factores abióticos (clima, topografía, tipo de suelo) y bióticos (Índices de Vegetación obtenidos mediante imágenes de satélite, diámetro de los árboles a 1,3 m, etc.) para identificar los principales factores que determinan las diferencias espaciales en el impacto de la sequía sobre el crecimiento forestal. La dinámica estacional de la actividad cambial y la formación de la madera en los bosques de P. halepensis se analizó mediante el muestreo en un bosque sometido a condiciones climáticas semiáridas. Para ello, se llevó a cabo el muestreo y preparación de muestras radiales de madera tomadas periódicamente (mini-cores) con el propósito de describir el proceso de crecimiento intraanual (xilogénesis) de forma detallada y para poder comprender cómo el crecimiento estacional puede estar respondiendo a los procesos de cambio climático. Esta información detallada ha Resumen xiv resultado crucial para corroborar los mecanismos que explican la respuesta del crecimiento radial a la variabilidad climática y a la sequía. En la zona de estudio se han encontrado dos patrones de crecimiento diferentes en respuesta a la variabilidad en la circulación atmosférica, principalmente como consecuencia de la variabilidad en el crecimiento entre las sub-zonas del noreste y sureste de la zona de estudio. Ello se aprecia tanto para las series de EW como para las de LW. La formación de EW y LW en las áreas más septentrionales está muy condicionada por la variabilidad que la NAO muestra en invierno y primavera, mientras que en los bosques meridionales el crecimiento se ve afectado por el índice WeMO de invierno. La formación de EW en los bosques del norte de la zona de estudio está negativamente correlacionada con los índices NAO de diciembre y abril, mientras que la formación de LW estaba asociada negativamente con el índice de la WeMO en septiembre. También se ha comprobado que la frecuencia de diferentes tipos de tiempo en invierno, verano y otoño ejerce un importante control sobre el crecimiento radial. La formación de EW en los bosques del norte se ve favorecida por una alta frecuencia de tipos de tiempo del sudoeste y oeste, mientras que una alta frecuencia de tipos del este y sureste tiene un papel negativo en la formación de EW en estas áreas. En los bosques situados en el sur, la formación de EW se ve reforzada por una elevada frecuencia de tipos del este y sudeste. La formación de EW en las zonas del norte también se ha visto favorecido por la frecuencia invernal de tipos de tiempo del sur, sudoeste y anticiclónicos, mientras que en los bosques del sur, dicha formación se relaciona negativamente con la frecuencia de tipos del sudeste y sur. El índice NAO de invierno mostraba una correlación negativa significativa con la precipitación en los bosques del norte del área de estudio, mientras que durante el invierno y el otoño el índice de la WeMO se asociaba negativamente con la precipitación en los bosques del sur. Teniendo en cuenta los registros de temperatura, la NAO de invierno muestra una relación positiva con las temperaturas mientras que en primavera y otoño se observan fuertes correlaciones negativas entre el índice de la WeMO y la temperatura en el sector septentrional del área de estudio. La formación de EW y LW aumentó en respuesta a las precipitaciones del invierno previo y de la Abstract xxi To calculate the impact of drought on growth of different tree species in north-eastern Spain, it was employed the Standardized Precipitation Index (SPI), calculated at different time scales (1-48 months). In addition, a set of abiotic (climate, topography, soil type) and biotic (Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), tree ring width, diameter at breast height) variables was used to identify the main factors determining the spatial differences in the drought impacts on forest growth. The seasonal dynamics of cambial activity and wood formation in P. halepensis in a semiarid site were explored by sampling and preparing minicores with the purpose of describing the process of intra-annual tree-ring formation (xylogenesis) and understanding the seasonal growth responses to climate. This mechanistic information is necessary to fully understand the radialgrowth responses to climate variability and drought stress. There were found two distinct growth patterns in response to atmospheric variability, reflecting the growth variability in northern and southern areas of eastern Spain, for both EW and LW growth series. The EW and LW growth in northern sites was determined by the winter-spring NAO variability whereas at the southern sites it was affected by the winter WeMO index. Considering monthly values, the EW growth at northern sites was negatively related to the previous December and the current April NAO indices whereas the LW growth was negatively associated with the current September WeMO. Winter, summer and autumn weather types also exerted a control on radial growth. The EW formation in the northern sites was enhanced by a high frequency of SW and W weather types whereas the high frequency of flows from the East and Southeast had a negative role on EW growth in these areas. In the southern sites, EW formation was enhanced by a high frequency of E and SE flows. The LW growth in the northern areas was also highly related to the winter frequency of S, SW and A weather types whereas in the southern sites it was negatively related with the frequency of SE and S weather types. Winter NAO showed a significant negative correlation with precipitation in northern sites whereas winter-autumn WeMO was negatively associated with precipitation in southern sites. Considering the temperature data, winter NAO Abstract xxii showed a positive relationship with temperatures in the northern areas while spring-autumn WeMO influence on the temperatures presented strong negative correlations in the northern parts of the study area. The EW and LW growth was enhanced by winter-spring precipitation of the current year in the entire study area. In addition the LW formation was also positively correlated with summerautumn precipitation whereas June-July temperatures constrained the development of both components of the tree ring, particularily in the north-eastern part of the study area. It was found that both atmospheric circulation indices and weather types exert significant impact on the EW and LW formation in P. halepensis forests. This influence was expressed through their control of surface climate conditions, temperature and particularly precipitation. The response of EW and LW growth to atmospheric circulation patterns and weather types was found to be geographically variable, indicating that the predicted changes in atmospheric circulation will result in contrasting tree growth responses in the northwest and southeast parts of the study area. Considering the temperature and precipitation influences on P. halepensis EW and LW growth, it was concluded that high summer temperatures and winter-spring water availability affected EW formation in negative and positive ways respectively, and to a lesser extent controlled the LW growth. This suggests that in a warmer future scenario, as predicted for the Western Mediterranean Basin, this species will show a more pronounced decrease in EW than in LW development causing a decline in tree-ring width and wood production, a reduction in hydraulic conductivity and, indirectly, a hampered carbon uptake. The impact of drought on growth varied noticeably among species and sites. Two distinct patterns were clearly observed considering spatial and temporal differences in the response of species to drought. Species growing in xeric sites (e.g., Pinus halepensis and Juniperus thurifera) showed the highest responses to SPI time-scales of 9-11 months while those located in mesic sites (e.g., Abies alba, Pinus sylvestris) responded more to SPI time scales shorter than 5 months. The SPI-growth correlations were significant, although weak, up to 30 months in xeric sites while no consistent association was observed at higher time scales. Important seasonal differences were Abstract xxiii noticed in the SPI-growth associations. Species growing in xeric areas responded to spring-summer SPI while those distributed in mesic sites responded more to summer SPI. Considering P. halepensis, EW and LW formation at inter-annual scales showed the strongest negative associations with mid-term cumulative drought stress recorded at time-scales from 10 up to 14 months, but these responses were observed in July and September, respectively, matching the phases of lowest EW and LW tracheid production at intra-annual scales. The associations between abiotic and biotic variables and the growth responses to drought in xeric and mesic sites indicated that tree-related (diameter at breast height, tree-ring width), remotesensing (NDVI, EVI), climatic (water balance, precipitation, temperature) and topographic variables (soil type, slope, elevation) were significantly and inversely correlated with the growth responses to drought in xeric and mesic sites, respectively. Most climatic variables (potencial evapotranspiration, mean maximum and minimum temperatures, July mean maximum and January mean minimum temperatures, solar radiation) were negatively related with the growth responses to drought in mesic sites. The growth responses to drought in xeric sites were strongly associated with water availability, temperature, elevation and NDVI. The regression analyses indicated that growth responses to drought in xeric forests were driven by the annual precipitation, inceptisol soils, AprilJune NDVI and slope while in mesic sites the major drivers of such responses were annual water balance, April-June EVI and inceptisol soils. However, the coefficients of the selected regression models showed that growth responses to drought in xeric forests were mainly driven by the annual precipitation while in mesic sites the annual water balance was the most important driver. The use of multi-scalar drought indices is particularly useful for monitoring the impact of climate variability on tree growth because the growth responses to drought are complex and timedependent. The time scales over which water deficit accumulates and affects noticeably tree growth, vary among species and sites within the same species. For this reason, drought indices must be associated with a specific time scale and assessed taking into account local conditions to be useful for monitoring impacts on tree growth. The high spatial and temporal variability in terms of growth Abstract xxiv responses to drought observed among species and sites in the Mediterranean forests was significantly driven by climatic, topographic and biotic variables, indicating that a combination of the above variables shaped the species behavior in response to drought. These findings suggest that warming-related drought stress might alter differently tree growth in mesic compared to xeric sites. Increasing aridity in the Western Mediterranean Basin is expected to cause selective growth decline in drought-sensitive species from mesic sites. However, disentangling the relative effects of warmer conditions and reduced precipitation on tree growth is an unsolved challenge which probably requires a multiproxy approach based on long-term data of radial growth, measures of carbon isotopic discrimination in wood and remote-sensing variables. Finally, these findings should be useful to understand forest growth responses to climate change, including an increasing frequency of severe droughts, and to adapt appropriate management strategies such as selective thinning to mitigate the impact of drought on tree growth. The management of Mediterranean forests under the forecasted warmer and drier conditions should focus on the main local factors differently modulating the negative impacts of drought on tree growth in xeric and mesic sites. Introducción 1 CAPÍTULO I INTRODUCCIÓN En este capítulo se muestra una breve introducción sobre la dendrocronología y sus aplicaciones, haciendo especial hincapié en los efectos del clima sobre el crecimiento de los árboles. Además, se proporciona información básica sobre la variabilidad del clima y la sequía en la cuenca Mediterránea, en general, y en la Península Ibérica, en particular, mostrando una revisión actualizada sobre el actual estado de la cuestión e incidiendo en la influencia del clima sobre la dinámica de los bosques de la región . La última parte de este capítulo justifica la realización de este estudio, la principal hipótesis de trabajo y los diferentes objetivos planteados. Introducción 2 Introducción 3 1.1. La dendrocronología El término dendrocronología deriva de las palabras griegas "dendro", (madera) y "cronología" (tiempo). El término hace referencia a una disciplina científica basada en un conjunto de principios, técnicas y métodos fundamentados en la datación precisa y en el estudio de los anillos anuales de crecimiento formados por plantas leñosas en zonas de clima estacional. A partir de ello se extrae, ordena e interpreta la información que contienen dichos anillos para inferir los diferentes factores que influyen en el crecimiento de los árboles y en la formación de la madera (Fritts, 1976). La dendrocronología se comenzó a desarrollar formalmente durante la primera mitad del siglo XX por el astrónomo A.E. Douglass (Douglass, 1940), que fue el fundador del Laboratorio de Investigación en Dendrocronología de la Universidad de Arizona (Tucson, Arizona, EE.UU.). Hay que tener en cuenta que la mayoría de los árboles ubicados en zonas frías y templadas, sometidas a un clima estacional, presentan un incremento periódico en el crecimiento, con la formación de anillos anuales. De hecho, de cada anillo se puede obtener información relacionada con su estructura y formación. Para ello, se pueden extraer parámetros que representan las características específicas de cada anillo: anchura, anatomía, densidad máxima, composición isotópica y otros parámetros visuales o analíticos que pueden hacer que se produzcan diferencias entre unos anillos en función de una serie de factores ecológicos y climáticos (Schweingruber, 1988). En las zonas con una marcada estacionalidad en el clima, el crecimiento de los árboles es discontinuo, pues éste se detiene en algunos periodos debido a la limitación impuesta por determinados factores climáticos externos. Ello explica la formación de anillos anuales de crecimiento. En muchas zonas del planeta con una marcada estacionalidad climática, el parón en el crecimiento de los árboles durante los períodos desfavorables (por ejemplo, el invierno frío en gran parte del hemisferio norte o el verano seco en zonas áridas) es habitual. De hecho, la actividad del cámbium se reinicia de nuevo cuando las condiciones climáticas vuelven a resultar favorables. Este patrón anual de actividad y crecimiento determina finalmente la estructura de la madera en forma de capas concéntricas anuales que, en sección transversal, se observan como anillos. En una escala Introducción 4 intra-anual, la producción de nuevas células del xilema (xilogénesis o formación de la madera) es rápida en la primera fase de crecimiento, disminuye a medida que avanza el verano y se detiene cuando las condiciones climáticas empeoran. Estas diferencias en la tasa de formación de células se reflejan también en las características de los anillos de los árboles, ya que se pueden apreciar dos bandas de crecimiento: la madera temprana (early wood, EW), formada a principios del período de crecimiento (primavera, verano) y la madera tardía (late wood, LW) formada a finales de la estación de crecimiento (verano, otoño). En las coníferas (pinos, abetos, etc.) la EW es clara y está formada por traqueidas con lúmenes transversales anchos y paredes celulares delgadas, mientras que la LW es oscura y está formada por traqueidas estrechas con paredes celulares gruesas (Schweingruber, 1988). Las diferencias de tamaño, densidad y color entre la parte lignificada oscura de un anillo y la clara permiten la identificación y datación de los anillos. La dendrocronología tiene aplicaciones en diferentes campos científicos debido a que los anillos contienen información sobre diferentes factores. Esta información se puede recuperar y analizar convenientemente, lo que permite el estudio y análisis de diferentes procesos ecológicos, geomorfológicos, climáticos, arqueológicos, etc., en el espacio y el tiempo a partir de una señal que pueda ser extraída de los anillos (Gutiérrez, 2008). Por ejemplo, los anillos de los árboles pueden registrar el efecto de los incendios forestales, terremotos, erupciones volcánicas, deslizamientos del terreno, caídas de rocas, avalanchas, inundaciones, el avance y retroceso de los glaciares, etc. Todos estos fenómenos pueden ser datados mediante las señales que dejan en los anillos, lo que permite determinar la frecuencia e intensidad con la que estos fenómenos ocurren en la zona afectada. En ecología y siguiendo un enfoque retrospectivo, la dendrocronología es muy útil para estudiar procesos que ocurren a escalas de tiempo largas e implican a organismos longevos como los árboles. Por ejemplo, el estudio de los anillos de los árboles permite datar el año de germinación y muerte de los árboles, que son los dos procesos clave en la dinámica de las poblaciones y en la sucesión de los bosques. El análisis de las series de crecimiento anual permite la determinación de Introducción 5 los regímenes de perturbaciones que pueden afectar la dinámica del bosque e incluso establecer la intensidad de la competencia entre árboles. El clima es el factor ambiental más importante que influye en el crecimiento radial de las plantas leñosas y en particular en los árboles (Fritts, 1976). Cuando los diferentes factores climáticos limitan el crecimiento de los árboles, los anillos formados son normalmente más estrechos en la mayoría de los árboles de una región en particular. Como resultado de la estrecha relación entre el crecimiento radial y las condiciones climáticas, las series de anillos anuales formados por árboles que crecen bajo las mismas condiciones climáticas muestran una cierta sincronía, siendo su patrón de variación temporal muy similar entre árboles (Fritts, 1976). Esta señal climática que permanece en la madera constituye también una “firma” temporal, debido a que resulta muy poco probable que un determinado patrón temporal específico se repite exactamente igual en otro período. La señal climática registrada en los anillos de los árboles se puede así utilizar para reconstruir la evolución del clima en el pasado en períodos sin registros meteorológicos instrumentales. En particular, resulta posible reconstruir las variables climáticas que más limitan el crecimiento de los árboles, y que muestran una influencia significativa en la formación de los anillos anuales. Por ejemplo, es de esperar que las temperaturas y precipitaciones sean los principales factores de crecimiento en los bosques de climas fríos (por ejemplo, bosques subalpinos) y secos (por ejemplo, bosques en zonas semiáridas), respectivamente. Los estudios dendrocronológicos llevan a cabo la elaboración de cronologías de crecimiento formadas a partir de un gran número de series obtenidas de un gran número de árboles de la misma especie que crecen en el mismo lugar con el objetivo de maximizar la señal climática común (Cook y Kairiuktsis, 1990). En resumen, la dendrocronología es una poderosa herramienta científica utilizada para mostrar cómo las condiciones climáticas determinan el crecimiento radial de los árboles (Schweingruber et al., 1988). Por esta razón, los anillos de árboles se han utilizado ampliamente como una aproximación válida en el estudio a largo plazo de los patrones de crecimiento de los árboles. Además, la anchura del anillo se puede utilizar como estima (proxy) de la producción primaria neta, Introducción 6 o de la capacidad de captar carbono por parte de los bosques, y para evaluar el impacto de los cambios ambientales sobre los patrones de crecimiento. Ello se basa en la suposición general de que la relación entre el crecimiento de los árboles y el clima permanece aproximadamente constante a lo largo del tiempo (Fritts, 1976). Además, la variabilidad temporal en el ancho de los anillos de los árboles se ha utilizado para: reconocer posibles cambios en la sensibilidad del crecimiento de los árboles al clima (Andreu et al., 2007), estudiar la variabilidad regional en la relación entre el crecimiento de los árboles y el clima dentro de una especie a partir de redes dendrocronológicas (Tardif et al., 2003; Piovesan et al., 2005; Carrer et al., 2007) o identificar la reciente disminución del crecimiento asociada al calentamiento global y la sequía en las regiones mediterráneas (Salto et al., 2006; Sarris et al., 2007; Camarero et al., 2011). 