Spatial and seasonal variability of the snowline elevation in the Langtang Valley in Nepal, using Landsat remotely sensed data from 1999-2013

PROJECTE O TESINA D’ESPECIALITAT Títol Spatial and seasonal variability of the snowline elevation in the Langtang Valley in Nepal, using Landsat remotely sensed data from 1999-2003 Autor/a Marc Girona i Mata Tutor/a Ernest Bladé i Castellet Departament Enginyeria Hidràulica, Marítima i Ambiental Intensificació Enginyeria Hidràulica, Marítima i Ambiental Data 16 de febrer de 2015 Thesis submitted in partial fulfillment of the requirements for the Degree in Civil Engineering Spatial and seasonal variability of the snowline elevation in the Langtang Valley in Nepal, using Landsat remotely sensed data from 1999-2013 presented by Marc Girona i Mata Student of the Degree in Civil Engineering Polytechnic University of Catalonia Supervisors Dr. Francesca Pellicciotti Institute of Environmental Engineering Chair of Hydrology and Water Resources Management, ETH Zurich Silvan Ragettli Institute of Environmental Engineering Chair of Hydrology and Water Resources Management, ETH Zurich Evan Miles Scott Polar Research Institute Department of Geography, University of Cambridge Responsible Prof. Dr. Ernest Blade Castellet Research group of Fluvial Dynamics and Hydrological Engineering Polytechnic University of Catalonia submitted on Zurich - February 16, 2015 ii Abstract The Langtang Valley -located in Nepalis part of the monsoon dominated region in the central Himalayas, and it is characterized by extreme orography and a high presence of glaciered areas. Nevertheless, this work focuses on the upper part of the Langtang River Basin, with outlet below the village of Kyanjing. Although the region has been object of several studies, little is known about the snowline behavior within the valley and the factors influencing it. This study proposes a new approach to extract the snowline data from remote sensed Landsat TM, ETM+ scenes, taking advantage of the high spatial resolution Landsat products o↵er. A point-scale description the snowline and its characteristics -altitude, slope, aspect, etc.- is obtained and analyzed. Two di↵erent spatial analysis of the snowline dynamics are carried out, each one preceded by a selection of scenes based on specific image-quality criteria. First, a monthly average analysis looks at typical and consistent annual snowline dynamics; and second, the topic of interseasonal variability of the snowline is addressed in greater detail. Both studies are statistically based, aiming to provide consistent and robust results. Results of this work show that climate -and specially precipitationplays a great e↵ect in the snowline altitude (SLA) characteristics and dynamics. Behavioral di↵erences of SLA for di↵erent periods of the year are clearly observable, confirming thus the existence of four di↵erentiated climatic seasons in the region. During winter and monsoon seasons precipitation events are stronger and more frequent, having a greater and dominating e↵ect in the SLA. Only during pre-monsoon and post-monsoon seasons, which are drier and more stable, the e↵ect of aspect in the SLA is clearly observed, with systematic SLA di↵erences between di↵erently orientated slopes. Interannual consistency of SLA dynamics is also shown in the results. Besides, the dominating e↵ect of precipitation in the SLA makes the latter a potentially precise indicator of the precipitation events, their frequency and magnitude. And last but not least, the upper Langtang Valley is shown to have not only temporal but also spatial variability of the SLA. Observed latitudinal and altitudinal SLA gradients are attributed to precipitation gradients within the catchment. Finally, Landsat TM and ETM+ products are proved to be a powerful instrument to study the snowline dynamics, either by analyzing the extracted data in statistically based studies, or by combining Landsat products with other sources of data. ii Resum La Vall de Langtang (Nepal) forma part de la regi´o central de l’Him`alaia dominada pel mons´o, i es caracteritza per la seva orografia extrema i l’alta pres`encia de glaceres. No obstant aix`o, aquest treball es centra en l’estudi de la part alta de la vall del riu Langtang, amb desembocadura a la vila de Kyanjing. Tot i que la regi´o ha estat objecte de diferents estudis, el comportament de la l´ınia de neu en la vall, aix´ı com els factors que la influencien, s´on questions poc estudiades i que encara no han estat resoltes. Aquest estudi proposa una nova manera d’extreure la l´ınia de neu a partir d’imatges de teledetecci´o Landsat TM i ETM+, aprofitant la seva alta resoluci´o espacial. Les caracter´ıstiques de la l´ınia de neu -altitud, pendent, orientaci´o, etc.- s´on obtingudes i analitzades a escala puntual, mitjanant dues an`alisis precedides, cadascuna, per una selecci´o d’escenes segons diferents criteris de qualitat d’imatge. En primer lloc, una an`alisi de mitjana mensual estudia les din`amiques anuals de la l´ınia de neu que tenen consist`encia interanual. En la segona an`alisi, es desgrana la variabilitat interestacional de la l´ınia de neu en detall. Tots dos estudis tenen una base estad´ıstica, amb l’objectiu d’obtenir resultats consistents i robustos. Els resultats d’aquest treball mostren que el clima -especialment la precipitaci´ot´e un gran efecte en la l´ınia de neu, les seves caracter´ıstiques i din`amiques. L’exist`encia de quatre estacions clim`atiques es confirma a partir de les difer`encies observades durant diversos per´ıodes de l’any en la l´ınia de neu. Durant les estacions d’hivern i mons´o, els esdeveniments de precipitaci´o s´on m´es forts i freqents i tenen, per tant, un efecte dominant sobre la l´ınia de neu. ´es nom´es durant les estacions del pre-mons´o i post-mons´o -m´es seques i estables climatol`ogicamentque l’efecte de l’orientaci´o de les pendents en la l´ınia de neu ´es clarament visible, amb difer`encies sistem`atiques entre diferents orientacions. L’alta consist`encia interanual de les din`amiques estacionals que segueix l´ınia de neu ´es tamb´e visible en els resultats. A m´es, l’efecte dominant de la precipitaci´o en la l´ınia de neu fa d’aquesta ´ultima un indicador potencial molt prec´ıs dels esdeveniments de precipitaci´o, la seva frequ`encia i magnitud. Per acabar, la variabilitat de la l´ınia de neu a la Vall de Langtang no nom´es t´e una vessant temporal estacional, sin´o que tamb´e existeix variabilitat espacial de la l´ınia de neu. Els gradients latitudinals i longitudinals observats en l’altitud de la l´ınia de neu s´on, en gran mesura, conseq`encia directa de gradients de precipitaci´o existents en la vall. Finalment, s’ha demostrat la validesa dels productes Landsat TM i ETM+ com a instrument per estudiar les din`amiques de la l´ınia de neu, ja sigui analitzant les dades en estudis estad´ıstics o combinant-les amb altres fonts d’informaci´o. iv Contents Abstract ............................................. ii Resum .............................................. iv List of Figures ......................................... viii List of Tables .......................................... xi 1 Introduction 1 1.1 Research objectives .................................... 2 1.2 Report outline ...................................... 3 2Studysite 5 2.1 Climate .......................................... 5 2.2 Division of the upper Langtang Valley in sub-catchments .............. 9 2.3 Characteristics of sub-catchments ............................ 13 3 Methodology and Data 19 3.1 Data ............................................ 19 3.1.1 Landsat imagery ................................. 19 3.1.2 Digital Elevation Model (DEM) ......................... 19 3.1.3 Radiation and climate data ........................... 20 3.2 Digital Elevation Model pre-processing to obtain sub-catchment division ...... 20 3.3 Determination of the snowline using GIS ........................ 20 3.3.1 Snow cover outline ................................ 20 3.3.2 Noise reduction .................................. 21 3.4 Quality analysis of the Landsat imagery ........................ 22 3.5 Database composition .................................. 22 3.5.1 Processing the snowline datasets ........................ 22 3.5.2 Division into aspect quarters .......................... 23 3.6 Spatial analysis of the snowline characteristics and dynamics ............ 23 3.6.1 Monthly average analysis ............................ 23 3.6.2 Seasonal snowline variability .......................... 24 vi Chapter 1 Introduction Water demand in Asia is expected to steeply increase with population growth, as well as with per capita food intake as regions become more developed [Immerzeel and Bierkens 2012]. Mountains are the water towers of the world, including Asia, where the Himalayas play a crucial role Viviroli et al. [2007]. Snow and glacial melt are important hydrological processes [Cruz et al. 2007, Immerzeel et al. 2009], and melt characteristics are expected to be greatly a↵ected by temperature and precipitation changes [Barnett et al. 2005]. These hydrological processes are definitively important players to determine water availability for other human uses. Climate change is a threat for Asia’s water towers. However, the e↵ect of climate change on water availability in Asia may di↵er substantially from one region to another and thus cannot be generalized [Immerzeel et al. 2010]. Therefore, regional study of climatic conditions and their e↵ect on melt are suggested. Taking advantage of remotely sensed images for monitoring snow and glacial melt is revealed to be a feasible and e↵ective approach for that purpose. However, the potential use of satellite imaging -and particularly regarding Landsat imageryis still to be explored and enhanced. Although remote sensing is an arising topic and widely used in glaciological studies, many of the latter are focused on mapping snow cover fraction (SCF). Among these, the great majority do not make use of Landsat data but other imagery sources, particularly MODIS remotely sensed data [Salomonson and Appel 2004, Vikhamar and Solberg 2003, among many others]. Indeed, it does seem reasonable to use a data source with higher temporal frequency -MODIS provides daily images while Landsat has a temporal resolution of 16 dayswhen studying a parameter -SCF- that is barely a↵ected by the spatial resolution -500 m in MODIS and 30 m for Landsat-. For that reason, MODIS snow cover products have been highly standardized [Salomonson and Appel 2004, Hall et al. 2002]. On the other hand, Landsat imagery -with 30 m spatial resolutionhas also been used to map SCF with more precision [Rosenthal and Dozier 1996], but there is always the downside of less temporal resolution. For that reason, the usage of Landsat imagery for snow cover analysis is basically restricted or addressed to the study of a few number of scenes, these corresponding either to a short period of time with high availability -minor SCF changes might be relevant when climate data is abundant- or to analog dates from di↵erent years -scenes with high quality are selected as representative of an entire period of time and compared with analog ones from di↵erent years-. Finally, less studies have addressed the study of another parameter, equally relevant and even more crucial for the regional study of climate. That is the snowline. A line on the earth surface intersected by an hypothetical surface on which ablation of snow and ice is balanced with snowfall is called snowline. Local variety of snowfall and ablation rates in alpine regions caused by extreme topographical conditions results in a uneven and rugged snowline, the 1 CHAPTER 1. INTRODUCTION 2 ”actual snowline”. Connecting the lower limits of the ”actual snowline”, the so-called ”orographic snowline” is obtained. However, snow may be dropped to lower elevations by climatic e↵ects or topographical conditions and survive there, even at elevations snow does not reach. The ”regional or climatic snowline” is defined as the snowline excluding the irregularities caused by orography. Thus, an estimation of the ”climatic snowline” might be obtained by averaging the elevations ”actual snowline” reaches [Nogami 1970]. The methodology proposed in this thesis to extract the snowline (SL) at the point scale from remotely sensed Landsat images has not been taken from any other study, but it is rather a new approach to the study of SL dynamics. 1.1 Research objectives In this chapter, the main research objectives of the thesis are presented and briefly discussed: 1. Assessing the potential use and reliability of Landsat TM, ETM+ imagery for the study of the snowline dynamics. Due to its unique characteristics -high spatial resolution but rather coarse temporal resolution-, the use of Landsat TM and ETM+ imagery might be very beneficial for certain purposes, whereas it is certainly not suitable for many other applications. It is thus a main objective of this study to evaluate the applicability of Landsat imagery on the study of the snowline dynamics. 