长江流域的非平稳SPI干旱时空特征分析

全球气候变化背景下,由于降水时间序列存在非平稳性,导致利用传统的标准化降水指数(SSPI)估计的干旱可能存在较大偏差。文章基于GAMLSS模型,以气候指数作为解释变量进行参数拟合,建立一种基于可变参数的非平稳伽玛模型,计算非平稳标准化降水指数(NSPI),并与SSPI指数对比分析长江流域1962～2016年干旱时空变化特征,结果表明:(1)NSPI与SSPI变化基本一致,但非平稳伽玛模型比平稳伽玛模型更好地重现降水量以捕获当前全球气候变化背景下的降水变化。(2)长江流域1962～2016年干旱有加重趋势。上游干旱烈度、干旱历时、干旱强度和烈度峰值的变化速率分别为每10年上升0.064、0.041、0.023和0.027,而中下游则分别为0.151、0.089、0.021和0.030;干旱强度以轻至中旱为主。长江源头和四川盆地西南部干旱较严重,而金沙江和雅砻江上游及鄱阳湖南部的干旱相对轻微。(3)与SSPI相比,NSPI估计的相同干旱烈度和干旱历时的重现期较大,且估计的干旱事件相对集中。NSPI的干旱风险表明长江源头、金沙江下游和洞庭湖流域中部是高风险集中区,而川江上游和鄱阳湖东南部是低风险集中区。(4)构建的非平稳伽玛模型估计的NSPI能较好的预测各干旱特征,且对干旱烈度和干旱历时的预测性能更好。

Spatio-temporal Characteristics of Drought in the Yangtze River Basin Using Non-stationary Standardized Precipitation Index

Under the background of global climate change, the drought estimated by the stationary standardized precipitation index (SSPI) may be greatly deviated due to the nonstationarity of precipitation time series. In this study, the non-stationary gamma model with variable parameters is established by fitting the model parameters with climate indices as explanatory variables based on GAMLSS model, and the non-stationary standardized precipitation index (NSPI) is calculated and compared with the SSPI. Then, the spatio-temporal change characteristics of drought in the Yangtze River Basin from 1962 to 2016 are investigated by using the NSPI. The results are as follows: (1) The changes of NSPI and SSPI are basically the same, but the non-stationary gamma model is better than stationary gamma model in reproducing precipitation to capture precipitation changes under the current global climate change. (2) The drought in the Yangtze River Basin tends to increase from 1962 to 2016. The drought severity, duration, intensity and peak increased, respectively, with average rates of 0.064, 0.041, 0.023 and 0.027 per ten years in the upper reaches of the Yangtze River Basin, while increased, respectively, by average rates of 0.151, 0.089, 0.021 and 0.030 per ten years in the middle and lower reaches. The intensity of drought is mainly mild to moderate drought. The drought is more serious in the source of the Yangtze River and the southwest of Sichuan Basin, while the drought is relatively mild in the upper reaches of Jinsha River and Yalong River and the south of Poyang Lake. (3) Compared with SSPI, the return period of the same drought severity and duration estimated by NSPI is larger, and the estimated drought events are more lumped. The drought risk of NSPI indicates that higher risk of drought occurs mainly in the source of Yangtze River, the lower reaches of Jinsha River and the middle of Dongting Lake Basin, while lower risk of drought covers mainly in the upper reaches of Sichuan river and the southeast of Poyang Lake. (4) The NSPI estimated by the non-stationary gamma model can better predict the drought characteristics, and the prediction performance of drought severity and duration is better.