1.2. Variabilidad climática y crecimiento forestal en la región mediterránea: el papel de las sequías. A lo largo del siglo XX se han observado numerosos cambios en la cuenca mediterránea con relación a la amplitud ecológica, composición, dinámica y fenología de los bosques (Walther et al., 2002; Peñuelas y Boada, 2003). Estos fenómenos se conectan de forma significativa con el cambio climático observado en la región (Parmesan y Yohe, 2003; Linares et al., 2010). De forma natural, el clima de la cuenca mediterránea presenta frecuentes y severos períodos de sequía intensas (Briffa et al., 1994; Demuth y Stahl, 2001; Lloyd-Huges y Saunders, 2002a,b; Vicente-Serrano, 2006; González-Hidalgo et al., 2009) y una elevada variabilidad interanual de las precipitaciones (Palutikof et al., 1994; Trigo et al., 2000; Hasanean, 2004; Zveryaev, 2004; Giorgi y Lionello, 2008; González-Hidalgo et al., 2009). Los escenarios de cambio climático predicen un notable aumento de la temperatura (+2-4°C) y una disminución de la precipitación (cerca del -20%) en la cuenca mediterránea a lo largo del siglo XXI (Gibelin y Deque, 2003; IPCC, 2007). La tendencia prevista hacia un incremento de las condiciones de aridez estará asociada a un incremento de la frecuencia de condiciones anticiclónicas y un desplazamiento hacia el norte en la trayectoria de las borrascas Introducción 13 importante, ya que interactúa con la precipitación para determinar la disponibilidad real de agua para el crecimiento de los árboles (Vicente-Serrano et al., 2010a). Unas elevadas temperaturas generalmente tienen un efecto positivo sobre el crecimiento bajo condiciones mésicas, pero su influencia resulta negativa si no se registra el correspondiente aumento en la precipitación, es decir en condiciones xéricas. La disminución de las precipitaciones, ya observada en la región mediterránea, está aumentando la frecuencia e intensidad de las sequías (Jones et al., 1996; Romero et al., 1998; Gibelin y Deque, 2003; Vicente-Serrano, 2006; IPCC, 2007; Giorgi y Lionello, 2008; García-Ruiz et al., 2011). Las sequías son el principal factor que limita el crecimiento y desencadena episodios relacionados con la mortalidad en los bosques, afectando de forma selectiva a diferentes especies y tipos de árboles (Allen et al., 2010; Koepke et al., 2010). La sequía produce una reducción en el crecimiento radial, conduce a alteraciones en la conductividad hidráulica (McDowell et al., 2008) y disminuye la productividad mediante una reducción en la actividad fotosintética y la captación de carbono (Hsiao, 1973; Flexas y Medrano, 2002). En particular, los déficits hídricos afectan el crecimiento del xilema en las coníferas reduciendo la turgencia celular durante la expansión inicial y la lignificación posterior de las traqeiadas (por ejemplo, Vysotskaya y Vaganov, 1989; Gindl, 2001), limitando la dinámica del cámbium y finalmente condicionando la formación de madera y el crecimiento secundario (Larson, 1963; Arend y Fromm, 2007). La Figura 2 muestra un modelo conceptual que describe cómo la disminución de la precipitación a lo largo de un gradiente de sequía creciente, afecta negativamente al crecimiento y altera las características de las series de anchura del anillo en árboles (Fritts, 1976; Scharnweber et al., 2011). La zona izquierda de la gráfica representa los árboles que crecen en sitios templados y húmedos (por ejemplo, los bosques pirenaicos de montaña), donde la anchura de los anillos muestra los valores medios más altos, pero la menor variabilidad interanual, es decir, son zonas menos sensibles a la variabilidad climática. Por otro lado, la región de la derecha representa los árboles que crecen en sitios secos (por ejemplo, bosques semiáridos de la Depresión Media del Ebro), donde la serie de Introducción 14 anchura de anillos muestra valores de crecimiento bajos y una mayor variabilidad interanual. Ello implica que son zonas muy sensibles a la variabilidad climática. Figura 2. Variación teórica de las características dendrocronológicas a lo largo de un gradiente de estrés debido a un aumento de la sequía (modificado a partir de Fritts, 1976; Scharnweber et al., 2011). La línea punteada indica el límite xérico en la distribución de los árboles. Sin embargo, los factores climáticos no son los únicos que condicionan los efectos de la sequía sobre el crecimiento del árbol. Las características geográficas de cada lugar, así como la elevación o las condiciones del suelo, determinan el grado en que los factores climáticos y la sequía afectan al crecimiento de los árboles (Fritts, 1976; Orwig y Abrams, 1997). Estos factores locales pueden suponer riesgos adicionales que incrementan el estrés causado por la sequía (suelos rocosos o pendientes pronunciadas) o que, en parte, pueden mitigar sus negativos efectos sobre el crecimiento de los árboles (suelos profundos o umbrías). Por ejemplo, los factores topográficos pueden determinar variaciones locales en la cantidad de agua retenida por los suelos y en la respuesta al clima (Tardif et al., 2003). Las diferencias en la exposición puede afectar la temperatura, el viento y la radiación, que, a su vez, influyen en la velocidad a la que la humedad se pierde por procesos de evapotranspiración (Fritts, 1976). Introducción 15 Por lo tanto, para comprender en profundidad cómo la variabilidad climática resulta determinante para explicar el crecimiento de los árboles en una región climáticamente compleja, como es el este de la Península Ibérica, resulta necesario seguir un enfoque metodológico que considere la variabilidad climática a diferentes escalas espaciales y temporales. Este enfoque debe permitir la comprensión de i) cómo los cambios en la circulación atmosférica y en el clima afectan el crecimiento de los árboles, ii) cómo es el impacto de la sequía sobre el crecimiento, iii) cómo este impacto varía en función de los tipos de bosques y las condiciones bioclimáticas de cada región y localidad, y, finalmente, iv) cómo las variaciones estacionales en la intensidad de la sequía determinan los procesos de formación de madera. 1.3. Justificación de la tesis, hipótesis y objetivos. La mayoría de los estudios acerca del impacto del clima sobre el crecimiento de los bosques se han centrado en la influencia de las variables climáticas de superficie (Fritts, 1976). Sin embargo, los patrones de circulación atmosférica determinan la variabilidad del clima a lo largo de grandes regiones (Garfin, 1998; Girardin y Tardif, 2005). Por otra parte, se espera que los principales signos de cambio climático se identifiquen inicialmente a partir de cambios en la circulación atmosférica (Giorgi y Mearns, 1991; Räisänen et al., 2004), y que los cambios previstos en la circulación atmosférica y las condiciones climáticas afecten al crecimiento de los bosques en la cuenca mediterránea directamente a través de cambios en los patrones de circulación a gran escala e indirectamente a través de modificaciones en los patrones atmosféricos regionales que determinan el clima a escala local. Por lo tanto, es particularmente importante investigar tanto los efectos directos como los indirectos de los principales patrones de circulación atmosférica sobre el crecimiento forestal, especialmente en áreas sujetas a diversas condiciones climáticas. Este es el caso de la Península Ibérica, donde el clima varía de oceánico a continental y de condiciones húmedas a semi-áridas, dando lugar a diversas condiciones climáticas y tipos de bosques sometidos a limitaciones de crecimiento muy diferentes (Nahal, 1981). Además, las importantes diferencias Introducción 16 espaciales en la respuesta de los bosques a la sequía (Andreu et al., 2007; Sarris et al., 2007; Martínez-Villalta et al., 2008; Sánchez-Salguero et al., 2010; Linares et al., 2010; Vicente-Serrano et al., 2010a), sugieren la importancia de analizar el crecimiento del bosque en relación a gradientes climáticos locales y la necesidad de considerar los bosques cercanos a su límite climático de distribución. Mientras que la tendencia prevista hacia condiciones progresivamente más secas es probable que cause una disminución en el crecimiento de los bosques mediterráneos, la extensión espacial y la magnitud del efecto de los factores atmosféricos y climáticos sobre el crecimiento de los árboles resulta muy incierta hoy en día (Andreu et al., 2007; VicenteSerrano et al., 2010a). Por otra parte, la elevada estacionalidad y variabilidad interanual que caracterizan las precipitaciones en la región del Mediterráneo y la diferente estacionalidad en el crecimiento forestal de los bosques de esta zona hacen difícil determinar los tiempos de respuesta en el crecimiento de los árboles a los déficits de precipitación. Pueden así aparecer desfases entre la escasez de agua y el crecimiento en función de diferentes estrategias funcionales o anatómicas y ajustes fenológicos y fisiológicos de los árboles para hacer frente a la sequía, pero también en función de la severidad de la sequía y en función del momento del año en que se produce el déficit hídrico más intenso. Todos estos mecanismos, ya sean aislados o actuando de forma sinérgica, pueden dificultar la identificación de los impactos de la sequía sobre el crecimiento del árbol. En la actualidad, resulta crucial ampliar el conocimiento existente sobre la respuesta de crecimiento de los árboles a la escasez de agua en la cuenca del Mediterráneo, pues los modelos generales de cambio climático predicen una gran reducción de las precipitaciones y un aumento de las tasas de evapotranspiración a finales del siglo XXI (Giorgi y Lionello, 2008; García-Ruiz et al., 2011). Existe también información contradictoria respecto a los impactos de la sequía sobre el crecimiento de las especies arbóreas a diferentes escalas temporales y en función de las condiciones bioclimáticas de cada lugar (Orwig y Abrams, 1997; Adams y Kolb, 2005). Por otra parte, la cuantificación del desarrollo intrae inter-anual de la madera en respuesta a las condiciones de estrés hídrico cuantificadas a diferentes Introducción 17 escalas de tiempo es crucial para entender la sensibilidad del crecimiento radial en respuesta a la variabilidad de las precipitaciones en la región (Andreu et al., 2007; Camarero et al., 2010). Teniendo en cuenta las cuestiones mencionadas, se plantearon las siguientes hipótesis: (i) las respuestas de crecimiento de los bosques a los patrones de circulación atmosférica y la variabilidad climática en el noreste de España debe ser espacialmente estructurada y (ii) deben existir respuestas contrastadas en el crecimiento de los árboles a la sequía entre diferentes especies pero también entre bosques de una misma especie sometidos a diferentes condiciones ambientales. En particular, se planteó la hipótesis de que los bosques ubicados en los sitios xéricos mostrarán una mayor capacidad de respuesta a la sequía mediante la reducción en la formación de la madera en comparación con zonas más húmedas donde se espera que la temperatura sea el principal factor limitante del crecimiento. Sin embargo, estas respuestas pueden depender también de las tendencias climáticas estacionales y de la alta plasticidad fenológica del xilema en especies ibéricas de árboles (De Luis et al., 2007; Camarero et al., 2010). Los objetivos de esta tesis son los siguientes: • (i) el uso de métodos dendrocronológicos para caracterizar los patrones espaciales y temporales de crecimiento radial en una red de bosques ubicados en el noreste de España y sometidos a un amplio gradiente climático que abarca desde condiciones climáticas semiáridas hasta frías y húmedas. Los lugares de muestreo incluyen diversos tipos de bosques tales como bosques mediterráneos en sitios semi-áridos de la cuenca del Ebro y bosques subalpinos en sitios húmedos de los Pirineos; • (ii) cuantificar la influencia de los patrones de circulación atmosférica a gran escala, la frecuencia de tipos de tiempo y las variables climáticas de superficie (temperatura y precipitación) sobre el crecimiento de los árboles; • (iii) describir y analizar las respuestas espacio-temporales de crecimiento a la sequía por medio de un índice de sequía multi-escalar; • (iv) detectar los principales factores geográficos y bioclimáticos que influyen en estas respuestas, Introducción 18 • (v) identificar el efecto de la clima y sequía en la formación de madera en bosques de Pinus halepensis, que es la especie forestal situada en los lugares más áridos de la región estudiada. El estudio incluye ocho especies de árboles que crecen a lo largo de un gradiente climático muy contrastado, y que muestran diferentes vulnerabilidades a la variabilidad del clima: cuatro especies de pino (Pinus halepensis, P. pinea, P. nigra, P. sylvestris), el abeto (Abies alba), la sabina albar (Juniperus thurifera), el quejigo (Quercus faginea) y la carrasca (Quercus ilex subsp. ballota). Estas especies representan bosques de sitios templados y húmedos (por ejemplo, A. alba), submediterráneos de transición (por ejemplo, Q. faginea) y sitios xéricos (por ejemplo, P. halepensis). Las respuestas de crecimiento a la sequía y los principales factores que influyen en este tipo de respuestas se analizaron teniendo en cuenta todas las especies antes mencionadas, mientras que la sensibilidad del crecimiento de los árboles a la variabilidad climática a escala general, regional y local, se evaluó mediante series de madera temprana (EW) y madera tardía (LW) en bosques de P. halepensis por varias razones. En primer lugar, esta conífera es una de las especies arbóreas dominantes en la cuenca del Mediterráneo Occidental y una de las especies ecológicamente más importantes en áreas semi-áridas (Ne'eman y Trabaud, 2000). En segundo lugar, las décadas finales del siglo XX se caracterizaron por una marcada variabilidad climática en la parte oriental de la Península Ibérica (De Luis et al., 2009), en cuyo ámbito el P. halepensis está bien representado, siendo la especie de coníferas dominante con una amplia distribución en el área de estudio. En tercer lugar, la aridez creciente en la región mediterránea podría tener importantes implicaciones para la dinámica de crecimiento a escalas intere intra-anual de los pinares de P. halepensis, sobre todo en zonas más secas. En cuarto lugar, ningún estudio ha investigado cómo diferentes factores climáticos afectan a la producción de EW y LW y cómo son las respuestas a largo plazo del crecimiento a la sequía, a pesar de las evidencias existentes sobre cómo las maderas temprana y tardía responden de manera diferente a diversas variables climáticas (De Luis et al., 2007; Camarero et al., 2010). Introduction 19 CHAPTER I INTRODUCTION In this chapter is briefly given an introduction on the dendrochronology and its applications, particularly focusing on climate effects on tree growth. In addition, this section provides basic information concerning climate variability and drought in the Mediterranean Basin in general and in the Iberian Peninsula in particular, based on the current state-of-the-art relevant literature, and their influence on forest growth in the region. The last part of this chapter justifies the undertaking of this study, the main hypothesis established and the objectives. Introduction 20 Introduction 21 1.1. Dendrochronology The term Dendrochronology, derived from the Greek words “dendro” –tree, wood– and “chronology” –time–, refers to a scientific discipline equipped with a set of principles, techniques and methods based on the precise dating and study of annual growth rings formed by woody plants to extract, sort and interpret information containing the different factors that influence their secondary growth, i.e. wood formation (Fritts, 1976). This science was formally developed during the first half of the 20th century by the astronomer A. E. Douglass (Douglass, 1940), the founder of the Laboratory of Tree-Ring Research at the University of Arizona (Tucson, Arizona, USA). Considering that the majority of trees form annual growth increments or tree rings in cold and temperate areas subjected to a seasonal climate, the information related to its formation and the factors influencing this process can be represented by the specific characteristics of each ring, namely width, anatomy, density, isotopic composition and other visual or analytical parameters that can differ from one ring to the other as a function of ecological and climatic constrains (Schweingruber, 1988). The growth of trees is not a continuous process in areas with seasonal climate since it stops at some point due to the limitation imposed by external climatic factors, thus explaining the formation of annual tree rings. In many areas of the planet with a marked climatic seasonality, the tree growth arrests during unfavorable periods, for instance in cold winter sites from the northern hemisphere or dry periods in arid areas, and re-starts again when weather conditions become favorable. This annual pattern of activity and arrested growth is marked in the structure of wood, in the form of annual concentric layers which, in cross section, are seen as rings. At intra-annual scale, the production of new xylem cells (xylogenesis or wood formation) is fast in the early growing season, slows down as the summer progresses and eventually stops when climatic conditions worsen. These differences in the formation rate of the cells are also reflected in the characteristics of the tree rings in conifers and ring-porous woods, composed of two bands: the earlywood (EW) formed at the beginning of the growing period (spring, early summer) and the latewood (LW) Introduction 22 formed at the end (late summer, autumn). In conifers (pines, firs, etc.) the EW is light and made up of tracheids with wide lumens and thin cell walls while the LW is dark and it is formed by narrow tracheids with thick cell walls (Schweingruber, 1988). The differences in size, density and color between the lignified dark LW part of a ring and the light EW of the next one allow identifying and dating of the rings. Dendrochronology has applications in different fields of the science because the rings contain information on several factors. This information can be retrieved and analyzed conveniently, allowing the study and analyzes of various ecological, geomorphological, climatic, archaeological, etc., processes in space and time whenever they leave a signal in rings which can be adequately extracted (Gutiérrez, 2008). For instance, tree rings can record the effects of fires, outbreaks, earthquakes, volcanic eruptions, landslides, rock falls, avalanches, floods, advance and the retreat of glaciers, etc. All these phenomena can be dated by the signs engraved on tree rings which allows to determine the frequency and intensinty with which these phenomena occur over the affected area. In ecology and following a retrospective approach, dendrochronology is very useful for studying processes that occur at long time scales and imply organisms with long lifespans such as trees. For instance, the study of tree rings allows dating the year of germination and death of trees, which are the two key processes of tree population dynamics and forest succession. The analysis of the annual growth series allows determining the disturbance regimes that might affect forest dynamics and infer tree-to-tree competition intensity. Climate is the most important environmental factor affecting radial growth in woody plants and particularly in trees (Fritts, 1976). When climatic drivers constrain tree growth, the rings formed are commonly narrow in most of the trees of a particular region. As a result of the close relationship between growth and climate, the series of annual rings formed by trees that grow under the same weather conditions show synchrony, its pattern of temporal variation in thickness is very similar among and within trees (Fritts, 1976). This “climatic signature in the wood” is also a “signature of time”, since it is unlikely that a specific Introduction 29 frequency of weather types prone to cause precipitation and linked to cyclonic activity, mainly SW and W weather types, cause warmer and more humid conditions in winter than weather types characterized by more stable conditions, related to anticyclonic activity, which are associated with low precipitation, leading to drought occurrence (Corte-Real et al., 1998; Goodess and Palutikof, 1998; Trigo and Da Camara, 2000; Martín-Vide, 2002; Goodess and Jones, 2002; Esteban et al., 2005; Vicente-Serrano, 2006, 2007; Vicente-Serrano et al., 2009). Drought severity and frequency will ultimately determine the interannual variations of tree growth, vigor and forest activity across a region (Martínez-Vilalta and Piñol, 2002; Vicente-Serrano et al., 2010a; Linares et al., 2010; Sánchez-Salguero et al., 2010; Pasho et al., 2011b). The spatial patterns of precipitation are among the main constraints for forest distribution and tree growth in the Iberian Peninsula, as evidenced by several tree-ring studies based on extensive dendrochronological networks (Macias et al., 2006; Andreu et al., 2007). The impact of water deficit on growth is supposed to be much higher in the most arid sites, where water availability largely constrains the physiological processes controlling tree growth, such as photosynthesis, carbon uptake and nitrogen use, than in mesic ones (e.g., Vicente-Serrano et al., 2006; Andreu et al., 