HYDROLOGICAL PROCESSES Hydrol. Process. 27, 2126–2143 (2013) Published online 11 May 2012 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/hyp.9306
Signature and hydrologic consequences of climate change within the upper–middle Brahmaputra Basin Biswajit Mukhopadhyay* Water Resources Group, Jacobs Engineering Group, Inc., 7950 Elmbrook Drive, Dallas, TX 75247-4951, USA
Abstract: A prevailing perception is that the glaciers and perennial snow and ice covered areas (SCAP) or in the Himalayan region are fast contracting. However, systematic studies providing the quantitative estimates of SCAP as a function of time within individual river basins are lacking. The importance of meltwater in river flows varies greatly from one river basin to another, yet the actual estimates of those contributions are largely unknown. This study bridges such knowledge gaps for the upper–middle Brahmaputra Basin by using best available digital cartographic and remotely sensed snow cover data. We find that when the entire basin is considered, SCAP decreased from 7637 764 in 1980 to 4298 1422 km2 in 1992. However, it has increased to 7160 2248 km2 in 2000. From 2000 to 2010, the SCAP has remained nearly constant around a mean of 10 052 1468 km2. The same trend is observed within individual physiographic zones of the basin. Such increase in SCAP is due to the increase in the precipitation over the middle Brahmaputra Basin and the Nyainquentanglha Mountains, as observed in station records. The incursion of moist air through the Brahmaputra valley to the higher elevations within the Nyainquentanglha Mountains causes snowfall during pre-monsoonal and post-monsoonal seasons and an expansion of the SCAP. Glacial expansions in the Nyainquentanglha Mountains have also been observed in other recent studies. In addition to the increase in precipitation and SCAP, another manifestation of climate change observed in this basin is the increasing temperature with a mean annual trend of +0.28 C/decade. The hydrologic consequences of the observed effects of climate change are expected to be an insignificant change in streamflows in the watersheds drained by the upper Brahmaputra River but a perceptible increase in river discharges in the watersheds drained by the middle Brahmaputra River and its tributaries, particularly within the upper and lower catchments of the middle Brahmaputra Basin. Copyright © 2012 John Wiley & Sons, Ltd. KEY WORDS
Brahmaputra Basin; climate change; snowmelt hydrology; snow cover change in the Himalayas
Received 9 November 2011; Accepted 7 March 2012
INTRODUCTION Virtually, all of currently available predictive models of climate change imply a warming trend for the atmosphere near the surface of the earth and thus make global warming almost unequivocal. The observations of the increases in global average air temperatures yield an estimate that is best described as a linearly increasing trend in global nearsurface air temperatures from 1960 to 2005. This trend shows a general warming in the likely range of 0.56 ºC to 0.92 ºC with an average of 0.74 ºC and a more rapid warming trend over the past 50 years (Bates et al., 2008). For example, according to Lozan et al. (2001), global temperatures have increased by 0.6 ºC 0.2 ºC since 1990. The fourth assessment report of the Intergovernmental Panel on Climate Change (IPCC, 2007) concludes that the average global temperature is very likely to increase between 1.8 C and 4 C by the year 2100. The warming of the global climate system is ascribed as the chief cause of the large-scale decrease of perennial snow and ice covered areas (SCAPs) and the extensive retreat of glaciers observed throughout the global land
*Correspondence to: Biswajit Mukhopadhyay, Water Resources Group, Jacobs Engineering Group, Inc., 7950 Elmbrook Drive, Dallas, TX 752474951, USA. E-mail:
[email protected];
[email protected] Copyright © 2012 John Wiley & Sons, Ltd.
