RUNOFF AND SEDIMENT MODELLING IN PART OF THE BRAHMAPUTRA BASIN

Abstract

The Brahmaputra river, termed as a moving ocean, is an antecedent snow-fed large Trans-Himalayan trans-boundary river which flows across four countries, namely, China, India, Bhutan and Bangladesh. This river system is one of the largest in the world, and majestic in multiple aspects: in the volumes of water and sediment that it gathers and passes on, the power with which these flows are routed, and the scale of changes that these powerful flows bring upon the landscape. Among the top ranking large rivers of the world, the ranking of the Brahmaputra river is fourth with respect to discharge, fourteenth with respect to drainage area, twenty fifth with respect to length, second with respect to sediment load and first with respect to specific discharge. The river system stands out in terms of its enormous and mostly untapped water resource potential, nearly 30% of India's water resources potential and 41% of the country's total hydropower potential are found in this basin. Unfortunately, this very same potential turns destructive with fatal regularity every year in the form of floods, migration of channels, erosion of land. The Brahmaputra river system drains a vast area with varying climatic and geomorphic conditions. The gradient of the river varies from very steep (1:385) near the source in the Tibetan plateau to very flat (1:37,700) in the lower part of Bangladesh. The climate of the basin varies from climatic Zone III (mountain climate: cold, dry and arid) in the north to climatic Zone I (tropical monsoon climate: relatively warm and humid) in the south. The basin experiences a wide variation of the average annual rainfall from less than 400mm to more than 6,000 mm. The discharge in the river between summer high flows and winter low flows fluctuates, on an average, by 12 times although in certain years it has been as high as 40 times. Sediment transport in the river is found to be more variable than its flow. During its course through the valley in the state of Assam, 19 maj or tributaries originating from the Himalayan mountain ranges join the river from the northern (right bank) and 13 major tributaries originating from southern (Assam) mountain ranges join the river from the south. Also, the Brahmaputra is a classic example of a braided river with multiple channels twining around numerous mid-channel and lateral sandbars. V Therefore, for this vast and dynamic alluvial river, which exhibits a high variability in flow and sediment yield, changing boundary conditions of the channel, and complex river morphology, the uncertainties involved in the mathematical idealization are significantly high. And to compound the problem, global warming and the impending climate change, especially in Himalayas has added another dimension of complexity. Better understanding of the drivers of stream flow and sediment discharge dynamics in the Brahmaputra River basin is needed for development of effective management strategies for the region. From a review of the literature, it is found that very few studies related to runoff and sediment modelling have been made so far. Also, there is a serious hurdle of data availability due to trans-boundary nature of the river system. 2.0 Objectives Based on the pertinent review of literature, the following objectives have been envisaged in the present study which are inter-related with a bearing on the proposed modelling: Study of the Brahmaputra river migration and estimation of bank erosion; Modelling of stage-discharge and sediment-discharge rating at river gauging sites; Runoff modelling in the largest sub-basin of Brahmaputra River including snowmelt runoff modelling; Modelling of sediment yield in the largest sub-basin of Brahmaputra River; Study of climate variability using meteorological parameters in the largest subbasin of Brahmaputra river; Study of probable impact of climate change on the hydrological regime of the largest sub-basin. Out of these six inter-related objectives, (i) and (ii) have been carried out for the Brahmputra river while objectives (iii), to (vi) have been carried out for the largest tributary of Brahmaputra river, namely, the Subansiri river basin. As described earlier, the Brabmaputra river basin is a very large basin with drainage area of 580,000 sq. km. (50.5% in China, 33.6% in India, 8.1% in Bangladesh and 7.8% in Bhutan). To model such a huge and complex basin with intricate variability in various parameters is not possible in a limited time. Therefore, it was considered prudent to take up the largest vi tributary, the Subansiri river which encompasses almost all the diverse intricate characteristics of the entire Brahmaputra basin to some measure. 3.0 Brahmaputra River Migration and Estimation of Bank Erosion Geometry of a stream channel is controlled by both water and sediment movement, which reflect regional climate, geology, and human land use in a given drainage basin. For planning and sustainable development, identification of morphological changes is essential. Using satellite images, this study is aimed at quantifying the actual bank erosion along the Brahmaputra River as well as major twenty two tributaries (12 north bank and 10 south bank) within India for a period of seventeen years (1990-2008). The entire course of Brahmaputra River in Assam from upstream of Dibrugarh up to the town Dhubri near Bangladesh border for a stretch of around 620 kms has been studied using an integrated approach of Remote Sensing (RS) and Geographical Information System (GIS). The channel configuration of the Brahmaputra River as well as the tributaries has been