Assessing global passive microwave soil moisture retrievals in mountainous terrain: insights from in situ validation and extended triple collocation

• Advanced ETC-based evaluation framework for mountain soil moisture retrievals. • Exposed terrain- and vegetation-dependent biases among SMAP, SMOS, and AMSR2. • Quantified dominant controls on retrieval sensitivity across environmental gradients. • Offered guidance for next-generation, terrain-aware soil moisture algorithms. Soil moisture (SM) conditions are closely linked to water availability, energy balance, and ecosystem processes, highlighting the importance of accurate SM data for hydrological modeling and environmental monitoring. Passive microwave products offer large-scale SM estimates, but their validation in mountainous areas remains difficult due to sparse in situ data and complex terrain. This study systematically evaluated three global passive microwave soil moisture products—SMAP, SMOS, and AMSR2—across mountainous regions by integrating in situ observations from the International Soil Moisture Network (ISMN) with the Extended Triple Collocation (ETC) method. Results indicated that SMAP and SMOS outperformed AMSR2 in both in situ and ETC-based evaluations, with higher median correlation coefficients (0.95 for SMAP, 0.84 for SMOS) and lower RMSEs (0.026 and 0.048 m3/m3, respectively). In the Rocky Mountains, where station density is highest, over 86% of sites showed consistent product rankings between in situ and ETC results, confirming the reliability of ETC in data-scarce areas. Further analysis revealed contrasting environmental controls: SMAP’s error increased with slope and vegetation density, while its sensitivity increased with elevation and surface roughness. SMOS showed more stable performance across gradients, whereas AMSR2 was more affected by terrain and vegetation complexity. These findings clarify uncertainty patterns and environmental controls on passive microwave soil moisture retrievals in mountainous regions, and provide insights for product selection, algorithm improvement, and data fusion in hydrological and ecological applications.