Scholarlink

<p>Reliable streamflow prediction in humid tropical basins is challenging due to multiple sources of uncertainty. This study aims to develop an integrated framework to simultaneously quantify and evaluate the effects of input, parameter, terrain, and structural uncertainties. Affecting daily streamflow prediction in the Chaliyar River Basin, a representative humid tropical river system in Kerala, India. Daily streamflow records (1988–2005) were used for model calibration (1988–2000) and validation (2001–2005). Precipitation gaps were addressed using Kalman smoothing with ARIMA. Parameter uncertainty was assessed through three algorithms: Sequential Uncertainty Fitting (SUFI-2), Particle Swarm Optimization (PSO), and Generalized Likelihood Uncertainty Estimation (GLUE). Terrain uncertainty was evaluated using SRTM (30&#xa0;m), ASTER (30&#xa0;m), and CARTOSAT (10&#xa0;m) Digital Elevation Models (DEMs). Structural sensitivity was tested using 300, 500, and 1000 calibration iterations. Streamflow predictions were evaluated with the coefficient of determination (R<sup>2</sup>) and Nash–Sutcliffe efficiency (NSE).The base SWAT model showed moderate performance for&#xa0;daily streamflow&#xa0;prediction&#xa0;(calibration: <i>R</i><sup>2</sup> = 0.58, NSE = 0.52; validation: <i>R</i><sup>2</sup> = 0.51, NSE = 0.42). Integrating input uncertainty reducing techniques (Kalman–ARIMA) and parameter optimization improved predictions: GLUE&#xa0;(calibration/validation <i>R</i><sup>2</sup>/NSE: 0.81/0.78, 0.68/0.66), PSO (0.80/0.79, 0.70/0.70), and SUFI-2 (0.82/0.81, 0.73/0.72).&#xa0;Compared to the base model, SUFI-2 achieved significant improvements of 56% in calibration NSE (from 0.52 to 0.81) and&#xa0;71% in validation NSE (from 0.42 to 0.72) for&#xa0;daily streamflow predictions.&#xa0;Cartosat&#xa0;DEM (10&#xa0;m) outperformed&#xa0;SRTM and ASTER (30&#xa0;m), with base&#xa0;model NSE values of 0.67/0.58 versus 0.52/0.42 and&#xa0;0.50/0.40, respectively. The optimal configurationSUFI-2 with CARTOSAT and Kalman–ARIMA achieved calibration <i>R</i><sup>2</sup> = 0.82, NSE = 0.81 and validation <i>R</i><sup>2</sup> = 0.74, NSE = 0.73.Results demonstrate that integrated uncertainty quantification combining high-resolution DEMs, advanced gap-filling, and SUFI-2 optimization significantly enhances SWAT performance in&#xa0;humid tropical basins, providing a robust framework for&#xa0;flood forecasting and&#xa0;water resource planning.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantification of uncertainties in the hydrological models for streamflow prediction in a humid tropical river basins in India: a case study of the Chaliyar river basin, Kerala

  • 期刊论文

  • 发表日期：

    2026年5月6日

查看全文
  • K. Naveena,
  • P. V. Shahanas,
  • T. M. Sharannya,
  • Baheerah Shada,
  • T. K. Drissia,
  • Chinni V. Naga Kumar Kommireddi,
  • U. Surendran,
  • N. K. Rajesh Kumar

摘要

Reliable streamflow prediction in humid tropical basins is challenging due to multiple sources of uncertainty. This study aims to develop an integrated framework to simultaneously quantify and evaluate the effects of input, parameter, terrain, and structural uncertainties. Affecting daily streamflow prediction in the Chaliyar River Basin, a representative humid tropical river system in Kerala, India. Daily streamflow records (1988–2005) were used for model calibration (1988–2000) and validation (2001–2005). Precipitation gaps were addressed using Kalman smoothing with ARIMA. Parameter uncertainty was assessed through three algorithms: Sequential Uncertainty Fitting (SUFI-2), Particle Swarm Optimization (PSO), and Generalized Likelihood Uncertainty Estimation (GLUE). Terrain uncertainty was evaluated using SRTM (30 m), ASTER (30 m), and CARTOSAT (10 m) Digital Elevation Models (DEMs). Structural sensitivity was tested using 300, 500, and 1000 calibration iterations. Streamflow predictions were evaluated with the coefficient of determination (R2) and Nash–Sutcliffe efficiency (NSE).The base SWAT model showed moderate performance for daily streamflow prediction (calibration: R2 = 0.58, NSE = 0.52; validation: R2 = 0.51, NSE = 0.42). Integrating input uncertainty reducing techniques (Kalman–ARIMA) and parameter optimization improved predictions: GLUE (calibration/validation R2/NSE: 0.81/0.78, 0.68/0.66), PSO (0.80/0.79, 0.70/0.70), and SUFI-2 (0.82/0.81, 0.73/0.72). Compared to the base model, SUFI-2 achieved significant improvements of 56% in calibration NSE (from 0.52 to 0.81) and 71% in validation NSE (from 0.42 to 0.72) for daily streamflow predictions. Cartosat DEM (10 m) outperformed SRTM and ASTER (30 m), with base model NSE values of 0.67/0.58 versus 0.52/0.42 and 0.50/0.40, respectively. The optimal configurationSUFI-2 with CARTOSAT and Kalman–ARIMA achieved calibration R2 = 0.82, NSE = 0.81 and validation R2 = 0.74, NSE = 0.73.Results demonstrate that integrated uncertainty quantification combining high-resolution DEMs, advanced gap-filling, and SUFI-2 optimization significantly enhances SWAT performance in humid tropical basins, providing a robust framework for flood forecasting and water resource planning.