中文摘要：

参考黑河上游4站点的观测数据,用包含多统计指标的全面评价方法评价了2004-2009年WRF制备的5 km,1 h降水数据和空间插值后的5 km,3 h GLDAS降水数据。结果表明两者对降雨的时空分布表现均优。 WRF产生的〈1 mm和〉8 mm降水的总日数、总量更符合实际降水情况,但缺少1～8 mm的降水,春季降水峰值偏大,总降水日数偏少。 GLDAS降水和观测的相关系数更高,但年降水量偏少；〈1 mm降水日数、总量过多,〉2 mm降水日数偏少,春季降水峰值过大。由此两组数据驱动模型模拟时,需考虑数据本身的特点,同时黑河流域仍迫切需要更精确的降水数据。

英文摘要：

A comprehensive comparative evaluating approach, consisting of a set of statistical indicators and comparisons against in situ observations from four established meteorological sites in the upper Heihe River basin, was proposed in this study. Hourly WRF precipitation and 3 h GLDAS precipitation in the same 5 km resolution, the latter being interpolated, were evaluated in the period of 2004-2009. It was shown that both the WRF and GLDAS data had well captured the time and spatial distribution characteristics of precipitation. WRF data were closer to observed precipitation in representing〈1 mm and〉8 mm precipitation in terms of annual rainy days and annual total amount. But WRF detected less 1～8 mm precipitation, larger peaks in spring, and fewer total rainy days. GLDAS precipitation had a higher correlation and less annual total amount to that observed, overestimated rainy days and the amount for〈1 mm events and generally underestimated rainy days for 〉2 mm. GLDAS data also detected larger peaks in spring. It is necessary to take the characteristics of the two datasets into account when they are used for the model. Meanwhile, it is urgent to produce more accurate precipitation data to meet the demands from eco-hydrological studies in HRB.