Non-stationary modeling of wet-season precipitation over the Inner Mongolia section of the Yellow River basin

Climate system anomalies and intensified human disturbances have different impacts on the hydrological cycle at different scales, chiefly leading to prominent spatio-temporal heterogeneities in precipitation distributions. It is of great significance to accurately detect the non-stationary changes of precipitation. Focusing on the Inner Mongolia section of the Yellow River basin which is considered as the key ecological barrier in northern China, the best probabilistic model with time-varying moments was established to fit the wet-season precipitation series from 1988 to 2017 for 38 meteorological stations. Considering four three-parameter distributions, time was used as covariate to describe the linear or nonlinear change of each parameter, and model optimization was performed by Akaike Information Criterion. Combined with conventional methods including the Trend-Free Pre-Whitening Mann–Kendall trend test and Sen’s slope estimator, the non-stationary behaviors of wet-season precipitation variability were quantitatively captured. Results showed that the generalized gamma distribution performed best in fitting the wet-season precipitation series in the study area, characterized with high skewness and heavy tails. The non-stationary characteristics of the wet-season precipitation were obvious in most areas during the past 30 years, especially in the central region. The non-stationarities of wet-season precipitation manifested in a downward trend in the mean value, an increase in dispersion degree and significant changes in distribution shape with time, consequently adding the uncertainty to wet-season precipitation process and raising the risk of extreme conditions. The findings of this study provide scientific references to water resources management, drought resistance, and disaster reduction.