surface (Oerlemans, 2005). The effect of a warming climate is distinctly manifested in the Himalayas and the Tibetan Plateau where the shrinking of ice-covered areas and SCAs in various parts has been detected by many satellite images and glacial mass balance studies (e.g. Berthier et al., 2007; Kulkarni et al., 2007; Kehrwald et al., 2008; Prasad et al., 2009; Raina, 2009). The Himalaya–Karakoram–Hindu Kush (HKH) mountain ranges along with the adjoining Tibetan Plateau and central Asian mountains such as the Tien Shan and Kunlun ranges of western China have the most highly glaciated areas and largest body of ice outside the polar region (Owen et al., 2002; Dyurgerov and Meier, 2005; Kulkarni et al., 2007). Several workers have claimed that the glaciers and snowfields of the HKH mountain belts and the Tibetan Plateau are the fastest receding glaciers and snowfields in the world (e.g. Dyurgerov and Meier, 2005; Prasad et al., 2009). In addition to glacial retreats, the widespread fragmentation of the glaciers has also degraded the total areal coverage of perennial snow and ice in this region (Kulkarni et al., 2007; Raina, 2009). Although these trends have been primarily attributed to global atmospheric warming, in some cases, a combination with decrease in precipitation, another probable sign of climate change, is also considered as the likely cause. The HKH region and the Tibetan Plateau, aptly called the Water Tower of Asia, is the source of ten largest rivers
UPPER–MIDDLE BRAHMAPUTRA BASIN CLIMATE CHANGE AND HYDROLOGIC EFFECTS
in Asia, namely, Ganga, Indus, Brahmaputra, Tarim, Mekong, Irrawady, Amu Darya, Salween, Yangtze Kiang and Huang Ho (Yellow River). Approximately 1.5 billion people, constituting more than one sixth of world population, inhibit the basins drained by these rivers and their tributaries. Consequently, fresh water availability in these river basins has a profound influence on the livelihood of a great human population on this planet. One of the main concerns in relation to the climate change in the Himalayan region is the reduction of perennial snow and ice, which reduces the water storage capacity of this stupendous source of fresh water that contributes at various degrees to the flows of these river systems during the melting season in the Himalayas. There are a limited number of direct measurements of glacier volumes and thicknesses in the Himalayan region. However, such measurements are essential to the calculations of the time required by a given glacier to possibly disappear or recede beyond restoration. Similarly, with the exceptions of a few well-known or benchmark glaciers, there are meager data on the quantitative estimates of changes in SCAP1 that have affected individual river basins in recent decades. Snow, ice and glacial melts do constitute a significant component of the river flows in all of the Himalayan river basins. However, the percentages of meltwater from perennial snow and ice in the river discharges vary from one river to other (Singh and Bengtsson, 2004; Barnett et al., 2005). For example, snow and ice melt contributions are estimated to be approximately 40% to 45% of river flows in the Indus and the Tarim Rivers, whereas 9% to 12% of the discharge originates as meltwater for the Ganga and the Brahmaputra Rivers (Jianchu et al., 2009). Interestingly, in most cases, how these estimates are made and hence the accuracy of these estimates are not known. Furthermore, nearly all Himalayan river basins have the peak discharge period during the summer months beginning in April/May and ending in August/September. Although the great rivers of the Himalayan region originate from certain glaciers, glacial melts are not the only melt components in streamflows. Meltwaters from the entire ice-covered areas and SCAs, existing above the equilibrium line, and monsoonal rain also make significant contributions to streamflows during the melting season. For the reasons stated earlier, a greater understanding of the perils of decline of SCAP in the Himalayan region is obtained if the quantitative estimates of changes in SCAP within individual river basins, as a function of time, are available. The quantitative estimates of the average retreat or shrinkage of perennial snow, ice and glacial covers provide a better representation of the changes in the cryosphere that have direct effects on the stream water availability. Thus, to understand the effects of climate change on river basin–scale hydrology, the estimation of 1
SCA includes both seasonal and perennial snow and glacier covers, whereas the SCAP refers only to that part of the SCA that is perennial. This part is most significant in contributions to river flows during the melting season (summer months).
Copyright © 2012 John Wiley & Sons, Ltd.
