Source apportionment and evolutionary characteristics of groundwater in the alpine gorge region of the tibetan plateau based on unsupervised machine learning and isotopic tracing: a case study of the zheduotang area, kangding

Groundwater in the alpine canyon regions of the Qinghai-Tibet Plateau is influenced by multiple recharge sources and geological structures, resulting in complex recharge processes and evolution mechanisms. This study focuses on the alpine canyon region of Zheduotang and integrates unsupervised machine learning with isotope tracers (δD, δ¹⁸O, and ³H) to systematically identify groundwater sources and their evolutionary processes. By integrating self-organizing maps (SOM) and K-means clustering, the water samples are classified into four groups, highlighting variations in hydrochemical facies, mineralization levels, and dominant ion sources among different water bodies. The results indicate that groundwater hydrochemical evolution is jointly governed by lithology and water-rock interaction intensity. Groundwater in the granite-dominated area is characterized by low-TDS Ca-HCO₃ and Ca-Mg-HCO₃, suggesting rapid shallow recharge, whereas groundwater in the metamorphic sandstone–slate area exhibits higher TDS with enriched Ca²⁺, Mg²⁺, and HCO₃⁻, indicative of longer flow paths and stronger water-rock interactions. Isotopic evidence indicates that groundwater and surface water are predominantly recharged by atmospheric precipitation and snow-ice meltwater, with recharge elevations ranging from 3594 to 5103 m, typically 400–800 m higher than their corresponding discharge elevations. Groundwater recharge in the study area occurs through three principal modes: direct recharge via high-elevation fractures, sustained input from alpine glacial lakes, and river leakage, and is strongly controlled by geological structures overall.