Incorporating long-term numerical weather forecasts to quantify dynamic vulnerability of irrigation supply system: A case study of Shihmen Reservoir in Taiwan
Journal
Agricultural Water Management
Journal Volume
306
Start Page
109178
ISSN
0378-3774
Date Issued
2024-12-20
Author(s)
Chia-Chuan Hsu
Abstract
Effective water resource management necessitates a comprehensive understanding of water supply and demand, as well as precise assessments of their balance. This is particularly critical as climate change intensifies, often resulting in water scarcity for many systems, especially in agriculture, where demands are substantial and time-sensitive. This study introduces a comprehensive framework for the dynamic vulnerability analysis of water resources, integrating long-term meteorological forecasts up to 6 months ahead. Techniques including K-nearest neighbors (KNN) for spatial downscaling, Long Short-Term Memory (LSTM) networks for simulating rainfall runoff, and Monte Carlo simulations (MCS) for analyzing system vulnerability are employed. By continuously updating model variables and incorporating uncertainties with historical hydrological and meteorological data, this approach enables the analysis of failure probabilities within the water supply system over a 6-month period. This model is applied to Taoyuan's Shihmen Reservoir area in northern Taiwan, under historical conditions and crop rotation scenarios, respectively. The simulation of historical conditions shows successful predictions of 6 out of 7 historical drought-induced irrigation stoppages during 2000–2021. Through dynamic analysis, the model provides advanced irrigation system vulnerability forecasts, and suggests that the crop rotation strategies can significantly enhance irrigation assurance even in dry years, and thus improving the resilience of irrigation systems. This study equips decision-makers with the ability to understand and manage forecast uncertainties, facilitating timely and effective water resource management planning.
Subjects
Agricultural water demand
Dynamic vulnerability analysis
Long-term weather forecasting
Machine learning
Water resources management
Publisher
Elsevier BV
Description
Article number: 109178
Type
journal article