资源环境科技发展态势分析平台

Comprehensive characterization of diversity in global patterns of precipitation variability and change is an important starting point for climate adaptation and resilience assessments. Capturing the nature of precipitation probability distribution functions (PDF) is critical for assessing variability and change. Conventional linear regression-based analyses assume that slope coefficients for the wet and dry tails of the PDF are consonant with the conditional mean trend. This assumption is not always borne out in the analyses of historical records. Given the relationship between sea surface temperature (SST) and precipitation, recent trends in global SST complicate interpretations of precipitation variability and risk. In this study, changes in the PDF of annual precipitation (1951-2011) at the global river basin scale were analyzed using quantile regression (QR). QR is a flexible approach allowing for the assessment of precipitation variability conditioned on the leading empirical orthogonal function (EOF) patterns of global SST that reflect El Nio-Southern Oscillation and Atlantic Multi-decadal Oscillation. To this end, the framework presented (a) offers a characterization of the entire PDF and its sensitivity to the leading modes of SST variability, (b) captures a range of responses in the PDF including asymmetries, (c) highlights regions likely to experience higher risks of precipitation excesses and deficits and inter-annual variability, and (d) offers an approach for quantifying risk across specified quantiles. Results show asymmetric responses in the PDF in all regions of the world, either in single or both tails. In one instance, QR detects a differential response to the leading patterns of SST in the Tana basin in eastern Africa, highlighting changes in variability as well as risk.