2. Estimating the seasonal and annual evolution of the snowline altitude (SLA) and their inter-annual variability. There are many di↵erent factors influencing the SLA. However, little is known about the direction and magnitude in which these factors influence SLA. Moreover, there exists no exhaustive and continued register of the SLA in the Langtang Valley. By estimating the seasonal and annual evolution of the SLA, assessing also their stability from year to year, generalized patterns of behavior might be determined, providing crucial information about which factors have a major influence into the snowline dynamics. 3. Studying the e↵ect of aspect in the SLA. From all the factors regulating the SLA, aspect is believed to have an important e↵ect. While it is true that other factors might be influencing SLA in greater measure, these remain usually constant over small spatial variations. To the contrary, aspect might change radically over short distances. The di↵erences of SLA between aspect ranges are studied in this thesis in order to determine the influence aspect in the same region, where other factors either remain constant or vary depending on aspect. 4. Assessing the spatial variation of climate -with special focus on precipitationin the upper Langtang Valley using the SLA as an indicator, and the e↵ect of climate in the SLA. CHAPTER 1. INTRODUCTION 3 The upper Langtang Valley is characterized by extreme topography. Due to this, climatic conditions have a high degree of spatial variation throughout the valley, being particularly visible in precipitation events. By looking at the spatial variation of the SLA, the magnitude of the climatic spatial variation and the e↵ect of climate -precipitation and temperaturein the SLA might be assessed. 5. Building a database with the extracted snowline data-sets to be used in further research. Finally, an important objective of this study is also to collect the extracted snowline data-sets and enhance their usability in order for further studies to take advantage of them. 1.2 Report outline This section provides a general overview of the report, presenting the contents included in each chapter. Chapter 2 is dedicated to the study site -the upper Langtang Valley-. The climate and some of its main characteristics are explained there, as well as the division of the catchment in smaller study regions. Chapter 3 includes detailed explanations of the methodology and data sources used in the study. Suitability of the chosen methods and data sources is also discussed in this chapter. Chapter 4 presents results obtained in this study. A quality analysis of the Landsat imagery is exposed in first place, followed by results of two di↵erent spatial analysis focused on annual and seasonal SLA dynamics respectively. Chapter 5 is dedicated to the discussion of results. The main behavioral patterns shown to be persistent over time are discussed and explanations for these patterns are either confirmed or hypothesized. This chapter also includes a brief discussion on the usefulness of Landsat imagery for the study of the snowline, and some suggestions for further research using the data extracted in this study. Chapter 6 summarizes main conclusions of the study and recommendations for further research. Chapter 2 Study site Introduction The Langtang River Basin is located in the monsoon dominated central part of the Himalayas, in Nepal (Figure 2.1). It lies about 130km north from Kathmandu Valley (Figure 2.2), close to the border with Tibet, China. The Langtang River flows through the main valley, which is typically U-shaped. Its drainage area is 585 km2of which 155 km2is glaciered, and the elevation ranges from 1406 m in Syafru Besi to the summit of Langtang Lirung at 7234 m. The forests in the region have temperate and sub-alpine vegetation. This study is focused on the upper Langtang River Basin, with outlet bellow the village of Kyanjing (Figure 2.3). From its 350 km2, about 31% is glaciered and 24% corresponds to debriscovered glaciers [Pellicciotti et al. 2014]. Langtang, Langshisha, Yala, Lirung or Shalbachum are some of the most important glaciers located in this catchment. 2.1 Climate The climate in the central part of Nepal Himalayas is dominated by monsoon circulation with easterly winds in the summer and westerly winds from October to May. In the Langtang Valley, 77% of annual precipitation occurs during the monsoon period (June to September; mean value over period from 1957 to 2002; [Uppala et al. 2005]. This percentage ranges from 70-85% depending on the location in Nepal [Ives and Messerli 1989, Singh 1985]. This suggests climate in the region to be characterized by two main periods: a dry period (October to May) and a wet period (June to September). More recent studies indicate the existence of four distinct seasons: pre-monsoon season lasts from March to mid-June and is characterized by gradual increase of air temperature with high diurnal temperature variation, while few and rather irrelevant precipitation events are recorded. During monsoon season (June to September), light daily precipitation events take place. Air temperature is positive but has low diurnal variability, mainly due to thick and persistent cloud cover. In the latter part of monsoon season, magnitude of precipitations events is considerably higher while their frequency decreases. The weather is fine during post-monsoon (October-November), with almost no precipitation and gradual air temperature decrease. Diurnal fluctuations of temperature increase, being comparable to those of pre-monsoon. The winter season (December to February) is dominated by the lowest temperatures, with few but extreme precipitation events, usually in form of snow and higher than daily precipitation values during monsoon [Immerzeel et al. 2014, Shiraiwa et al. 1992]. This results in a large amount of snowfall in the valley, and has 5 CHAPTER 2. STUDY SITE 6 Figure 2.1 – Map of Nepal and its bordering countries. significant e↵ect on the snowline [Morinaga et al. 1987]. Precipitation in winter is caused by westerly troughs but also associated with the existence of large scale snow cover [Seko 1987], and it is characterized by very large inter-annual variability [Seko and Takahashjl 1991]. The starting time of pre-monsoon temperature increase is also subjected to high inter-annual fluctuations [Morinaga et al. 1987]. Due to the extreme topographical variation, substantial spatial variation of the precipitation events -magnitude and frequencyis observed not only at macro scale but also at local scale [Shrestha and Aryal 2010]. Indeed, small latitude variation turns into a significant divergence in the ratio from monsoon to non-monsoon precipitation in Nepal (0.14 in Kathmandu and 0.66 in the Langtang Valley [Seko 1987]. On the other hand, precipitation is also strongly dependent on altitude [Seko 1987, Immerzeel et al. 2014]. The strong local wind circulations induced by the orographic changes in the large scale mountain valley explain this significant variation of the local scale climatic conditions [Yasunari and Inoue 1978]. At the same time, the altitudinal and horizontal precipitation gradients -spatial variabilityhave a high degree of seasonal variability [Immerzeel et al. 2014]. According to Seko [1987], in the Langtang Valley, precipitation decreases with altitude from June to September whereas this correlation is positive in from December to March. Lower precipitation and higher air temperature is registered in the upper parts of the Langtang Valley. This suggests a negative precipitation gradient from central parts of the valley towards the North-east, attributed to the extreme topography, acting as a natural barrier. Consequently, glaciers in the upper part of the valley show less accumulation during monsoon season, and a drier winter in the North accentuates the equilibrium line altitude (ELA) regional di↵erences [Benxing CHAPTER 2. STUDY SITE 7 Figure 2.2 – Langtang Valley, located about 130 km north of Kathmandu, in the central part of Nepal. CHAPTER 2. STUDY SITE 8 Figure 2.3 – Upper Langtang Valley with respect to the entire catchment. CHAPTER 2. STUDY SITE 9 et al. 1984, Shiraiwa et al. 1992]. The upper Langtang Valley shows an heterogeneous spatial pattern of mass balance and ice volume, which confirms that heterogeneity is large also within a relatively small catchment [Pellicciotti et al. 2014]. Meteorological data used in this thesis is annexed in Appendix B. 2.2 Division of the upper Langtang Valley in sub-catchments In order to identify meaningful patterns in such an heterogeneous catchment [Pellicciotti et al. 2014], a division of the upper Langtang Valley in sub-areas becomes necessary. The snowline (SL), highly dependent on climate, is expected to behave di↵erently among sub-regions in the valley. Therefore, more and clearer patterns are expected to appear when looking at the data in a sub-regional -thus more detailedscale. A reasonable way to split the upper Langtang Valley in smaller regions is according to some of its morphological characteristics. Taking into account the fact that the valley is highly glaciered, it is convenient that derived sub-regions correspond, as much as possible, to the di↵erent glaciers in the catchment. A sub-catchment based division thus seemed to be a reasonable option, since it would potentially enclose entire glaciered areas within the same sub-region. Besides, the range of values that topographic parameters (altitude, aspect and slope) took was to be similar from one sub-catchment to another. The catchment was divided in seven di↵erent sub-catchments. Figure 2.4 shows the derived sub-catchments in the upper Langtang Valley and, at the same time, smaller regions representing draining areas within each sub-catchment. Glaciered regions of the basin appear green-colored in the figure. Sub-catchments as di↵erentiated study regions More than 70 di↵erent glaciers have been identified within the Langtang Valley [Shiraiwa and Yamada 1992]. However, its relevance is variable and only those which account for a substantial part of the catchment surface have been considered as a subject of study. Some of the most relevant glaciered regions of the upper Langtang River Basin appear in Figure 2.5. There is a high degree of coincidence between the glaciered areas in Figure 2.4 -glaciered regions in green- and actual glaciers in Figure 2.5. At the same time, the previous observation is to be extended at the sub-catchment level since main glaciers correspondence with sub-catchments shows to be evident, allowing to formulate the following statements: — sub-catchments 1 and 2 correspond to the Langtang glacier area; — sub-catchment 3 coincides with Langshisha glacier; — part of sub-catchment 6 corresponds to Lirung glacier; — sub-catchment 7 corresponds to Shalbachum glacier. While Langtang and Langshisha glaciers account for the entire eastern side of the Upper Langtang Catchment, a more exhaustive description of the glaciered areas in the western side of the valley is required. A detailed map of the western region is shown in Figure 2.6. Revisiting sub-catchment 6, two di↵erent sub-areas are observed. While it is true that Lirung glacier accounts for a big of it, Figure 2.6 exhibits the fact that Kimoshung glacier is also entirely contained in sub-catchment 6, but in its eastern side. Sub-catchments 4 and 5 do not account for a big part of glaciered area. Besides, the glaciered fraction observed in the southern part of both sub-catchments does not correspond any relevant glacier. However, Figure 2.6 shows how Yala glacier is located in the northern side of the subcatchment 5. CHAPTER 2. STUDY SITE 10 Figure 2.4 – Sub-catchments of study within the upper Langtang Valley. Green regions correspond to glaciered areas; streamline is grey-colored; dark blue lines surround the di↵erent sub-catchments -study regions-, numbered from 1 to 7; light blue lines represent smaller hydrological regions within each sub-catchment. CHAPTER 2. STUDY SITE 17 Figure 2.8 – Slope map of the Langtang Valley derived from a Digital Elevation Model of 100m grid size. Chapter 3 Methodology and Data 3.1 Data 3.1.1 Landsat imagery The study of seasonal changes in the snowline (SL) at a sub-catchment scale requires a good equilibrium between temporal and spatial resolution of the data. Both due to low accessibility and size of the study site, satellite remote sensing is an optimal solution for this purpose. However, diverse satellite imagery products with di↵erent characteristics are accessible and, therefore, a trade-o↵that optimally fulfills both preliminary conditions was to be found. Landsat, representing the world’s longest continuously acquired collection of space based moderateresolution land remote sensing data, is highly suitable for this case study. Landsat products Thematic Mapper (acquired with Landsat 4 and 5) and Enhanced Thematic Mapper Plus have a cell resolution of 30 m and a new Landsat image is available every 16 days approximately. Landsat TM scenes were obtained in Landsat missions 4 and 5, while ETM+ is a Landsat 7 product [United Stated Geological Survey 2014]. All available Landsat TM and ETM+ images for the period 1999-2013 were accessed. Unfortunately, the sensors are unable to penetrate cloud cover, so actual availability is lower, especially during monsoon period. In addition, some images have partially reduced quality due to local cloudiness. Landsat TM and ETM+ data used in this study have been previously atmospherically-corrected [Miles et al. 2013] via the LandCor implementation of the 6S radiative transfer model [Kotchenova and Vermote 2007, Zelazowski et al. 2011]. 