2127
shrinkage or expansion of basin-scale SCAP, integrated over a certain period, is more significant and useful than the information on individual glaciers. The examination of SCAP at scales of individual river basins offers various advantages in the assessments of the effects of climate change on regional water resources. A general large area, such as the entire Himalayan Range or the Tibetan Plateau, covers various portions of several river basins. On the other hand, the importance of snow and ice melt contributions to streamflows varies from basin to basin, as noted earlier. The degree of snow and ice melts to river flows depends on the area of a basin that is covered by perennial snow and ice, the energy input the basin primarily receives from atmospheric heat, the solar and terrestrial radiations and the input of precipitation mass (rain and seasonal snow) to the basin. Furthermore, quantitative assessments of temporal changes in SCAP within a particular river basin help to address certain questions that are integral to the development of adaptation and mitigation policies for regional water resources management in a changing climate. Such questions include the following: (i) Will the relative importance of snow and ice melt compared with rainfall change with global warming? (ii) To what extent are the river flows and flow regimes affected by snow and glacial melts? (iii) Within each river basin, how long will it take for the SCAP to melt out and how different will be the seasonal patterns of the flow regimes of the rivers within those basins while the decline in SCAP takes place? One of the impediments in the development of scientifically sound models of cryospheric states within the Himalayan region is the shortage of pertinent high quality data. Several factors have contributed to the paucity of knowledge concerning the hydrologic cycle of ten large river basins. One major factor is surely the challenging and remote nature of the terrain where systematic field-based observations and measurements are extremely difficult to obtain. Accessibility to certain parts of the terrain is also limited by regional geopolitics. For these reasons, reliance should be placed on remotely sensed data to detect the effects of climate change on the cryosphere of this region. The objective of the present investigation was to use the best available digital cartographic and remotely sensed snow cover data at moderate spatial resolutions to assess the changes in SCAP in one of the major Himalayan river basins that originate in the Tibetan Plateau. The aim was also to explore the implications of these changes within the context of the current understanding of the climate change and their effects on the hydrologic regimes of this river basin.
UPPER–MIDDLE BRAHMAPUTRA BASIN The river basin selected for the present investigation is the upper–middle Brahmaputra Basin (UMBB; Figure 1). This basin is drained by the mountainous section of one of the greatest rivers in Asia, namely, the Brahmaputra Hydrol. Process. 27, 2126–2143 (2013)
2128
B. MUKHOPADHYAY
Figure 1. Location and boundary of UMBB (shown in red) in relation to the Himalayas and central Asian highlands. Other mountain ranges and locations of certain townships are shown. The courses of two rivers, namely, the Ganga and the Brahmaputra Rivers, are shown up to the point of their confluence. Country boundaries are also shown. The length of the upper–middle Brahmaputra River up to the township of Goalpara is 2844 km
River. The Brahmaputra River originates from the Jima Yangzong (or Chamyungdung) Glacier at an elevation of approximately 5300 m in the Kailash Range of southwestern Tibet and flows to the east along a narrow valley straddling the Himalayan Range to the south and the Gangdise Mountains to the north (Figure 1). The basin widens to the east and straddles the eastern Himalayan Ranges to the south and the Nyainquentanglha Shan (Shan = Mountains) to the north. At its eastern limit, the river makes a sharp U-turn near Mount Namcha Barwa (Figure 1). Here, it forms a deep long canyon, only recently explored, known as Yarlung Tsangpo Grand Canyon or Big Bend of Tsangpo George, considered to be the deepest, steepest and longest gorge in the world. After leaving this canyon, the river cuts through the foothills of the eastern Himalayas and enters the plains of the state of Assam, India. After leaving India near the border town of Dhubri, it enters Bangladesh and joins the Ganga River. In this study, UMBB is considered up to the point near the township of Goalpara in India (Figure 1). Downstream of Goalpara, several major tributaries from the mountains of Bhutan and India, such as the Manas and the Teesta, join the Brahmaputra River, and the river basin is quite complex. Thus, downstream of Goalpara, the basin is considered as the lower Brahmaputra Basin. Further downstream, from the confluence of the Ganga and the Brahmaputra Rivers, the basin becomes vast and deltaic. At its lower reaches, flows in the Brahmaputra and its tributaries are heavily dominated by monsoon rainfall–runoff (pluvial regime). In the lower Brahmaputra Basin, although contributions from snow and ice melts (nival regime) come from the eastern Himalayan Ranges of Bhutan and India, their contributions relative to those coming from the south Asian monsoonal precipitation are lower. Thus, for the purpose of assessment of snow and glacial melt contributions to the flows in the Brahmaputra River, the upper to the Copyright © 2012 John Wiley & Sons, Ltd.
