地形复杂山区相对湿度空间插值方法对比研究

Comparison on Spatial Interpolation Methods of Relative Humidity in Complex Mountainous Terrain

在气象站点稀疏、气象数据少的山区,气象数据离散、精度低,不能满足山区开展精细化气象服务的需求。针对复杂地形条件的气象数据插值,不同插值方法显著影响相对湿度等气象数据精度及适用性。以玉溪山区为研究区域,选取2011—2019年10个训练站点的相对湿度实测数据,使用多元回归与残差分析相结合的AMMRR法与基于薄盘样条理论的ANUSPLIN插值软件进行空间插值处理,并模拟研究区域内相对湿度的空间分布,进行模拟精度检验,对比2种方法的插值结果。结果表明:(1)AMMRR中多元线性回归模型模拟效果可以满足建模精度要求,但插值结果仍存在部分地区与实际情况出入较大的问题。(2)ANUSPLIN的MAE为3.34、MRE为0.05、RMSE为3.91,3种精度评价指标均更低,空间插值结果更加稳定且精度更高,能更好地反映相对湿度在山区的分布情况。(3)玉溪山区相对湿度分布趋势总体表现为西南部与北部较高而中部较低。该研究可补充玉溪山区相对湿度的数据空缺,为市县区域经济作物病害防治及产量精准预测提供基础数据支持,并为站点较少的复杂山区选择空间插值方法提供参考。

In the mountainous areas with complex terrain,the meteorological data collected at sparse meteorological stations appears to be discrete with low accuracy,which cannot meet the demand for refined meteorological services in rural construction;therefore,it need proper mathematical treatment of interpolation for availability before utilization.Different interpolation methods significantly affect the accuracy and applicability of the discrete meteorological data,such as relative humidity.In this study,it took the Yuxi area of Yunnan province of China as a case study,where the distinctive feature of the landform is its steep terrain in low latitude plateau but equipped with very few of weather stations.Relative humidity data were collected at 10 training stations in Yuxi from 2011 to 2019;Both the AMMRR method,which combined with multiple regression and residual analysis,and the ANUSPLIN method based on the theory of Partial Thin Plate Smoothing Splines were separately used to conduct spatial interpolation fit;Then the spatial distribution of relative humidity in Yuxi was simulated,followed by accuracy inspection of the two methods and result comparison.The results show that:(1)Multiple linear regression model from AMMRR method could satisfy the expected accuracy for simulation,but the interpolation values still had distinct discrepancy as compared with field observations at some places.(2)MAE,MRE and RMSE obtained by ANUSPLIN method were 3.34,0.05 and 3.91,respectively,achieving better precision indexes,with more stable and accurate spatial interpolation results,suggesting better reflection of the distribution of relative humidity in hilly areas.(3)The distribution trend of relative humidity in the Yuxi hilly area was generally higher in the southwest and north,and lower in the middle.This study could solve the shortage in the data of relative humidity in the Yuxi hilly area,which provides basic data support for the prevention and control of economic crop diseases and accurate prediction of yield in the region,and gives also a reference for choosing spatial interpolation methods in complex hilly areas with few weather stations.