Solute concentrations and accumulation in shallow groundwater can trigger a negative domino effect of environmental and economic problems. Accurately tracking the solute chemistry of groundwater is essential for assessing their adverse impacts and pinpointing hotspots of contaminants that require conservation measures. However, existing mapping methods have been greatly limited by the peculiarities of shallow groundwater, such as its challenging hydraulic connection, uneven distribution, and complex driving factors prevalent in arid regions. To address these challenges, we designed a novel framework that integrates species distribution models (SDMs) with traditional hydrological models and advanced machine learning algorithms to predict the spatial distribution of groundwater solutes in a multidisciplinary effort. We carried out a systematic collection of shallow groundwater, deep groundwater, and surface water samples from three adjacent hydrological units in the arid regions of northwest China. By employing the SDMs framework originally utilized in ecology to assess biological species suitability, we could simulate and predict solute concentrations in groundwater. The results emphasized that solutes in surface water were important variables in the final models for Na+, K+, SO42-, and Cl-, while deep groundwater influenced Ca2+. In addition, integrating predictor variables into the SDMs enabled the discovery of additional valuable information. Our results highlighted the improved power for mapping groundwater by combining SDMs with multidimensional driving factors. This novel framework could not only clarify the solute movement mechanisms but also reveal their spatial patterns via a transdisciplinary approach, offering a versatile tool for groundwater management and policies. Plain Language Summary This study examines solute contamination in shallow groundwater,
which can lead to serious environmental and economic issues. Accurately tracking solutes is essential for understanding their impact and identifying areas that need protection. However, current methods are constrained by the sparse field samples of shallow groundwater, including its uneven distribution, data collection difficulties, and complex movement patterns, especially in arid regions. To address these challenges, we developed a new approach combining species distribution models (SDMs), hydrological models, and machine learning. This method improves predictions of solute distribution in groundwater with limited water samples from shallow and deep groundwater, as well as surface water, in northwest China. Using the SDM framework, we successfully predicted solute concentrations in shallow groundwater, and found that surface water influenced solutes like Na+, K+, SO4, and CI, while deep groundwater affected Ca2+. Integrating multiple factors into the model provided new insights into the complex interactions between groundwater and surface water. Our proposed innovative framework enhances the mapping of groundwater contamination and offers a valuable tool for groundwater management, especially in water-scarce regions.