Significant differences in terrestrial water storage estimated by four common methods

The Yellow River headwaters, located in the northeastern part of the Tibetan Plateau. Terrestrial water storage can be estimated by multiple approaches. However, the limited quantification of these methods regarding terrestrial water storage stocks limits the assessment of their applicability. Here, we quantified and compared water storage and its spatial patterns by four common methods: SWAT, InVEST, WB (based on water balance theory), and RSI (remotely sensed inversion). The results showed that SWAT, InVEST, and WB captured remarkable spatial heterogeneity of water storage, with CV (coefficient of variation) being 57.8 %, 41.2 %, and 85.2 %, respectively, whereas the CV of RSI was only 12.5 %, with WB exhibited the most spatial heterogeneity. RSI showed a pronounced distinct spatial pattern compared to the other three methods. Precipitation and NDVI (p < 0.01) are the common main drivers for all methods except RSI. The discrepancies in water storage can be attributed to the differences in models or methods response to influencing factors, e.g., the effects of topography and land use on water storage are considered to varying degrees. The biases or errors in the average water storage caused by different methods across various years range from 90.3 mm to 136.3 mm. Consequently, it is critical to consider the applicability of the methodology, especially considering different climatic, land use, soil, and topographic environments. • Differences in water storage (WS) estimated by different methods were compared. • The biases or errors of WS estimated by different methods are up to 136.3 mm. • WS estimated by water balance reflects the most spatial heterogeneity. • Remote sensing inversion is not suitable for reflecting spatial variability. • Precipitation and NDVI are common driving factors except for remote sensing.