地理因素辅助的黄土高塬典型流域面雨量制图效果比较与评价

为了获取适宜的黄土高塬流域面雨量计算方法,选择算术平均(Am)、泰森多边形(Tsn)、普通克里金(OK)、泛克里金(UK)、反距离权重(IDW)、简单克里金(SK)、考虑海拔的简单克里金(SK_EL)以及将海拔和空间坐标信息作为辅助因子的简单克里金(SK_EXY)和地形插补法(Tr)等9种方法计算了环江下游流域2006—2011年面雨量。结果表明:(1)各年面雨量Am法全流域只有1个数值,无空间分布,Tsn法呈斑块状分布,IDW法在站点周围空间趋势表达不连续,其余6种基于克里金方法的面雨量空间变化趋势基本一致,均是从东南向西北连续递减的等值带,其中OK的插值曲面最平滑,Tr与SK法的结果更加符合降雨随地形的变化趋势;(2)24站测值为基础的单年和多年交叉验证结果均表明各种方法插值所得面雨量的精度以考虑高程和站点空间信息的Tr法最高,SK_EXY其次,并没有一种插值方法在所有的精度检验指标中都最差;(3)当只考虑面雨量值不考虑其空间分布时,可以选择Am与Tsn法相结合,其值接近各种方法的均值,而当综合考虑面雨量值及其空间分布时,则宜选择Tr和SK_EXY法。

Comparison and Evaluation of Areal Precipitation Mapping Effectiveness with Consideration of Geographic Factors in the Loess Plateau

In order to obtain optimal areal rainfall calculating method of watershed in Loess Plateau, methods of Arithmetic mean (Am), Thiessen (Tsn), Ordinary Kriging (OK), Universal Kriging (UK), Simple Kriging (SK), Simple Kriging with elevation auxiliary (SK_EL), Simple Kriging with elevation and spatial coordinate of rain gauge auxiliary (SK_EXY), as well as Topography Regression (Tr), were employed to estimate areal rainfall of Huanjiang watershed in Loess Plateau from 2006 to 2011. Efficiencies of 9 methods were tested. Results showed that: (1) Areal precipitation from Am method was a value each year without spatial information. Tsn method generated patchy distribution, and IDW resulted in discontinuous trend around the site. Areal precipitation from the other 6 kriging based methods showed similar spatial pattern. All of them were equivalent band decreasing from the southeast to the northwest. Interpolation surface from OK was the smoothest and results from TR and SK expressed the changing trend of precipitation along terrain. (2) Both single-year and multi-year cross-validation based on the 24 rain gauges revealed that Tr and SK_EXY had the higher and SK had the lowest interpolation accuracy. (3) It is suggested that the combination of Am and Tsn method had the advantage when spatial distribution was not necessary, and the value of precipitation were close to the mean value of all methods, but Tr and SK_EXY had obvious advantage when both distribution and value of precipitation were necessary.