SDSM模型在海河流域统计降尺度研究中的适用性分析

Suitability Analysis of SDSM Model in the Haihe River Basin

全球气候变暖对陆地水循环会产生重大影响，统计降尺度方法是解决大尺度气候信息和小尺度水文响应的空间尺度不匹配问题的有效方法之一。选取NCEP／NCAR再分析日资料（简称NCEP）和HadCM3在A2和B2情景下的大气变量日资料（简称H3A2和H382），并选取海河流域上11个站点1961年～2000年日平均温度、蒸发皿蒸发量以及降水的实测资料，采用SDSM（the Statistical Down—Scaling Model）方法，进行海河流域气候变化特征量的降尺度研究。研究表明：①使用SMLR（Stepwise Multi—Line Regression）方法，可以在广阔的空间范围内优选出具有一定物理机制的适用于不同预报量的预报因子；②SDSM方法对日平均温度、蒸发皿蒸发量以及降水都能较准确模拟，其观测与模拟值的确定性系数分别可达99％，92％和73％以上；③SDSM方法在模拟极端事件时存在一定系统偏差，而且模拟的峰值略滞后。

Statistical downscaling （SD） methods are often used to fill the gap between large-scale climate change information and fine-scale hydrological impact studies. Among them, the Statistical DownScaling Method （SDSM） is widely used for its simplicity and superior capability. Although plenty of SD has been used in western countries, there is a paucity of such work on Chinese river basins. In this article, applicability of SDSM in the Haihe River basin was evaluated, and its strengths and weaknesses in downscaling air temperature, evaporation and precipitation are discussed. The atmospheric data used in the study are daily reanalysis data from the National Center for Environmental Prediction and the National Center for Atmospheric Research （NCEP/NCAR）, and emissions scenarios A2 and B2 of the HadCM3 model from the Hadley Centre for Climate Prediction and Research. Daily mean temperature, pan evaporation and precipitation was provided from 11 weather stations in the Haihe River basin for 1961 2000. It is a big challenge to select predictors due to the strong influence of the East Asia monsoon and complex climate conditions. We found that different sets of predictors for different predictands can be selected from a wide space using the SMLR method, and these predictors have some dynamic meaning. Mean temperature is more sensitive to near surface atmospheric variables, the airflow from northwest is transferred mainly by vorticity and horizontal wind, and the southeast airflow is affected mainly by sea surface mean pressure. For pan evaporation, northern zonal wind at 500 hPa level and humidity in the central and middle areas have influence; and for precipitation, in the north it is mainly affected by meridional airflow above 700 hPa and horizontal airflow below 700 hPa, and in the south it is mainly influenced by the vorticity at 850 hPa level. The results showed that： 1 ） using the Stepwise Multi-Linear Regression （SMLR） method, different predictors can be selected from a wide space and for different predictands; 2） whether in the calibration period （1961 - 1983 and 1994- 2000） or in the validation period （1984 ~ 1993 ）, the amount and change pattern of mean temperature, pan evaporation and precipitation can be reasonably simulated; the determination coefficient （R2） between observed and downscaled mean temperature, pan evaporation and precipitation attained values of 99%, 93 % and 73 %, respectively; and 3） there are systematic errors in simulating extreme events with SDSM. Although SDSM and the SMLR method can simulate mean values and trends, they have some limitations with unique terrain and in conditions affected by special local weather events.