2007; Martínez-Vilalta et al., 2008). However, precipitation variability is not the only factor driving growth patterns; temperature as well, plays an important role because it interacts with precipitation to determine actual water availability for tree growth (Vicente-Serrano et al., 2010a). High temperatures generally have a positive effect on growth under mesic conditions, but their influence can be negative if there is no corresponding increase in precipitation, as drought stress may occur. The decrease in precipitation, already observed in the Circum-Mediterranean region, is increasing the frequency and intensity of droughts (Jones et al., 1996; Romero et al., 1998; Gibelin and Déqué, 2003; Vicente-Serrano, 2006; IPCC, 2007; Giorgi and Lionello, 2008; García-Ruiz et al., 2011) which are believed to be the main drivers of growth decline and related mortality episodes in forests, affecting selectively species, stands and trees (Allen et al., 2010; Koepke et al., 2010). Introduction 30 Drought causes reduction in radial growth, alterations in hydraulic conductivity (McDowell et al., 2008) and decreases productivity due to limitations in water use and photosynthesis (Hsiao, 1973; Flexas and Medrano, 2002). In particular, water deficit affects xylem growth by influencing initially cell turgidity during expansion (e.g., Vysotskaya and Vaganov, 1989; Gindl, 2001), reducing or preventing cell metabolism and indirectly limiting cambial dynamics, wood formation and secondary growth (Larson, 1963; Arend and Fromm, 2007). Figure 2 shows a conceptual model describing how the decreasing precipitation along a climatic gradient increases drought stress, negatively affecting growth and altering the dendrochronological characteristics of tree-ring width series (Fritts, 1976; Scharnweber et al., 2011). The left area of the graph represents trees growing in mesic sites (e.g., humid mountain forests) where tree-ring width series show high mean values but the lowest year-to-year variability (mean sensitivity), i.e. they are complacent series, while the area to the right represents trees growing in dry sites (e.g., semi-arid woodlands) where tree-ring width series show low values but the highest mean sensitivities and variability, corresponding to sensitive series, and also the highest correlation in growth among co-existing trees and the highest response to climate. Figure 2. Theoretical variation of dendrochronological characteristics along a gradient of increasing drought stress (modified from Fritts, 1976; Scharnweber et al., 2011). The right dashed line indicates the xeric limit of tree distribution, i.e. the forest border. Introduction 31 However, climatic factors are not the only drivers which condition drought effects on tree growth. Site factors as well, such as elevation or topography or soil conditions, affect the degree to which climatic drivers and drought impact tree growth (Fritts, 1976; Orwig and Abrams, 1997). These local drivers may impose additional risks exacerbating drought stress (e.g., rocky soils or steep slopes) or they may partially mitigate its negative effects on tree growth (e.g., deep soils or northern aspects). For example, the topographic factors may impose local variations in the amounts of water retained by soils (Tardif et al., 2003). The differences in site exposure can affect temperature, wind and radiation, which, in turn, influence the rate at which moisture is lost by evapotranspiration processes leading to different drought conditions (Fritts, 1976). Therefore, to understand in depth how climate variability processes are determining tree growth in a region so complex in climate and forest composition like the eastern Iberian Peninsula, it is necessary to follow a methodological approach that consider the climate variability on different spatial and temporal scales. This approach must allow understanding i) how the general atmospheric circulation drives changes in surface climate and affects tree growth, ii) how is the impact of drought on growth, iii) how this impact varies as a function of forest type and bioclimatic conditions of the region and, finally, iv) how seasonal variations of drought intensity impact wood formation. 1.3. Justification of the thesis, hypotheses and objectives. Most analyses of climate impact on forest growth have focused on the influence of surface climate factors (Fritts, 1976). Nevertheless, atmospheric circulation patterns affect climate variability over large regions (Garfin, 1998; Girardin and Tardif, 2005). Moreover, it is expected that the main signs of climate change will be identified earlier through changes in atmospheric circulation (Giorgi and Mearns, 1991; Räisänen et al., 2004). The predicted changes in atmospheric circulation and climatic conditions are expected to affect forest growth in the Mediterranean Basin directly through changes in large-scale circulation patterns but also indirectly through modifications in regional atmospheric Introduction 32 patterns and proximate local climate factors such as wind patterns and precipitation. Therefore, it is particularly important to investigate both the direct and indirect influences of large-scale atmospheric circulation paterns on tree growth, particularly in areas subjected to diverse climatic conditions. This is the case of the Iberian Peninsula, where climate ranges from mild to continental and from humid to semi-arid types, resulting in diverse climatic conditions and forest types subjected to different growth constraints (Nahal, 1981). In addition, the large spatial differences found at local scales in the response of forests to drought (Sarris et al., 2007; Andreu et al., 2007; Martìnez-Villalta et al., 2008; Sánchez-Salguero et al., 2010; Linares et al., 2010; Vicente-Serrano et al., 2010a), suggest the importance of analyzing forest growth in relation to local climatic gradients, and the need to consider forests near their limit of distribution, as the first impacts of climate change processes are expected to be observed in these ecotones (Neilson, 1993). While the predicted trend towards progressively drier conditions is likely to cause a decline in the growth of Mediterranean forests, the spatial extent and the magnitude of the effect of atmospheric and climatic drivers on tree growth is uncertain (Andreu et al., 2007; Vicente-Serrano et al., 2010a). Moreover, the large seasonality and year-to-year variability that characterize precipitation in the Mediterranean region and the different site-dependent seasonality of tree growth in forests from this area could make very difficult to determine the time response of tree growth to the precipitation deficit. Lags between water shortages and growth can appear as a function of different anatomical and physiological adjustments of trees to cope with drought stress but also in response to drought severity and duration, and to the season in which water deficit occurs. All these mechanisms, either isolated or acting synergistically, can challenge the identification of drought impacts on tree growth. Currently, a deeper knowledge on the tree growth responses to water shortages in the Mediterranean Basin is a crucial task since General Climate Change Models predict a large reduction of precipitation and an increase of the evapotranspiration rates by the end of the twentieth-first century (Giorgi and Lionello, 2008; García-Ruiz et al., 2011). There exists also contradictory information regarding the drought impacts on growth of tree species at different time scales and across Introduction 33 contrasting site conditions (Orwig and Abrams, 1997; Adams and Kolb, 2005). Moreover, the quantification of intraand inter-annual wood development in response to drought stress assessed at different time scales is crucial for understanding the increased sensitivity of radial growth in response to the amplified precipitation variability detected in Mediterranean region during the past century (Andreu et al., 2007; Camarero et al., 2010). Considering the aforementioned issues, it was hypothesized (i) that the growth responses of forests to the atmospheric circulation patterns and climate variability in north-eastern Spain must be structured, (ii) that there must be contrasting growth responses to drought among and within tree species and also among sites with different local conditions (xeric vs. mesic sites). In particular, it was hypothesized that forests located in xeric sites will show a higher responsiveness to drought by reducing the wood formation compared to mesic locations where temperature is expected to be the major constrain of growth. However, these responses may also depend on seasonal climatic trends and on the high phenological plasticity of xylogenesis in Iberian tree species (De Luis et al., 2007; Camarero et al., 2010). The objectives of this thesis were:  (i) to use dendrochronological methods to characterize the spatial and temporal patterns of radial growth in a network of forests located in north-eastern Spain, where a wide climatic gradient exists ranging from semi-arid to cold and wet climatic conditions. The sampled sites include diverse forest types such as Mediterranean woodlands under semi-arid conditions in the Middle Ebro Basin to mountain forests under humid conditions in the Pyrenees;  (ii) to quantify the influence of large-scale circulation patterns, regional weather types and the surface climate variables (temperature and precipitation) on tree growth;  (iii) to describe and analyze the spatio-temporal growth responses to drought by means of a multi-scalar drought index;  (iv) to detect the main geographic and bioclimatic drivers influencing these responses; Introduction 34  (v) to identify the effect of climate and drought on seasonal wood formation in Pinus halepensis which is the tree specie located in the most arid sites of the studied region. The study includes eight tree species growing along a wide climatic gradient, showing contrasting vulnerability to climate variability: four pine species (Pinus halepensis, P. pinea, P. nigra, P. sylvestris), silver fir (Abies alba), Spanish juniper (Juniperus thurifera), and two oak species (Quercus faginea, Q. ilex). These species represent species typically associated with mesic sites and humid conditions (e.g., A. alba), transitional sub-Mediterranean locations (e.g., Q. faginea) and xeric sites (e.g., P. halepensis). The spatio-temporal growth responses to drought and the main drivers influencing such responses were analyzed considering all of the above mentioned species, while the sensitivity of tree growth to climatic variability at large, regional and local scales was evaluated on earlywood (EW) and latewood (LW) width series of P. halepensis, for several reasons. First, this conifer is one of the dominant tree species in the Western Mediterranean Basin and the most ecologically important species in semi-arid woodlands (Ne´eman and Trabaud, 2000). Second, the late 20th century was characterized by marked climatic variability in the eastern Iberian Peninsula (De Luis et al., 2009) and in all these areas P. halepensis is present, being the dominant conifer species with a broad distribution across the study area. Third, the increasing aridity in the Mediterranean region could have important implications for inter and intra-annual growth dynamics of P. halepensis, particularly in drought-prone areas. Fourth, no studies have investigated how climatic drivers affect EW and LW production and the long-term responses to drought of these components, despite ample evidence that each of these components of P. halepensis growth responds differently to diverse climatic variables (De Luis et al., 2007; Camarero et al., 2010). Methodology 35 CHAPTER II METHODOLOGY This chapter provides general information about the study area, the methodological possibilities to carry out the research and indicates the specific methods applied to investigate the responses of forests to climatic drivers and drought. In particular, this section describes the available dense network of tree ring samples, the climate dataset, the drought indices and the general statistical analysis employed. Methodology 36 Methodology 37 2.1. Study sites and species The study area includes forests in Aragón and Valencia regions, eastern Spain (Figure 3, Table 1). The Aragón region is subjected to continental Mediterranean conditions with a typical summer drought and it is characterized by a strong climatic gradient ranging from a semiarid climate in the Middle Ebro Basin (mean annual temperature 13.1 ºC, total annual precipitation 558 mm) to humid climates in the Pyrenees and Pre-Pyrenees (mean annual temperature 8.5 ºC, total annual precipitation 1750 mm). In the Middle Ebro Basin the average temperatures in January and July are about 6.7 ºC and 26.0 ºC, whereas in Pyrenean stations the means for both months are about 1.5 ºC and 16.8 ºC, respectively (Cuadrat et al., 2007). Figure 3. Distribution of forests (gray areas) in eastern Spain and the location of study sites grouped in five major regions (Pyrenees, Pre-Pyrenees, Middle Ebro Basin, Iberian Range and Valencia region). Different symbols represent different forest species (see sites names in Table 1). Methodology 38 In Valencia region the climate is typically Mediterranean with high daily, seasonal and interannual precipitation variability (De Luis et al., 2001). There, the mean annual temperature is about 15.7ºC and the mean annual sum of precipitation about 466mm (De Luis et al., 2009). Seasonal variability in precipitation is less pronounced in the Pyrenees and Pre-Pyrenees than in the Ebro Basin and Valencia region, although summer is the predominant dry season in all study sites. In the Middle Ebro Basin and Valencia region there is a negative water balance (precipitation minus evapotranspiration), as a consequence of the high potential evapotranspiration (PET) that occurs in summer. Annual PET reaches 1300 mm in some sectors of the Middle Ebro valley. Moreover, in the Middle Ebro Basin and Valencia region, the high temporal variability in precipitation constrains tree growth, as severe droughts are frequent (Vicente-Serrano and Cuadrat, 2007), and periods of more than 80 days without precipitation are common (Vicente-Serrano and Beguería, 2003). In the Middle Ebro Basin the lithology is characterized by millestones and gypsums (Peña et al., 2002), which contribute to aridity because there is poor retention of water by these soils (Navas and Machín, 1998). In the Pyrenees, soils are usually deep and basic and develop over limestone, sandstones and granites. The Pre-Pyrenees form a mountain chain characterized by a transitional sub-Mediterranean climate with a mean annual temperature of 11.1 ºC and mean annual precipitation of about 950 mm (Cuadrat et al., 2007). In the Valencia region the landscape shows widely contrasting geological features represented by a diversity of soils. The bedrock material is dominated by limestone, dolomites, calcareous substrates which produce shallow soils and marls characterized by deep soils (Kazanis et al., 2007). The study area contains very different forest types (Figure 4) in terms of the dominant forest species and vegetation communities (Costa et al., 2005). Most of the studied forests are located in the mountain chains (Pyrenees and Pre-Pyrenees in the north, and the Iberian range in the south) forming pure conifer (e.g., silver fir, Abies alba Mill.) or mixed conifer-hardwood forests (e.g., silver fir-beech forests). In the northern study area, the Pyrenees and Pre-Pyrenees constitute transitional and mountainous areas between more humid conditions northwards or upwards and Methodology 45 to what extent the sample size is representative of a theoretical infinite population (Wigley et al., 1984) and the mean correlation (Rbar) among individual residual series within each site. Dendrochronologycal methods provide useful information on climate-growth relationships based on year-to-year correlative approaches but these approaches are unable to capture intraannual growth variability and delayed growth responses to climate (Camarero et al., 1998). Therefore, dendrochronological information is necessarily to be compared with xylogenesis studies which describe the process of xylem formation and capture the intra-annual growth responses to climate (Camarero et al., 1998; Rossi et al., 2006b). This analysis provides the necessary information to fully understand the dendrochronological patterns and provide mechanisms explaining radial-growth responses to climate variability and drought stress. Xylogenesis studies are now widely employed to provide new data to explain previously established climate–growth relationships (Deslauriers et al., 2003; Rossi et al., 2006b; Vaganov et al., 2006; De Luis et al., 2007; Camarero et al., 2010). Therefore, xylogenesis was also used in this study (see a more detailed methodological description of this approach in section 3.3). 2.3. Climatic data Different climate information has been used in this study. Firstly, to characterize atmospheric circulation patterns on different spatial scales, several atmospheric circulation indices and weather types were employed. Secondly, to describe the spatial and temporal variability of the surface climatic conditions in the study area, precipitation and temperature data from different climatic databases were used. The following sections describe the climatic data available for this study and the importance of using them to study climate-growth relationships. 2.3.1 Atmospheric circulation The large scale atmospheric circulation determines the regional atmospheric variability (frequency of weather types) and the surface climate conditions (spatio-temporal variations of temperature and Methodology 46 precipitation) in the Iberian Peninsula (Muñoz-Díaz and Rodrigo, 2004b; Vicente-Serrano and López-Moreno, 2005; Martín-Vide and López-Bustins, 2006; Vicente-Serrano and Lopez-Moreno, 2006). Various studies have investigated the direct influence of the spatio-temporal variations of temperature and precipitation on tree growth variability in the Mediterranean region (Macias et al., 2006; Andreu et al., 2007). However, as the variations in temperature and precipitation are controlled by the atmospheric circulation, it is expected that apart from the local climatic conditions, atmospheric circulation as well should exert some influence on the spatial patterns of tree growth in the region. According to Bijak (2009), temperature and precipitation do not explain the whole variability observed in tree-ring width series. This may suggest that tree-ring formation could be partially under the influence of more general factors. In other words, it remains uncertain what is the magnitude of the effect of atmospheric circulation variability on tree growth in the Mediterranean region. Most of the existing studues regarding the influence of atmospheric circulation on tree growth in the Mediterranean region have been focused on the North Atlantic Oscillation (NAO). For example, Piovesan and Schirone (2000) found high signals of NAO on tree-ring growth of Fagus sylvatica in Italy. Tree-ring widths of Pinus sylvestris trees in Fennoscandia were also significantly related to variations in the winter NAO (D’Arrigo et al., 1993). Solberg et al. (2002) found negative relationships between Picea abies growth and the NAO in central Norway. Moreover, links between tree rings and the NAO index have been used to reconstruct the NAO index using tree-rings as proxies (Cook et al., 1998; Jones et al., 2001; Cullen et al., 2001; Schultz et al., 2008). In the Iberian Peninsula, only a few studies have investigated the impact of broad scale atmospheric patterns (mostly NAO) on forest growth (Bogino and Bravo, 2008; Roig et al., 2009; Rozas et al., 2009). In particular, Roig et al. (2009) used tree ring series from two deciduous species of western Iberia to reconstruct the NAO index. Rozas et al. (2009) studied the responses of two Pinus and Quercus species to NAO variations in north-western Spain. Bogino and Bravo (2008) Methodology 47 analyzed the impact of NAO on Pinus pinaster growth in eastern Spain. However, the eastern Spain is also affected by other atmospheric circulation patterns such as Western Mediterranean Oscillation (WeMO) and the Mediterranean Oscillation (MO) (González-Hidalgo et al., 2009; Vicente-Serrano et al., 2009). In addition, the general atmospheric circulation variability is propagated regionally by means of a series of weather types which control the intensity and spatial distribution of precipitation at local scales (Goodess and Jones, 2002; Vicente-Serrano and López-Moreno, 2006). Therefore it is crucial to consider all these circulation patterns and the frequency of regional weather types as possible drivers of spatio-temporal variability of forest growth in the study area. For this purpose, P. halepensis forests, covering a wide climatic gradient in eastern Spain, were selected to investigate the sensitivity of tree growth to climate variability at broad and local scales. Details of the specific methodological approach implemented to obtain the atmospheric circulation indices used in this study are given in section 3.1. 