摘要
The Ganga River, the longest river in India, is stressed by extreme anthropogenic activity and climate change, particularly in the Varanasi region. Anticipated climate changes and an expanding populace are expected to further impede the efficient use of water. In this study, hydrological modeling was applied to Soil and Water Assessment Tool （SWAT） modeling in the Ganga catchment, over a region of 15 621.612 km2 in the southern part of Uttar Pradesh. The primary goals of this study are： ① To test the execution and applicability of the SWAT model in anticipating runoff and sediment yield; and ② to compare and determine the best calibration algorithm among three popular algorithms-sequential uncertainty fitting version 2 （SUFI-2）, the generalized likelihood uncertainty estimation （GLUE）, and par-allel solution （ParaSol）. The input data used in the SWAT were the Shuttle Radar Topography Mission （SRTM） digital elevation model （DEM）, Landsat-8 satellite imagery, soil data, and daily meteorological data. The watershed of the study area was delineated into 46 sub-watersheds, and a land use/land cover （LULC） map and soil map were used to create hydrological response units （HRUs）. Models utilizing SUFI- 2, GLUE, and ParaSol methods were constructed, and these algorithms were compared based on five cat-egories： their objective functions, the concepts used, their performances, the values of P-factors, and the values of R-factors. As a result, it was observed that SUFI-2 is a better performer than the other two algo-rithms for use in calibrating Indian watersheds, as this method requires fewer runs for a computational model and yields the best results among the three algorithms. ParaSol is the worst performer among the three algorithms. After calibrating using SUFI-2, five parameters including the effective channel hydraulic conductivity （CH_K2）, the universal soil-loss equation （USLE） support parameter （USLE_P）, Manning＇s n value for the main channel （CH_N2）, the surface runoff lag time （SURLAG）, and the available water capac-ity of the soil layer （SOL_AWC） were observed to be the most sensitive parameters for modeling the pre-sent watershed. It was also found that the maximum runoff occurred in sub-watershed number 40 （SW#40）, while the maximum sediment yield was 50 t.a ^1 for SW#36, which comprised barren land. The average evapotranspiration for the basin was 411.55 mm.a ^1. The calibrated model can be utilized in future to facilitate investigation of the impacts of LULC, climate change, and soil erosion.
恒河是印度最长的河流,目前正遭受人类活动和气候变化的影响,其中瓦拉纳西区域所受的影响尤为严重。预计气候变化和不断扩张的人口会进一步影响水资源的有效利用。本研究利用水土评价工具(soil and water assessment tool,SWAT)模型进行水文模拟,研究区选择恒河流经的印度北方邦南部地区,覆盖面积达15 621.612 km^2。本研究主要目标如下:(1)检验SWAT模型在径流和产沙量预测方面的可行性和适用性,(2)对序贯不确定性分析(SUFI-2)、普适似然不确定性估计(GLUE)和并行求解(ParaSol)这3种常规校准算法进行对比分析,并确定出最佳校准算法。SWAT模型中输入数据的来源为航天飞机雷达地形测绘任务(SRTM)获取的数字高程模型(DEM)、Landsat-8卫星图像、土壤数据和逐日气象数据。首先,本研究将研究流域划分为46个子流域,使用土地利用/土地覆被(LULC)图和土壤图创建了水文响应单元(HRU)。然后,构建了基于SUFI-2、GLUE和ParaSol算法的模型,并基于5类标准对这些校准算法进行了比较,标准包括目标函数、使用概念、性能、P因子和R因子值。研究结果表明:(1) SUFI-2算法需要的运行时间较少,同时校准结果最好,ParaSol算法最差;(2)使用SUFI-2算法对模型进行校准后,获得对模拟结果影响最显著的5个敏感参数——主河道水力传导率(CH_K2)、USLE方程水土保持因子(USLE_P)、主河道曼宁系数值(CH_N2)、地表径流滞后时间(SURLAG)以及土壤有效含水量(SOL_AWC);(3) 40号子流域(SW#40)产生的径流量最大,土地利用类型为荒地的36号子流域(SW#36)的产沙量最大,达到50 t·a^(–1)。研究区的平均蒸散量为411.55 mm·a^(–1)。校准后的模型可以用于探究未来LULC的变化、气候变化和土壤侵蚀的影响。
