不同气象插值方法在新疆草地NPP估算中的可靠性评价

Evaluating reliability of grassland net primary productivity estimates using different meteorological interpolation methods

气象因子对研究草地生产力、植被长势、灾害评估等都有着重要的意义,本研究采用不同空间插值方法(协同克里格法Cokriging、反距离加权法IDW和ANUSPLIN法)对新疆地区90个气象站点2000-2011年多年的7月平均降水和气温数据进行空间插值分析,使用均方根误差法(RMSE)以及平均绝对误差法(MAE)对插值的结果进行评价,讨论不同方法对该地区降水和气温插值结果的影响。利用不同插值方法,基于CASA模型进行新疆草地NPP的估算,结合实测生物量数据,对3种插值方法下的估算结果进行评估,结果表明,1)降水和气温数据都是基于ANUSPLIN法的插值结果最优(MAE_(降水)=6.45,RMSE_(降水)=8.77,MAE_(气温)=2.11,RMSE_(气温)=3.52)。2)基于不同插值方法得到的气象要素估算的新疆草地NPP精度不同,将实测数据与同时期CASA模型模拟值相关性进行分析,基于ANUSPLIN法插值的气象要素估算NPP的精度最高(R^2=0.794 7),NPP实测值与模拟值有良好的线性关系,比基于Cokriging插值的气象要素估算精度提高了13.23%,比IDW提高了20.13%。说明提高气象要素的插值精度有利于新疆草地NPP的估算研究。

Meteorological factors are significant in researching grassland productivity,vegetation growth,and disaster assessment.This study used different spatial interpolation methods including Cokriging（CK）,inverse distance weighting（IDW）,and ANUSPLIN,to analyse the average July precipitation and temperature datasets of 90 meteorological stations in Xinjiang from 2000 to 2011.In addition,the mean absolute error（MAE）and root mean square error（RMSE）were used to evaluate the interpolation results,and we discussed the effects of different interpolation methods on spatial variation of precipitation and temperature.Furthermore,the spatial variations in precipitation and temperature obtained using various interpolation methods were used to calculate the grassland net primary productivity（NPP）in Xinjiang using the Carnegie-Ames-Stanford（CASA）model and verified their accuracy using field measured biomass data.We obtained the following results,1）The interpolation results of the precipitation and temperature using the ANUSPLIN was better than those of the other methods（MAE_（precipitation）=6.45,RMSE_（precipitation）=8.77,and MAE_（temperature）=2.11,RMSE_（temperature）=3.52）,which indicates that the ANUSPLIN is a superior method for interpolating meteorological factors in Xinjiang.2）The accuracy of estimating the Xinjiang grassland NPP differed between the various methods.The calculation of the coefficients between the field measured biomass data and the simulated values obtained using CASA model,showed that the ANUSPLIN had the highest accuracy,with an R^2 of 0.794 7.There was a good linear relationship between the measured and simulated values of the grassland NPP.Compared with the Cokriging and IDW,the accuracy of ANUSPLIN was higher by 13.23% and 20.13%,respectively.These results show that improving the accuracy of the interpolation results of meteorological factors could enhance the estimation of grassland NPP.