mapped for the years 1990, 1997 and 2008 using IRS 1A LISS-I for 1990 and IRS-P6 LISS-Ill satellite images for 1997 and 2008. It is observed that there has been significant erosion in the river during a period of 18 years as compared to deposition. Erosion is more pronounced in the southern bank and overall rate of erosion and deposition for 1990-2008 work out to be 62.7 Km2/year and 12.6 Km2/year respectively. For more recent period of 1997-2008, overall rate of erosion and deposition work out to be 72.5 Km2/year and 30.37 Km2/year respectively. The erosion rate has been calculated after excluding avulsion of southern bank. The results indicate sharp increase in land lost due to river bank erosion in recent years. The total erosion in twenty three tributaries for the period of 1990-2008 has been found to be 565.68 Km2 and for the period 1997-2008 has been found to be 476.44 Km2 which is also registering sharp increase in recent past. Erosion is more prominent in the northern tributaries compared to the southern tributaries. Among the tributaries, highest erosion for all the years has been found in the Subansiri river. The study amply suggested that three to four major geological channel control points are present along the Brahmaputra in Assam flood plains which are by and large holding the river around the present alignment. These control points are in the vicinity of Jogighopa near Goalpara, Pandu near Guwahati, Tezpur and Bessamora in Majuli. Many reaches along the Brahmaputra river have been perceived as suffering from high erosion that endanger nearby VII settlements and infra-structure. The reaches have been prioritized with respect to the land area lost in seventeen years. The stream bank erosion resulting in land area loss is surmised to be closely linked to high sediment inflows mainly contributed from the catchment areas receiving heavy precipitation. 4.0 Modelling of Stage-discharge and Sediment-discharge Rating Assessment of stream flow and volume of sediments being transported in rivers is important for a variety of hydrologic applications such as water resources planning and operation, water and sediment budget analyses, hydraulic and hydrologic modelling, design of storage and conveyance structures, hydroelectric power generation and water supply, determination of the effects of watershed management, and environmental impact assessment. Rating curves based on regression analysis are widely used by field engineers to estimate the stream flow as well as sediment load. The regression and curve-fitting techniques are not adequate in view of the complexity of the hydrological processes. A problem inherent in the rating curve technique is the high degree of scatter, which may be reduced but not eliminated. The artificial neural networks (ANNs) concept and its applicability to various problems of water resources have been amply demonstrated by various investigators. The ANN approach for modelling the discharge and sediment rating relation has already produced very encouraging results. In the present study, ANN technique has been used for modelling stage-discharge and sediment-discharge rating relationship. The developed ANN models have been trained, validated and cross-validated for two important gauging sites, namely, Panchratna, on the main stem of Brahmaputra river and Chouldhuaghat, at the outlet of the Subansiri river basin. Daily data of stage, discharge and sediment concentration at the gauging sites for fifteen years (1993-2007) have been used for modelling. Out of this, nine years (1993-2001) have been used for training, three years (2002-2004) for validation and three years (2005-2007) for cross-validation. Three different learning algorithms (back propagation, genetic algorithm and Levenberg-Marquardt) have been used for training of five types of ANN models (with varying input data) and their performance have been compared. Correlation coefficient is seen to vary from 0.78 to 0.99 for various models at the two gauging sites. The performance of the ANN models have also been compared with the conventional rating curve technique. It has been found that discharge as well as VIII sediment concentration estimations at both the gauging sites obtained by ANNs were superior to the sediment rating curve ones. 5.0 Runoff Modelling for Subansiri River basin In this study, runoff modelling using two approaches has been carried out for the Subansiri basin. One is a conceptual snowmelt runoff modelling approach and the second is a black box ANN technique. Subansiri river has its origin in the Central Himalayas in Tibet at an approximate altitude of 5340 in. Traversing through the snow clad mountains of Great Himalayas, it enters India in its North-East corner in Arunachal Pradesh. The river bed level of it drops from more than 4000m height in the mountainous region to less than 1 00m in the foothills before it enters the plains of Assam promising stupendous hydropower potential (22 projects having potential of 15,191 MW already proposed/in progress). The contribution of the Subansiri river is estimated to be about 10.66% of the total discharge of Brahmaputra River at Pandu near Guwahati. The total catchment area of Subansiri basin upto the Chouldughat gauging site is 26419 sq.km, out of which 16181.76 sq.km (61.25% of total) lies in India and 10237.24 sq.km (38.75% of total) lies in Tibet. 