2.3.2. Temperature and precipitation data Various temperature and precipitation climatic datasets based on AEMET (Agencia Estatal de Meteorología, Spanish Meteorological Agency) data are available for Spain. Most of them consist of precipitation series (Romero et al., 1998; González-Rouco et al., 2001; González-Hidalgo et al., 2004, 2011; Brunetti et al., 2004; De Luis et al., 2009) and less deal with temperature data (Morales et al., 2005; Brunet et al., 2006, 2007). The main purpose behind these studies was to develop complete and homogeneous climatic datasets with improved quality. However, several problems such as spatial density, temporal coverage, quality control and homogenization are usually associated with the available climatic datasets in the region. For example, Brunet et al. (2006) developed a new daily adjusted dataset of 22 observatories of maximum and minimum temperatures in the whole Spain but the spatial coverage was inadequate for several spatial studies. Romero et al. (1998) created 410 complete daily precipitation series for the Spanish Mediterranean area, using information derived from 3366 individual series but the homogeneity of the resultant series was not Methodology 48 checked and the dataset is temporally limited (period 1964-1993). To overcome these problems and to construct reliable databases of precipitation and temperature series, a full process of reconstruction, quality control and homogenization of the climatic data is needed (Vicente-Serrano et al., 2010b). Recently, this approach was followed by Vicente-Serrano et al. (2010b) and El Kenawy et al. (2011) to develop two complete, dense, reliable and homogeneous databases of daily precipitation and temperature series for north-eastern Spain. Different climatic datasets, either based on local observatories or transformed into gridded values, were used to study climate-growth relationships in the study area. In this regard, for the sampled forests located near the populated areas (Middle Ebro Basin), where a considerable high number of local meteorological stations are available, two homogeneous and spatially dense datasets of daily precipitation and temperature series were employed (Vicente-Serrano et al., 2010b; El Kenawy et al., 2011). Since these datasets are covering only north-eastern Spain, additional climatic data of precipitation and temperature series obtained from the MOPREDAS dataset (González-Hidalgo et al., 2011) and the AEMET were used for the forest sites located in Valencia and Alicante provinces. Moreover, considering that many of the sampled forests are located in mountainous areas (Pre-Pyrenees and Pyrenees) where only few meteorological stations are available, the existing local precipitation data were interpolated at a spatial resolution of 1000 m (gridded data) and converted to monthly data to have a regular grid with information in each one of the sampled forest (Vicente-Serrano et al., 2010b). To take into account the effect of elevation on precipitation, and to have more reliable estimations for each forest, the interpolation was done using a Digital Terrain Model and a Geographic Information System (GIS)-assisted regression-based approach (Ninyerola et al., 2000,2007; Vicente-Serrano and Beguería, 2003; Vicente-Serrano, 2007). The precipitation in each 1000 m grid point was estimated, for each month between 1950 and 2006, by means of a stepwise-regression model, in which the independent variables were the elevation, the latitude and the longitude of each site. The residuals, i.e. the differences between the observed and modeled precipitation, were also included in the estimations by means of a local Methodology 49 interpolation procedure (splines with tension - Mitasova and Mitas, 1993) to include the local precipitation features recorded each month, which were not well represented by the regression models. The validation of the grid layers was done for each monthly layer by a jackknifing method, based on withholding, in turn, one station out of the network, estimating regression coefficients from the remaining observatories and calculating the difference between the predicted and observed value for each withheld observatory (Phillips et al., 1992). This method has been frequently used in climatology (e.g. Daly et al., 1994; Holdaway, 1996; Hofstra et al., 2008). The average Root Mean Square Error for the different months and years was 15.2 mm, being lower in summer (6.5 mm) than in winter (22.3 mm). The D agreement index (Willmott, 1982) showed an average of 0.94 for the different monthly layers, with a range between 0.82 and 0.99, which indicates a high reliability between the observed and modeled precipitation data. 2.4. Drought indexes The drought severity is frequently quantified by drought indices which consider the complexity to determine the magnitude, duration and surface extent of droughts (Wilhite and Glantz, 1985; Redmond, 2002). For these reasons, numerous efforts have been made to develop methods for quantifying drought severity. The main efforts consisted of developing drought indices that enable earlier identification of droughts, quantification of their severity and spatial extent. Several drought indices, using diverse variables and parameters for drought quantification, were developed during the 20th century (Du Pisani et al., 1998; Heim, 2002). Most studies related to drought analysis have been conducted using either (i) the Palmer Drought Severity Index (PDSI; Palmer, 1965), based on a soil water balance equation, or (ii) the Standardized Precipitation Index (SPI; McKee et al., 1993), based on a precipitation probabilistic approach. The PDSI has numerous deficiencies (VicenteSerrano et al., 2011) but its main shortcoming for the identification of drought impacts is the fixed Methodology 50 temporal scale used to calculate it (Guttman, 1998). This is not in agreement with the frequently accepted fact that drought is a multi-scalar, i.e. time-dependent, phenomenon considering that the period from the water shortages to impacts in a given system differs noticeably. Consequently, drought indices must be associated with a specific time scale to be useful for monitoring drought impacts and be comparable in time and space (Guttman, 1998; Hayes et al., 1999). Among the existing drought indices (Heim, 2002; Mishra and Singh, 2010) only the SPI and the Standardized Precipitation Evaporation Index (SPEI) can be obtained at different time scales. The SPI is calculated using precipitation data to identify the varied times of response of different hydrological systems to precipitation deficits in a better way than other indices like the PDSI (McKee et al.,1993). The SPEI was developed by Vicente-Serrano et al. (2010c) to include both precipitation and temperature influence on droughts by means of the evapotranspiration processes. Both indices have the advantage of allowing the determination of duration, magnitude and intensity of droughts and can be calculated at different time scales. The later is important in determining ecological impacts of droughts, considering the different time-dependent physiological strategies of vegetation to deal with water deficit (Hsiao, 1973). However, in this study, only the SPI was employed to analyze the influence of drought on tree growth (all species) since the responses of forests to long-term temperature anomalies can be very different in mesic vs. xeric sites. For example, in mesic sites, where water availability is high, high temperatures can favor forest growth by increasing photosynthetic activity, whereas in xeric sites, characterized by frequent water deficit, the same process increases drought stress leading to growth decline and even die off (Jump et al., 2006; Martínez-Vilalta et al., 2008; Vicente-Serrano et al., 2010a). Therefore, depending on water availability, the effect of temperature increase can be positive or negative for forest growth in the respective sites. This is why the SPEI was not included in the drought-growth analysis, since the study considers both xeric and mesic sites. In addition, precipitation is a very important variable explaining the frequency, duration and severity of droughts (Chang and Cleopa, 1991; Heim, 2002) and its variations can markedly influence spatio- Methodology 51 temporal patterns of tree growth, particularly in water-limited environments (Macias et al., 2006; Andreu et al., 2007). 2.5. Factors affecting growth-drought responses: climate, topography and vegetation activity Remote sensing data and GIS-techniques were used in this study to evaluate the potential roles of the differences in vegetation activity and topographical variables in explaining the spatial differences in the growth responses to drought in each sampled forest. Remote sensing offers a feasible tool to objectively and systematically monitor vegetation condition by estimating photosynthetically active vegetation throughout the growing season. The major interest of remote sensing imagery lies on the possibility of extrapolating acquired data at pixel resolution, to get spatially continuous information less costly than ground surveys and in a relatively short time. The utility of remote sensing for vegetation monitoring is based on the response of vegetation cover to radiation in the visible and near-infrared regions of the electromagnetic spectrum (Myneni et al., 1995). Visible radiation is mainly absorbed by vegetation in photosynthesis processes while near infrared radiation is principally reflected, owing to the internal structure of leaves (Knipling, 1970). High vegetation activity is characterized by low reflectivity of solar visible radiation and high reflectivity in the near infrared region of the spectrum. GIS-related technologies have been widely used in a variety of ecological applications. According to Booth and Tueller (2003), GIS is a powerful tool for integrating and analyzing data derived from remotely sensed imagery interpretations, soil surveys, vegetation maps, land ownership maps, utility maps, water resources, geology and many other potential themes that can be presented spatially. These geographically referenced data sets are spatially registered so that multiple themes of data can be quickly compared and analyzed together. According to Buchan (1997), satellite images are used to identify what is growing, while the GIS component is used to further analyze its position on the earth, measure area, etc, providing in this way a complete record of the site. Methodology 52 Different indices have been developed for monitoring and measuring vegetation status using spectral data (Bannari et al., 1995). Among them, the most widely used is the Normalized Differences Vegetation Index (NDVI) (Rouse et al., 1973). Numerous authors have pointed out the close relationship between NDVI, vegetation activity and tree radial growth (e.g., Wang et al., 2004; Pettorelli et al., 2005; Lopatin et al., 2006; Kaufmann et al., 2008; Khabarova et al., 2010; Vicente-Serrano et al., 2010a; Julien et al., 2011). The NDVI measures the fractional absorbed photosynthetically active radiation (Myneni et al., 1995) and exhibits a strong relationship with the green leaf-area index (Carlson and Ripley, 1997). In addition, the Enhanced Vegetation Index (EVI) enhances the vegetation signal with improved sensitivity in regions with high biomass (e.g, dense forests) and allows improved vegetation monitoring through a decoupling of the canopy background signal and a reduction in atmosphere influences (Huete et al., 2002). NDVI is an excellent measure of the photosynthetic activity but it has some limitations to analyze vegetation activity in dense forests since the relationship between vegetation parameters (leaf area, annual net primary production (ANPP), vegetation coverage, etc.) and the NDVI are sometimes non-linear because the NDVI saturates before the maximum biomass is reached (Carlson et al., 1990). On the contrary, the EVI was developed to improve the vegetation signal and provide a more accurate measure of vegetation activity in dense biomass regions. Since the canopy cover of the forests located in the most arid sites of the study area (mainly P. halepensis) is not very dense while the opposite occurs with the Pyrenean silver fir stands, it is justified the use of both indices to provide complementary information as the best measure of photosynthetic activity in the respective areas. The data was obtained from the products of Moderate Resolution Imaging Spectroradiometer (MODIS 13A1 product, 16-day at 500 m resolution; available at http://www. daac.ornl.gov /MODIS/modis.html in HDF format; see Huete et al., 2002) which provide significant refinements in spectral, radiometric, and geometric properties compared to previously available data sets with similar spatial resolution (Justice & Townshend, 2002; Zhang et al., 2004). Data processing included images re-projection from a Sinusoidal to a UTM-30N-S/IGN projection, Methodology 53 images stacking to provide a full coverage of the study area and crossing of the images with forest sites location to extract the NDVI and EVI values at each sampled forest. 2.6. Statistical analyses In this section are described the main statistical analysis applied in the framework of this study, providing also a justification of their use to investigate the relationships between the climatic variables, drought and the spatio-temporal variations of tree growth in the study area. 2.6.1. Principal component analysis (PCA) The main purpose of the PCA analysis is to reduce the dimensionality of the data into a few, uncorrelated components, leading to more understandable and clearly interpretable dataset (Vicente-Serrano et al., 1999). There are six possible modes of the PCA called O, P, Q, R, S, and T (Richman, 1986). The modes differ according to which parameter is chosen as variable, which of them is considered as individual case and which as a fixed entity. In the case that the parameter has been fixed, there remain two options: the S and T modes. The T-mode is the result of choosing the individual observations as variables and sites as cases of those variables. When rotated, T-mode identifies subgroups of observations with similar spatial patterns. The S-mode considers the sites as variables and the observations as cases. The S-mode compares the sites and identifies those which show similarity in a particular observation. In other words, this method enables common features to be identified and specific relevant local characteristics to be detected (Richman, 1986). This is why the S-mode PCA was selected as the most appropriate one to be applied for this study to summarize the spatio-temporal variability of forest growth in the large data set used. 2.6.2. Correlation analyses Correlation analysis is a widely used statistical technique for studying climate-growth relationships (Briffa and Cook, 1990). Among the several correlation coefficients available, in this study it was Methodology 54 employed the parametric Pearson correlation because the climate and growth data used, fullefilled the requirements of normal distribution and homogeneity of the variance (Clark and Hosking, 1986). Therefore, this correlation coefficient was used for several purposes as detailed in the specific sections of the results chapter. 2.6.3. Superposed epoch analysis (SEA) Superposed epoch analysis (SEA) is a non-parametric technique used to examine how punctual events modifiy the values of a continuously derived variable such as growth (Hoenig, 1989; Haurwitz and Brier, 1981). Since the statistical significance is determined by a randomization test, SEA is a robuts approach which does not rely on the usual assumptions (normality, homogeneity of variance, independence of observations) of parametric testing. In a previous dendrochronological study (Martín-Benito et al., 2008), SEA was used to assess the relationship between extreme climatic events (e.g. drought) and tree growth response in the corresponding and following years. For each drought occurrence, the tree ring data is selected from a window of years preceeding, including, and after the event occurs. In this regard, SEA was employed in this study to investigate the impacts of drought severity on EW and LW growth in P. halepensis, during the current and following (post-drought) two years of growth. 2.6.4. Regression analyses Linear regression analysis is a statistical technique used to predict the response of a variable (predictand) as a function of another one (predictor) assuming they are linearly related. To explore such relationships, the data is organised into variables of interest and regression is applied to estimate the quantitative effect of the causal variables upon the variable that they influence. Regression analysis with a single explanatory variable is termed “simple regression”, while “multiple regression” is a technique that allows additional factors to enter the analysis separately so that the effect of each of them can be estimated. Multiple regression analysis has shown to be Results 61 Figure 6. Distribution of P. halepensis in the Mediterranean Basin (map was taken from EUFORGEN, http://www.euforgen.org/distribution_maps.html) and eastern Spain (gray area) and the location of the study sites (black points). The sites codes are explained in Table 1. Specific methods Dendrochronological methods The samples preparation, cross-dating and standartization were performed following standard dendrochronological methods (see section 2.2). The common period 1960-2003 was selected because all sites residual chronologies showed EPS values above the 0.85 threshold, which is widely used in dendrochronological studies (Wigley et al., 1984). Atmospheric circulation patterns The NAO, MO and WeMO atmospheric circulation patterns, which affect autumn and winter climatic conditions (particularly precipitation) over eastern Spain, were selected following VicenteSerrano et al. (2009) and González-Hidalgo et al. (2009). Autumn (September to November), spring Results 62 (April to May), summer (June to August) and winter (December to March) indices were calculated. To calculate the seasonal circulation indices were used the monthly SLP grids from the NCEPNCAR ds010.1 Monthly Northern Hemisphere Sea Level Pressure Grids (http://dss.ucar.edu/ datasets/ds010.1/; Trenberth and Paolino, 1980). This dataset contains complete records for the study period (1960–2003), with a spatial resolution of 5°. The atmospheric circulation indices were calculated monthly from the differences between the series of standardized SLPs recorded at the two points closest to the sites most used to calculate these indices: Gibraltar (south of the Iberian Peninsula; 35° N, 5° W) and Rejkiavic in Iceland (65° N, 20° W) in the case of the NAO (Jones et al., 1997); Gibraltar and Lod in Israel (30° N, 35° E) in the case of MO (Palutikof, 2003); and Gibraltar and Padova (Italy) (45° N, 10° E) in the case of the WeMO (Martín-Vide and LopezBustins, 2006). Seasonal atmospheric circulation indices were obtained from the average of the monthly series. Classification of weather types The general atmospheric circulation, well represented in East Spain by means of the general atmospheric circulation patterns cited above, is propagated regionally by means of different weather types that represent pressure fields and winter flows with a noticeable role on the surface climate conditions (e.g., precipitation and temperature) (Yarnal et al., 2001). On one hand, a high frequency of weather types prone to cause precipitation would tend to produce humid conditions. On the other hand, weather types characterized by stability conditions will be the direct cause of droughts. The influence of the frequency of weather types on the surface climate in eastern Spain (e.g., VicenteSerrano and López-Moreno, 2006) justifies their use to investigate the possible influence on tree radial growth. Several attempts have been made to develop classification methods based on different categories of weather type (see review in Yarnal et al., 2001). Among these, automatic methods allow the construction of homogeneous daily or monthly series of atmospheric climatic conditions, Results 63 at local and regional scales. The most widely used automatic method to classify weather types is that formulated by Jenkinson and Collison (1977), which is based on the Lamb (1972) catalogue. This has been widely used to classify weather types in the Iberian Peninsula (Spellman, 2000; Trigo and DaCamara, 2000; Goodess and Jones, 2002; Vicente-Serrano and López-Moreno, 2006; LópezMoreno and Vicente-Serrano, 2007). To obtain a daily classification of weather types it was used a sea surface pressure grid of 16 points centered over the Iberian Peninsula (see Figure 1 in VicenteSerrano and López Moreno, 2006). From daily pressure data at these points over the period 19602003 were calculated the type and direction of winds (cyclonic/anticyclonic, directional