摘要
在农业气候资源研究中,站点数据的区域化问题是进行资源优化配置和高效利用的一个重要环节。通过采用逐步回归分析与空间插值相结合的方法,以甘肃省及其相邻省区的112个站点1970~2001年31年的月平均温度和降水数据以及计算得到的月平均太阳辐射和潜在蒸散量为数据源,对甘肃省气候资源进行了区域化。对每种气象要素都采用了两种空间插值方法,并对插值结果运用了绝对验证和相对验证两种方法进行了验证和对比。结果表明:温度残差的平均绝对误差(MAE)是Spline<I,其值分别为:0.744℃和0.754℃,平均相对误差(RME)分别为:9.56%和9.66%。降水的平均绝对误差是Kriging<I,其值分别为:5.39mm和6.12mm,平均相对误差分别为:20.61%和23.45%,作物生育期3~11月分别为:17.19%和20.01%。太阳辐射的平均绝对误差是Spline<I,其值分别为:16.22MJ和16.44MJ,平均相对误差分别为:3.40%和3.44%。潜在蒸散量平均绝对误差也是Spline<I,其值分别为:7.80mm和7.96mm,平均相对误差分别为:10.57%和10.65%。根据相对误差结果分析4种气象要素的区域化结果由好到差的排序是:太阳辐射>温度>潜在蒸散量>降水,但都达到了较高的精度。
In research of agricultural climate resources, the zonal digitization of climate data is very important for optimization, collocation and high efficient utilization of agricultural resources. Using 11-year of monthly mean air temperature, precipitation, solar radiation and evapotranspiration potential at 112 station in Gansu province and adjacent regions, we combined the methods of stepwise regression analysis and spatial interpolation to regionalize climate variables in Gansu province. Based on stepwise regression analysis,we chosen the methods of zonal digitization. The two methods of spatial interpolation were applied for every climate variables, and farther the results were validated and contrasted by absolute errors from cross-validation test and relative errors from proportion of mean absolute errors (MAE) and corresponding monthly mean climate variables. The results of mean absolute errors were ranked as Spline<IDW for the residual of air temperature, Spline gives lower mean absolute errors which averaged 0.744 ℃ and relative mean errors(RME) averaged 9.56%. The rank of precipitation is Kriging<I, Kriging gives lower mean absolute errors which averaged 5.39mm and relative mean errors averaged 20.61%, yet kriging had lower relative mean errors of 17.19% during crop growth seasons.The rank of solar radiation is Spline <I,Spline gives lower mean absolute errors which averaged 16.22MJ and relative mean errors averaged 3.40%. The results of mean absolute errors were also ranked as Spline <IDW for evapotranspiration potential, Spline gives lower mean absolute errors which averaged 7.80mm and relative mean errors averaged 10.57%. Compared with these methods of spatial interpolation, Spline is optimal to apply for interpolating remnant errors of monthly mean temperature, solar radiation and evapotranspiration potential,Kriging is suitable for monthly mean precipitation. On basis of relative mean errors,the four climate variables were ranked as solar radiation> air temperature> evapotranspiration potential> precipitation, but the zonal digitization of the four climate variables all achieved good precision.
出处
《地理科学》
CSCD
北大核心
2004年第4期444-451,共8页
Scientia Geographica Sinica
基金
中国科学院知识创新工程重要方向项目(KZCX3-SW-333)
关键词
气候资源
区域化
回归分析
空间插值
climate resources
zonal digitization
regression analysis
spatial interpolation