3.1.2 Digital Elevation Model (DEM) The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) is an imaging instrument and one of the five sensors onboard Terra -a satellite of NASA’s Earth Observing System-, launched in December 1990. ASTER data is used to create detailed maps of land surface temperature, reflectance, and elevation, since February 2000. A 30 m resolution ASTER Global Digital Elevation Model resampled to 100 m resolution and slightly topographically corrected was used in this study. Its vertical accuracy is between 30 and 40 meters in areas with slopes less than 30. Although Landsat imagery has higher spatial resolution, 19 CHAPTER 3. Methodology and data 20 the precision required in this study with respect to topographic parameters -derived from the DEM- was perfectly fulfilled by the 100 m resampled DEM. The biggest source of error is found in the categorization of Landsat imagery cells and it cannot be prevented anyway. 3.1.3 Radiation and climate data Measurements of short-wave solar incoming radiation, air temperature and precipitation recorded with automatic weather stations (AWS) during di↵erent field campaigns in the upper Langtang Valley are used in this study. Potential clear-sky global irradiance simulated by Ragettli et al. [2014] -with a non parametric model based on Iqbal [1983] accounting for the position of the sun relative to every model grid cell at each time stepis used in this study. The vectorial algebra approach proposed by Corripio [2003] was used for the interaction between the solar beam and terrain geometry. More detailed description of the methods is to be found in Ragettli et al. [2014]. 3.2 Digital Elevation Model pre-processing to obtain subcatchment division The sub-regions of study were determined using a hydrological GIS extension tool, which -based on topographical informationcomputes the stream routing of the basin, considering as part of the streamline every cell draining more than a certain amount of upstream cells. The number of cells is fixed by setting a threshold and therefore the number of sub-catchments determined depends on this threshold. A small threshold returns a large number of small sub-catchments whereas a larger threshold returns less but bigger sub-catchments. In order for the study to be not only precise but also feasible, the number of sub-catchments derived was to be rather small. After an iteration process, a threshold of 1000 cells was set. With that threshold, the valley was divided into seven main sub-catchments, which were afterwards used as study regions. A map of the derived sub-catchments in the upper Langtang Valley is previously shown in Figure 2.4. 3.3 Determination of the snowline using GIS Spatial analysis and mapping of Landsat imagery was required to study the SL behavior and the behavioral di↵erences throughout the Langtang Valley. The use of GIS tools, concretely ArcGIS software , allowed the determination of the SL shape and the extraction of its characteristics at the point-scale, resulting in a database that contains SL features and attributes for each Landsat image. The process with which the SL was determined has three major and di↵erentiated steps, conceptually explained below. 3.3.1 Snow cover outline As already mentioned in Chapter 1, the snowline (SL) is theoretically equivalent to the snow cover outline. Unfortunately, due to the presence of partial cloud cover, deep shadow areas and CHAPTER 3. Methodology and data 21 cells containing no information, the snow cover outline is not fully visible in satellite imagery. Having said that, partial determination of the snow cover outline is still possible, and it turns to be a first step to extract SL information from the Landsat scenes. Therefore, a GIS routine was set up to extract the snow cover outline from each image. Snowcategorized cells adjacent to a non-snow-categorized cell are selected and converted to a closed polygon shape. As a result, this accounts both for the visible fraction of the snow cover outline and also for cells bordering cloud cover, deep shadowing and non-recorded areas. The other two steps of the routine are enclosed in the noise reduction process. 3.3.2 Noise reduction Cloud cover, deep shadowing and NaN filters A point-scale treatment of the closed polygon allows the removal of misclassified cells in the snow cover outline estimation. By clipping the Landsat image (raster) with the closed polygon, a layer of points containing Landsat information corresponding to the snow cover outline was obtained. Filters detecting cloud cover, deep shadowing and NaN (non-recorded area) closeness to the points were applied to extract the visible snow cover outline. Based on a threshold distance, points too close to a cloud cover, shadowing or non-recorded cell were taken as misclassified and, thus, not extracted as snow cover outline. 30 meters were set as threshold distance to assure that only those points whose corresponding cell was adjacent to an erroneous-value cell were eliminated. Outcrops and nunataks removal An outcrop is that part of a rock formation which is exposed at the Earth’s surface. A nunatak is a rocky summit or mountain range that stands above a surrounding ice sheet in an area that currently is being glaciated [Oxford University Press 2014] Occasional outcrops or nunataks in high altitude ranges are, in great measure, a direct consequence of terrain’s orography. Although the SL position is influenced by the topography, it certainly depends on other factors as well. Wind, precipitation and other climatic phenomena have, for instance, a big e↵ect on SL dynamics. It is not the point of this study to analyze outcrop’s and nunatak’s presence or behavior. Besides, due to its di↵erent behavior and high altitude range, considering them as a part of the SL would have caused results to be biased towards higher snowline altitudes (SLAs), which do not represent its actual behavior. As a matter of fact, keeping these regions in the analysis might had been a potential source of error, since misclassification of cells is more likely to appear in areas with extremely sheer topography. Therefore, removal of outcrops and nunataks was justified. In order to exclude outcrops and nunataks from the SL delineation, the GIS routine includes a third step, consisting on a filter that selects the closed areas within the snow cover fraction (SCF) that are not covered by snow. By removing the snow cover outline points surrounding them, suitable SL delineation is obtained. CHAPTER 3. Methodology and data 22 3.4 Quality analysis of the Landsat imagery The coarse temporal resolution of Landsat imagery becomes an impediment to a detailed and robust analysis of SL temporal dynamics. Contrariwise, due to its high spatial resolution (in comparison to other satellites), the information extracted from the scenes acquires more reliability. However, not only the temporal and spatial resolution of Landsat scenes a↵ect the results. Actual reliability of the extracted SL values depends also on the fraction of the Landsat scene containing useful information. Cloud cover and deep shadowing areas are considered useless for the study. The same applies for the non-recorded cells. All of them reduce the quality of Landsat scenes. For that reason, a ratio concerning image quality was calculated for all images: Image quality ratio = Cells containing useful information Total number of cells (3.1) As regards to the image quality e↵ect on the visible snow cover outline, which determines the SL shape afterwards, the snow cover visible fraction was obtained also for each Landsat image: Snow cover fraction = Cells categorized as Snow Total number of cells (3.2) While it is true that other values were taken into consideration, both ratios served in great manner the validation process of the imagery. 3.5 Database composition 3.5.1 Processing the snowline datasets The output of the GIS routine that derives the SL of the Langtang Valley is, for each Landsat scene, a set of points defined by their position in the basin and a wide range of attributes adding information. The attributes correspond to parameters that are object of study or, at least, might have influence in SL characteristics and dynamics. The following parameters, defined at point-scale and corresponding to the SL, were extracted and included into the database: — Altitude — Slope — Aspect — Sub-catchment — Potential radiation — Clear sky factor Note that slope and aspect information was directly extracted from the digital elevation model (DEM). The average number of snowline points (SLPs) derived from each image -considering all selected Landsat images from 1999-2013- is approximately 14300 points, and each of these datasets needed to be processed afterwards. As previously mentioned in Chapter 2, the Langtang Valley was CHAPTER 3. Methodology and data 23 divided in sub-catchments. Each sub-catchment was expected to show a relatively homogeneous behavior as far as the SL dynamics. Randomly picked Landsat scenes were processed to test the methodology and, as far as the extracted datasets, even by looking at one single sub-catchment, the range of values SL parameters took -especially the snowline altitude (SLA)- was quite large. Consequently, a further division of the SL data was reasonable and justified. 3.5.2 Division into aspect quarters A more detailed analysis of the SL required sorting the dataset according to another filter or classification. As previously explained in Section 2.3, aspect plays an important role in SL dynamics. Besides, in a potential correlation analysis among parameters -which was an option but did not take place in the end-, aspect was to be treated as a qualitative or dummy variable, setting aspect ranges -aspect has not a linear but circular nature (360 equals 0)-. Both reasons made aspect a good candidate for reaching a more detailed classification of the data. In high-mountain regions, areas exposed to similar climatic and topographic conditions but with di↵erent orientation might behave substantially di↵erent. That is due to the fact that mountainous regions are characterized by sheer topography and, thus, shadow plays a big role there, limiting the amount of incoming radiation received the terrain. In order to make the analysis sensitive to this phenomenon, four aspect ranges were set: East (45-135), South (135-225), West (225-315) and North (315-45). 3.6 Spatial analysis of the snowline characteristics and dynamics In this section, the methods used for the di↵erent analysis of the SL data are explained. Two types of analysis were considered. 3.6.1 Monthly average analysis The point of the following analysis is to extract consistent patterns on the annual evolution of the snowline altitude (SLA) in the upper Langtang Valley, with special attention to the divergences and similarities between di↵erent parts of the valley, represented by previously derived sub-catchments. Although temporal resolution of Landsat imagery is rather low, the studied time period (1999- 2013) is considered large enough to unequivocally state that the observed average behavior of the overall period is representative of the general annual patterns of the SLA in the Langtang Valley. For that reason, a monthly average analysis of the Landsat extracted data was carried out to determine the average annual SLA evolution and dynamics. The mentioned analysis consists on averaging the extracted SLA of Landsat images corresponding to the same month, during the entire time period 1999-2013. On one hand, the study had to take advantage of the maximum amount of images considered valid for the analysis and available for the period. Thus, all the valid Landsat scenes were added to the analysis. One the other hand, each year had to be equally weighted in the monthly average, independently of the number of images, so that the results are not biased towards the periods with more data available. That is the reason why the SLA values CHAPTER 3. Methodology and data 24 corresponding the same year and month were previously averaged, resulting in a single monthly SLA value for each year and month, subsequently added to the following equation: (MA SLA value)month =P2013 y=1999 [(SLA value)y]month [No. of years represented]month (3.3) As a result, a monthly average (MA) of the SLA value is calculated for each month, allowing an annual evolution analysis. Di↵erent SLA statistics -mean, median, minimum and standard deviationwere extracted from each Landsat image. This analysis was carried out for all of them -the statistics-, applying exactly the same explained methodology to obtain monthly average values of each one of them. Landsat imagery was previously filtered in order to assure a minimum image quality. Although the number of snowline points (SLPs) extracted from a scene is not a direct indicator of image