and hybrid) on which to base a classification of weather types. For this purpose it was used again the NCEPNCAR Northern Hemisphere Sea Level Pressure Grids, but at a daily time scale (http://dss.ucar.edu/datasets/ds010.0). Quantitative monthly series can be derived from the daily weather types using the sum of the number of weather types in each class during the month (CorteReal et al., 1998). The 26 weather types obtained using Jenkinson and Collison’s method were summarized by the elimination of hybrid types, which were reclassified at 50% to cyclonic (C), anticyclonic (A) or directional weather types (N, north; NE, northeast; E, east; SE, southeast; S, south; SW, southwest; W, west; and NW, northwest) (Trigo and DaCamara, 2000; Vicente-Serrano and López-Moreno, 2006). Seasonal series of the frequency of the 10 weather types from 1960 to 2003 were related to P. halepensis growth. Climate data To explain the mechanisms driving the influence of the atmospheric circulation processes on the EW and LW growth, were used data of monthly precipitation and temperature from 1960 to 2003 for each sampled forest. The monthly climatic data were grouped seasonally following the same approach as for the atmospheric circulation patterns: Autumn (September to November), spring (April to May), summer (June to August) and winter (December to March). Results 64 Statistical analyses The PCA analysis (S-mode) was applied to determine the spatial patterns in the inter annual variability of EW and LW growth in P. halepensis forests across the study area for the period 19602003. The areas represented by each component, showing the spatial patterns of EW and LW growth, were identified by mapping the factorial loadings. The PCA was performed on a covariance matrix calculated among the chronologies (Legendre and Legendre, 1998). The number of components was selected using the criterion of an eigenvalue > 1, the components were rotated (Varimax) to redistribute the final explained variance and to obtain more stable and robust spatial patterns (Richman, 1986; Garfin, 1998). The spatial classification of EW and LW growth was carried out using the factorial loading values obtained for each component, with the forests being grouped using the maximum loading rule. Each forest was assigned to the component with the greatest loading value. This method has been applied in many climatic classification studies (e.g. Comrie and Glenn, 1998). To explain the influence of variability in atmospheric circulation on the spatio-temporal patterns of EW and LW growth, correlation analyses were carried out (Briffa and Cook, 1990). To compare with the de-trended forest growth series and to avoid the possibility that atmospheric circulation trends could disrupt potential relationships, prior to assessing the correlations, the trend in each of the atmospheric circulation series was removed by fitting a linear trend in each series. Correlation analyses were performed between EW and LW residual indices, de-trended monthly and seasonally atmospheric circulation indices, and the de-trended series of the frequency of weather types for the period 19602003. The joint influence of climate and EW on LW was obtained by carrying out partial correlation analysis. Finally, to determine climate processes that drive the influence of atmospheric circulation on forest growth, it was calculated the correlation between the atmospheric circulation patterns and the surface climate as well as the correlation between EW, LW and the seasonal precipitation and temperature at each sampling site. Results 65 Results Earlywood and latewood chronologies Initially, the statistical characteristics of the EW and LW series in each P. halepensis forest used in this section are shown. EW growth varied more than LW growth among sites (EW, 0.62-2.29 mm; LW, 0.15-0.66 mm) (Table 2, Figure 7). The average values of first-order autocorrelation (AC1) and mean senistivity (MSx) were higher for the EW (AC1 = 0.64, MSx = 0.39) than for the LW (AC1 = 0.49, MSx = 0.33) chronologies. Similar results were obtained for the mean correlation (Rbar) among individual series within each site; and the expressed population signal (EPS), which were also higher for the EW (Rbar = 0.60, EPS = 0.97) than for the LW (Rbar = 0.40, EPS = 0.93) series. Table 2. Dendrochronological statistics of earlywood (EW) and latewood (LW) P. halepensis chronologies for the common period 1960-2003. The column EW-LW shows the Pearson correlation coefficient calculated between the series of both variables. EW LW Site Trees (radii) Period EWLW MW (mm) SD (mm) AC1 MSx Rbar EPS MW (mm) SD (mm) AC1 MSx Rbar EPS AY 19 (33) 1946-2006 0.56 2.07 0.91 0.60 0.21 0.34 0.93 0.58 0.39 0.33 0.28 0.28 0.92 GR 15 (30) 1946-2006 0.46 1.78 1.02 0.69 0.29 0.54 0.97 0.51 0.29 0.35 0.32 0.33 0.93 VA 17 (31) 1925-2009 0.65 1.4 0.91 0.70 0.36 0.61 0.96 0.41 0.28 0.66 0.31 0.43 0.95 CV 13 (23) 1928-2009 0.42 1.79 1.07 0.74 0.28 0.48 0.92 0.56 0.38 0.70 0.29 0.33 0.90 CS 13 (23) 1900-2009 0.64 0.90 0.59 0.57 0.33 0.69 0.98 0.25 0.19 0.50 0.32 0.40 0.93 VL 15 (29) 1878-2006 0.68 0.94 0.65 0.55 0.44 0.55 0.96 0.27 0.23 0.48 0.38 0.39 0.93 PU 15 (22) 1943-2009 0.58 1.12 0.74 0.65 0.34 0.67 0.98 0.36 0.20 0.55 0.30 0.39 0.92 CP 14 (27) 1927-2006 0.62 1.58 1.05 0.70 0.36 0.57 0.96 0.42 0.26 0.55 0.26 0.31 0.91 FR 16 (29) 1844-2006 0.57 0.72 0.63 0.68 0.51 0.63 0.97 0.25 0.21 0.61 0.34 0.37 0.93 VM 17 (30) 1959-2009 0.57 1.43 0.83 0.62 0.34 0.8 0.99 0.40 0.23 0.55 0.28 0.52 0.96 CA 16 (28) 1845-2007 0.65 0.62 0.51 0.55 0.64 0.71 0.98 0.15 0.11 0.39 0.42 0.41 0.94 DA 14 (28) 1934-2006 0.54 1.62 1.0 0.48 0.46 0.82 0.99 0.44 0.28 0.25 0.43 0.65 0.98 OL 15 (27) 1960-2006 0.40 2.29 1.98 0.77 0.37 0.78 0.99 0.66 0.36 0.46 0.30 0.47 0.95 AL 15 (31) 1888-2006 0.63 1.11 0.83 0.69 0.43 0.66 0.98 0.34 0.25 0.60 0.23 0.33 0.93 OR 16 (30) 1921-2003 0.53 1.45 1.18 0.63 0.41 0.58 0.97 0.58 0.45 0.51 0.35 0.46 0.95 RE 15 (30) 1789-2003 0.40 0.93 0.64 0.58 0.41 0.39 0.94 0.36 0.31 0.47 0.37 0.26 0.89 JA 15 (35) 1863-2003 0.56 1.05 0.62 0.48 0.44 0.59 0.97 0.37 0.30 0.26 0.47 0.40 0.95 FN 15 (24) 1863-2006 0.68 1.06 0.75 0.69 0.40 0.53 0.93 0.27 0.20 0.51 0.35 0.41 0.9 GU 36 (75) 1912-2006 0.52 1.12 1.43 0.77 0.43 0.55 0.99 0.37 0.37 0.60 0.33 0.40 0.97 Statistics for raw-data series: MW, mean ring width; SD, standard deviation; AC1, first order autocorrelation. Statistics for residual series: MSx, mean sensitivity; Rbar, mean interseries correlation; EPS, expressed population signal. Sites names as in Table 1. Consequently, EW growth showed a greater year-to-year persistence (AC1), a higher change among consecutive years (MSx) and a higher common signal (Rbar, EPS) than LW formation. The change in EW and LW width among consecutive years (MSx) increased as latitude decreased but Results 66 these trends were not significant (EW, p = 0.15; LW, p = 0.10). The correlation between EW and LW series decreased significantly as the LW width increased (r = 0.67, p = 0.002). Earlywood Residual indices 0.0 0.5 1.0 1.5 2.0 Latewood Year 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 0.0 0.5 1.0 1.5 2.0 Figure 7. Residual chronologies of earlywood (EW) and latewood (LW) (gray lines) widths for the P. halepensis network developed in eastern Spain, and the overall mean for each variable (black lines). Spatio-temporal patterns of P. halepensis growth The PCA was used to summarise the EW and LW growth series and to retain the main patterns of spatio-temporal variability across the study area. PCA analysis revealed four and five components (PCs) for EW and LW, which accounted for 72% and 71% of the variance, respectively (Table 3). This indicates a slightly higher spatial variability for the LW formation, although the components retained, represent a high percentage of the total variance, which shows that tree growth in the region is not very heterogeneous and that coherent temporal patterns, representative of large regions can be found. Results 67 Table 3. Correlations (Pearson coefficient) calculated between the main principal components (PC) of EW and LW widths and atmospheric circulation indices (NAO: North Atlantic Oscillation; MO: Mediterranean Oscillation; WeMO: Western Mediterranean Oscillation) for winter (W), spring (Sp), summer (Su) and autumn (A). The third raw is the variance explained by each principal component. EW LW PC1 PC2 PC3 PC4 PC1 PC2 PC3 PC4 PC5 Variance (%) 33.58 15.62 12.9 10.1 20.9 17.97 11.8 12 8.6 NAO-W -0.32* 0.05 -0.3 0.18 -0.28 0.05 -0.21 -0.13 0.17 MO-W -0.25 -0.08 -0.21 -0.03 -0.29 0.02 -0.19 -0.21 -0.20 WeMO-W 0.12 -0.43** -0.04 0.02 -0.04 -0.18 0.02 -0.22 -0.23 NAO-Sp -0.45** 0.10 0.25 -0.20 -0.14 0.12 0.05 -0.16 -0.02 MO-Sp 0.10 -0.12 0.19 0.14 -0.31* -0.16 0.16 -0.20 0.10 WeMO-Sp -0.20 -0.30 -0.12 0.26 0.17 -0.13 0.16 -0.09 0.10 NAO-Su −−− −−− −−− −−− 0.14 0.23 0.33* 0.04 0.26 MO-Su −−− −−− −−− −−− -0.01 -0.22 -0.24 0.04 -0.16 WeMO-Su −−− −−− −−− −−− 0.14 0.23 0.33 0.07 0.26 NAO-A −−− −−− −−− −−− -0.13 -0.06 0.07 0.02 0.10 MO-A −−− −−− −−− −−− -0.11 -0.35* -0.06 -0.04 0.04 WeMO-A −−− −−− −−− −−− -0.15 -0.36* -0.03 0.19 0.08 Significance levels: ** p < 0.01, * p < 0.05. Overall, EW chronologies showed greater correlation among nearby sites than LW series did. The spatial extend of these relationships was significant up to 300 and 380 km for EW and LW chronologies, respectively (Figure 8). Figure 8. Spatial extent of earlywood (EW) and latewood (LW) chronologies. The diagonal lines represent the indicated linear regressions, whereas the dashed horizontal lines represent significance thresholds (p < 0.05) for the correlation values. Results 68 The spatial distribution of the PCA loadings, corresponding to the 4 and 5 retained components for EW and LW respecively, shows clear geographical patterns in tree growth across the study region (Figure 9 A). Figure 9. Spatial patterns of P. halepensis growth (earlywood (EW) and latewood (LW) chronologies) as revealed by PCA based on the loadings of the first four and five principal components, for earlywood (EW) and latewood (LW), respectively (A), and spatial classification based on the maximum loadings for each component (B). A B Results 69 According to the PCA loadings of the first components of EW and LW production, two main growth patterns, corresponding to PC1 and PC2, were found in northwestern and southeastern sites, respectively. These components group the 49.2% and the 38.9% of the total EW and LW variability, respectively (Table 3). This indicates that a high percentage of the tree-growth variance of the region is represented by these components. Additional sites were represented by the other components, but they account for a lower percentage of the total variance and they are found in transitional areas between the northwestern and southeastern locations, and in the northeastern part of the study area. The spatial classification of EW and LW variability based on the maximum loading rule shows clearly the distinction between the northwest and southeast sectors, both for EW and LW formation, with few differences among them (Figure 9 B). Influence of atmospheric circulation on P. halepensis growth The influence of atmospheric circulation patterns on P. halepensis growth was assessed by means of correlation analysis. In general, very few significant correlations were found between the seasonal atmospheric circulation patterns and the EW and LW series of the retained principal components (Table 3). Significant and negative relationships were found between the winter, spring NAO index and EW PC1, whereas the EW PC2 was negatively related to the WeMO winter index. For the LW, significant correlations were found between the PC1 and the MO in spring, between the PC3 and the summer NAO and between the autumn MO and WeMO and the LW PC2. These results clearly indicate that the two main components, which represent the major percentage of growth variability, are significantly correlated to some of the atmospheric circulation patterns at a seasonal scale. In addition, it was found a high temporal agreement between the variability of the atmospheric circulation patterns (winter and spring NAO, winter and autumn WeMO) and tree growth as represented by the first and second components of EW and LW (Figure 10). Such temporal agreement is coherent with spatial analyses between EW and LW growth and the seasonal atmospheric circulation patterns (Figure 11). The EW PC1 is significantly related to the winter and Results 70 spring NAO in north and northwestern forests. In contrast, the effects of the winter and autumn WeMO indices on PC2 EW and LW formation were greater at southeastern sites. -3 -2 -1 0 1 2 3 4 Winter NAO -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 PC1 EW NAO -2 -1 0 1 2 3 Spring NAO -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 NAO Standardized score -2 -1 0 1 2 3 Winter WeMO -1.0 -0.5 0.0 0.5 1.0 1.5 PC2 EW WeMO Year 1960 1965 1970 1975 1980 1985 1990 1995 2000 -2 -1 0 1 2 3 Autumn WeMO -1.0 -0.5 0.0 0.5 1.0 1.5 PC2 LW Figure 10. Temporal evolution of the first two principal components (PC1, PC2) of EW and LW P. halepensis chronologies, and related atmospheric circulation indices (NAO, WeMO) calculated from the previous winter to the current autumn. Results 77 There were also significant correlations between winter and spring precipitation and the LW width in most sites, which could suggest a lagged response of the winter climate on the forest growth in late summer and autumn, possibly driven by soil water storage. Nevertheless, for the forests showing significant correlations between LW and winter precipitation, it was found that partial correlations (considering the influence of EW on LW) between winter/spring precipitation and LW were not significant. In 11 of the 12 sites with significant correlation between winter precipitation and LW, no significant partial correlation was found, whereas 11 sites showed significant partial correlations (p < 0.05) between the EW and the LW variability. This would provide further evidence that the influence of the winter precipitation on LW is mainly driven through a direct influence on the EW formation. Nevertheless, in the southern study area, the analysis also revealed a direct influence of the surface climate on the LW formation. Therefore, in the southern sites there was a significant correlation between the autumn precipitation and the LW formation. These sites also showed a significant association between the autumn precipitation and the WeMO, with a clear different behavior as compared with the rest of the sites. In addition, no significant correlations were found between LW PCs and temperature. This would explain that although spring and autumn WeMO are correlated with temperatures in northern sites, both EW and LW formation in these sites are not correlated with the WeMO, since tree growth in those sites is not mainly driven by temperatures. For the southern sites, the influence of the autumn WeMO on LW was directly driven by precipitation. Therefore, in the southern region both EW and autumn precipitation played a significant role in explaining the LW variability. EW and autumn precipitation showed significant partial correlations with the LW in the southern areas (r = 0.65 and r = 0.43, p < 0.01, for the EW and autumn precipitation, respectively). This indicates that EW formation and also autumn precipitation are affecting LW development. Results 78 Discussion In P. halepensis the production of EW showed a greater year-to-year variability and a higher treeto-tree common variance than did LW; these results are similar to those reported by De Luis et al. (2009). Thus, EW and LW covaried to some extent, indicating a strong temporal association, as suggested by xylogenesis studies of P. halepensis (De Luis et al., 2007; Camarero et al., 2010). EW and LW responded to atmospheric circulation patterns and weather types in different ways, although for both variables were detected at least two groups of forests (those located in northwestern and southeastern sites) with contrasting responses. Such a geographically structured climatic response suggests that northwestern sites are more influenced by southwestern flows, which was indicated by the stronger relationships with the winter NAO index relative to that which occurred for southeastern sites near the Mediterranean coast, where P. halepensis growth was mainly affected by the winterautumn WeMO and the autumn MO indices. These findings are consistent with the associations found between atmospheric patterns and weather types. For instance, the winter NAO and MO indices were negatively (positively) associated with the frequency of cyclones (anticyclones) in winter. The winter WeMO index was positively correlated with the frequency of N, W and NW weather types. High levels of the WeMO index in autumn were also associated with greater occurrence of N, NW and anticyclonic weather types, and a low frequency of E, SE, S and cyclonic weather types. These results confirm that the effects of atmospheric circulation patterns on P. halepensis growth in the study area are probably an indirect expression of their effects on local weather types and climatic factors, including precipitation, temperature and radiation (Garfin, 1998; Mäkinen et al., 2003). In agreement with the results obtained by Vicente-Serrano and López-Moreno (2006) and González-Hidalgo et al. (2009) for the same region, the same spatial gradient in the effect of NAO and WeMO on tree growth has been found for precipitation. Although NAO mainly affects the southwestern Iberian Peninsula, the southwest flows are reactivated when they reach the Pre-Pyrenean chains, explaining why the forests located in the Ebro valley are affected by this pattern. In contrast, the humid influence of the Results 79 southwestern flows associated with the NAO does not reach the Mediterranean coastal region (Rodó et al., 1997; Rodríguez-Puebla et al., 1998), where the humid influences come from the east, as reflected by the WeMO index. Thus, the spatial pattern of P. halepensis forests in eastern Spain reflects the influence of atmospheric circulation patterns on precipitation, and clearly demonstrates that these effects act through the control of variability of local precipitation in the region. Several studies have reported that, in addition to precipitation, the winter NAO also exerts a marked influence on surface air temperature, and that its association with tree growth varies from northern (positive) to southern (negative) Europe (Piovesan and Schirone, 2000; Linderholm et al., 2003; Schultz et al., 2008), but in eastern Spain the results indicate an insignificant influence of temperature variability on both EW and LW growth and support the view that most of the influence of the atmospheric circulation variability on tree growth is mainly driven through changes in precipitation. The spatially constrained negative relationship between the winter NAO index and EW growth at northwestern sites in eastern Spain is consistent with results from earlier studies of Iberian conifer forests (Bogino and Bravo, 2008; Rozas et al., 2009). However, growth at xeric southwestern sites is also negatively related to the winter WeMO index, which can be explained by more winter precipitation, and consequent enhanced radial growth in spring during negative phases of the winter WeMO index (Martín-Vide and Lopez-Bustins, 2006; González-Hidalgo et al., 2009; Vicente-Serrano et al., 2009). A similar result has been reported for the WeMO index in spring and P. halepensis forests in the middle Ebro basin (Ribas et al., 2008). These results highlight the marked spatial variability in the effect of general atmospheric circulation patterns as drivers of tree growth, which may be a consequence of distance to the sea and the landscape relief. LW formation in P. halepensis was negatively associated with the Mediterranean indices (WeMO, MO) at southeastern sites, suggesting a dependence on autumn precipitation. This finding suggests that autumn precipitation may be important for LW growth in P. halepensis, which is consistent with phenological observations that cambial activity in this species can be greatly Results 80 enhanced after summer because of sporadic autumn rainfall, which is usually produced by Mediterranean cyclonic activity (De Luis et al., 2007; Camarero et al., 2010). An important finding of this study is the distinct influence of winter atmospheric circulation patterns on LW formation in P. halepensis. Two explanations can be proposed for this observation. First, EW formation in P. halepensis is partly dependent on photosynthetic activity and moisture reserves supplied by precipitation prior to spring growth (Kagawa et al., 2006), and the amount of winter rainfall is inversely correlated with the NAO (Piovensan and Schirone, 2000). In P. halepensis, EW is usually formed between March and June whereas LW starts developing in July and ends its maturation in November (De Luis et al., 2007; Camarero et al., 2010). Thus, EW formation may be very dependent on the water