Integrating Upstream–Downstream Connectivity and Analytical Methods to Prioritize a Potential High-Risk Watershed for Iodine Monitoring

Olumide Ajulo; Kelly Good; John Sivey; Alina Ebling; Shital Vaidya

Understanding how upstream activities influence downstream drinking-water quality is critical for watershed management and public health protection. This study builds on the sub-basin flow-path framework developed by Good et al. (2025) to identify a high-priority drinking water treatment plant (DWTP) and associated source water for iodine monitoring. Using watershed connectivity and flow path tracing, we identified watersheds with potential contributions that may affect disinfection byproduct (DBP) precursor inputs into DWTP intakes. Unlike existing iodine and bromide studies that relied on random site selection or existing concentration data, our study links hydrologic flow paths and watershed characteristics to source water vulnerability. The selected DWTP serves as the focus for ground-truthing this prioritization through field sampling and laboratory analysis. Analytical methods include ion chromatography (IC) for iodide and iodate speciation and inductively coupled plasma triple quadrupole mass spectrometry (ICP-QQQ) for total iodine quantification. Together, these measurements will evaluate whether upstream source contributions align with the predicted high-priority ranking. This in-progress work establishes a defensible, watershed-based framework that connects hydrologic source areas, chemical monitoring, and drinking water management. The approach demonstrates how flow-informed site selection and advanced analytical methods can be integrated to improve understanding of I-DBP precursor dynamics and support proactive source water protection and treatment strategies.

Using Mental Models to Bridge the Gap Between Human and Natural Systems: An Application to Land and Water Management in a Transboundary Arid Basin

Raquel Neri-Barranco; Saurav Kumar

Managing water resources in coupled human–natural systems (CHNS) remains a persistent challenge, especially in transboundary arid regions where social, political, and ecological boundaries overlap. To address this complexity, we introduce AIMM (Artificial Intelligence-driven Mental Modeler)—an open-source software platform that integrates stakeholder knowledge with empirical time-series data to support watershed management and decision-making. AIMM enables users to construct mental models representing system components and hypothesized causal relationships, while an embedded machine learning engine based on Dynamic Double Machine Learning (EconML) estimates the direction and magnitude of these connections using historical data.

An initial application focused on the El Paso–Ciudad Juárez region, where rapid expansion of pecan orchards is transforming land use and intensifying water demand in an already water-scarce basin. Five academic stakeholders constructed mental models identifying key drivers and feedbacks influencing agricultural expansion and water availability. AIMM quantified the structure of each model through graph-theoretic indices (e.g., node density, centrality, and hierarchy) and compared them using a modified distance ratio metric to assess areas of consensus and divergence. A composite model was generated through matrix aggregation, providing a transparent, data-grounded representation of stakeholder understanding.

This integrative framework demonstrates a scalable approach for linking qualitative stakeholder perspectives with quantitative hydrologic and land-use data, offering a novel pathway to co-produce actionable knowledge for watershed management. Future work will extend AIMM to include a broader group of regional actors—government agencies, NGOs, and producers—to identify shared priorities and intervention points for sustainable land and water governance. By combining cognitive and empirical insights, AIMM advances inclusive, evidence-based decision support for managing complex socio-environmental systems.

Climate-Driven Landslide Susceptibility Modeling Using Slope Units and Machine Learning in a Himalayan Watershed

Tulasi Ram Bhattarai; Netra Prakash Bhandary; Kalpana Pandit

Landslide incidence in the Nepal Himalaya is increasing as changes in rainfall intensity and seasonality influence slope stability across steep watersheds. Effective prediction of these hazards requires modeling approaches that capture the landscape's geomorphic structure and can integrate future climate information. This study assesses current and future landslide susceptibility in the Seti River Basin by integrating slope-unit–based terrain segmentation with a Random Forest model trained on 11 conditioning factors representing topographic, hydrological, geological, and climatic conditions. Model performance for the historical period showed strong agreement between predicted susceptibility and observed landslide locations, indicating that slope units provide a coherent spatial framework for watershed-scale hazard analysis. Future precipitation inputs were derived from the ensemble mean of 32 CMIP6 climate models under the SSP2–4.5 and SSP5–8.5 scenarios for the historical, near-future, and far-future periods. The baseline susceptibility map identified high-risk zones aligned with deeply incised valleys and rapidly uplifting terrain. Future projections indicate an expansion of moderate and high susceptibility classes, driven by increased seasonal and extreme rainfall, while the very high susceptibility class shows a slight decline. This suggests that the hazard may diffuse over a larger portion of the watershed rather than intensify at existing hotspots. These results demonstrate that climate-informed, slope-unit-based machine learning models can improve the detection of evolving landslide patterns in mountainous watersheds. The approach offers a transferable framework for regions facing similar hydroclimatic stress and supports long-term watershed management planning under changing environmental conditions.