5.1 Snowmelt runoff modelling Very limited snowmelt runoff studies have been carried in eastern Himalayan region, and no study has yet been done for Subanbsiri basin even though it has a very huge potential of hydropower development. Therefore, in order to simulate the snowmelt runoff in the Subansiri basin, SNOWMOD, a snowmelt runoff model has been applied which uses a temperature index approach of snowmelt computation. It simulates all components of runoff, i.e. snowmelt runoff, rainfall-induced runoff and base flow, using limited data. Various inputs required for SNOWMOD are snow cover area from satellite data, elevation from DEM to divide the basin into elevation zones, observed rainfall and average temperature in all the elevation zones & observed discharge at the outlet of the basin. Snow cover area (SCA) is one of the important input parameters in modelling. In eastern part of the Himalayan region, due to cloud cover it is very difficult to get cloud free satellite data for most part of the year. Therefore, snow covered area in the basin has been determined from Moderate Resolution Imaging Spectroradiometer (MODIS - a NASA space satellite) in the form of eight day composite snow cover data Ix (MODIS/Terra-MOD 1 0A2 products). The chances of getting cloud-free scenes in case of MODIS are higher due to higher temporal resolution. Besides, MODIS has an automated snow-mapping algorithm, which reduces the time and errors incorporated 41 during processing satellite data manually. For the present study, weekly MODIS data have been downloaded and processed for a period of eight years (2000-08) to obtain the SCA depletion curves for the Subansiri basin. Further, the SRTM-DEM has been used to prepare elevation map of the study area. The basin has been divided into 10 elevation zones with an altitude difference of 600m. The area covered in each elevation zone of the basin have been computed and used in the snowmelt runoff model. Besides SCA, daily rainfall and temperature data of observed stations within the basin in Indian part have been used. For the Tibetan part, high resolution gridded rainfall data at 0.25 degree grid from APHRODITE has been used. Discharge data available at the outlet of the basin (Chouldhuaghat) have been used. With all the input data processed and prepared in the SNOMOD format, the model was calibrated using the dataset for a period of three years (2000-2003) and model parameters for stream flow routing were optimised. Using the optimised parameters, the model was validated for a period of three years (2004-2007). From simulations, it is clear that the volumes and peaks of stream flow have been reproduced well by the model. For all the years, snowmelt runoff and rainfall runoff have been computed separately. The accuracy of the stream flow verification has been determined using three criteria, viz, shape of the outflow hydrograph, efficiency and difference in volume. Efficiency has been computed using non-dimensional Nash-Sutcliffe 'R2' value. For the three year periods i.e., for 2000-2003 and 20004-2007, efficiency varied from 73% to 79%. In all the years, high flow occurred in the month of July and August. A close observation of rainfall and stream flow data indicates that most of the peaks in the stream flow were because of rainfall. It has been found that annual snowmelt runoff contribution varied from 11-16% for calibration and validation years. While annual runoff due to rainfall and base flow contribution varied from 58-62% and 25-27% respectively. 5.2 Runoff Modelling using ANN technique Runoff modelling is of prime importance in many water resources planning, design, and management activities. The involvement of many often-interrelated physiographic and climatic factors makes the rainfall-runoff process not only very complex to understand but x also extremely difficult to model. Artificial Neural Networks (ANNs) have been proposed as efficient tools in developing nonlinear systems theoretic models of the rainfall-runoff process. In the present study, multi-layer feed forward ANN technique has been applied for developing runoff models for the Subansiri basin using available daily data of rainfall, temperature and discharge at various locations as well as the computed SCA in the basin. ANN modelling has been carried out for the same period as used for snowmelt runoff modelling so that a comparison can be made. Four groups of ANN models have been developed. The first group considers only rainfall and temperature as input variables; the second group considers rainfall, temperature and snow cover area as input variables(has exactly the same input data as that of SNOWMOD); in the third group, one day lagged runoff is added to the models of first two groups designed with cross-correlation analysis; and the fourth group considers input variables in forecasting mode, i.e., the same input variables as that of group three models without the input variables at time t, in other words, all the input variables are from previous time step/s. The results of first group indicate the value of correlation as 90% and 88% in training and validation respectively. The second group ANN models perform slightly superior compared to first group with value of correlation as 93% and 92% in training and validation respectively which is higher than that of snowmelt runoff models. 