quality, it is directly influenced by the image quality and, thus, it becomes an indirect indicator of it. 3.6.2 Seasonal snowline variability Snowline altitude (SLA) is influenced by a lot of factors, including climatic conditions and events. Although climatic conditions follow a more or less consistent annual evolution, seasonal distribution and relevance of meteorological events is not consistent from one year to another. Therefore, since SLA is a↵ected by them, neither SLA time patterns -at seasonal scaleare consistent from one year to the next. Nevertheless, mean SLA amplitude -with SLA amplitude being understood to be the di↵erence between the maximum and the minimum value reached by the mean SLA of the considered Landsat sceneswithin a season is no longer dependent on the distribution of the meteorological events within the season and it is thus expected to have higher consistency between analog periods from di↵erent years. A seasonal study of the SLA variability requires a great level of detail, which is just not provided by the method previously proposed for the monthly average analysis. As regards to the purpose of characterizing SLA amplitude within a short period of time, there are two fundamental requirements Landsat data has to fulfill. Since the SLA amplitude is to be calculated within a single season of a certain year, the amount of Landsat images selected to calculate it -SLA amplitudein one season of a given year will be rather poor -Landsat imagery has a return period of 16 days and actual availability is even lower-. Therefore, the quality of the imagery is essential for having all the upper Langtang catchment well represented. As discussed in Section 4.1.2, Image Quality Ratio is a good indicator of actual image quality, and the threshold above which good image quality is insured is 0.7. For that reason, the condition IQR 0.7 was used as a first filter for Landsat imagery. After assuring image quality, periods of study are to be determined. It is obvious that not all seasons of each year have enough image availability to be included in the analysis. It is also expected that not the same amount of periods will fulfill a given set of conditions -regarding image availabilityfor each season. And, by trying to standardize the process to an equal number of periods studied for each season, either poorly represented periods would have to be included in CHAPTER 3. Methodology and data 25 the analysis or greatly represented ones would be excluded -reducing in both cases the quality of the analysis-. That is the reason why, for each season, only the periods which fulfill the following conditions were added to the analysis: — “Image per month” ratio 1 — The entire season is represented -that is, images are greatly distributed all over the period, so that each month is represented-. As a result, di↵erent amount of periods were selected for each season, based on the fulfillment of the aforementioned conditions. This way, the results obtained for di↵erent seasons are trustable and might be contrasted, allowing a comparative analysis of the SLA amplitude variability among di↵erent seasons. Finally, the seasons in which the year was split are: — Winter (December 1st to Feburary 28th) — Pre-monsoon (March to June 14th) — Monsoon (June 15th to September 30th) — Post-monsoon (October 1st to November 30th) Monsoon season was not included into the analysis due to the lack of Landsat qualified imagery during that period. CHAPTER 4. RESULTS 33 di↵erences are observed. While it is true that the same reasoning is not valid for Western regions 6 and 7, it is a fact that sub-catchment 6 is almost always at the lower extreme, while sub-catchment 7 shows the transition between sub-catchment 6 and 4. Recalling the sub-catchment distribution, displayed in Figure 2.4, note that sub-catchment 7 is adjacent to both regions 4 and 6 and all three account for the same portion -a 10% approximately- of the Langtang Valley area (Table 2.2). Results concerning sub-catchment 6 are, therefore, to be treated carefully, since it is, specially during summer period, substantially underrepresented. Alhough Image Quality Ratio is not used as a filter criterion in this analysis, the importance of assuring a certain image quality level is discussed in Section 4.1. However, by setting a certain number of snowline determined points as a threshold, average quality of the selected imagery is expected to increase. An IQR around 0.8 is observed in Figure 4.4 for each month and subcatchment except for August and September, when IQR reaches its minimum around 0.6. Figure 4.4 – Average Image Quality Ratio for the Landsat images studied in the monthly average analysis for each month and sub-catchment. Finally, Table 4.2 displays the average number of SLPs corresponding to each aspect range. Only Landsat scenes selected for the current analysis were averaged, so these values provide information about actual consistency of the results found in Section 4.2.3, where aspect quarter behavior is studied. 4.2.2 Main SLA statistics: mean, median, minimum and standard deviation The values analyzed and discussed in this section result from a monthly average of the statistics determined in the studied Landsat images. Therefore, these “monthly averaged statistics” are based on as many values as Landsat scenes selected for the corresponding sub-catchment and month. The way how the average is carried out is widely explained in Section 3.6.1. CHAPTER 4. RESULTS 34 Table 4.2 – Average number of snowline points (SLPs) for each aspect quarter in each subcatchment. Sub-catchment Average number of SLPs East South West North 1&2 1336.72 1325.82 2122.62 1094.19 3473.15 837.36 930.42 1142.81 4567.10 436.28 727.94 535.38 5481.10 478.53 460.49 1255.60 6265.64 736.67 341.29 58.95 7435.56 429.96 566.08 168.78 In order to simplify and clarify the explanation of the results, the terminology “mean” or “average” is never used -from here onwardswhen referring to the “monthly averaged statistics”. Otherwise, the term “Mean” is implemented to denote the statistic itself, from which a monthly average value is also obtained. Mean Monthly average values of the Mean SLA are plotted for each sub-catchment in Figure 4.5. While it is true that, overall, the SLA ranges from roughly 4600 up to 5500 meters, the range of values reached by a certain sub-catchment -no matter which oneis, at least, reduced by half. Moreover, there are sustained di↵erences between SLAs of di↵erent sub-catchments, which indicate that SL variability has to be studied for each sub-catchment separately. Figure 4.5 – Annual evolution of the monthly average Mean SLA for each sub-catchment. CHAPTER 4. RESULTS 35 Table 4.3 – Statistics of the monthly average Mean SLA for each sub-catchment. Sub-catchment 1&2 3 4 5 6 7 Mean 5381.82 5297.86 5115.54 4856.25 5040.81 5149.86 St. Dev. 82.72 103.88 110.87 187.30 152.87 157.74 Some patterns -regarding Mean SLA di↵erences between sub-catchments-, which concern the entire Langtang Valley, might be appreciated in the figure. The fact that these SLA di↵erences are in most cases sustained throughout the year suggests three commentaries. First, the change of SLA behavior between sub-regions is not radical, but progressive across adjacent sub-catchments. Table 4.3 shows how the annual average SLA -mean statistic- and its variability -standard deviationrange from sub-catchments 1&2 -located in the North-Eastern side of the Langtang Valley-, where highest SLA and lowest variability are observed, to subcatchment 5 -South-Western and opposite side of the valley-, where lowest SLA and extremely high variability takes place. Although the fact that adjacent sub-catchments behave similar might seem an obviousness, it is not at all. Adjacent regions might have -and that is the case, for instance, of sub-catchments 5 and 6- very di↵erent aspect quarters compositions. In a regions where climatic conditions are similar, aspect is expected to have a greater e↵ect in SL position. Nevertheless, and although aspect does have an influence on SLA as seen in further steps, this is not explaining the behavior of sub-catchments in the chart. Therefore, from the spatial variation of SLA -absolute values and variabilityin the Langtang Valley, a potential spatial gradient in climatic conditions might be inferred. Second, and picking up the point made in the previous paragraph, two particular gradients might be observed as regards to the SLA spatial variation. On one hand, Eastern sub-catchments -1&2 and 3- have the highest SLAs while Western regions -sub-catchments 5 and 6- have the lowest elevations. Central regions 4 and 7 have SLAs ranging in between both East and West extremes. On the other hand, and within the Weat-East gradient, a North-South gradient is, to a lesser extent, identified. Comparing North-South SLAs within the Eastern -1&2 and 3-, central -7 and 4- and Western -6 and 5- sides of the catchment, Northern sub-catchments -1&2, 7 and 6- show, generally, higher SLAs than their analogous in the Southern side -3, 4 and 5-. In conclusion, a dominant East-West SLA gradient together with a -less important but also visible- North-South SLA gradient explain the sustained snowline elevation di↵erences in the Langtang Valley. Third and last point regards SLA variability, quantified by the standard deviation values in Table 4.3. While sub-catchments 1&2, 3 and 4 have lower standard deviations, sub-areas 5, 6 and 7 exhibit a rather high variability, which suggests a careful observation of their annual evolution (Fig. 4.5). A sudden and extremely pronounced drop in August characterizes all three mentioned sub-catchments. This unexpected behavior becomes specially suspicious for the monsoon period, where image availability is rather poor. Indeed, revisiting Fig. 4.3, only one image is considered the monthly average analysis of sub-catchment 6 in August, which excludes this sub-catchment to be analyzed in August. As regards to the other two sub-catchments, also few images corresponding to CHAPTER 4. RESULTS 36 August were selected, which increases the possibility that one single value is skewing the monthly average towards a lower value. However, Table 4.4 shows how, for both regions, there is a really consistent SLA in most of the images, rejecting the possibility of outliers skewing the average. Table 4.4 – Mean SLA values for the August Landsat scenes selected in sub-catchments 5 and 7. Sub-catchment 5 Sub-catchment 7 Date Mean SLA Date Mean SLA 2002.08.08 5122 2002.08.08 4974 2003.08.27 4745 2003.08.27 4806 2004.08.29 4515 2004.08.29 4905 2006.08.19 5276 2008.08.24 4485 2008.08.24 4910 2011.08.01 4427 2011.08.17 4380 Due to the reduced amount of studied images, it is possible that exceptional climatic conditions took place during the years when those Landsat scenes were collected. For that reason, a second step to contrast the behavior of Western sub-catchments in August is to look at climate data - monthly accumulated precipitation and mean monthly temperature- of those years where Landsat images are selected in sub-catchments 5 and 7. Figures 4.6 and 4.7 compare the climate data from those years to the average -and standard deviationvalues of the period 1999-2010, for which reliable climate data is available. It is important to note that climate data is only available for the entire Langtang Valley, and no data is registered separately for the sub-catchments derived from this study. Precipitation values for the month of August throughout the di↵erent considered years are, indeed, not far away from the average. Only year 2002 has specially high precipitation in August due to late monsoon period. Besides, SLA in August 2002 happens to be higher than the rest of years, which dismisses precipitation as a possible explanation for the observed SLA behavior. The same reasoning applies for the monthly temperature chart, where, again year 2002 is the only one showing a substantially high value in August. However, temperature is specially high for the whole year during 2002, which, again, proofs that temperature is not influencing SLA drop. In conclusion, thus, neither precipitation nor temperature were exceptional for the years where images are selected, not causing the lower peak of August SLA in sub-catchments 5, 6 and 7. CHAPTER 4. RESULTS 37 Figure 4.6 – Monthly precipitation values for selected years (2002, 2003, 2004, 2006 and 2008) contrasted with average and standard deviation values of the period 1999-2010. Figure 4.7 – Monthly temperature values for selected years (2002, 2003, 2004, 2006 and 2008) contrasted with average and standard deviation values of the period 1999-2010. In order to be sure that this phenomenon is not caused by the utilization of the mean as statistic, monthly average median SLA values were plotted instead. Although median SLA is analyzed further on in this report, the same behavior is observed in August -when using the median as statistic- (Figure 4.9). Therefore, the SLA drop in August is not a consequence of using the mean as statistic. A factor potentially causing this lower peak in August could be the fact one of the aspect ranges CHAPTER 4. RESULTS 38 would be clearly di↵erentiated