- krige.cv(formula = VALUE~1, locations = ~X+Y, model = vgm(61.39472, "Exp", 5326.663 , 10), data = data.observed ,nfold= nrow(data.observed))
head(kriging)
注意,krige.cv()函数本质上是进行很多次克里金插值,然后我们就可以分析被拿出的已知点的值和预测值,估计克里金插值的可信性。这里我们每次拿出一个点,所以nfold的值设置为和data.observed的行数一样,可以看到结果:
var1.pred var1.var observed residual zscore fold X Y
1 19.020820 31.29921 12 -7.020820 -1.2549348 1 -318466.5 3794841
2 11.220626 27.62193 10 -1.220626 -0.2322499 2 -304466.5 3794841
3 10.758057 24.82611 9 -1.758057 -0.3528407 3 -303466.5 3794841
4 9.200462 24.85426 11 1.799538 0.3609613 4 -302466.5 3794841
5 10.840395 27.07824 7 -3.840395 -0.7380158 5 -301466.5 3794841
6 12.446044 29.40208 14 1.553956 0.2865826 6 -297466.5 3794841
结果包含了预测值,预测值的方差,已知值,残差,z统计量值等,我们可以绘制出z统计量图:
ggplot(data = kriging, aes(x=X,y=Y))+
geom_raster( aes(fill= zscore)) +
scale_fill_gradient2( name="zscore",low = "green",
mid = "grey", high = "red", midpoint = 0,
space = "rgb", na.value = "grey50")
结果如下:
或者其直方图来评价插值方案:
par(mar=c(2,2,2,2))
hist(kriging$zscore)
欢迎大家留言讨论,转载请注明出处 http://www.cnblogs.com/ABMRG/p/5186727.html 。
参考文献:
[1] http://desktop.arcgis.com/en/arcmap/10.3/tools/3d-analyst-toolbox/how-kriging-works.htm
关于R基本栅格数据处理我推荐一篇博文:http://rstudio-pubs-static.s3.amazonaws.com/1057_1f7e9ac569644689b7e4de78c1fece90.html