1-lowever, the ANN models simulate only the total runoff. The third and fourth group of rainfall-runoff ANN models perform better than first two groups with higher values of correlation as 97% & 96% and 96% & 95% in training and validation respectively. However, the third and fourth group models use antecedent discharge data as input, therefore, have limited applications. Whereas, the first two type of ANN models are more generalized and simpler for studying the impact of climate change as they have only rainfall, temperature and SCA as input. The performance of the ANN models have also been compared with multiple linear regression (MLR) models. It has been found that runoff estimations in all four groups of models obtained by ANNs were superior to the MLR ones. 6.0 Sediment Yield Modelling Soil erosion and sediment yield is one of the major problems in Himalayan region. The rivers emerging out from the Himalayan region transport the sediment at a very high rate. Himalayan and Tibetan region cover only about 5% of the Earth's land surface, but they supply about 25% of the dissolved load to the world oceans. The models for sediment xi simulation vary from a simple regression relationship to complex simulation models. Regression and curve fitting methods are inadequate to model the non-linearity in the relationship. On the other hand, the application of physics-based distributed process computer simulation is often problematic, due to the use of idealized sedimentation components, or the need for massive amounts of detailed spatial and temporal environmental data, which are not available. Neurocomputing provides one possible answer to the problematic task of sediment transfer prediction. Therefore, in this study, ANN models have been developed for simulation of daily, ten-daily and monthly suspended sediment flux from the Subansiri river basin. Daily data of rainfall, temperature, SCA and discharge for a period of 2001-07 have been used to develop daily sediment ANN models. For weekly and monthly ANN models, weekly and monthly average of suspended sediment concentration, discharge, temperature and SCA have been derived from the daily observed data. Similarly, total rainfall on weekly and monthly basis as well as the corresponding rainfall intensity have been derived from daily observed rainfall data. Again, four groups of each, daily, weekly and monthly sediment simulation ANN models have been developed for the Subansiri basin. The performance has been tested based on RMSE, correlation coefficient and Nash efficiency. The performance of ANN models have also been compared with multiple linear regression (MLR) models. The correlation between observed and simulated suspended sediment concentration varies from 70% to 90% for all the ANN models which is higher than the corresponding MLR models. 7.0 Trend Analysis The climate of the Himalaya is changing, particularly meteorological variables like, rainfall and temperature. Trend analysis in the annual, seasonal and daily rainfall, rainy days, rainfall intensity and temperature (maximum, minimum, mean and diurnal) for Subansiri river basin have been carried out. The data used in this study consists of daily gridded rainfall data (APHRODITE) at 0.25°x0.25° resolution for a period of 37 years (1970-2007) and daily gridded temperature data (IMD high resolution data) at 1 0x10 resolution for a period of 36 years (1969-2005). Seven grid points of daily rainfall spatially well distributed within the basin and lying in different climatic zones of the basins (ranging from cold and to tropical humid) have been taken up for the rainfall trend analysis. For temperature trend analysis, eight grid points lying inside and close to the XII basin have been taken which represent the different climatic regions of the basin. Historical trends have been examined using parametric (regression analysis) and nonparametric (modified Mann-Kendall statistics and Sen's slope estimator). The results show that many of these variables demonstrate statistically significant changes occurring in last three decades. The annual trend of rainfall and rainy days is increasing in the three grid points falling in the upper part of the Subansiri basin which represents part of cold arid desert of Tibet and part of the sub Himalayan thick forest area. The temperature trend analysis shows that minimum temperature in general is increasing at all of the grid points and diurnal temperature is decreasing at all the grid points considered for analysis, however, maximum temperature shows no trend. The mean temperature shows no significant trend in annual, pre-monsoon and post-monsoon seasons, however, there is rising trend during monsoon and winter seasons. 8.0 Probable Impact of Climate Change on the Hydrological Regime of Subansiri River basin Any simulation or forecast of stream flow is not complete without considering the impact of climate change. Therefore, the probable impact of climate change has been analyzed using hypothetical climate scenarios to understand the behavior of snowmelt runoff under the changed climatic conditions. The climate change impact study has been carried out for Subansiri basin using data of 2000-2003 and 2004-2007. For this purpose, ten hypothetical scenarios such as (T + 1°C, P + 0%), (T + 2°C, P + 0%), (T + 1°C, P + 5%), (T + 2°C, P + 5%), (T + 1°C, P + 10%), (T + 2°C, P + 10%), (T + 1°C, P - 5%), (T + 2°C, P - 5%), (T + 1°C, P - 10%), (T + 2°C, P - 10%) with respect to the baseline scenario (T + 0, P + 0%) have been created to study the impact of climate change on stream flow. The change in computed stream flow due to change in climate scenario provided an indication of the influence of climate change. It has been observed that total stream flow as well as snowmelt runoff increases when temperature is increased. The reason behind this increase is the increased snowmelt runoff which increases because of more melting of snow with the rise in temperature. It has been observed that the snowmelt runoff increased by about 5% and 12% for the increase of 1°C and 2°C in temperature respectively. However, there is not much change in snowmelt runoff with the changed precipitation scenarios. It has been found from this study that stream flow changed for all