from the others in August -due to any reason-, resulting in a general drop in SLA. However, Figure 4.8 shows that aspect quarters behavior during the studied period is similar. Although the magnitude of the SLA drop is di↵erent in each aspect range, the only face actually diverging from the overall behavior is the North face, where SLA for sub-catchments 5 and 7 does not fall, but rises in August. Anyway, East, West and South orientation ranges have a consistent behavior as far as the discussed drop. Figure 4.8 – Monthly average Mean SLA of aspect quarters in each sub-catchment. Having discarded the aforementioned potential causes, and since mean SLA for each Landsat image is the result of averaging the elevation of all points determining the SL, variability of SLPs elevation might provide a hint of the underlying cause of this behavior. Monthly average SLA Standard deviation -regarding a single imageis exhibited in Figure 4.14. While it is true that sub-catchment 5 has its maximum standard deviation value in August, this is not the case for subcatchment 7 -or any other sub-catchment-, where August is not characterized by high variability. Finally, Figure 4.12, which contains monthly average Minimum SLA values, indicates that Minimum SLA also decreases in August for the Western sub-catchments suggesting, thus, that probably the SL as a whole is su↵ering an elevation range decrease. All in all, it is a fact that this SLA drop existed in the selected images, but no explanation regarding possible causes was found. A last step missing into this direction is the discussion on whether the SLA August lower peak can be extrapolated. For this purpose, Landsat scenes selected for sub-catchments 5 and 7 in August are also examined in the rest of sub-catchments. Table 4.5 shows that mean SLA values CHAPTER 4. RESULTS 39 of the mentioned Landsat scenes for sub-catchments 1, 3 and 4 are continuously slightly below the average SLA value for August. Although most of these values are not far away from the monthly average -less than 1- there is a clear enough lowering tendency to consider those Landsat scenes not completely representative of what happens in the Western sub-catchments of the Langtang Valley during the month of August. Table 4.5 – Mean SLA values of available August Landsat scenes for sub-catchments 1&2, 3, 4, 5 and 7. Annual and August average and standard deviation values of Mean SLA for the mentioned sub-catchments. Date Sub-catchment 1&2 3 4 5 7 2002.08.08 5201 4913 5032 5122 4974 2003.08.27 5124 5138 5181 4745 4806 2004.08.29 5132 5010 4830 4515 4905 2006.08.19 5293 5393 - - 5276 2008.08.24 5219 5019 5070 4485 4910 2011.08.01 - 4769 4938 4427 - 2011.08.17 5475 5091 - 4380 - August Avg. 5371 5258 5167 4654 4974 August 299.10 324.01 273.04 413.22 356.64 August Avg. - 15071 4933 4893 4240 4617 Annual Avg. 5381 5297 5115 4856 5149 Annual 381 410 310 412 379 Annual Avg. - 15000 4887 4805 4444 4770 Eventually, no explanation why the lower SLA peak in August exists for sub-catchments 5, 6 and 7 was found. While it is true that it is not a result of the data characteristics but an actual drop at least in the selected images, these ones are not completely representative of the Langtang Valley SLA behavior, reason why no conclusions can be drawn. Median As already mentioned before in this Section, mean and median have very similar annual dynamics. In fact, it is reasonable that SLA changes a↵ect mean and median the same way. Figure 4.9 exhibits monthly average Median SLA values for the sub-catchments, where these similarities might be observed by comparing it with Figure 4.5. CHAPTER 4. RESULTS 40 Figure 4.9 – Annual evolution of the monthly average Median SLA for each sub-catchment. Table 4.6 – Statistics of the monthly average Median SLA for each sub-catchment. Sub-catchment 1&2 3 4 5 6 7 Mean 5326.94 5252.11 5097.15 4799.38 4995.22 5122.57 St. Dev. 80.23 106.78 98.25 177.56 166.51 157.30 What is not so obvious, though, is the relationship their absolute values have or their sensitivity towards the multiple factors a↵ecting SLA. In order to better visualize the relationship between these two statistics, the di↵erence between mean and median SLA is shown in Figure 4.10. A clear pattern of behavior -consistent across the basin and throughout the yearshows how the mean almost always takes higher values than the median. There is a clear explanation to that behavior and it has to do with the distribution of the snowline points (SLPs) in the elevation histogram of each Landsat image. Normally, SLA histograms -for a single imagehave one single peak, which is closer to the lowest part of the histogram than to the highest SLA. This indicates a higher density of SLPs at lower altitudes than the mean, and it is caused because of the existence of cues in the higher elevation bands, as shown in Figure 4.11. Detailed aspect quarters analysis on this point is carried out in Section 4.3. CHAPTER 4. RESULTS 41 Figure 4.10 – Di↵erence between monthly average Mean SLA and Median SLA for each subcatchment. Figure 4.11 – Snowline points (SLPs) elevation histogram corresponding to the Landsat image from 16/04/2013 in sub-catchments 1&2. Minimum Based on the typical shape of SLPs elevation histogram, the presence of outliers in high elevation bands seems quite evident. However, the lowest part of the SL is characterized by a dense presence of SLPs with a consistent altitude, reason why the study of the minimum SLA becomes of interest for this study. In Figure 4.12, some of those characteristics and similarities with the mean SLA CHAPTER 4. RESULTS 42 dynamics in August -and the entire monsoon periodhave already been mentioned. A sustained minimum SLA di↵erence between sub-catchments is present, although they are not as clear as the ones from mean SLA. On the other hand, from January until May minimum SLA increases while mean normally decreases; and from October until February, minimum decreases while mean SLA increases. Figure 4.13 helps in noticing such diverging behaviors, representing the di↵erence between mean and minimum SLA. Finally, annual average minimum SLA is exhibited in Table 4.7. Sub-catchments 1&2, in the North-Eastern side of the valley, take an average minimum SLA 500 meters higher than subcatchment 5, located in the South-Western part. Almost the same di↵erence, in absolute values, is observed for the mean SLA (Table 4.3). Mean and minimum seem to have, then the same spatial variance throughout the Langtang Valley. Figure 4.12 – Annual evolution of the monthly average Minimum SLA for each sub-catchment. Table 4.7 – Statistics of the monthly average Minimum SLA for each sub-catchment. Sub-catchment 1&2 3 4 5 6 7 Mean 4550.30 4392.22 4367.60 4040.84 4200.99 4304.40 St. Dev. 64.94 132.43 207.76 256.11 136.36 202.87 CHAPTER 4. RESULTS 49 tude of all considered periods -SLA amplitude being understood to be the di↵erence between the maximum and the minimum value reached by the mean snowline altitude (SLA) of the considered Landsat scenes in a certain period-. The methodology used to perform this analysis is described in detail in Section 3.6.2. Note that monsoon season is not studied due to lack of Landsat scenes. 4.3.1 Pre-monsoon season Pre-monsoon season lasts from March 1st until June 14th, when monsoon is starting. Premonsoon periods considered valid for the seasonal SLA variability analysis are listed in Table 4.13. Six periods -corresponding to di↵erent yearswere selected, all of them with a similar number of scenes resulting in ”images per month” ratios around 1.5. Table 4.13 – Periods considered for the SLA variability analysis during pre-monsoon season. Period Year Number of images ”Images per month” ratio 12005 5 1.43 22008 5 1.43 32009 6 1.71 42010 6 1.71 52012 6 1.71 62013 5 1.43 Figure 4.20 shows the mean SLA amplitude of the aspect ranges for each sub-catchment - the average for the entire sub-catchment is also plotted-. It corresponds to the mentioned subcatchment approach. The errors bars represent two times the standard error of the values corresponding to di↵erent years and averaged to obtain the plotted mean value. Most values in the figure have a rather low standard error. This is partially due to the fact that six years met the criteria and were added to the analysis -the other periods are less well representedbut it is also consequence of the climatic conditions generally a↵ecting this season, when the weather is more stable than, for instance, in winter period. Having noted that, lower amplitudes are observed towards the East side of the Langtang River Basin. East sub-catchments (1&2 and 3) have less SLA amplitude -average values around 400- 450 metersthan West ones (5 and 6) -with values ranging from 8000 until 1200 meters-. Subcatchments 4 and 7 -located in the center part of the valleyevidence the transition of behavior between both sides. There is also a North-South gradient within the East-West gradient -regarding SLA amplitude-, in which Northern sub-catchments have substantially higher amplitudes than the Southern ones. Although SLA amplitude in sub-catchment 3 (South-Eastern) is only slightly lower than in 1&2 (North-Eastern), the di↵erence is evident when comparing sub-catchments 4 (South-central) and 5 (South-Western) to 7 (North-central) and 6 (North-Western) respectively. The fact that the CHAPTER 4. RESULTS 50 Langtang Valley reaches higher latitudes in its Eastern part (see Figure 2.5) might partially explain the lack of di↵erence between sub-catchments 1&2 and 3. Finally, standard deviation of the SLA amplitude is rather low (Figure 4.20), due to the fact that weather during pre-monsoon season is more or less consistent during di↵erent years (see Figures B.1,B.2 and Tables B.2,B.2 in Appendix B). Figure 4.20 – Mean SLA amplitudes for each sub-catchment and aspect quarter in Pre-monsoon season. Error bars correspond to 2 times the standard error of each SLA amplitude value. The second approach to look at pre-monsoon SLA amplitude throughout the Langtang Valley is grouping the data into aspect quarters. This configuration is displayed in Figure 4.20,where complementary observations are to be pointed out. On one hand, East, South and West ranges have comparable spatial evolutions, inducing the entire catchment to have -on averagealso a similar behavior. However, the magnitude of East range SLA amplitude values is consistently higher than in West face. To a lesser extent, there is also a contrast between South and North aspect ranges, being the South face normally less stable -higher amplitudesall over the pre-monsoon season. On the other hand, North face is behaving divergently to the others. Having excluded subcatchment 6 for being underrepresented (see Table 4.14), the spatial di↵erences in North aspect quarter are minimal. CHAPTER 4. RESULTS 51 Table 4.14 – Average number of snowline points (SLPs) from Landsat imagery added to the premonsoon seasonal analysis. Sub-catchment East South West North 1&2 1018.17 1200.88 1983.36 770.17 3372.90 723.34 924.41 1126.45 4464.01 378.87 624.51 472.55 5444.68 452.34 414.84 1331.76 6143.06 482.16 239.37 70.06 7387.39 354.42 429.98 225.84 Figure 4.21 – Mean SLA amplitudes for aspect quarter and sub-catchment in Pre-monsoon season. Error bars correspond to 2 times the standard error of each SLA amplitude value. 