资源环境科技发展态势分析平台

Precipitation in Xinjiang, China, was modelled with covariates, such as time and climate indices, using quantile regressions. Compared to a frequentist quantile regression, a Bayesian quantile regression tended to generate smoother and narrower band confidence intervals of quantile regression coefficients, especially at extremely high and low quantile levels. A full picture of temporal trends at quantile levels from 0.01 to 0.99 indicates that the wet season (May to August) precipitation in Northern Xinjiang and the western regions of Southern Xinjiang showed statistically significant increases with different magnitudes over all quantile levels. However, the wet season precipitation in South-eastern Xinjiang decreased at some quantile levels. The Eastern Atlantic/Western Russia (EAWR) pattern was the most significant large-scale climate pattern that influenced wet season precipitation when compared to other studied patterns, i.e. the El Nino-Southern Oscillation (ENSO), the Atlantic Multidecadal Oscillation (AMO), the Pacific Decadal Oscillation (PDO), the Northern Oscillation (NO), the Arctic Oscillation (AO) and the North Atlantic Oscillation (NAO). The quantile regression coefficients associated with the EAWR index positively increased from low to high quantile levels. The ENSO significantly affected the extremely high wet season precipitation in Xinjiang. El Nino increased and La Nina decreased wet season precipitation in Northern Xinjiang, with different magnitudes at different quantile levels. Other climate patterns, i.e. the AMO, PDO, NO, NAO and AO, did not evidently affect the wet season precipitation conditional on the ENSO and EAWR. These findings suggest that the predictability of seasonal precipitation over Xinjiang can be improved by incorporating indices associated with the ENSO and EAWR as model predictors.