middle sections of Brahmaputra River are the most critical sections. The Brahmaputra River is indeed a unique river that drains such diverse environments as the cold dry plateau of Tibet, the rain-drenched Himalayan slopes, the landlocked alluvial plains of Assam and the vast deltaic lowlands of Bangladesh (Goswami, 1985). It should be noted here that the upper and middle sections of the Brahmaputra River, as defined in this study, are also in climatically different zones. The upper reaches of the Brahmaputra flow through extremely arid terrain. Although the upper Brahmaputra Basin receives significantly less monsoonal moisture than the middle Brahmaputra Basin, it also receives moisture recycled from the Tibetan Plateau. The middle Brahmaputra Basin on the other hand has pronounced influence of the South Asian Monsoon and contains alpine as well as temperate forests.
DATA AND METHODS Data
One primary requirement of the present investigation is to establish a baseline condition when the SCAP in the river basin under investigation is considered to be minimally affected by warming of the climate system in recent decades – a phenomenon that is considered undeniable by most workers in the field of climate change (see Bates et al., 2008). Such a condition is established from the Digital Chart of the World (DCW) produced by the Environmental Systems Research Institute for the US Defense Mapping Agency (DMA) from the Operational Navigation Charts, also known as aeronautical charts, as vector maps at a scale of 1:1 000 000 (DMA, 1992a). The DCW is generally considered to provide the most comprehensive and consistent global digital hydrographic data currently Hydrol. Process. 27, 2126–2143 (2013)
UPPER–MIDDLE BRAHMAPUTRA BASIN CLIMATE CHANGE AND HYDROLOGIC EFFECTS
available at a cartographic small scale. The hydrographic data sets include global SCAP. Although the information in the Operational Navigation Charts dates from the 1960s to the 1980s (Danko, 1992), it was primarily based on the photogrammetric analyses of the Department of Defense Corona imagery acquired in the 1960s. Thus, the SCAP for a river basin, derived from DCW, is considered to represent pre-1980 conditions. In this study, 1980 is assigned as the latest year from which the hydrographic features of the Himalayan region are represented in the DCW. The metadata of DCW are given by DMA (1992b). All data are found to be topologically correct. No duplicate features are present. All areas are completely described as depicted on the source manuscripts. Detailed horizontal accuracy figures are developed by comparing the positions of well-defined points in road, railway, utility line and drainage features against sources of higher accuracy, by measuring offsets and by expressing differences as a ‘circular map accuracy’ figure at a 90% CI. On the basis of this assessment, 10% error is assigned to the estimates of the SCAP derived from the DCW. Another digital source from which quantitative estimate of SCAP in a basin can be derived is a land cover map, produced at a certain time, by remote sensing. For a regional river basin with large areal extent where basinspecific land cover maps are not readily available, such data sets, developed from data derived from satellite observations, can be used. Two different sets of global land cover (GLC) data that are currently available as most reliable data sets are used in the present investigation. The first set of GLC data, produced by the United States Geological Survey National Center, the University of Nebraska at Lincoln and the Joint Research Center of the European Commission, from the data collected by the Advanced Very High Resolution Radiometer instruments on board the polar-orbiting satellites of National Oceanic and Atmospheric Administration, spanning a 12-month period from April 1992 to March 1993 (Loveland et al., 2000), is considered to represent the GLC characteristics for the years around 1992. The area-weighted accuracy of this GLC data set is estimated to be 66.9% (Scepan, 1999). This accuracy figure weights the importance of each of the land cover classes’ accuracy based on the land area occupied by that class. The second set of GLC data are obtained from the Global Environment Monitoring unit of the Institute of Environment and Sustainability of the European Commission of Joint Research Center as raster maps as derived from the data acquired by the VEGETATION instrument on board the SPOT 4 satellite, launched on 24 March 1998. The Global Environment Monitoring GLC 2000 data are considered to be the internationally standardized land cover data, producing the land cover information of the earth for the years around 2000 using the land cover classification system produced by the Food and Agriculture Organization of the United Nations Environment Program. The overall accuracy of the GLC 2000 land cover areas is 68.6% (Mayaux et al., 2006). Copyright © 2012 John Wiley & Sons, Ltd.