4.3.2 Post-monsoon season Five periods happened to be valid for post-monsoon -the shortest season-, which lasts for two months (October-November). The “images per month” ratio ranges from 1.5 to 2.5, as displayed in Table 4.15. Table 4.15 – Periods considered for the SLA variability analysis during post-monsoon season. Period Year Number of images ”Images per month” ratio 12006 5 2.5 22008 3 1.5 32009 4 2 42012 3 1.5 52013 5 2.5 CHAPTER 4. RESULTS 52 Looking at Table 4.16, North face of sub-catchment 6 has to be -similarly as for the previous seasonexcluded from the analysis for not being adequately represented -the average number of snowline points is lower than 20-. Table 4.16 – Average number of snowline points (SLPs) from Landsat imagery added to the postmonsoon seasonal analysis. Sub-catchment East South West North 1&2 1787.16 1374.84 2252.27 1470.19 3677.27 939.62 991.07 1265.44 4708.90 481.92 761.84 514.50 5500.22 281.17 363.01 1150.97 6307.21 654.51 268.33 18.02 7447.09 336.76 636.99 115.55 Figure 4.22 evidences the fact that post-monsoon season is characterized by low SLA amplitudes, ranging from slightly less than 200 up to 400 meters -on average for the entire subcatchments-. The East-West gradient is maintained, being the SLA amplitudes in the Western side (subcatchments 5 and 6) around 150 meters higher than in the East (sub-catchments 1&2 and 3). A remarkable fact in this figure is the behavior of sub-catchment 4, exhibiting the lowest SLA amplitude values. Comparing its behavior with its adjacent region in the North (sub-catchment 7), the hypothesis of a North-South gradient -within the same longitude range; that is East, Center or West side of the basinis reinforced. The same consideration pointed out in the previous Section 4.3.1 regarding the East side of the basin applies in this case. CHAPTER 4. RESULTS 53 Figure 4.22 – Mean SLA amplitudes for aspect quarter and sub-catchment in Post-monsoon season. Error bars correspond to 2 times the standard error of each SLA amplitude value. Two observations are to be made about Figure 4.23 -where SLA amplitudes are grouped into aspect quarters-. First, East and West faces behavior emphasizes the fact that there exists a North-South amplitude di↵erence. Second, North face is, again, showing less SLA amplitude than the other aspect ranges -especially when comparing it to its opposite range, South face-. Figure 4.23 – Mean SLA amplitudes for aspect quarter and sub-catchment in Post-monsoon season. Error bars correspond to 2 times the standard error of each SLA amplitude value. CHAPTER 4. RESULTS 54 4.3.3 Winter season As far as winter season, only four periods could be included in the study. The relative number of Landsat images is, however, higher than for the other seasons, reaching ratios higher than 2 scenes per month on average (Table 4.17). Table 4.17 – Periods considered for the SLA variability analysis during post-monsoon season. Period Year Number of images ”Images per month” ratio 12004-2005 6 2 22008-2009 8 2.67 32009-2010 10 3.33 42012-2013 5 1.67 The first consequence of having less periods to average is a higher degree of irregularity with respect to the standard error. While it is true that this statistic is already fluctuating all over the basin -depending on the actual magnitude of the SLA amplitude valuesduring pre-monsoon and post-monsoon seasons, its behavior in winter is even more unsteady. Observing Figure 4.24, sub-catchment 1&2 exhibits extremely low standard error values whereas remarkably large values involve sub-catchment 5 -on the opposite side of the basin with respect to sub-catchment 1&2-. In terms of absolute value, SLA amplitudes in winter are higher than during post-monsoon season and minor than the ones found for pre-monsoon season. On the other hand, regarding relative spatial variation of the SLA amplitude -relative spatial variation being understood to be the SLA amplitude di↵erence all over the basin with respect to the absolute value-, winter season shows also higher spatial variation than post-monsoon season but very similar to pre-monsoon period. The spatial variation analysis is detailed in the following section 4.3.4. One more time, the spatial gradient East-West -with higher SLA amplitudes in the Western part (sub-catchments 5 and 6) whereas lower values towards the East (sub-catchments 1&2 and 3)- seems quite obvious, and the North-South gradient for same longitude ranges is, again, less evident but still visible in the central part of the valley (sub-catchment 4 in the South compared to 7 in the North). The poor number of considered periods might be a partial explanation of this less conclusive results. However, climatic inconsistency along years (see Appendix B) is, for sure, causing results to be more confusing. CHAPTER 4. RESULTS 55 Figure 4.24 – Mean SLA amplitudes for aspect quarter and sub-catchment in winter season. Error bars correspond to 2 times the standard error of each SLA amplitude value. As in the other seasons, North face in sub-catchment 6 is underrepresented and, therefore, it is not object of analysis. Table 4.18 – Average number of snowline points (SLPs) from Landsat imagery added to the winter seasonal analysis. Sub-catchment East South West North 1&2 1418.97 1327.39 2272.18 1299.60 3526.37 797.57 1018.96 1065.90 4581.16 473.81 690.15 510.60 5530.10 560.75 460.93 1277.89 6285.13 637.41 407.60 56.11 7364.69 373.07 700.58 103.88 Figure 4.25 exhibits the behavior of di↵erent aspect ranges regarding the Mean SLA amplitude. CHAPTER 4. RESULTS 56 Figure 4.25 – Mean SLA amplitudes for aspect quarter and sub-catchment in winter season. Error bars correspond to 2 times the standard error of each SLA amplitude value. 4.3.4 Comparative analysis of the studied seasons After having analyzed each season separately, the figures and tables contained this section aim to provide some evidences of the di↵erentiated SLA amplitude behavior between seasons, as well as the spatial variation between them. A clear di↵erence SLA amplitude di↵erences Figures 4.26, 4.27 and 4.28 show the di↵erences between SLA amplitudes corresponding to di↵erent seasons. Looking at the average values for entire sub-catchments in Figure 4.26, SLA amplitude di↵erences between winter and pre-monsoon are rather stable across the Eastern (sub-catchments 1&2 and 3) and central (sub-cachments 4 and 7) parts of the valley, being the SLA amplitudes between 100 and 200 meters higher in pre-monsoon season, whereas the behavior of Western sub-catchments (5 and 6) is really contrasting. However, sub-catchment 5 is characterized by having normally the lowest di↵erence in SLA amplitude between both seasons while, on the other hand, sub-catchment 6 exhibits rather high distinction between seasons. In general, though, there is a low level of spatial consistency throughout the valley -with regard to the di↵erences in SLA altitude between the mentioned seasonsacross aspect quarters. CHAPTER 4. RESULTS 57 Figure 4.26 – Di↵erences between Pre-monsoon and Winter SLA amplitudes. Amplitude = Pre-monsoon SLA Amplitude Winter SLA Amplitude. Winter SLA amplitude is continuously higher than post-monsoonal. A sustained di↵erence for in each sub-catchment -ranging, in most cases, from 100 to 250 metersmight be observed in Figure 4.27. Regarding spatial stability in di↵erent aspect ranges, it seems clear that largest values are normally held by the North-Eastern part of the basin (sub-catchment 1&2), whereas minimal values are usually recorded in the South-West (sub-catchment 5). Therefore, the aspect quarters SLA amplitude di↵erences are, to some extent, correlated. CHAPTER 4. RESULTS 58 Figure 4.27 – Di↵erences between Winter and Post-monsoon SLA amplitudes. Amplitude = Winter SLA Amplitude Post-monsoon SLA Amplitude. Finally, pre-monsoon and post-monsoon SLA amplitude di↵erences are plotted in Figure 4.28. Note that these are not adjacent seasons and the di↵erences are, thus, expected to be potentially larger in this case. Moreover, monsoon period -taking place in betweenis characterized for being an unexplored season with a lot of unknowns, concerning climatic conditions and snow cover dynamics. The values observed in the figure are, indeed, greater than for the cases previously analyzed, taking values -on average and for the entire sub-catchments- from 250 up to 400 meters. The largest values are recorded in sub-catchment 6 -partially due to the pronounced orography (Table 2.3) but also due to the lack of data in this sub-catchment during both seasons (Tables 4.14 and 4.16)- whereas there is not a clear region with sustained lower values for each orientation quarter. Spatial consistency of SLA amplitude di↵erences is hence not clear in this case. CHAPTER 4. RESULTS 65 Figure 4.19 – Di↵erence between aspect quarters Mean and Median SLA for each sub-catchment. Chapter 5 Discussion As formerly discussed in Chapter 1, the methodology followed in this study to extract the SL at the point scale from remotely sensed Landsat images does not have clear precedent in a previous study, but it is a slightly new approach -with evident affinities to similar methods- to the study of SL dynamics. In any case, none of the studies which were found to have analogies with the one proposed in this thesis correspond to the Langtang Valley. After using Landsat imagery for the current study, usefulness of Landsat when mapping and monitoring the SL is briefly discussed in this chapter. Due to diverse reasons -lack of resources and low accessibility of the region among others-, there is not a continuous and enough exhaustive register -of any kind- of the snowline altitude (SLA) annual evolution in the Langtang Valley. Despite that fact, there exist few studies that include SL measurements in certain sectors of the valley. Leaving aside single and isolated observations that doesn’t have a continuity over time, a daily register of the SLA based on visual observations was recorded in the South-western sector of the upper Langtang Valley from July 1985 to June 1986 [Morinaga et al. 1987]. Although the very low scope of the observations, partial validation of the methodology is achieved by comparing the results obtained in this study with the mentioned register. After validating -as far as possiblethe methodology, a discussion on the generalized and main topics arisen in the Results chapter (Chapter 4) is carried out. Recalling and briefly summarizing the principal conclusions achieved, three di↵erentiated points are found. These are: — the spatial variability of the SLA within the Langtang Valley; — the e↵ect of aspect as an influencing parameter to the SLA; — and the inter-seasonal evolution of the SLA amplitude variability. Understanding this chapter as a continuation of the previous one, the main di↵erence between them lies on the fact that observations stated in Chapter 4 are taken one step further, trying to provide feasible explanations to those behavioral patterns shown to be persistent. 67 CHAPTER 5. DISCUSSION 68 The last purpose of this chapter is not to provide a single or definitive argument that unequivocally explains the observed behaviors, but to shed light into a topic that has been barely studied -due to its technical complexityuntil recent times. Consequently, di↵erent possible explanations regarding the patterns previously observed are exposed and discussed in this chapter. Eventually, potential uses -alternative or complementary to this study- of the SL extracted data sets are proposed. 5.1 Usefulness of Landsat imagery for the study of the snowline dynamics As previously mentioned in Chapter 3, Landsat imagery has a high spatial resolution -50 mbut a low temporal resolution -the return period is equal or higher than 16 days-. Depending on the season, image quality and therefore also image availability might drastically reduce due to cloud cover. The study of the SL requires high spatial resolution, and that is the reason why Landsat imagery becomes a powerful tool when mapping the SL. On the other hand, the SL is greatly a↵ected by cloud cover -much more than the snow cover fraction (SCF)- and considering low quality images -with big portion of cloud coveringmay lead to interpretation errors. Therefore, high quality of the scenes is also required, reducing significantly the amount of images valid for this type of study. With a rather modest number of scenes, there are basically two strategies to follow. The first methodology is the one used in this study, consisting on a statistical analysis of the output obtained from the Landsat imagery. By doing this, a high-resolution temporal study of the SL is not possible, but a seasonal or inter-annual trend analysis might be achieved. Secondly, there is another way of using the Landsat data, and that way is combining it with other sources of data -such as other satellites imagery, meteorological data and SL or SCF visual observations-. Landsat images provide an good source of reliability -due to the high level of detail they achieve-, and it could be used to validate results obtained with another source of data. For instance, validating daily MODIS scenes -which could derive the SCF and indirectly the SL- with Landsat data is an approach to be considered. In conclusion, Landsat data is at the moment an indispensable resource to be used when studying the SL and its dynamics. However, it often requires other sources of data in order to complement the analysis. 5.2 Partial validation of the methodology Morinaga et al. [1987] studied the seasonal variation of SLA in the Langtang Valley for the whole year, from July 1985 to June 1986, based upon the photograph observations. Pictures from CHAPTER 5. DISCUSSION 69 the surrounding landscape were taken from the Base House (BH), 3920 m. Two reference slopes (North and South facing) were selected for the study. Figure 5.1 shows a map of the zone, with the BH and the slopes. Figure 5.1 – Location of the observation site (Base House). 1: North-facing slope. 