Improving Explainable AI in Attributing Hydrological Responses to Climate Variabilities in Snow-dominated Watersheds – A new publication by Jinyu Hui, Dr. Adam Wei, and Dr. Yiping Hou.
Hydrological attribution in the era of AI: A new framework for understanding climate-runoff relationships
Watershed Ecosystems Project researchers Jinyu Hui and Dr. Adam Wei are co-authors on a new study highlighting how explainable AI improves attribution of hydrological responses to climate variability in snow-dominated watersheds. In the era of AI, predicting runoff under a changing climate is no longer the biggest challenge – explaining why they respond that way has become the next frontier. In a new paper, “Improving explainable AI in attributing hydrological responses to climate variabilities in snow-dominated watersheds” published in Applied Soft Computing, Watershed Ecosystems Project researchers proposed an explainable AI framework (CF-BootSVR) to make hydrological attribution more transparent, efficient, and reliable.
Applied to the Upper Penticton Creek watershed (241) and the Greata watershed, CF-BootSVR successfully uncovered the climate drivers behind runoff changes. For each watershed, it clarified the relative contributions of seasonal water and energy conditions to multiple runoff indicators. In cross-basin comparisons, it identified differences between the two studied watersheds in their hydrological sensitivity to a changing climate.
These results highlight the complexity of hydrological responses under changing environmental conditions.
This study represents an important step forward in advancing explainable AI for hydrological attribution study. By connecting AI-driven model interpretation with real-world hydrological processes, it makes climate-runoff attribution more transparent and adaptable. The framework supports localized watershed decision-making and shows strong potential for ecohydrological attribution studies under changing climate conditions.
Find the full publication here:
Hui, J., Wei, X., Hou, Y., 2025. Improving explainable AI in attributing hydrological responses to climate variabilities in snow-dominated watersheds. Applied Soft Computing183, 113682. https://doi.org/10.1016/j.asoc.2025.113682.
Forest Loss and Landscape Pattern Change Cause Watersheds to Release More Young Water – A new publication by Ming Qui, Xiaohua Wei and Yiping Hou.
Forest loss and landscape change accelerate young water export from watersheds
Watershed Ecosystems Project researcher Dr. Adam Wei is co-author on a new global study published in Proceedings of the National Academy of Sciences examining how forest loss and landscape configuration influence watershed hydrology. The paper, led by Ming Qiu, synthesizes data from 657 watersheds worldwide to better understand how forests regulate the storage and release of precipitation across diverse hydroclimatic settings.
The study finds that forest loss significantly increases the proportion of “young water” in streamflow—meaning recent precipitation moves more rapidly through watersheds, reducing natural water storage capacity and potentially heightening risks related to flash flooding, drought vulnerability, and contaminant transport.
The analysis also demonstrates that hydrological responses depend not only on the amount of forest cover but, critically, on how forests are spatially arranged. In particular, forest edge density can meaningfully influence evapotranspiration and watershed water retention, especially in sparsely forested landscapes.
Together, these findings have implications for forest and watershed management, especially in areas of intensified timber extraction. The study advances a global understanding of forest–water interactions and reinforces the importance of integrated forest conservation and landscape planning for sustaining watershed function under increasing climate and disturbance pressures.
Find the full publication here:
Qiu, M., Wei, X., Hou, Y., 2026. Forest loss and landscape pattern change cause watersheds to release more young water. Proceedings of the National Academy of Sciences 123 (5), e2522502123. https://doi.org/10.1073/pnas.2522502123.
Forest Disturbance Thresholds on Summer Low flows in the Interior of British Columbia, Canada – a new publication by Yiping Hou, Xiaohua Wei, Jinyu Hui, Ziprnh Xu, Ming Qui, Minfang Zang, Qiang Li, and Qi Chen.
Forest disturbance and summer low flows: A new framework to identify “how much is too much”
In British Columbia’s interior, summer low flows are more than a hydrological statistic: they support community water supply and sustain aquatic ecosystems during the driest time of the year. As climate change intensifies drought risk and forest disturbances increase, watershed managers face a pressing question: At what point does forest disturbance begin to trigger significant changes in summer low flows? In our paper published in CATENA, Watershed Ecosystems Project researchers Dr. Adam Wei and Jinyu Hui worked with others toward the development of a practical approach to answer that question by identifying forest disturbance thresholds, above which significant changes in low flows can be detected.
The team introduced a seasonal hydrological response curve method and applied it to 20 forested watersheds in the interior of British Columbia. We found that summer low-flow thresholds spanned a wide range of disturbance levels, from 8% to 52% CECA. These thresholds were shaped by both climate and watershed properties.
Importantly, the hydrological response was not unidirectional; forest disturbance can cause both increases and decreases in summer low flows, depending on the watershed context.
By providing a method applicable to individual watersheds, this study offers a practical tool to support local watershed planning, community water security, and aquatic ecosystem protection under changing climate and disturbance regimes.
Figure 1. The 20 study watersheds in BC.
Find the full publication here:
Hou, Y., Wei, X., Hui, J., Xu, Z., Qiu, M., Zhang, M., Li, Q., Chen, Q., 2024. Forest disturbance thresholds on summer low flows in the interior of British Columbia, Canada. CATENA 243, 108173. https://doi.org/10.1016/j.catena.2024.108173.