the scenarios of temperature and precipitation. XIII The observed maximum % increase in mean annual stream flow is about 6% for (T+2°C & P+lO%) scenario and the minimum % decrease in mean annual stream flow observed is about 11% for (T+1°C & P-1O%) scenario. Results of seasonal analysis of stream flow under the scenario of warmer climate and enhanced precipitation (T+2°C, P+lO%) indicate that the annual water availability is increased marginally with a reduction in water availability during pre-monsoon, winter and post-monsoon seasons, however, there is substantial increase in water availability during monsoon season which will have an increased risk of floods in the already flood prone basin. The impact of a warmer climate and reduced precipitation, i.e., (T+1°C, P-lO%) climate scenario shows comparatively significant reduction in annual water availability in the basin with reduction in water availability in all the four seasons. 9.0 Concluding Remarks This research endeavour contributed to gaining an insight into the complex river hydrodynamics of the Brahmaputra and its tributaries by conducting series of implicitly interrelated studies namely, study of the bank erosion and migration, development of ANN based rating curves for main river as well as Subansiri River, development of ANN based runoff and sediment models as well as snowmelt runoff models for Subansiri basin, study of the climate variability in the Subansiri basin and study of the probable impact of climate change on hydrological regime of the Subansiri basin. The present study in a large braided river basin like the Brahmaputra with complex and diverse behaviour has been a daunting task by all measures. Key findings and observations of the study are briefly narrated below: The study of river dynamics through analysis of multi-date satellite data has provided not only the information on the channel configuration of the river system but also has brought out several significant facts about the changes in river morphology, relative stability of reaches of the river banks and changes in the main channel as well as its tributaries. The results provide latest and reliable information on the dynamic fluviogeomorphology of the Brahmaputra River for design and implementation of drainage development programmes and erosion control schemes in the north eastern region of the country. The spatio-temporal analysis of planform changes has brought to the fore the intensely transient nature of fluvial landform of Brahmaputra river networks which xiv warrants investigation of the implicitly related drivers through hydrologic process modelling. The ANN based rating curve models have simulated the river flow and sediment characteristics which profoundly influence morphological variability. The formulation of rating relationship models on daily basis (for important gauging stations at Pancharatna of Brahmaputra river and Chouldhuaghat of Subansiri river) by capturing the inherent trend have thrown light on the variability pattern of the discharge and sediment at these points. Such daily rating curves are useful for deriving rating curves at sub-daily time scales required for real time flood forecasting. Snowmelt runoff modelling for the Subansiri river basin having a complex and diverse hydrometeorology with upper 40% area in cold and region of Tibetan plateau and lower area joining the floodplains of Assam is a first study of its kind. The SNOWMOD simulated all the three components of runoff, i.e. snowmelt runoff, rai nfall- induced runoff and base flow. It is envisaged that the information generated out of this study will add substantially towards planning and development of hydropower projects in the region. The ANN based runoff and sediment models for the trans-boundary Subansiri basin with highly complex hydro-meteorological parameters have simulated the stream flow and sediment flux from the basin at various time scales with a high degree of performance indicators, even with limited data availability due to its trans-boundary nature. This study also is a first of its kind for whole of the Subansiri basin and is expected to contribute immensely to better planning, design, and development of water resources of the basin. The application of a semi-distributed conceptual model in the form of snowmelt runoff model (SNOWMOD) and a data based model in the form of artificial neural network models has also provided a comparison of the two techniques for modelling rainfall-runoff response in a large Himalayan basin. Although, SNOWMOD performs inferior compared to ANN model, the fact that it is a simulation method based around physics-based conceptualisations and estimates all the three components of runoff, it is potentially a more flexible method for different types of application. ANN models are expected to perform better for real-time forecasting. The study of climate variability within the Subansiri basin which originates in the Himalayan region, also known as "the roof of the world" has provided a critical analysis on the latest trends of rainfall and temperature in the basin. No such study for this basin has been reported so far. Furthermore, a study of the probable impact of climate change on the hydrological regime of Subansiri basin has been done which reflects the state of xv stream flow including snowmelt in an enhanced temperature scenario. It is now well known that climate change poses a real threat to the Himalayan region and its large rivers and to the inhabitants of their basins. The present study aims to provide information for planning of climate change adaptation strategies for the Subansiri sub basin of the Brabmaputra river.