Abstract

Present study assesses the effect of finer land-use classification in simulating the rainfall-runoff response of Kangsabati reservoir catchment (3,627 km²) and command (7,112 km²) by considering cropland heterogeneity in variable infiltration capacity (VIC) model. High resolution LISS-IV satellite imageries were used for the land-use classification. Global sensitivity analysis was performed using VIC-ASSIST to identify the most and least influential parameters based on the sensitivity index of elementary effects. A fully distributed calibration approach was employed using 16 (detailed) and 8 (lumped) vegetation classes. Low flows during lean periods were over-estimated and peak flows were under-estimated by both the model setups at Kangsabati reservoir site. Detailed land-use classification resulted in the reduction in streamflow over-estimation (Percent Bias (PBIAS) from −20.99 to −14.41 during calibration and from –22.83 to –7.17 during validation) at daily time step. It further demonstrates the improvement in simulating the peak flows; hence, highlighting the importance of detailed land-use classification for vegetation parameterization in VIC model setup. River discharge regulation at Kangsabati reservoir resulted in poor model performance at Mohanpur, downstream site of Kangsabati reservoir. Therefore, calibration for Mohanpur was performed after updating the VIC simulated streamflow with routed reservoir spillage using Hydrologic Engineering Center-River Analysis System (HEC-RAS) model. Streamflow updation employing HEC-RAS at Mohanpur improved the modelling efficiency (Nash–Sutcliffe efficiency (NSE) from 0.50 to 0.65 during calibration and from 0.55 to 0.67 during validation) and reduced bias (PBIAS from 6.25 to –2.23 during calibration and from 15.06 to 7.40 during validation) considerably for daily flows. Model performance with reasonable accuracy was achieved at both the calibration locations which demonstrates the potential applicability of VIC model to predict streamflow in the monsoon dominated Kangsabati reservoir catchment and command.

Highlights

• LISS-IV satellite imageries were classified using ground truth survey data obtained for different crop types in the study area

• Crop specific vegetation parameterization was used in setting up VIC modeling framework

• Modeling efficacy was assessed for two vegetation parameterization schemes using single crop type and multiple crops

• Global sensitivity analysis and fully distributed automatic calibration was performed using VIC-ASSIST software package

• Utility of HEC-RAS was shown in routing reservoir spillage to the downstream gauging point in VIC modeling framework in the absence of integrated reservoir module

中文翻译：

---

利用VIC模型模拟季风为主的水库集水区的水文响应与异种种植模式。

摘要

本研究评估了更精细的土地利用分类在模拟康萨巴蒂水库集水区（3,627 km ²）和命令区（7,112 km ²）的降雨径流响应中的作用。），并在可变渗透能力（VIC）模型中考虑了农田的异质性。高分辨率的LISS-IV卫星图像用于土地用途分类。使用VIC-ASSIST进行全局敏感性分析，以根据基本效应的敏感性指数确定影响最大和最小的参数。使用16种（详细）和8种（集总）植被类别的完全分布式校准方法。康萨巴蒂水库站点的两种模型设置都过高地估计了稀水期的低流量和峰值流量。详细的土地利用分类导致每日时间步长的流量过高估计（校准期间的偏差百分比（PBIAS）从-20.99降低到-14.41，而在验证期间从–22.83降低到–7.17）。它进一步证明了在模拟峰值流量方面的改进。因此，在VIC模型设置中强调了详细的土地利用分类对于植被参数化的重要性。康萨巴蒂水库的河流流量调节导致康萨巴蒂水库下游站点Mohanpur的模型性能不佳。因此，使用水文工程中心-河流分析系统（HEC-RAS）模型更新了VIC模拟流，并通过路由的水库溢流进行了Mohanpur的校准。在Mohanpur使用HEC-RAS进行的流量更新将建模效率（校准期间的纳什-萨克利夫效率（NSE）从0.50降低至0.65，验证期间从0.55降低至0.67）和减小的偏差（PBIAS从校准期间的PBIAS从6.25降低至–2.23，从15.06降低至–15。 7。验证期间为40），对于每日流量而言相当可观。在两个校准位置均获得了具有合理精度的模型性能，这表明VIC模型可用于预测季风主导的康萨巴蒂水库集水区和指挥部的流量。

强调

• 使用研究区域内不同作物类型获得的地面真相调查数据对LISS-IV卫星图像进行分类

• 作物特定植被参数化用于建立VIC建模框架

• 使用单一作物类型和多种作物对两种植被参数化方案的建模效果进行了评估

• 使用VIC-ASSIST软件包执行全局灵敏度分析和完全分布式的自动校准

• 在没有集成储层模块的情况下，HIC-RAS的效用在VIC建模框架中将储层溢流路由至下游测量点时得到了展示