2: South-facing slope [Morinaga et al. 1987]. The location of the BH corresponds to the study region sub-catchment 5 (South-West side of the valley). Sub-catchment 5 might be divided into two main sides. Northern side of the subcatchment is almost entirely South-facing whereas Southern side of the sub-catchment is dominated by North-facing slopes (Figure 5.2). It is thus reasonable to compare South and North facing SLAs extracted from Landsat imagery to the SLA in the slopes observed in the study. Monthly average annual evolution of mean, minimum and standard deviation statistics of North and South SLAs in sub-catchment 5 are compared to the SLAs reported by Morinaga et al. [1987]. Figure 5.4 shows the annual evolution of the SLAs in North and South-facing slopes observed by Morinaga et al. [1987] and the daily precipitation and temperature. When SLA is lower than the BH at 3920 m, circles are plotted in the line of the chart. To the contrary, when they are higher than the peaks (4957 m for North-facing slope; 4986 m for South-facing slope), circles are plotted in the upper line of the figure. Significant information gaps are observed in August and February. On the other hand, monthly average annual evolution of North and South-facing mean SLAs are plotted in Figure 5.3. The figure also contains the minimum SLAs, since visual observations from a lower point (BH) might be mainly showing the behavior of the lower part of the SL. Detailed CHAPTER 5. DISCUSSION 70 Figure 5.2 – Aspect distribution in sub-catchment 5. figures including annual evolution of standard deviation in both faces are shown in the Appendix A (Figures A.1 and A.2). Four main periods -with great correspondence to the four-season division to the seasons defined by Immerzeel et al. [2014], Shiraiwa et al. [1992]- are identified in Figure 5.4 and, to a lesser extent, in Figure 5.3. Monsoon season (June to August) observations are higher than the peaks in both slopes and thus plotted on the upper line. To the contrary, SLA was observed at lower altitudes than the BH (3920 m) during winter season (December to February). During these periods, North-South SLA di↵erence cannot be observed (Figure 5.4). The same periods in Figure 5.3 are shown to have slightly higher SLAs in North-facing than in South-facing slopes. This might be attributed to lack of data during monsoon period and important snowfalls during winter. For this reason, monsoon and winter season are not really helpful for validating the Landsat extracted data. Morinaga et al. [1987] observations for late monsoon (September) and post-monsoon season (October-November) show a remarkably consistent di↵erence between South and North SLAs. The same applies for the pre-monsoon season (March to May), excluding March, where SLA in the South-facing slope is below the BH probably due to observed frequent precipitation events until the beginning of April. While the SLAs recorded for both slopes during post-monsoon season -where no precipitation events are listedare very similar to the mean SLAs in Figure 5.3, pre-monsoon season shows less SLA in Figure 5.4 than in mean SLAs in Figure 5.3, being closer to the minimum SLAs. In both cases, though, the dynamics followed by SLA coincide. Regardless absolute values, the di↵erence between North and South-facing SLAs obtained during the year 1985-1986 and the analog mean values for the period 1999-2013 (Figure 5.5) are CHAPTER 5. DISCUSSION 71 comparable for pre-monsoon and post-monsoon periods, when precipitation is dominating over all other factors. On the other hand, winter and monsoon periods are more a↵ected by precipitation events, which hide the SLA di↵erences. Besides, lack of data in monsoon and SLA values lower than BH in winter make those periods less suitable for the comparison. As far as pre-monsoon and post-monsoon, the magnitude of North-South SLA di↵erences is similar for both data-sets (compare Figures 5.4 and 5.5). Concluding, results have been partially validated with an incomplete and really localized but reliable source of data, which proves the method used to extract the SL from Landsat scenes is valid and leads to consistent results. Figure 5.3 – Monthly average annual evolution of North and South-facing SLAs in sub-catchment 5. Mean and minimum statistics represented. 5.3 Spatial variability of the SLA in the upper Langtang Valley The upper Langtang Valley is known to be a rather small region a↵ected by climate spatial and altitude variability [Pellicciotti et al. 2014]. Previous studies have recorded and concluded the existence of precipitation and temperature gradients throughout the catchment [Shiraiwa et al. 1992, Immerzeel et al. 2014]. Since snowline altitude (SLA) is greatly influenced by precipitation and air temperature, spatial variability of these factors is also expected to have a significant e↵ect on SLA [Morinaga et al. 1987]. While much has been concluded about precipitation and temperature dependence on altitude [Shiraiwa et al. 1992, Immerzeel et al. 2014], significant knowledge gaps CHAPTER 5. DISCUSSION 72 Figure 5.4 – SLA observations from July 1985 to June 1986 from the Base House (3920 m) in the Langtang Valley [Morinaga, 1992]. are still to be revealed of the spatial variation of climate and its e↵ect on the snowline. This study has focused on exploring spatial -latitude and longitudevariability of the SLA. Two di↵erent methodologies were used to study the SLA dynamics. Monthly average values were computed to look at annual evolution of the snowline in di↵erent sub-catchments, whereas seasonal SLA variability was analyzed by selecting only those years with more image availability for a certain season. Although both methodologies were meant to study di↵erent phenomena, results from both analysis reveal the existence of strong spatial variability of the SLA within the upper Langtang River Basin. Two patterns are observed in the results. According to Shiraiwa et al. [1992], North-eastern side of the catchment received less precipitation than downstream in the South-western part of the catchment. Not only this has been observed, but two di↵erentiated and complementary SLA spatial gradients suggest the confirmation of that hypothesis. First, longitudinal changes in the catchment are regarded. As far as the sub-catchments position in the basin, three longitudinal ranges can be identified. Sub-catchments 1&2 and 3 are located in the East side of the Valley; sub-catchments 4 and 7 in the central part; and the West side is occupied by sub-catchment 5 and 6. Results in Chapter 4 show persistent di↵erences in SLA between sub-catchments that last for the entire year. In this figures, SLA in the Eastern side is always higher than SLA in the Central side. At the same time, SLA in the Central part of the catchment is generally higher than in the Western side. It is thus evident the existence of a West-East positive spatial gradient on the SLA. Given the correlation between the ELA and the SLA -SLA is often used as a proxy for ELA-, these results coincide with previous observations by CHAPTER 5. DISCUSSION 73 Figure 5.5 – Monthly average annual evolution of North-South SLA di↵erence in sub-catchment 5. Benxing et al. [1984], where ELA is reported to decrease from North-East to the Western part of the valley. Secondly, the latitudinal changes of the SLA were analyzed. Retaking the longitudinal categorization of the sub-catchments, each longitude range contains two sub-catchments, being always one clearly above the other. This way, sub-catchments 1&2 (East), 7 (Central), and 6 (West) are placed in the Northern side of the valley, whereas sub-catchments 3 (East), 4 (Central) and 5 (West) are located below in the Southern side. Figures from Chapter 3 (previously listed) show the sustained SLA di↵erences between sub-catchments in the same longitudinal range (East, Central, West), evidencing the existence of a South-North positive SLA, being this vertical gradient subjected to the East-West gradient stated before. In conclusion, two SLA spatial gradients are clearly reported in this study. Previous studies had noted the existence of precipitation and SLA gradients from North-East to western parts of the upper Langtang Catchment, but the description of the spatial variability was coarse. Remotely sensed mapping of the SL allows this study to shed light into the topic. Although there are no explicit evidences proving it, given the big influence precipitation has on SLA, precipitation positive gradients from North to South and East to West are hypothesized. The sub-catchment based division -being each sub-catchment enclosed by mountainous topographycomplemented with the extreme topography of the upper Langtang Valley support the hypothesis of orography acting as a natural barrier for stratus and moisture, as reported in Shiraiwa et al. [1992]. CHAPTER 5. DISCUSSION 74 5.4 E↵ect of aspect in the SLA Aspect influence into snow cover (SC) dynamics has been stated by several studies [Jain et al. 2008, Rowland and Moore 1992, Murray and Buttle 2003, Jost et al. 2007, Hendrick et al. 1971, Berndt 1965, Golding and Swanson 1986, D’Eon 2004, Varhola et al. 2010, among others]. The e↵ect of aspect on snow accumulation and ablation is principally a function of exposure to solar radiation. In the northern hemisphere, more snow is expected to accumulate on North-facing slopes, due to reduced melting and sublimation rates during the accumulation period [Golding and Swanson 1986]. Unlike elevation, aspect is shown to have a similar e↵ect from year to year [Jost et al. 2007]. Due to the direct e↵ect of snow cover fraction (SCF) into the SLA, aspect is thus expected to have a substantial influence on SLA. The results from this study confirm the e↵ect of aspect into SLA and two main patterns are observed. In first place, West-facing SLA is observed to be sustainedly higher than East-facing SLA. This West-East (W-E) SLA di↵erence is generalized throughout the entire catchment and it ranges principally from 0 to slightly more than 200 m, excluding outliers. The maximum di↵erences are recorded in winter (up to 200 m) whereas these generally decrease afterwards, reaching its minimum during monsoon (down to 100 m). The SL retreat during pre-monsoon might explain the drop in W-E SLA di↵erence. As far as the spatial di↵erences, the East side of the basin shows more higher and more consistent W-E SLA di↵erences. Since the Eastern part of the catchment is less exposed to precipitation events, this might be an indicator of important and prevalent e↵ect precipitation has into SLA. Future research should look into the mentioned West-East SLA di↵erence in greater detail and the factors influencing it. The second pattern regards the di↵erence between North and South-facing SLAs. While South SLA is generally higher than Northern in the Western side of the valley, alternated dominance is observed in the Eastern sub-catchments, being North-facing SLA slightly higher during winter and monsoon seasons. However, South SLA turns to be substantially higher than North SLA during pre-monsoon and post-monsoon seasons, where climatic conditions are not characterized by precipitation events. Regardless the di↵erence in the range of values, the same applies for Western sub-catchment, where the South SLA is higher than North SLA during the entire year but specially during pre-monsoon and post-monsoon seasons. Lower N-S SLA di↵erences -or even North SLA being higher than South SLA- might be attributed to the e↵ect of precipitation, prevailing over any other condition during monsoon and winter. It can be suggested from both patterns that slopes with di↵erent orientation might be subjected to slightly di↵erent climatic conditions. These results are thus in the same direction of those previously obtained by Shiraiwa et al. [1992], who concluded that South and North-facing slopes in the Western reaches of the catchment could be exposed to distinct moisture conditions due to CHAPTER 6. CONCLUSIONS 81 Outlook and recommendations The relationship between climatic conditions and the snowline -altitude- has been clearly proven in this thesis. In particular, a double statement can be inferred: the SL is greatly a↵ected by climate -being precipitation events the dominant factor- and thus, at the same time, the SL becomes a great indicator of climatic conditions. Future studies should look into this relationship and, eventually, try to quantify the e↵ect of main factors influencing SL dynamics. In line with the previous point, wind and avalanching are expected to have a great influence on SL dynamics. Therefore, these phenomena should be taken into consideration for any furhter study looking at SLA changes in the short term. Combined use of Landsat selected scenes and climate data -regarding wind and avalanchesfor a short-period of time might be a good approach. Results before presented strongly suggest a great climate variability within the Langtang Valley. Working in this line, further research is encouraged on the analysis of cloud cover in di↵erent subregions of the Langtang Valley, since it is a direct indicator of precipitation. Cloud cover data is easily extracted from the Landsat imagery, and results from such a study are expected to be promising. In great measure, this work was intended to assess the reliability and suitability of Landsat data for the study of the SL. Mapping of the SL at the daily scale could potentially be achieved complementary using Landsat TM, ETM+ and MODIS imagery. Landsat scenes should be used as a validation method for daily results obtained with MODIS product. Finally, further research could be devoted to the e↵ect of aspect in the SLA, particularly during periods of low precipitation. Interactions between aspect and climate could be relevant for a better understanding of climate dynamics within the Langtang Valley, which has turned to be one of the main findings of this work. Bibliography Barnett, T. P., Adam, J. C., and Lettenmaier, D. P. (2005). Potential impacts of a warming climate on water availability in snow-dominated regions. Nature, 438(7066):303–9. Benxing, Z., Watanabe, O., and D. Mulmi, D. (1984). Glacier Features and Their Variations in the Langtang Himal Region of Nepal. Bulletin of Glaciological Research, 2:121–127. Berndt, H. W. (1965). Snow Accumulation and Disappearance in Lodgepole Pine Clearcut Blocks in Wyoming. Journal of Forestry, 63(2):88–91. Corripio, J. G. (2003). Vectorial algebra algorithms for calculating terrain parameters from DEMs and solar radiation modelling in mountainous terrain. Cruz, R. V., Harasawa, H., Lal, M., and Wu, S. (2007). Climate change 2007: impacts, adaptation and vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the IPCC. Cambridge University Press, Cambridge and New York. D’Eon, R. G. (2004). Snow depth as a function of canopy cover and other site attributes in a forested ungulate winter range in southeast British Columbia. Journal of Ecosystems and Management, 3(2):1–9. Golding, D. L. and Swanson, R. H. (1986). Snow distribution patterns in clearings and adjacent forest. Water Resources Research, 22(13):1931–1940. Hall, D. K., Riggs, G. A., Salomonson, V. V., DiGirolamo, N. E., and Bayr, K. J. (2002). MODIS snow-cover products. Remote Sensing of Environment, 83:181–194. Hendrick, R. L., Filgate, B. D., and Adams, W. M. (1971). Application of Environmental Analysis to Watershed Snowmelt. Journal of Applied Meteorology, 10(3):418–429. Immerzeel, W. W. and Bierkens, M. F. P. (2012). Asia’s water balance. Nature Geoscience, 5:841–842. 