supply and photosynthetic activity in late winter and early spring. Relative to southwestern sites, which are more subjected to Mediterranean cyclonic activity, a high winter NAO index may be associated with low winter rainfall and reduced EW formation at northwestern sites, under the strong influence of Atlantic westerlies. The greater degree of EW formation may result in an increase in hydraulic conductivity and the synthesis of more carbohydrates for LW formation in summer and autumn (Camarero et al., 2010). A second and much less plausible explanation is that in those years with abundant winter rainfall (low winter NAO index) water is stored in deep soil layers, providing surface soil water for LW growth in summer. The first explanation appears more likely, as xylogenesis studies support a functional link between EW and LW growth, which show plastic responses to climatic conditions. For instance, at coastal sites cambial activity may occur for longer than at inland sites, where trees are subject to low winter temperatures and summer droughts (De Luis et al., 2007; Camarero et al., 2010). According to these authors, the transition between EW and LW is linked to the temperature rise and the decline in soil water reserves between late spring and early summer (June-July), but xylogenesis processes are plastic. For instance, these studies indicate that EW may start late (e.g., April-May) in cold northwestern sites whereas LW maturation may last up to December in warm southeastern sites. This cambial plasticity is consistent with other features of P. halepensis, which is a species Results 81 well adapted to drought because of its ability to rapidly reduce transpiration by stomatal closure (Borghetti et al., 1998), and its capacity to make efficient use of water reserves that become available several months prior to growth (Sarris et al., 2007). The first explanation is also consistent with findings of allometric relationships between tree ring width and maximum LW density (Kirdyanov et al., 2007). The analyses indicate that weather types also exert a relatively strong influence on EW and LW growth, probably because of their indirect effect on local patterns of ultimate climatic drivers, including precipitation, solar radiation and temperature (Garfin, 1998). In winters characterized by extended periods of high pressure (i.e. a high frequency of N and anticyclonic weather types), EW formation was less than in other years. Explanations for these associations include the relationship of anticyclone weather types to dry and cold conditions in winter over the eastern Iberian Peninsula (Goodess and Jones, 2002). Very cold and dry winters may result in less photosynthetic activity in P. halepensis, and reduce the amount of carbohydrates available for EW growth during the following spring (Kagawa et al., 2006). In contrast, cyclonic activity linked to an increase in Atlantic westerlies (SW and W weather types) caused warmer and more humid conditions in winter than during anticyclonic periods, which enhanced EW formation and, probably through indirect effects, LW development. In autumn, LW development was enhanced by cyclonic conditions probably related to Mediterranean convection, as both E and SE (N and NW) weather types were positively (negatively) related to LW growth. In the eastern half of the Iberian Peninsula, autumn rainfall is usually associated with easterly and southeasterly winds, and Mediterranean storms (Goodess and Jones, 2002; Beguería et al., 2009). An increase in autumn precipitation may enhance cambial activity in P. halepensis and co-occurring tree species after the summer drought (Camarero et al., 2010), and mild conditions may provide a longer growing season, particularly at coastal southeastern sites (De Luis et al., 2007). In conclusion, the results indicate that large-scale atmospheric patterns (NAO, MO, WeMO) and local weather types influence EW and LW formation in P. halepensis forests of eastern Spain. Results 82 However, across this dendrochronological network both EW and LW responded differently to atmospheric circulation patterns and weather types, suggesting that the predicted changes in atmospheric circulation will result in contrasting tree growth responses in the northwest and southeast areas of this Mediterranean region. Therefore, the forecasted trend of increasing winter anticyclonic conditions and reduced activity of westerlies (high NAO and WeMO indices) are expected to reduce EW production in northwestern sites, whereas an increase of autumn cyclonic Mediterranean conditions (low WeMO and MO indices) will enhance LW formation in southeastern sites. Results 83 3.2. Impacts of drought at different time scales on forest growth across a wide climatic gradient in north-eastern Spain Introduction Water availability is one of the main climatic constraints for tree growth in the CircumMediterranean forests. Thus, several studies have shown a strong correlation between precipitation and radial growth in different Mediterranean forests and tree species (Tardif et al., 2003; Macias et al., 2006; Andreu et al., 2007; Sarris et al., 2007; De Luis et al., 2009; Linares et al., 2010; Carrer et al., 2010; Lebourgeois et al., 2010; Mérian et al., 2010). The impact of water deficit on growth is much higher in the most arid sites, where water availability largely constrains the main physiological processes of vegetation (growth, photosynthesis, carbon and nitrogen use), than in mesic sites (e.g., Jump et al., 2006; Vicente-Serrano et al., 2006, 2010a; Sarris et al., 2007; Martínez-Vilalta et al., 2008). Although the main patterns of precipitation-growth relationships are well known, the large seasonality and year-to-year variability that characterize precipitation in the Mediterranean region and the different site-dependent seasonality of tree growth in forests from this area may make very difficult to determine the response times of tree growth to the precipitation deficit. Furthermore, lags between water shortages and growth can appear as a function of different anatomical and physiological adjustments of trees to cope with drought, but also in response to drought severity and to the season in which water deficit occurs. All these mechanisms, either isolated or acting synergistically, can challenge the identification of drought impacts on tree growth. Commonly drought indices are used with the purpose of solving the current problems of quantifying drought severity since it is very complex to determine the magnitude, duration and surface extent of droughts (Wilhite and Glantz, 1985; Redmond, 2002). Different drought indices have been developed to quantify the water deficit in an objective way, which is usually better than using the precipitation information itself (Keyantash and Dracup, 2002; Heim, 2002; Mishra and Singh, 2010). Drought indices are based on the quantification of the cumulative water shortages Results 84 over a period of time. Some of the indices are based on soil water balance equations. The best example of this type of indices is the Palmer Drought Severity Index (PDSI, Palmer, 1965; Wells et al., 2004). Different studies have analyzed the influence of the drought conditions on tree growth using the PDSI (Orwig and Abrams, 1997; Kempes et al., 2008; Bhuta et al., 2009; Mundo et al., 2010). Nevertheless, although the PDSI can be useful to determine the severity of a drought, the index has several deficiencies (Alley, 1984; Weber and Nkemdirim, 1998), being its main shortcoming that it can only be calculated at a unique time scale (Guttman, 1998; Vicente-Serrano et al., 2011). On the contrary, drought is a multi-scalar phenomenon, given the great variety of response times found in different hydrological, agricultural and environmental systems to the occurrence of water deficits (e.g., McKee et al., 1993; Ji and Peters, 2003; Vicente-Serrano and López-Moreno, 2005; Lorenzo-Lacruz et al., 2010; Quiring and Ganesh, 2010). The problems involved with the use of PDSI have motivated the development of drought indices that can be calculated at different time-scales such as the Standardized Precipitation Index (SPI) (McKee et al., 1993) and the Standardized Precipitation Evapotranspiration Index (SPEI) (Vicente-Serrano et al., 2010c). The quantification of droughts at different time scales is crucial to determine their ecological impacts, given the different physiological strategies of vegetation to cope with water deficit. Studies analyzing the drought impacts on vegetation activity using the SPI have shown contrasting responses according to the time scales at which drought affected vegetation activity and also depending on vegetation types (Vicente-Serrano, 2007) and environmental conditions for the same vegetation type (e.g., Ji and Peters, 2003; Quiring and Ganesh, 2010). At present, most studies quantifying the vegetation response to different drought time-scales have been carried out using remote sensing data for mid-term (10-30 years) datasets, which is related to the potential photosynthetic activity of the canopy or the leaf area of the forest (Vicente-Serrano, 2007). Nevertheless, currently there are lack of studies analyzing the response of tree secondary growth to different time-scales of drought, which are quantified by means of multi-scalar drought indices. Results 85 This methodological approach could improve the knowledge of the long-term responses of tree growth to water availability better than using precipitation data itself or other drought indices. Currently, a deeper knowledge of the tree growth responses to water shortages in the Mediterranean Basin is a crucial task. Forests are particularly sensitive to climate change because the long-life span of trees does not allow for a quick adaptation to rapid environmental changes such as current climate warming (Andreu et al., 2007). As a matter of fact, various studies have provided evidence on the direct effect of drought on forest decline, particularly in Mediterranean forest ecosystems where water shortage is the main factor constraining growth (Sarris et al., 2007; Linares et al., 2010; Sánchez-Salguero et al., 2010). In this study it was analyzed the response of tree growth to different time scales of drought, quantified by means of the Standardized Precipitation Index (SPI), in forests from the Aragón region, in North-eastern Spain (Figure 14). Forests located in this region were particularly selected because they grow under a strong climatic gradient ranging from semiarid conditions in the Middle Ebro Basin to humid conditions in the Pyrenees. Furthermore, a trend towards more arid conditions was observed during the late 20th century in the study area (López-Moreno et al., 2010). Therefore, these forests represent an appropriate model to test the drought impact and the expected contrasting vulnerability to drought stress of species growing in the area. The objective of this study was to determine whether the use of multi-scalar drought indices is an effective way to determine the impact of water deficit on growth, and whether this approach may detect different responses, in terms of magnitude and seasonality, in a variety of tree species and locations. The study includes eight tree species growing along the mentioned climatic gradient showing contrasting vulnerability to drought stress: four pine species (Pinus halepensis, P. pinea, P. nigra, P. sylvestris), silver fir (Abies alba), Spanish juniper (Juniperus thurifera), and two oak species (Quercus faginea, Q. ilex). Results 86 Figure 14. Distribution of forests (gray area) in the north-eastern Spain and the location of study sites. Different symbols represent different forest species. See sites names in Table 1. These species represent species typically associated with mesic sites and humid conditions (e.g., A. alba), transitional locations (e.g., P. sylvestris) and xeric sites (e.g., P. halepensis). This is why all these species were considered in the study, to test if the responses to drought vary among species, sites and contrasting climate conditions found in the study area. Results 93 In general, important differences were found in the responses of tree growth to the different time scales of the SPI (Figures 16 and 17). 0.2 0.4 0.4 0.6 0.6 0.6 0.4 0.2 0.6 0.6 0.0 0.4 0.0 510 15 20 25 30 35 40 45 Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Correl. coef. 0.2 0.4 0.6 0.4 0.2 0.0 0.4 0.4 0.4 0.2 Time-scale 510 15 20 25 30 35 40 45 Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 0.0 0.2 0.4 0.6 0.2 0.2 0.2 0.2 0.4 0.2 0.4 0.2 0.4 0.6 0.2 0.4 0.4 0.4 0.0 0.2 0.0 0.0 0.0 Time-scale 510 15 20 25 30 35 40 45 Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 0.0 0.2 0.8 J. thurifera 0.0 0.2 0.2 0.0 0.2 0.0 0.4 0.4 0.4 0.6 0.2 0.0 0.4 0.4 0.0 0.2 Time-scale 510 15 20 25 30 35 40 45 Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 0.0 0.2 0.4 0.6 P. nigra P. pinea Time-scale 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.0 0.2 0.0 0.0 Time-scale 510 15 20 25 30 35 40 45 Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2 0.2 0.0 0.2 0.0 0.0 0.4 0.0 Time-scale 510 15 20 25 30 35 40 45 Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec A. alba P. sylvestris P. halepensis 0.2 0.2 0.2 0.2 0.0 0.4 0.4 0.4 0.4 0.2 0.4 0.4 0.4 0.2 0.4 0.4 0.2 0.0 0.2 0.2 0.0 0.4 0.0 Time-scale 510 15 20 25 30 35 40 45 Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 0.2 0.2 0.4 0.2 0.0 0.2 0.0 Time-scale 510 15 20 25 30 35 40 45 Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Q. ilex Q. faginea Figure 17. Mean correlation coefficients between ring-width chronologies and monthly SPI series at different time scales for the eight studied species. Bold lines frame significant correlations (p≤0.05). Results 94 In P.halepensis forests, significant correlations were revealed at time scales from 1 to 48 months. Nevertheless, at short time scales (1 to 3 months) significant correlations only appeared for the period May - July. In this specie the highest correlations (r = 0.60-0.70) were observed at time scales between 12 and 16 months during June, July and August. High correlations were also found between the SPI and the chronologies of P. pinea (maximum r = 0.70), P. nigra (maximum r = 0.62) and J.thurifera (maximum r = 0.61) considering spring (May), summer (June, July, August) and early-fall (September) months, particularly at time scales from 9 to 15 months. The remaining conifers (A. alba, P. sylvestris) did not show strong associations between growth and SPI, presenting low and significant correlation coefficients (r = 0.30-0.40) during summer months for time scales ranging between 2 and 5 months. A higher growth response to drought was observed for Q. faginea and Q. ilex based on the significant correlations they showed at time scales up to 25 months, mostly during summer and autumn months. Q. ilex indicated lower growth-SPI correlations than Q. faginea, with association being significant only at time scales between 8 and 25 months. Geographically structured growth-drought relationships Here is shown the spatial variability of the associations between forest growth and drought. A high site-to-site variability was detected in the growth response to drought within the same species across the study region (Figure 18). Consequently, a clear South-North gradient was observed in the maximum correlations between growth and SPI. This gradient corresponded to maximum correlation values for sites located in the Middle Ebro Basin in the case of Mediterranean conifers (P. nigra, r = 0.84; P. halepensis, r = 0.83; P, pinea, r = 0.81; J. thurifera, r = 0.75) and oaks (Q. faginea, r = 0.74; Q. ilex, r = 0.67), intermediate values in sub-Mediterranean P. sylvestris forests from the Pre-Pyrenees (r = 0.66) and minimum values in Pyrenean mesic A. alba forests (r = 0.64). Results 95 Figure 18. Geographic variability of the growth-drought associations showing (1) maximum Pearson correlations between ring-width chronologies for all sites and the monthly SPI series (the symbols are proportional to the absolute correlation value), (2) time scale (in months) at which the maximum correlation was achieved; and (3) month of the year at which the highest correlation was reached. Most species reached the maximum growth-drought correlations at time scales varying from 9 to 11 months, except A. alba and P. sylvestris which showed the maximum association at time scales lower than 5 months in most of their sampled forests. Nevertheless, large differences were also noted in the time scales at which maximum correlation were observed for each sampled species without any clear spatial pattern. Finally, irrespective of the tree species, the study site, the magnitude of the SPI-tree growth correlations and the characteristic time scale, the maximum correlations were commonly obtained in summer months (June to August), which indicates the relevance of water availability for tree growth during the late growing season. Results 96 Discussion This study evaluated the impact of droughts on eight tree species forming forests distributed along a wide climatic gradient in north-eastern Spain, by means of dendrochronological methods and using the Standardized Precipitation Index (SPI). This is the first investigation exploring the relationship between tree growth and drought using multi-scalar drought indices. Particularly, the research focused on the impact of different time scales characterizing drought intensity on radial growth, given that droughts may act on growth at different characteristic time scales. For instance, at short time scales dry and moist periods alter with a high frequency, while at long time scales droughts are less frequent but longer in duration (Hayes et al., 1999; Vicente-Serrano, 2006). The analyses revealed two distinct patterns in terms of the growth sensitivity to drought, depending on the time scale of drought and the studied species. Species growing in the Middle Ebro Basin under Mediterranean semiarid conditions (all Pinus species excepting P. sylvestris, Quercus species, J. thurifera) showed stronger growth responses to drought compared to those growing in mountainous areas from the Pre-Pyrenees and Pyrenees (A. alba, P. sylvestris) characterized by a humid and cold climate. It has been found previously that in xeric Mediterranean areas tree growth is mainly limited by low precipitation, while in mesic Mediterranean areas the main factors constraining growth are low temperatures (Richter et al., 1991; De Luis et al., 2007; VicenteSerrano, 2007; Camarero et al., 2010). The high level of dependence on water availability of northeastern Spanish forests has been also reported before mainly in pine and oaks species (Corcuera at al., 2004a,b; Andreu et al., 2007; Montserrat-Martí et al., 2009; Gutiérrez et al., 2011). Overall, these studies reported marked spatial variations in the response of forests to drought as a function of climate conditions, confirming that forests located in the driest sites are the most sensitive to drought occurrence. Tree species growing in the driest sites of the study area, i.e. the Middle Ebro Basin, showed robust relationships (r > 0.60) with the SPI drought series at time scales between 9 and 11 months, which is a remarkable finding since it indicates that cumulative precipitation conditions during one Results 97 year impact tree growth as illustrated by Sarris et al. (2007). The response of growth to drought (SPI) time scales longer than 11 months decreased gradually but correlations were significant up to 30 months (r > 0.30). Beyond this threshold, growth responses to drought were very low, indicating that precipitation recorded for periods longer than 30 months is not significantly affecting radial growth. Other studies that analyzed the relationships between precipitation and growth in semiarid sites showed that the correlations between two variables increased when considering the cumulative precipitation over a period of consecutive months (De Luis et al., 2009; Linares and Tíscar, 2010). Thus, forest growth variability in similar drought-prone areas is determined by the precipitation recorded during the year of tree-ring formation but also by the precipitation that fell in the previous year (Sarris et al., 2007). The use of a multi-scalar drought indicator allowed confirming this question in the analyzed forests. In the semiarid Middle Ebro Basin, the previous-winter soil water reserves are crucial for supporting tree growth during spring (Pasho et al., 2011a). It has been found previously that spatio-temporal variations in soil moisture and related rainfall pattern determine the growth response to climate in most Iberian forests (Andreu et al., 2007). Soil types (limestone and gypsum) in the Ebro Basin valley may additionally intensify the effects of drought conditions on tree growth because they have low water holding capacity (Guerrero et al., 1999). All the studied species found in the Middle Ebro Basin showed the strongest growth response to drought during spring and summer months, which indicates that forest growth in