2129
The global snow cover data sets used in this investigation are the ones produced by the Earth Science Division of the Science Mission Directorate of the National Aeronautics and Space Administration (NASA). NASA launched a coordinated series of polar-orbiting and low inclination satellites for the long-term global observations of the land surface, biosphere, solid earth, atmosphere and oceans as part of an umbrella program known as Earth Observing System (EOS). As a part of NASA’s EOS program, global snow cover data are being collected by two satellites, named Terra and Aqua, since 2000 using Moderate Resolution Imaging Spectroradiometer (MODIS) instruments, which are 36-channel visible to thermal infrared sensors. The first MODIS instrument was launched on board the Terra satellite on 18 December 1999, and the second was launched on board the Aqua satellite on 4 May 2002. Terra’s orbit around the Earth is timed so that it passes from north to south across the equator in the morning, whereas Aqua passes south to north over the equator in the afternoon. Terra MODIS and Aqua MODIS are viewing the entire Earth’s surface every 1 to 2 days. The MODIS instrument provides high radiometric sensitivity (12 bit) in 36 spectral bands ranging in wavelengths from 0.4 to 14.4 mm. MODIS obtains measurements with spatial resolutions of 250 m (bands 1 and 2), 500 m (bands 3–7) and 1000 m (bands 8–36) using a continuously rotating double-sided scan mirror. The MODIS instruments provide calibrated, geo-referenced radiance data from individual bands. Hall and Riggs (2007) provide a detailed account of the MODIS products. In short, the MODIS snow product suite begins with a 500-m resolution, 2330-km swath snow cover map, which is then girded to a sinusoidal grid. The sequence proceeds to a latitude–longitude grid called climate-modelling grid (CMG), with a spatial resolution of 0.05 0.05 and cylindrical equidistant projection. Monthly average snow cover is calculated from the 8-day composites for the month. Two sets of MODIS-derived snow cover data, collected from both Terra and Aqua, are obtained from National Snow and Ice Data Center (NSIDC). The MODIS/Aqua Snow Cover Monthly L3 Global 0.05 CMG (MYD10CM) data set (Version 5) and the MODIS/Terra Snow Cover Monthly L3 Global 0.05 CMG (MOD10CM) data set (Version 5) contain snow cover data in Hierarchical Data Format–Earth Observing System (HDF-EOS). Both these data sets consist of 7200 column 3600 row global arrays of snow cover in a 0.05 CMG. The MYD10CM data set covers the period starting from July 2002 to the present. The MOD10CM data set covers the period starting from March 2000 to present. However, because of insufficient data received, the MOD10CM time series does not contain the monthly granules for August 2000, June 2001 and March 2002. For these missing monthly averages, 8-day composite data are obtained. For these 3 months, each monthly data set contains four 8-day composites of which typically one granule contains insufficient data and is discarded in the Hydrol. Process. 27, 2126–2143 (2013)
2130
B. MUKHOPADHYAY
development of monthly averages. In the present study, MODIS-derived snow cover data are used for the period March 2000 to December 2010. Other data sets used for developments of the topology, climatology and hydrology of the river basins are described by Mukhopadhyay and Singh (2011). In particular, monthly average precipitation and temperature data sets developed by Matsuura and Willmott (2009a,2009b), covering the period 1900–2008, have been used to develop long-term (109-year) monthly average climatic grids for the entire river basin. Values at specific locations are extracted from these climatic grids. At certain locations, monthly average climatic data are also derived from the CLIMWAT database developed by the Water Development and Management Unit of the Food and Agriculture Organization (FAO, 2006) and some published literature. The CLIMWAT database has been developed from station records. The period of record is typically 1971–2000 or at least 15 years of record. River flows are derived from microwave date collected by the Advanced Microwave Scanning Radiometer Earth Observing System on NASA’s Aqua satellite (Brakenridge et al., 2007; Dartmouth Flood Observatory, 2011). At this point, it should be noted that no-gauge-based streamflow data for the Brahmaputra Basin are available to the public because of the policies of the governments, albeit such policies detriment advancements of knowledge by deterring dissemination of information that are critical for understanding the hydrological processes of this important river basin. Long-term average monthly discharge and associated statistics, covering the period 1956–1979 at only one gauging station (Pandu), are available and obtained from the Global Runoff Data Center (2011). General methods