83 BIBLIOGRAPHY 84 Immerzeel, W. W., Droogers, P., de Jong, S. M., and Bierkens, M. F. P. (2009). Large-scale monitoring of snow cover and runo↵simulation in Himalayan river basins using remote sensing. Remote Sensing of Environment, 113:40–49. Immerzeel, W. W., Petersen, L., Ragettli, S., and Pellicciotti, F. (2014). The importance of observed gradients of air temperautre and precipitation for modeling runo↵from a glacierized watershed in the Nepalese Himalayas. Water Resources Research, 50(3):2212–2226. Immerzeel, W. W., van Beek, L. P. H., and Bierkens, M. F. P. (2010). Climate change will a↵ect the Asian water towers. Science (New York, N.Y.), 328:1382–1385. Iqbal, M. (1983). An introduction to solar radiation. Ives, J. D. and Messerli, B. (1989). The Himalayan Dilemma: Reconciling Development and Conservation. Jain, S. K., Goswami, A., and Saraf, a. K. (2008). Role of Elevation and Aspect in Snow Distribution in Western Himalaya. Water Resources Management, 23(1):71–83. Jost, G., Weiler, M., Gluns, D. R., and Alila, Y. (2007). The influence of forest and topography on snow accumulation and melt at the watershed-scale. Journal of Hydrology, 347(1-2):101–115. Kotchenova, S. Y. and Vermote, E. F. (2007). Validation of a vector version of the 6S radiative transfer code for atmospheric correction of satellite data. Part II. Homogeneous Lambertian and anisotropic surfaces. Applied optics, 46(20):4455–64. Kuhn, M. (1984). Mass budget imbalances as criterion for a climatic classification of glaciers. Geografiska Annaler. Series A. Physical Geography, 66:229–238. Miles, E. S., Willis, I., and Pellicciotti, F. (2013). Monsoonal Variations of Supraglacial Lakes , Langtang Khola , Nepal. Morinaga, Y., Seko, K., and Takahashi, S. (1987). Seasonal variation of snowline in Langtang Valley, Nepal Himalayas, 1985-1986. Bulletin of Glaciological Research, (5):49–53. Murray, C. and Buttle, J. (2003). Impacts of clearcut harvesting on snow accumulation and melt in a northern hardwood forest. Journal of Hydrology, 271(1-4):197–212. Nogami, M. (1970). The Snowline: its Definition and Determination. The Quaternary Research, 9(1):7–16. BIBLIOGRAPHY 85 Ohmura, A., Kasser, P., and Funk, M. (1992). Climate at the equilibrium line of glaciers. Journal of Glaciology, 38:397–411. Oxford University Press (2014). A Dictionary of Geology and Earth Sciences. Pellicciotti, F., Stephan, C., Miles, E. S., Immerzeel, W. W., and Bolch, T. (2014). Mass balance changes of the debris-covered glaciers in the Langtang Himal in Nepal between 1974 and 1999. 00(000):1–14. Ragettli, S., Pellicciotti, F., Immerzeel, W. W., and Miles, E. S. (2014). Unraveling the hydrology of a Himalayan watershed through integration of high resolution in-situ data and remote sensing with an advanced simulation model. under review in Advances in Water Resources. Rosenthal, W. and Dozier, J. (1996). Automated mapping of montane snow cover at subpixel resolution from the Landsat Thematic Mapper. Water Resources Research, 32:115–130. Rowland, J. and Moore, R. (1992). Modelling solar irradiance on sloping surfaces under leafless deciduous forests. Agricultural and Forest Meteorology, 60(1-2):111–132. Salomonson, V. V. and Appel, I. (2004). Estimating fractional snow cover from MODIS using the normalized di↵erence snow index. Remote Sensing of Environment, 89:351–360. Seko, K. (1987). Seasonal variation of altitudinal dependence of precipitation in Langtang Valley Nepal Himalayas. Bulletin of Glaciological Researc, 5:41–47. Seko, K. and Takahashjl, S. (1991). Characteristics of Winter Precipitation and its E↵ect on Glaciers in the Nepal Himalaya. Bulletin of Glaciological Research, 9:9–16. Seltzer, G. O. (1994). Climatic Interpretation of Alpine Snowline Variations on Millennial Time Scales. Quaternary Research, 41:154–159. Shi, Y., Zheng, B., and Li, S. (1992). Last glaciation and maximum glaciation in the Qinghai- Xizang (Tibet) Plateau: A controversy to M. Kuhle’s ice sheet hypothesis. Chinese Geographical Science, 2:293–311. Shiraiwa, A., Ueno, K., and Yamada, T. (1992). Distribution of mass input on glaciers in the Langtang Valley, Nepal Himalayas. Bulletin of Glaciological Research, 10:21–30. Shiraiwa, A. and Yamada, T. (1992). Glacier Inventory of the Langtang Valley , Nepal Himalayas. Technical report. Shiraiwa, T. and Watanabe, T. (1991). Late Quaternary glacial fluctuations in the Langtang valley, Nepal Himalaya, reconstructed by relative dating methods. Arctic and Alpine Research, 23:404–416. Shrestha, A. B. and Aryal, R. (2010). Climate change in Nepal and its impact on Himalayan glaciers. Regional Environmental Change, 11(S1):65–77. Singh, I. L. (1985). Rainfall distribution. In Majupuria TC (ed) Nepal: nature’s paradise: insight into diverse facets of topography, flora and ecology. White Lotus Co. Ltd, Bangkok, pages 56–58. United Stated Geological Survey (2014). Landsat Enhanced Thematic Mapper. Uppala, S., Kallberg, P., Simmons, A., Andrae, U., Bechtold, V., M., F., Gibson, J., Haseler, J., Hernandez, A., Kelly, G., Li, X., Onogi, K., Saarinen, S., Sokka, N., Allan, R., Andersson, E., Arpe, K., Balmaseda, M., Beljaars, A., Van De Berg, L., Bidlot, J., Bormann, N., Caires, S., Chevallier, F., Dethof, A., Dragosavac, M., Fisher, M., Fuentes, M., Hagemann, S., Holm, E., Hoskins, B., Isaksen, L., Janssen, P., Jenne, R., McNally, A., Mahfouf, J., Morcrette, J., Rayner, N., Saunders, R., Simon, P., Sterl, A., Trenberth, K., Untch, A., Vasiljevic, D., Viterbo, P., and Woollen, J. (2005). The ERA-40 re-analysis. Q. J. R. Meteorol. Soc., 131:2961–3012. Varhola, A., Coops, N. C., Weiler, M., and Moore, R. D. (2010). Forest canopy e↵ects on snow accumulation and ablation: An integrative review of empirical results. Journal of Hydrology, 392(3-4):219–233. Vikhamar, D. and Solberg, R. (2003). Snow-cover mapping in forests by constrained linear spectral unmixing of MODIS data. Remote Sensing of Environment, 88:309–323. Viviroli, D., D¨urr, H. H., Messerli, B., Meybeck, M., and Weingartner, R. (2007). Mountains of the world, water towers for humanity: Typology, mapping, and global significance. Water Resources Research, 43(7):1–13. Yasunari, T. and Inoue, J. (1978). Characteristics of Monsoonal Precipitation around Peaks and Ridges in Shorong and Khumbu Himal. Seppyo, pages 26–32. Zelazowski, P., Sayer, A. M., Thomas, G. E., and Grainger, R. G. (2011). Reconciling satellitederived atmospheric properties with fine-resolution land imagery: Insights for atmospheric correction. Journal of Geophysical Research D: Atmospheres, 116. Appendix A Additional figures and tables 87 APPENDIX A. ADDITIONAL FIGURES AND TABLES 88 Figure A.1 – Monthly average annual evolution of North facing SLAs in sub-catchment 5. Mean, minimum and standard deviation statistics represented. Figure A.2 – Monthly average annual evolution of South facing SLAs in sub-catchment 5. Mean, minimum and standard deviation statistics represented. APPENDIX A. ADDITIONAL FIGURES AND TABLES 89 Table A.1 – SLA amplitude values -for each aspect quarter within each sub-catchment- of the periods included in the winter seasonal analysis. Period Sub-catchment SLA Amplitude [m] Entire catchment East South West North 1 1&2 385.40 408.57 547.75 472.24 222.24 3578.48 604.87 610.50 863.38 416.90 4657.71 579.52 1199.85 446.87 591.48 51123.21 1046.45 1413.40 1158.05 943.56 6998.56 1019.45 846.31 1215.15 1304.37 7795.93 970.90 1105.22 605.70 964.42 2 1&2 217.75 325.50 378.25 274.48 113.68 3222.66 214.13 195.03 282.04 202.30 466.49 92.03 71.91 39.44 78.46 5175.70 205.38 188.30 224.49 184.49 6164.79 830.21 184.58 262.98 589.06 7274.98 272.77 329.44 307.61 155.45 3 1&2 262.77 445.31 483.92 352.85 110.85 3438.37 285.18 580.25 712.99 351.63 4401.82 356.04 718.04 362.72 435.56 5667.96 462.00 1114.04 703.00 632.86 6696.95 736.87 567.91 831.67 1212.79 7412.86 427.39 884.96 515.32 832.35 4 1&2 260.91 441.19 426.46 419.84 89.42 3338.71 229.48 428.62 497.98 335.05 4488.26 102.36 1058.75 392.39 572.21 51129.07 1217.42 1382.81 1274.34 761.62 6773.00 705.97 545.07 1094.70 1417.03 7735.57 359.81 1089.50 472.46 1149.11 APPENDIX A. ADDITIONAL FIGURES AND TABLES 90 Table A.2 – SLA amplitude values -for each aspect quarter within each sub-catchment- of the periods included in the pre-monsoon seasonal analysis. Period Sub-catchment SLA Amplitude [m] Entire catchment East South West North 1 1&2 497.78 1315.61 1096.72 352.23 914.72 3 724.14 1992.03 688.64 851.04 833.05 4 445.30 681.54 1520.49 456.94 462.00 5 346.77 1482.79 1864.82 681.77 391.09 6 1826.83 2377.67 2030.42 1488.11 2312.40 7 677.53 2068.38 2038.34 939.54 568.85 2 1&2 594.18 406.85 547.72 874.09 841.37 3 659.56 565.75 801.07 762.74 562.34 4 469.78 272.51 538.02 682.26 660.50 5 497.12 398.08 715.83 699.54 505.01 6 900.83 1229.43 945.47 916.30 671.17 7 462.03 289.72 437.85 612.43 527.28 3 1&2 597.55 867.20 808.67 500.43 752.07 3 819.56 680.13 888.86 1149.05 679.53 4 675.01 707.22 773.84 718.66 732.57 5 749.54 627.58 932.07 935.09 775.94 6 1049.48 1453.36 926.57 1213.65 1420.09 7 989.51 708.45 1034.73 926.55 987.42 4 1&2 380.04 723.10 810.33 344.46 563.35 3 486.43 1060.83 362.91 603.24 586.44 4 395.97 465.93 578.22 477.55 521.11 5 438.67 1036.12 700.27 421.95 490.50 6 1247.87 1273.07 1130.98 1148.66 1807.53 7 564.12 1242.95 810.13 467.62 838.23 5 1&2 504.55 645.95 488.13 628.71 678.83 3 311.97 734.03 998.81 691.05 731.15 4 881.71 613.89 969.61 750.23 1004.85 5 1072.28 852.88 1282.92 1176.31 1126.34 6 1189.69 1128.12 1175.23 1456.74 1202.65 7 1095.51 1022.02 1044.11 1262.41 948.33 6 1&2 366.33 446.30 297.32 253.33 699.76 3 174.30 307.02 349.37 270.29 456.99 4 663.76 483.79 629.48 416.03 981.63 5 920.02 1217.82 1465.48 885.81 675.36 6 566.14 622.20 502.26 846.28 700.94 7 548.85 589.47 623.38 384.61 660.52 APPENDIX B. METEOROLOGICAL DATA 97 Table B.2 – Mean monthly temperature values for the period 1999-2010, and the monthly statistics Mean and . Month Year 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Mean  Jan -2.51 -1.49 -5.71 -0.46 2.81 -2.24 -3.79 0.62 -0.15 -2.47 0.27 -2.17 -1.44 2.14 Feb 0.42 -3.58 -7.39 1.65 -0.54 -1.50 -3.21 2.75 -2.38 -3.28 0.05 2.82 -1.18 2.86 Mar 1.95 -0.13 -6.68 3.69 2.58 3.33 0.41 0.76 1.47 0.82 0.91 4.97 1.17 2.78 Apr 4.34 5.50 -3.72 6.47 7.23 3.75 3.64 3.79 5.24 3.64 4.48 5.55 4.16 2.63 May 5.95 8.23 -0.30 9.51 8.35 6.71 5.65 7.35 7.03 6.35 5.95 7.98 6.56 2.35 Jun 7.91 10.18 -0.93 11.58 10.72 8.63 9.34 9.22 9.32 9.58 - 9.40 8.63 3.16 Jul 8.47 11.13 10.01 12.45 10.70 9.87 10.52 11.21 - 10.22 11.11 8.63 10.39 1.10 Aug 8.11 10.98 11.31 12.36 10.14 10.58 10.37 10.14 10.60 10.23 10.34 7.40 10.21 1.26 Sept 7.04 7.75 10.44 10.10 9.07 9.18 9.01 8.73 8.80 7.95 8.38 4.45 8.41 1.49 Oct 4.95 3.98 8.36 7.24 6.29 4.65 4.33 4.24 5.63 4.09 5.02 2.13 5.08 1.57 Nov 1.44 0.67 5.77 4.63 3.10 1.93 1.94 1.33 1.89 3.44 2.99 -0.74 2.37 1.68 Dec -3.87 -2.50 3.39 2.90 -0.69 4.07 0.04 1.17 0.37 2.03 0.07 -0.58 0.53 2.25 Annual 681.92 730.60 641.00 798.50 701.30 692.30 759.80 683.90 911.75 643.32 592.90 769.02 - - Acknowledgements First of all, I would like to thank the three supervisors of this thesis, Dr. Francesca Pellicciotti, Silvan Ragettli and Evan Miles for their great support and enormous contribution to this work. I am grateful for their always constructive inputs, comments and discussions but, above all, it was their enthusiasm for the topic which inspired me the most. Secondly, I want to thank the tutor of this thesis in the UPC, Prof. Dr. Ernest Blade i Castellet, for accepting my proposal and thus o↵ering me the opportunity to write the Master Thesis in collaboration with the Institute of Environmental Engineering, Chair of Hydrology and Water Resources of ETH Zurich. And last but not least, special thanks go to my family and friends. To my parents -Merc`e and Joanfor their both personal and academic support during all these years; and to my grandparents -Paquita, Rosa, Joan and Francescfor being always the greatest source of inspiration. 99