the area is highly dependent on spring and summer cumulative water deficit. First, most of these species show their maximum radial-growth rates between May and June (Camarero et al., 2010). Second, water deficit starts to affect markedly vegetation activity and plausibly growth in the study area as early as June (VicenteSerrano, 2007). In addition, in the driest sites it was found that the growth-drought correlation decreased at time scales lower than 9 months, indicating that these forests may show certain adaptative capacity in response to short droughts. The response of tree growth to drought greatly varied among species, being the maximum growth-drought correlation very high for P. halepensis, P. pinea, P. nigra and J. thurifera (r = 0.60- Results 98 0.80), moderate for Q. ilex and Q. faginea (r = 0.50-0.60) and low for P. sylvestris and A. alba (r = 0.40-0.50). The variability of species responses to drought may indicate very different strategies and functional threshold in coping with droughts. The lower tree growth–drought correlations found in drought-tolerant oaks (e.g., Q. ilex) as compared with drought-avoiding pines (e.g., P. halepensis) could be related to the more efficient conductive elements (vessels in oaks vs. tracheids in conifers), a more conservative water and deeper root systems of the former as compared with the later species, which might mitigate the negative effects of short-term water shortages on tree growth (Hacke and Sperry, 2001; Willson et al., 2008). However, the results indicate that Q. ilex showed a higher growth plasticity in response to drought, i.e. low growth-SPI correlations, in comparison to Q. faginea, suggesting a greater resistance to water constrains of the former as compared with the later species, which is in agreement with the low phenological activity of Q. faginea in summer (Corcuera et al., 2004a,b; Montserrat-Martí et al., 2009). J. thurifera also appeared to be affected moderately by drought despite this species is considered a drought-resistant species among the Iberian conifers and its radial-growth dynamics are very plastic in response to drought and to episodic rains (Camarero et al., 2010). Considering Pinus species from the most arid study sites (P. halepensis, P. pinea, P. nigra), it was detected a stronger response to cumulative droughts over a 11-months period, during spring and summer months in comparison to other co-existing species in the area. This indicates that tree growth in these pine species is sensitive to mid-term water deficits which agrees with the findings of Linares et al. (2010), who found that P. halepensis growth in south-eastern Spain can be limited by drought during the summer prior to growth. Although pine species as P. halepensis are considered as drought-avoiding species, they may show functional growth thresholds in response to lasting and severe drought leading to growth decline and death (Novak et al., 2011). For instance, P. halepensis is adapted to the scarcity of soil water in the short-term due to summer drought by ceasing secondary growth and recovering it rapidly when water becomes available (Borghetti et al., 1998; Nicault et al., 2001; Rathgeber et al., 2005; De Luis et al., 2007, 2011). However, its growth Results 99 may be much vulnerable to midand long-term droughts as the results support. Moreover, considering the fact that some of the studied sites represent the southernmost populations of some of the studied species in Europe (e.g., A. alba) and these stands are growing near the species’ climatic tolerance, they may be affected severely by strong and lasting droughts leading to forest decline (Macias et al., 2006; Camarero et al., 2011). The results indicated no impact of long-term droughts on radial growth of mountain conifers from mesic sites (A. alba, P. sylvestris), which responded to short-term droughts of a duration lower than five months during summer. This association may be explained by the low water-use efficiency of A. alba, which is a species whose photosynthetic rates are very sensitive to atmospheric drought (Guehl et al., 1991) and its growth rates respond to short-term cumulative water deficit in late summer (Camarero et al., 2011). In the case of P. sylvestris, previous studies have clearly indicated that summer drought contrains growth and xylogenesis in P. sylvestris (Camarero et al., 2010; Gruber et al., 2010) and severe water deficit may even lead to droughtinduced mortality (Martínez-Vilalta and Piñol, 2002; Sánchez-Salguero et al., 2010). The response to drought was site-dependent and this variability among sites was greater in P. nigra and Q. faginea forests as compared with the other species. However, the number of sites sampled to capture the variability of both species was low in comparison to other well-replicated species such as P. halepensis and A. alba. In general, sites located in the driest areas of the Middle Ebro Basin showed a higher response to drought compared to those located in mesic mountainous areas where water availability is high. For example, it was observed that P. halepensis showed higher growth-drought correlations in the driest sites than in northern less xeric sites receiving more precipitation. Given the large intra-specific differences, in the section 3.4 are investigated the relative roles of local conditions (topography, soil type, etc.) on the growth responses to drought at different time scales. Therefore, multi-scalar drought indices are particularly useful for monitoring the impact of climate variability on forest growth because the response of tree growth to droughts is complex. The Results 100 time scales over which precipitation deficits accumulate affecting noticeably forest growth vary among species and sites within the same species. For this reason, drought indices must be associated with a specific time scale and assessed taking into account local conditions to be useful for monitoring impacts on forest growth as has been done with remote-sensing assessments of vegetation activity (Ji and Peters, 2003; Vicente-Serrano, 2007; Quiring and Ganesh, 2010). In the current context of climate warming, several climate models have indicated that drought frequency and intensity are expected to increase in the Western Mediterranean Basin (Giorgi and Lionello, 2008). Increasing aridity is expected to cause growth decline and enhance mortality particularly in drought-sensitive species (Linares et al., 2010; Gruber et al., 2010; Koepke et al., 2010). The approach used in this study for examining growth responses to drought at different time scales and considering multiple tree species and sites across a wide climatic gradient in north-eastern Spain may represent a first step in understanding and forecasting forest responses to future climate change. If climate warming causes more frequent and severe droughts in the future, many forests in the study area will be adversely and selectively affected. Results 101 3.3. Climatic impacts and drought control of radial growth and seasonal wood formation in Pinus halepensis Introduction In Circum-Mediterranean forests, drought is considered as the main driver of the tree growth variability (Sarris et al., 2007). Climate change models project a decrease in annual mean precipitation and rising air temperatures over the Mediterranean Basin for the late 21st century leading to an increase in evapotranspiration (IPCC, 2007; Giorgi and Lionello, 2008; García-Ruiz et al., 2011). These trends towards increasing arid conditions in the region are expected to cause more frequent episodes of forest growth decline and mortality events as those already observed by some authors (Macias et al., 2006; Linares et al., 2009; Galiano et al., 2010; Sánchez-Salguero et al., 2010). Various studies indicate contrasting tree growth responses to drought (Martín-Benito et al., 2008; Koepke et al., 2010; Linares et al., 2010; Sánchez-Salguero et al., 2010). For instance, some authors (Abrams et al., 1998; Sarris et al., 2007) noted that trees growing in xeric sites showed declining radial growth trends and increased crown dieback and mortality in response to longlasting droughts. However, other researchers argued that species growing in xeric locations develop adaptive features to withstand the negative effects of drought on growth by reducing water loss or increasing hydraulic efficiency, and by showing lagged growth responses or recovering rapidly after the drought (Bréda et al., 2006; McDowell et al., 2008; Eilmann et al., 2009). Dendrochronology is a useful tool for providing information on climate-growth relationships based on year-to-year correlative approaches. However, these relationships must be compared with xylogenesis studies for capturing short term climatic influences on radial tree growth (Camarero et al., 1998; Rossi et al., 2008). Therefore, information regarding the phenology of xylem growth is required to better understand the processes underlying the effects of climate and drought stress on seasonal wood formation. Results 102 Pinus halepensis forests provide a valuable system to explore how drought measured at different time scales constrains radial growth. This conifer is one of the dominant tree species in the Western Mediterranean Basin and the most ecologically important species in semi-arid woodlands (Ne´eman and Trabaud, 2000). P. halepensis is considered as a species well adapted to withstand drought by reducing growth as water availability decreases (Serre-Bachet, 1992; Borghetti et al., 1998; Nicault et al., 2001; Rathgeber et al., 2005; De Luis et al., 2007; Camarero et al., 2010). The increasing aridity in the Mediterranean region could have important implications for the intraannual growth dynamics of P. halepensis which could be reflected also in long-term (inter-annual) growth trends in this species. In addition, there are no studies analyzing jointly the intra and interannual long-term responses of earlywood (hereafter abbreviated as EW) and latewood (hereafter abbreviated as LW) formation to climate and drought in P. halepensis forests from semi-arid areas. In Circum-Mediterranean drought-prone forests, a deeper knowledge on the EW and LW responses to recurrent water shortages at several time scales is important for understanding how amplified precipitation variability will drive radial growth trends (Andreu et al., 2007). In a previous study on growth-drought relationships assessed at multiple time scales, it was found that P. halepensis growth is highly sensitive to mid-term cumulative drought stress, particularly during summer (Pasho et al., 2011b). This study aims at understanding how drought and climate variability affects the patterns of seasonal wood formation (EW and LW widths) of P. halepensis forests from central (Middle Ebro Basin) and southern (Teruel) Aragón, north-eastern Spain (Figure 19). In addition, the research aims at providing a mechanistic basis concerning the long-term impacts of drought and climate on EW and LW production by studying xylogenesis processes (periods and production rates of different tracheid types) in a semi-arid P. halepensis forest. Moreover, the current and post-drought (up to two years later) effects on growth of both wood types are investigated to quantify possible lagged effects of water deficit on seasonal wood production. Results 109 Table 7. Dendrochronological statistics of earlywood (EW) and latewood (LW) P. halepensis chronologies for the common period 1970–2005. Earlywood Latewood Site No. trees (No. radii) Period Correlation EW-LW MW ± SD (mm) AC1 MSx Rbar EPS R2adj (%) MW ± SD (mm) AC1 MSx Rbar EPS R2adj (%) VM 17 (30) 1959-2009 0.61 1.39 ± 0.81 0.59 0.44 0.80 0.99 80.30 0.39 ± 0.21 0.52 0.47 0.56 0.95 72.45 CV 13 (23) 1928-2009 0.47 1.75 ± 0.94 0.75 0.38 0.47 0.89 82.85 0.54 ± 0.33 0.71 0.34 0.50 0.86 64.06 CS 13 (23) 1900-2009 0.75 0.68 ± 0.37 0.41 0.47 0.64 0.97 40.04 0.18 ± 0.10 0.39 0.44 0.62 0.91 18.33 VA 17 (31) 1925-2009 0.74 1.28 ± 0.75 0.62 0.43 0.68 0.92 54.01 0.37 ± 0.22 0.58 0.43 0.51 0.86 41.92 TA 14 (27) 1927-2006 0.57 1.91 ± 1.22 0.59 0.50 0.78 0.97 44.68 0.43 ± 0.24 0.35 0.48 0.52 0.85 24.60 PU 15 (22) 1943-2009 0.58 0.85 ± 0.43 0.60 0.38 0.69 0.85 50.10 0.30 ± 0.16 0.56 0.40 0.54 0.88 29.10 MA 15 (28) 1963-2009 0.60 1.49 ± 1.27 0.75 0.53 0.81 0.94 64.51 0.38 ± 0.26 0.53 0.46 0.69 0.93 46.54 PL 15 (30) 1961-2009 0.38 1.35 ± 1.08 0.73 0.49 0.79 0.94 70.67 0.39 ± 0.25 0.53 0.44 0.61 0.92 50.87 MP 14 (29) 1965-2009 0.52 1.53 ± 1.26 0.81 0.37 0.68 0.93 77.61 0.38± 0.22 0.61 0.37 0.49 0.90 59.49 Mean −−− −−− 0.58 1.36 ± 0.86 0.65 0.44 0.70 0.93 62.75 0.37± 0.22 0.53 0.43 0.56 0.89 45.26 Statistics: EW–LW, Pearson correlation coefficient calculated between the residual earlywood and latewood chronologies for each site. Raw-data series: MW, mean width; SD, standard deviation of width; AC1, first order autocorrelation. Residual series: MSx, mean sensitivity; Rbar, mean interseries correlation; EPS, expressed population signal; R2adj, adjusted R2 obtained relating monthly climatic variables and EW, LW residual chronologies through stepwise linear regressions. The last line provides the mean values for all sites and considering statistics calculated for raw ring-width data (MW, SD, AC1) and residual chronologies (MSx, Rbar and EPS). Results 110 Figure 20. Residual chronologies of P. halepensis earlywood (EW) and latewood (LW) widths (gray lines) for the studied sites and their overall means (lines with empty circles). Black lines with triangles show the evolution of the drought index (Standardized Precipitation Index, SPI, mean of all sites) at time scales of 12 months in July and September, the scale at which EW and LW growth series showed the strongest response to the SPI drought index. Note that positive and negative SPI values indicate wet (high EW and LW indices) and dry conditions (low EW and LW indices), respectively. Selected dry years (1967, 1986, 1994 and 2002) for the study of growth departures are encircles in the upper graph. Considering the variability in climate-growth associations among sites, strong positive correlations were found between May precipitation and EW growth in the sites MA, MP and PL whereas for the sites VA, VM and PU summer precipitation was found to be more important (Figure 22). The rest of sites did not show any particular correlation with spring-summer precipitation (except June). However, the growth in all sites was influenced considerably by winter precipitation (December-January). Considering the LW, there were found positive correlations between previous October precipitation and LW growth in the sites MA and PL. Results 111 Figure 21. Mean (± SE) correlation coefficients calculated between earlywood and latewood width chronologies of the nine study sites and monthly climatic variables (mean temperature, mean maximum and minimum temperatures, and total precipitation). Growth is related with climate data from the previous August to current October, i.e. of the prior year (months abbreviated by lowercaste letters) and the year of tree-ring formation (months abbreviated by uppercaste letters). The significance level (p≤0.05) is indicated by dashed horizontal lines. Results 112 The remaining sites did not present any significant correlation with the previous autumn current winter precipitation. The spring precipitation was particulary important for the sites MP, VM and CS whereas the summer precipitation was positively and significantly correlated with almost all sites. The autumn precipitation in general did not show any particular influence on LW growth except September month which affected positively the growth in sites MP, VA and PU. With regard to temperature, generally positive correlations were found between EW growth and minimum winter temperatures, particularly for the sites MA, MP, PL, CS and CV. The same can be commented for the mean and maximum temperature in winter but not for the summer months in which temperature excerted a high negative influence on EW growth in almost all sites (except site VA). The LW growth was enhanced by the winter minimum and mean temperatures, particularly at the sites MP, PL, VA, SC and CS whereas spring and autumn temperatures did not have any significant influence. Summer mean and maximum temperatures generally constrained the LW growth, specifically at the sites VM, CS and CV. Results 113 Figure 22. Relationships (Pearson correlation coefficients) between earlywood and latewood width indices and monthly climatic variables (mean, minimum, maximum temperature and total precipitation) for the nine study sites. Growth is related with climate data from the previous August to October of the year of tree-ring formation. The significance level (p≤0.05) is indicated by dashed horizontal lines. Results 114 Intra-annual dynamics of earlywood and latewood formation In this section is described the intra-annual development of EW and LW formation in P. halepensis (site PU) as revealed by the xylogenesis analysis. In the selected site, climatic conditions during 2010 were within the range of the local long-term climatology with mean annual temperature of 15.3 ºC (long-term mean = 15.0 ºC) and total precipitation of 267 mm (long-term mean = 327 mm) (Figure 23). Considering monthly values in 2010, April and July temperatures and April rainfall amount were above long-term mean values whereas May precipitation was below historical means. J F M A M J J A S O N D Temperature (ºC) 0 5 10 15 20 25 30 1951-2011 2010 Month J F M A M J J A S O N D Precipitation (mm) 0 20 40 60 80 100 120 140 160 J F M A M J J A S O N D Precipitation (mm) 0 20 40 60 80 100 Temperature (ºC) 0 20 40 Figure 23. A. Climatic conditions (mean monthly temperature and total precipitation) at the site PU during the year 2010 as compared with historical data (period 1951-2011, boxplots showing the median values and outliers) from the Zaragoza-Aeropuerto meteorological station (latitude 41° 39’ N, longitude 1° 00’ W, elevation 263 m a.s.l.) located at ca. 24 km. B. Climatic diagram of Zaragoza-Aeropuerto station based on 1951-2011 data. A B Results 115 Xylem formation had already started in March and xylogenesis was active until November (Figure 24). The cambial activity followed a unimodal pattern, with two major peaks (May-June and mid July-August) of tracheid formation corresponding respectively to the enlargement and wallthickening phase. Figure 24. Number of cells of P. halepensis according to their development phase (cambial cells, radially enlarging tracheids, wall-thickening tracheids and mature tracheids −earlywood and latewood tracheids are shown as bars with different fill types) formed during the year 2010 in the site PU. The image shows the earlywood (EW) and latewood (LW) of a cross-section of a ring (total width = 1.22 mm) from a wood sample taken in mid November. Data are means (n = 10 trees). Results 116 These two phases match with the periods of maximum EW and LW formation in that order (Figure 25 B). The rate of EW tracheid production was higher (0.38 cells day-1) than that of LW (0.22 cells day-1). The EW tracheid formation started in late March and finished in September in most trees whereas the first LW tracheids were formed in July and the last ones were observed in November (Figure 24). Drought-growth relationships In this section are shown the results regarding drought-growth relationships as revealed by the SEA and SPI analysis. The SEA indicated differences in the annual deviations of EW and LW widths in response to drought. Significant decreases in EW width were observed in five out of nine sites during the year of drought whereas only one site showed significant reduction in LW width one year after the drought (Table 8). The mean deviations in EW and LW widths during the drought year were -0.48 and -0.31, respectively, whereas they showed similar responses up to two years after the drought. Table 8. Annual earlywood (EW) and latewood (LW) width departures observed up to two years after the occurrence of a severe drought (in year “0”). Departures were calculated in response to selected severe droughts (1967, 1986, 1994 and 2002; see Figure 20). Significant (p ≤0.05) values are in bold. EW LW Years after drought Years after drought Site 0 1 2 0 1 2 VA −0.35 −0.21 −0.14 −0.28 −0.52 0.17 CV −0.34 −0.24 0.16 −0.30 −0.05 0.24 CS −0.46 −0.04 0.23 −0.37 0.05 0.18 PU −0.20 −0.26 0.09 −0.11 −0.26 −0.08 TA −0.74 −0.42 0.01 −0.28 −0.18 −0.01 VM −0.56 −0.18 0.28 −0.25 −0.12 −0.06 MP −0.45 −0.15 −0.25 −0.38 −0.26 0.23 PL −0.85 −0.43 −0.28 −0.44 −0.20 0.08 MA −0.41 0.35 0.3 −0.35 −0.07 0.02 Mean −0.48 −0.18 0.04 −0.31 −0.18 0.09 Results 117 The correlations between EW and LW growth indices and the different SPI time-scales reached maximum values at similar time-scales (10-14 months), despite significant correlations were found up to 48 (EW) and 35 (LW) months (Figure 25 A). The strongest responses of EW to drought intensity (r = 0.70–0.72) were observed in July, whereas LW responded strongly (r = 0.54) to SPI September values. The months when the highest responses of EW and LW width to drought at inter-annual