The principal tool used in the data analysis is the geographic information system (GIS) technology. The GIS software system used for this purpose is ArcGISArcInfo, developed by the Environmental Systems Research Institute. Procedures used in the development of river basin topology and other hydrologic characteristics are described by Mukhopadhyay and Dutta (2010). Map projection
All of the data described earlier are available in various formats that are initially stored in a database with a geographic coordinate system, which is a spherical coordinate system. These data represented on an ellipsoidal earth must be projected on a planar coordinate system so that linear and planer features such as stream lengths, catchment, watershed and basin areas as well as SCA can be properly quantified. Finlayson and Montgomery (2003) showed that a careful selection of map projection in the creation of a digital elevation model (DEM) can minimize the length and area distortion when analysing large portions of the earth in a two-dimensional plane of a DEM. Thus, it is important to establish a standard projection scheme that Copyright © 2012 John Wiley & Sons, Ltd.
yields minimal distortion in area. We have selected the Lambert Equal Area Conformal Conic projection using the parameters given by Mukhopadhyay and Singh (2011). This is one of the best projection systems for middle latitudes. The characteristics of the DEM for the river basin are given in next section. Data processing and derivation of snow cover grids
The spatial data described earlier, originating in different data formats, are brought into the GIS through appropriate conversions. Subsequently, the data are processed to develop raster grids with uniform cell sizes and same map projection. All other grids, such as the SCA and other climatic grids, have been resampled to have these same spatial resolutions for the reasons given by Mukhopadhyay and Dutta (2010). The general methodology used in the development of a geodatabase with uniform spatial scale, resolution and coordinate system from data obtained from varied sources with varied format has been described by Mukhopadhyay and Dutta (2010). Additional details on processing and usage of MODIS data are given in the following paragraphs. Each of the cells in a snow cover grid derived from MODIS data has an integer value. Values that range from 0 to 100 denote the percentage of area in that cell which is snow covered. However, in a grid derived from MODIS data, there are cells with values higher than 100. These values denote other features or attributes such as cloud cover (250), indecision (253), night (211), water mask (254) or simply a gap in the data (255). Not all of the values that are higher than 100 are necessarily present for a certain month in a specific region that is being considered. However, the monthly snow cover grids for a river basin generally have some cell values that do not lie between 0 and 100. Thus, we first classify those cells with values higher than 100 as cells with no data on snow cover from MODIS observation. However, in reality, some or all of those cells can potentially represent fully or partly SCAs. For this reason, those cells are compared with the snow cover grid that is prepared from GLC. The 2000 GLC data set is used in the absence of any other better reference data set. Thus, if any of the no data cell is found to be snow covered in the GLC grid, then that cell is converted to a cell with a value of 100. To minimize uncertainties that can arise from this approach, we have used both MYD10 CM (Aqua data) and MOD10CM (Terra data) time series. For a given month of a year, when both Aqua and Terra data are available, the set with lower number of no data cells is selected for final analysis. For the estimation of the area of SCAP, the most critical months are the summer months, which represent the period of snow melting. Fortunately, for these months, the number of no data cells in both Aqua and Terra MODIS data is very few (
