Integration of Watershed eco-physical health through Algorithmic game theory and supervised machine learning

• Published in: Groundwater for sustainable development Vol. 26; p. 101216
• Main Authors: Nasiri Khiavi, Ali, Tavoosi, Mohammad, Khodamoradi, Hamid, Kuriqi, Alban
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.08.2024
• Subjects: Condorcet algorithm; Ecological sustainability; Integrated watershed management; Water scarcity; Watershed health; Integrated watershed management; Ecological sustainability; Condorcet algorithm; Water scarcity; Watershed health
• ISSN: 2352-801X; 2352-801X
• DOI: 10.1016/j.gsd.2024.101216

The eco-physical health assessment of watersheds is crucial for sustainable water resource management and ecosystem services. This study quantifies the eco-physical health of the Talar watershed in Iran using the geometric mean method (GMM), game-theoretic algorithm (GTA), and machine learning algorithms including Random Forest (RF), Support Vector Machine (SVM), Simple Linear Regression (SLR), and K-Nearest Neighbor (KNN) for distributed and semi-distributed monitoring. The results show that the RF algorithm performed better than other models, as indicated by MAE, MSE, RMSE, and AUC statistics with values of 0.032, 0.003, 0.058, and 0.940, respectively. The watershed health index prioritization with different approaches showed that the pattern of watershed health changes positively from upstream to downstream. Based on the watershed health prioritization using GMM, it can be said that sub-watersheds Int6 and Int5 are the healthiest sub-watersheds in the studied watershed, with values of 0.93 and 0.90, respectively. Based on the watershed prioritization using the GTA approach, it can also be said that sub-watersheds Int6, Int5, and Int01 are the healthiest ones. In the case of the RF algorithm, the average values of the pixels in each sub-watershed showed that sub-watersheds Int6 and Int01 were recognized as the healthiest sub-watersheds with values of 0.91 and 0.88, respectively. Int6 consistently emerged as the healthiest sub-watershed across all methods, attributed to high TWI and NDVI values and low slope, DEM, erosion, and CN values. In general, it can be said that the watershed health index in the studied catchment fully followed the factors affecting the catchment's health and that the spatial patterns of change of this index were consistent with the physiographic and hydroclimatic conditions in all three semi-distributed and distributed approaches. The study's implications underline the importance of multi-criteria and multi-algorithm approaches in accurately assessing and managing watershed health for sustainable development. [Display omitted] •Watershed health index displayed consistent spatial patterns across various approaches.•Watershed health changes positively from upstream to downstream.•Watershed health index is consistent with physiographic and hydroclimatic conditions.•RF algorithm outperformed other models with superior MAE, MSE, RMSE, and AUC statistics.