Simulation of Design Flood Discharge under Projected Land Cover Scenarios Using ANN–MOLUSCE and HEC-HMS in the Cijangkelok Watershed

Overview

This study quantifies the impact of projected land cover changes on design flood discharge in the Cijangkelok Watershed through integrated hydrologic and spatial modeling. Land cover transitions between 2009 and 2022 are documented and extended to 2035 using machine learning-based prediction, with hydrologic parameters derived from these scenarios and applied to rainfall-runoff simulations.

Methods and approach

Land cover classification data spanning 2009 and 2022 were obtained from governmental sources. Projected 2035 land cover was modeled using QGIS MOLUSCE with an artificial neural network algorithm. Curve Number, impervious surface fraction, and initial abstraction values were extracted and parameterized for each temporal scenario. Design flood discharge was simulated using HEC-HMS with both SCS and Snyder Unit Hydrograph methods. Model validation employed overall classification error assessment and Kappa coefficient analysis for land cover prediction, while design flood simulation accuracy was evaluated using PBIAS and Nash-Sutcliffe efficiency metrics.

Key Findings

Land cover modeling achieved minimum overall error of 0.0332 and Kappa coefficient of 0.765. Dominant land transitions by 2035 involved conversion from dryland to rice paddies, built-up areas, and plantation forests. Composite Curve Number increased progressively from 67.9 (2009) to 68.0 (2022) to 68.4 (2035). Impervious surface fraction rose from 5.6% to 5.7% to 6.4%, while initial abstraction declined from 24.0 to 23.9 to 23.5 mm. Design discharge simulations produced peak flows of 617.2, 623.8, and 641.3 m³/s using SCS methodology, with Snyder methodology yielding 621.3, 621.6, and 630.5 m³/s respectively. Statistical validation generated PBIAS values of 0.1–0.2 and NSE of 1.0, indicating very good model performance. Projected discharge increases of 1.1–2.8% represent moderate changes attributable to land cover transformation.

Implications

Incremental increases in hydrologic parameters and design discharge reflect the cumulative effect of land cover changes on catchment water partitioning. The modest magnitude of discharge escalation despite documented land conversion indicates that conversion to paddy rice cultivation mitigates runoff generation through temporary ponding and infiltration mechanisms, partially offsetting the hydrologic effects of increased impervious surfaces and reduced forest cover. Results suggest land cover composition and functional hydrologic properties warrant differentiation in flood risk assessment rather than area-based metrics alone.