Ecosystem water use efficiency (WUE) is a key indicator for understanding the response of carbon–water processes to environmental changes. However, as affected by uncertainty of WUE estimation and attribution, the individual contributions of environmental drivers to WUE changes and the underlying mechanisms remain unclear on the Tibetan Plateau (TP). Here, a theory-based analytical WUE model was modified by introducing an optimal stomatal behavior model, and used to estimate monthly WUE and their drivers during 1982–2018 for main vegetation types on the Tibetan Plateau (TP). Rationality of three mainstream analytical attribution methods—analytical-based partial derivative method (APDM), regression-based partial derivative method (NPDM), and climate elasticity method (CEM)—was examined. Furtherly, the contributions of various environmental drivers to WUE variation trend were estimated for various vegetation types. Results indicate that: (1) The modified model performed well in estimating WUE against with observed WUE at five eddy-covariance (EC) flux stations, which decreased the uncertainty of WUE estimation. WUE increased significantly in the TP with a slope of 0.49 × 10-2 g C kg−1 H2O yr−1 during 1982–2018 with forests increased most rapidly; (2) Only APDM captured both the variation and magnitude of WUE trends, and both NPDM and CEM failed to reproduce the magnitude of WUE trends, though CEM was better than NPDM benefited from its nonlinear structure. Through comparing these methods, we found that the nonlinear direct effects of these forcing variables on WUE change could be higher than their indirect effects caused by their interactions. (3) There are some differences in the main drivers of WUE trends for various vegetation types. Increase in WUE was mainly driven by the increase in leaf area index (LAI) for both steppes and meadows, while decreased air water vapor pressure deficit (VPD) and elevated CO2 showed higher positive effects for broad-leaf forests. The distinct response and extremely scarce EC observations of forests suggest that more attention should be paid to forests in the TP. This study could help in selection of attribution methods and understanding the mechanisms of carbon–water coupling of alpine ecosystems under a changing climate.