scales were detected, coincided with those in which were observed low production rates of EW (July) and LW (August-September) tracheids at intra annual scales (Figure 25 B). Finally, the maximum production rates of EW (May-June) and LW (mid July-August) tracheids occurred approximately one month before the strongest impact of drought on inter-annual EW and LW growth series was detected. A) B) Figure 25. A) Mean correlation coefficients of earlywood (EW) and latewood (LW) width chronologies and the SPI drought index calculated at different time scales (1-48 months, x axis) from January to December (y axis) for the nine study sites. Bold lines frame significant correlations (p≤0.05). B) Calculated daily rates of EW and LW tracheids production in P. halepensis during the year 2010. Results 118 Certain variability among sites regarding EW and LW growth responses to drought was also detected (Figure 26), but these local responses matched the general patterns described before. However, the strongest response of EW growth to drought was found for the sites VM, CV, CS, VA whereas the sites TA, PU, MA, PL and MP were less affected by drought. The same pattern was also observed with respect to the LW. 0.2 0.2 0.4 0.4 0.2 0.0 -0.2 0.2 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.4 0.4 0.6 0.6 0.4 0.2 0.6 0.4 0.4 0.0 0.2 510 15 20 25 30 35 40 45 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.4 0.4 0.6 0.6 0.6 0.4 0.2 0.0 0.4 0.4 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.2 0.4 0.6 0.6 0.4 0.4 0.2 0.0 0.4 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.4 0.6 0.6 0.4 0.6 0.2 0.4 0.0 0.2 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.2 0.4 0.4 0.4 0.4 0.6 0.4 0.6 0.4 0.2 0.4 0.2 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.2 0.2 0.2 0.4 0.2 0.4 0.2 0.2 0.0 0.2 0.4 0.4 0.4 0.4 0.2 0.6 0.4 0.2 0.6 0.4 0.2 510 15 20 25 30 35 40 45 0.0 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.4 0.4 0.4 0.2 0.2 0.2 0.0 0.0 0.2 0.2 0.2 0.0 0.0 0.0 0.4 0.4 0.6 0.2 0.4 0.2 0.4 0.0 0.4 0.6 0.2 0.6 0.4 0.2 0.0 0.4 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 0.2 0.2 0.2 0.2 0.2 0.0 0.2 0.2 0.2 0.2 0.0 0.2 0.4 0.4 0.2 0.4 0.0 0.2 0.4 0.4 0.4 0.2 0.4 0.6 0.6 0.6 0.2 0.6 0.4 0.2 0.0 0.0 0.2 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.2 0.0 0.0 0.4 0.4 0.2 0.4 0.0 0.4 0.4 0.6 0.6 0.4 0.4 0.6 0.2 0.4 0.0 0.4 0.2 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.0 0.2 0.4 0.6 0.2 0.4 0.2 0.2 0.4 0.2 0.2 0.2 0.4 0.4 0.2 0.2 0.2 0.2 0.2 0.4 0.4 0.4 0.4 0.2 0.6 0.4 0.6 0.2 0.0 0.4 0.2 0.2 0.2 0.2 0.2 0.2 0.4 0.2 0.4 0.4 0.4 0.2 0.2 0.0 0.2 0.2 Time scale (months) Months Earlywood Latewood MA MP PL VA VM TA CS CV PU Figure 26. Pearson Correlation coefficients obtained relating earlwyood and latewood width indices vs. the monthly drought index (SPI) calculated at different time scales (1-48 months, x axis) from January up to December (y axis) in the nine study sites. Bold lines frame significant correlations (p≤0.05). This is in line with the findings of Fekedulegn et al. (2003) who stressed that forest responses to drought may be affected by a combination of many factors including precipitation, temperature, inherent species’ characteristics, and site topographic features. These factors may impose additional constrains on the growth responses to drought. As revealed in this study, the growth sensitivity to drought was also affected by soil and topographic conditions. For instance, the lithology (millstones and gypsums) that characterize xeric environments in the study area enhance the negative effects of water deficit on growth since soils are generally shallow, characterized by limited ability to hold adequate moisture (Guerrero et al. 1999; Vicente-Serrano 2007), which increase the negative effects of drought on tree growth. In the Pyrenees, the dominant soils (e.g., inceptisol) had an inverse association with the growth responsiveness to drought, suggesting the presence of mitigating effects of the negative impacts of water deficit on growth, most likely due to the high capacity of these soils to hold water in the deeper layers. This is the case of many A. alba forests located in valley bottoms of the Pyrenees, characterized by deep soils and humid climatic conditions (Macias et al. 2006; Camarero et al. 2011). We expected that topographic factors such as elevation and slope would exert contrasting influences on the species responses to drought in xeric and mesic forest sites, respectively. However, we found that the growth responses to drought were only affected by slope in the case of xeric sites (PC1), most likely due to its local influence on surface runoff and water retention by soils. The studied forests in xeric sites, mostly located in low elevation but topographically complex areas (plain localities have been mostly converted to croplands), grow under an exacerbated drought stress in undulating or moderate slopes since in these areas trees are more sensitive to the rapidly occurring water shortages (Fekedulegn et al. 2003). Topographic position has already been shown to drive the soil water availability in areas subjected to seasonal droughts affecting tree growth and the species composition of diverse ecosystems as tropical forests (Engelbrecht et al. 2007). Xeric sites are also characterized by low precipitation and high temperatures which enhance the drought impacts on growth (Camarero et al. 2010). In these areas, high temperatures in summer most likely deplete soil water reserves causing drought-induced embolism. Slope and aspect can also influence growth locally through changes in the radiation received by trees (Tardif et al. 2003; Leonelli et al. 2009). Generally, south-facing slopes in xeric areas with low water holding capacity increase the drought impacts on tree growth (Sa ´nchez-Salguero et al. 2010). The opposite was observed in A. alba forests located in northoriented slopes with deep soils. Overall, topography acts as a local modulator of the effects of drought on growth but further research on this subject in Circum-Mediterranean forests is desirable. The tree features such as tree size (diameter) or the growth rate (mean tree-ring width) and vegetation activity indices (NDVI, EVI) showed positive and negative associations with the growth responsiveness to drought in xeric and mesic sites, respectively. These associations indicate that low water availability associated with prolonged and intense drought lead to a decreasing canopy growth (e.g., shoot extension, leaf production) and probably a decline in photosynthetic activity causing a reduction in cambial activity and hydraulic conductivity (Corcuera et al. 2004a, b; Linares et al. 2009; Montserrat-Martı ´et al. 2009; Vicente-Serrano et al. 2010). Such growth decline may further enhance the species’ vulnerability against drought stress. In conclusion, our study highlights that Mediterranean forests show high spatial and temporal variability in terms of growth responses to drought. This variability observed among species and sites was significantly driven by climatic, topographic variables, and biotic variables indicating that a combination of variables shape the species’ behavior in response to drought. Since climate models predict rising temperatures and enhanced evapotranspiration for the Mediterranean Basin (Giorgi and Lionello 2008), our findings suggest that warming-related drought stress might affect growth dynamics on different time scales in mesic than in xeric forests. However, disentangling the relative effects of warmer conditions and reduced precipitation on tree growth is an unsolved challenge which probably requires a multiproxy approach based on longterm data of radial growth, isotopic discrimination in wood, and remote sensing variables. Acknowledgments Edmond Pasho thanks the financial support of the Albanian Ministry of Education and Science. This work has been supported by the research projects CGL2008-01189/BTE, CGL200804847-C02-01/BOS financed by the Spanish Commission of Science and Technology and FEDER, and EUROGEOSS (FP7-ENV-2008-1226487) and ACQWA (FP7-ENV-2007-1212250) financed by the VII Framework Program of the European Commission. JJC thanks the support of ARAID. References Abrams MD, Ruffuer MC, Morgan TA (1998) Tree-ring responses to drought across species and contrasting sites in the ridge and valley of central Pennsylvania. For Sci 44:550–558 Adams HD, Kolb TE (2005) Tree growth response to drought and temperature in a mountain landscape in northern Arizona, USA. J Biogeogr 32:1629–1640 Allen CD, Macalady AK, Chenchouni H, Bachelet D, McDowell N, Vennetier M, Kitzberger T, Rigling A, Breshears DD, Hogg EH et al (2010) A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. For Ecol Manage 259:660–684 Eur J Forest Res 123 Author's personal copy Andreu L, Gutie ´rrez E, Macias M, Ribas M, Bosch O, Camarero JJ (2007) Climate increases regional tree-growth variability in Iberian pine forests. Glob Ch Biol 13:1–12 Aussenac G (2002) Ecology and ecophysiology of circum-Mediterranean firs in the context of climate change. Ann For Sci 59: 823–832 Bannari A, Morin D, Bonn F, Huete AR (1995) A review of vegetation indices. Rem Sens Rev 13:95–120 Bogino S, Bravo F (2008) Growth response of Pinus pinaster Ait. to climatic variables in central Spanish forest. Ann For Sci 65:506–518 Borghetti M, Cinnirella S, Magnani F, Saracino A (1998) Impact of long-term drought on xylem embolism and growth in Pinus halepensis Mill. Trees 12:187–195 Briffa K, Cook ER (1990) Methods of response function analysis. In: Cook ER, Kairiuktsis LA (eds) Methods of dendrochronology. Kluwer, Dordrecht, pp 240–247 Camarero JJ, Olano JM, Parras A (2010) Plastic bimodal xylogenesis in conifers from continental Mediterranean climates. New Phytol 185:471–480 Camarero JJ, Bigler C, Linares JC, Gil-Pelegrı ´n E (2011) Synergistic effects of past historical logging and drought on the decline of Pyrenean silver fir forests. For Ecol Manag 262:759–769 Carlson TN, Ripley DA (1997) On the relation between NDVI, fractional vegetation cover, and leaf area index. Rem Sens Environ 62:241–252 Cook ER (1985) A time series approach to tree-ring standardization. Dissertation, University of Arizona, Tucson Corcuera L, Camarero JJ, Gil-Pelegrin E (2004a) Effects of a severe drought on Quercus ilex radial growth and xylem anatomy. Trees 18:83–92 Corcuera L, Camarero JJ, Gil-Pelegrı ´n E (2004b) Effects of a severe drought on growth and wood-anatomical properties of Quercus faginea. IAWA J 25:185–204 Costa M, Morla C, Sainz H et al (2005) Los bosques Ibe ´ricos: una interpretacio ´n geobota ´nica. Planeta, Madrid Cuadrat JM, Saz MA, Vicente-Serrano SM (2007) Atlas Clima ´tico de Arago ´n. Gobierno de Arago ´n, Zaragoza De Luis M, Gric ˇar J, C ˇufar K, Ravento ´s J (2007) Seasonal dynamics of wood formation in Pinus halepensis from dry and semi-arid ecosystems in Spain. IAWA J 28:389–404 De Luis M, Novak K, C ˇufar K, Ravento ´s J (2009) Size mediated climate-growth relationships in Pinus halepensis and Pinus pinea. Trees 23:1065–1073 Engelbrecht BMJ, Comita LS, Condit R, Kursar TA, Tyree MT, Turner BL, Hubbell SP (2007) Drought sensitivity shapes species distribution patterns in tropical forests. Nature 447: 80–82 Fekedulegn D, Hicks RR, Colbert JJ (2003) Influence of topographic aspect, precipitation and drought on radial growth of four major tree species in an Appalachian watershed. For Ecol Manage 177:409–425 Flexas J, Medrano H (2002) Drought inhibition of photosynthesis in C3 plants: stomatal and non-stomatal limitations revisited. Ann Bot 89:183–189 Fritts HC (2001) Tree rings and climate. Academic Press, London Garcı ´a-Ruiz JM, Lo ´pez-Moreno JI, Vicente-Serrano SM, Lasanta T, Beguerı ´a S (2011) Mediterranean water resources in a global change scenario. Earth Sci Rev 105:121–139 Giorgi F, Lionello P (2008) Climate change projections for the Mediterranean region. Glob Planet Ch 63:90–104 Guehl JM, Aussenac G, Bouachrine J, Zimmermann R, Pennes JM, Ferhi A, Grieu P (1991) Sensitivity of leaf gas exchange to atmospheric drought, soil drought, and water-use efficiency in some Mediterranean Abies species. Can J For Res 21: 1507–1515 Guerrero J, Alberto F, Hodgson J, Garcı ´a-Ruiz JM, Montserrat G (1999) Plant community patterns in a gypsum area of NE Spain. Interactions with topographic factors and soil erosion. J Arid Env 41:401–410 Gutie ´rrez E, Campelo F, Camarero JJ, Ribas M, Munta ´n E, Nabais C, Freitas H (2011) Climate controls act at different scales on the seasonal pattern of Quercus ilex L. stem radial increments in NE Spain. Trees 25:637–646 Hayes M, Wilhite DA, Svoboda M, Vanyarkho O (1999) Monitoring the drought using the standardized precipitation index. B Am Meteorol Soc 80:429–438 Holmes RL (1983) Computer-assisted quality control in tree-ring dating and measurement. Tree-Ring Bull 43:69–78 Houghton JT, Ding Y, Griggs DJ, Noguer M, Van der Linden PJ, Dai X, Maskell K, Johnson CA (eds) IPCC (2001) Climate change 2001: the scientific basis. Contribution of working group I. Cambridge University Press, Cambridge Hsiao TC (1973) Plant responses to water stress. Ann Rev Plant Physiol 24:519–570 Huete A, Didan K, Miura T, Rodriguez EP, Gao X, Ferreira LG (2002) Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Rem Sens Env 83:195–213 IGN (Instituto Geogra ´fico Nacional) (2006) Atlas Nacional de Espan ˜a. Geologı ´ a. Geomorfologı ´ a y Edafologı ´ a. Ministerio de Fomento, Madrid, Spain Knipling EB (1970) Physical and physiological basis for the reflectance of visible and near-infrared radiation from vegetation. Rem Sens Env 1:155–159 Koepke DF, Kolb TE, Adams HD (2010) Variation in woody plant mortality and dieback from severe drought among soils, plant groups, and species within a northern Arizona ecotone. Oecologia 163:1079–1090 Kuparinen A, Savolainen O, Schurr FM (2010) Increased mortality can promote evolutionary adaptation of forest trees to climate change. For Ecol Manage 259:1003–1008 Legendre P, Legendre L (1998) Numerical ecology. Elsevier, New York Leonelli G, Pelfini M, Battipaglia G, Cherubini P (2009) Site-aspect influence on climate sensitivity over time of a high-altitude Pinus cembra tree-ring network. Clim Ch 96:185–201 Linares JC, Tı ´scar PA (2010) Climate change impacts and vulnerability of the southern populations of Pinus nigra subsp. salzmannii. Tree Physiol 30:795–806 Linares JC, Camarero JJ, Carreira JA (2009) Plastic responses of Abies pinsapo xylogenesis to drought and competition. Tree Physiol 29:1525–1536 Linares JC, Camarero JJ, Carreira JA (2010) Competition modulates the adaptation capacity of forests to climatic stress: insights from recent growth decline and death in relict stands of the Mediterranean fir Abies pinsapo. J Ecol 98:592–603 Macias M, Andreu L, Bosch O, Camarero JJ, Gutie ´rrez E (2006) Increasing aridity is enhancing silver fir Abies alba (Mill.) water stress in its south-western distribution limit. Clim Ch 79: 289–313 Martı ´nez-Vilalta J, Pin ˜ol J (2002) Drought-induced mortality and hydraulic architecture in pine populations of the NE Iberian Peninsula. For Ecol Manage 161:247–256 McDowell N, Pockman WT, Allen CD, Breshears DD, Cobb N, Kolb T, Plaut J, Sperry J, West A, Williams DG, Yepez EA (2008) Mechanisms of plant survival and mortality during drought: why do some plants survive while others succumb to drought? New Phytol 178:719–739 McKee TBN, Doesken J, Kleist J (1993) The relationship of drought frequency and duration to time scales. In: Eight conference on applied climatology. American meteorological society, Anaheim, CA, pp 179–184 Eur J Forest Res 123 Author's personal copy Mitasova H, Mitas L (1993) Interpolation by regularized spline with tension. Math Geol 25:641–655 Montserrat-Martı ´G, Camarero JJ, Palacio S, Pe ´rez-Rontome ´C, Milla R, Albuixech J, Maestro M (2009) Summer-drought constrains the phenology and growth of two coexisting Mediterranean oaks with contrasting leaf habit: implications for their persistence and reproduction. Trees 23:787–799 Myneni RB, Hall FG, Sellers PJ, Marshak AL (1995) The interpretation of spectral vegetation indexes. IEEE Trans Geosci Rem Sens 33:481–486 Navas A, Machı ´n J (1998) Spatial analysis of gypsiferous soils in the Zaragoza province (Spain), using GIS as an aid to conservation. Geoderma 87:57–66 Ninyerola M, Pons X, Roure JM (2007) Monthly precipitation mapping of the Iberian Peninsula using spatial interpolation tools implemented in a geographic information system. Theor Appl Climatol 89:195–209 Orwig DA, Abrams MD (1997) Variation in radial growth responses to drought among species, site, and canopy strata. Trees 11:474–484 Pasho E, Camarero JJ, de Luis M, Vicente-Serrano SM (2011) Impacts of drought at different time scales on forest growth across a wide climatic gradient in north-eastern Spain. Agr For Meteorol 151:1800–1811 Pen ˜a JL, Pellicer F, Julia ´n A, Chueca J, Echeverrı ´a MT, Lozano MV, Sa ´nchez M (2002) Mapa Geomorfolo ´gico de Arago ´n. Consejo de Proteccio ´n de la Naturaleza de Arago ´n, Zaragoza Pereira JS, Chaves MM (1995) Plant responses to drought under climate change in Mediterranean-type ecosystems. In: Moreno JM, Oechel WC (eds) Global change and mediterranean-type ecosystems. Ecological Studies, vol. 117. Springer, New York, pp 140–160 Phillips DL, Dolph J, Marks D (1992) A comparison of geostatistical procedures for spatial analysis of precipitation in mountainous terrain. Agr For Meteorol 58:119–141 Pichler P, Oberhuber V (2007) Radial growth response of coniferous forest trees in an inner Alpine environment to heat-wave in 2003. For Ecol Manage 242:688–699 Pons X (2011) MiraMon. Geographic information system and remote sensing software. UAB-CREAF, Barcelona Pons X, Ninyerola M (2008) Mapping a topographic global solar radiation model implemented in a GIS and refined with ground data. Int J Climatol 28:1821–1834 Richman MB (1986) Rotation of principal components. J Clim 6:29–35 Rolland CR, Michalet C, Desplanque A, Petetin A, Aime ´S (1999) Ecological requirements of Abies alba in the French Alps derived from dendro-ecological analysis. J Veg Sci 10:297–306 Rouse JW, Haas RH, Shell JA, Deering DW (1973) Monitoring vegetation systems in the Great Plains with ERTS-1. Third Earth Res Technol Satellite Symp 1:309–317 Sa ´nchez-Salguero R, Navarro-Cerrillo RM, Camarero JJ, Ferna ´ndezCancio A (2010) Drought-induced growth decline of Aleppo and maritime pine forests in south-eastern Spain. For Syst 19:458–469 Sarris D, Christodoulakis D, Ko ¨rner C (2007) Recent decline in precipitation and tree growth in the eastern Mediterranean. Glob Ch Biol 13:1–14 Sarris D, Christodoulakis D, Ko ¨rner C (2011) Impact of recent climatic change on growth of low elevation eastern Mediterranean forest trees. Clim Ch 106:203–223 Tardif J, Bergeron Y (1997) Comparative dendroclimatological analysis of two black ash and two white cedar populations from contrasting sites in the Lake Duparquet region, northwestern Quebec. Can J For Res 27:108–116 Tardif J, Camarero JJ, Ribas M, Gutierrez E (2003) Spatiotemporal variability in tree growth in the central Pyrenees: climatic and site influences. Ecol Monogr 73:241–257 Vicente-Serrano SM (2006) Differences in spatial patterns of drought on different time scales: an analysis of the Iberian Peninsula. Water Res Manage 20:37–60 Vicente-Serrano SM (2007) Evaluating the impact of drought using remote sensing in a Mediterranean, semi-arid region. Nat Hazards 40:173–208 Vicente-Serrano SM, Beguerı ´a S (2003) Estimating extreme dry-spell risk in the middle Ebro valley (Northeastern Spain): a comparative analysis of partial duration series with a General Pareto distribution and Annual maxima series with a Gumbel distribution. Int J Climatol 23:1103–1118 Vicente-Serrano SM, Saz MA, Cuadrat JM (2003) Comparative analysis of interpolation methods in the middle Ebro valley (Spain): application to annual precipitation and temperature. Clim Res 24:161–180 Vicente-Serrano SM, Cuadrat JM, Romo A (2006) Aridity influence on vegetation patterns in the middle Ebro valley (Spain): evaluation by means of AVHRR images and climate interpolation techniques. J Arid Env 66:353–375 Vicente-Serrano SM, Lanjeri S, Lo ´pez-Moreno JI (2007) Comparison of different procedures to map reference evapotranspiration using geographical information systems and regression-based techniques. Int J Climatol 27:1103–1118 Vicente-Serrano SM, Lasanta T, Gracia C (2010) Aridification determines changes in leaf activity in Pinus halepensis forests under semi-arid Mediterranean climate conditions. Agr For Meteorol 150:614–628 Wigley TML, Briffa KR, Jones PD (1984) On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. J Clim Appl Meteorol 23:201–203 Willmott CJ (1982) Some comments on the evaluation of model performance. Bull Am Meteorol Soc 63:1309–1313 Xoplaki E, Luterbacher J, Gonza ´lez-Rouco JF (2006) Mediterranean summer temperature and winter precipitation, large-scale dynamics, trends. Il Nuovo Cimento 29:45–54 Eur J Forest Res 123 Author's personal copy Doctoral dissertation