plotting issue with autofitVariogram in automap

Question

I am a newbie in R. I am using autofitVariogram to daily rainfall data of 50 stations.The sample data is provided below.Some of stations have missing values represented by "NaN" values.

My question is regarding the variogramfit. The variogram covers only a distance of 60,000m. Why are the points in bins beyond 60Km not plotted. I had seen from spatial correlation plot maximum distance from lon-lat information is >200Km.

The summary of latitide and longitude information is provided below. summary(lonlat) lon lat
Min. :74.78 Min. :15.77
1st Qu.:75.14 1st Qu.:16.04
Median :75.56 Median :16.33
Mean :75.54 Mean :16.37
3rd Qu.:75.94 3rd Qu.:16.66
Max. :76.31 Max. :17.23

$ Sample data  given below:
dput(rain[140:145,])

structure(list(Col0 = c(0, 0, 1, 9, 6.5, 0), Col1 = c(1.5, 36, 
21, 44, 4, 0), Col2 = c(0, 0, 24.5, 21.5, 7.5, 1), Col3 = c(0, 
1, 45, 3, 0, 0), Col4 = c(2, 0, 5, 54.5, 13.5, 0), Col5 = c(0.5, 
2, 0, 3.5, 13.5, 0), Col6 = c(0.5, 0, 0, 59, 15.5, 0), Col7 = c(0, 
0, 2.5, 1, 0, 0), Col8 = c(0, 6, 24, 2, 5.5, 0), Col9 = c(0, 
3, 6, 1, 0, 7), Col10 = c(0.5, 1, 64, 20, 1, 0.5), Col11 = c(NaN, 
NaN, NaN, NaN, NaN, NaN), Col12 = c(0, 11, 75, 19, 15.5, 0), 
    Col13 = c(0, 4, 57.5, 50.5, 8.5, 0), Col14 = c(1.5, 0.5, 
    127, 33.5, 34.5, 0), Col15 = c(0, 7, 0.5, 13, 1, 0), Col16 = c(0, 
    0.5, 81.5, 15, 49, 0), Col17 = c(0, 0, 4.5, 17, 5.5, 1), 
    Col18 = c(0, 3, 2.5, 0.5, 0, 0), Col19 = c(NaN, NaN, NaN, 
    NaN, NaN, NaN), Col20 = c(0, 0, 0, 0, 7, 0), Col21 = c(0, 
    1, 0, 5, 3.5, 0), Col22 = c(0, 0, 11.5, 28, 3.5, 0), Col23 = c(0, 
    0, 48.5, 0, 24.5, 0), Col24 = c(0, 0, 0, 10, 0.5, 14), Col25 = c(NaN, 
    NaN, NaN, NaN, NaN, NaN), Col26 = c(0, 7.5, 16, 28.5, 20.5, 
    0), Col27 = c(1.5, 0.5, 38, 28.5, 50, 0), Col28 = c(NaN, 
    NaN, NaN, NaN, NaN, NaN), Col29 = c(NaN, NaN, NaN, NaN, NaN, 
    NaN), Col30 = c(2.5, 0, 0, 80.5, 28, 13.5), Col31 = c(1, 
    0, 17, 85.5, 3.5, 0), Col32 = c(0, 0.5, 8, 101, 20, 4), Col33 = c(NaN, 
    NaN, NaN, NaN, NaN, NaN), Col34 = c(4, 3, 17, 122, 2, 2), 
    Col35 = c(0, 15.5, 14.5, 20, 3.5, 0), Col36 = c(0, 6.5, 8.5, 
    21, 7, 0), Col37 = c(0, 0, 1.5, 14.5, 0, 1.5), Col38 = c(0, 
    28, 30, 4, 0, 73), Col39 = c(28.5, 0, 4.5, 9.5, 1, 0), Col40 = c(1.5, 
    11.5, 32.5, 55, 0, 1), Col41 = c(0, 14.5, 0, 19, 12.5, 47.5
    ), Col42 = c(0, 28, 29, 17, 0.5, 20.5), Col43 = c(NaN, NaN, 
    NaN, NaN, NaN, NaN), Col44 = c(0, 19, 3.5, 42, 0, 0), Col45 = c(0, 
    0, 85, 15.5, 1, 0), Col46 = c(0, 0.5, 8, 24, 0.5, 0), Col47 = c(0, 
    1.5, 7, 12, 8.5, 0), Col48 = c(0, 0, 0, 43.5, 0, 1.5), Col49 = c(0, 
    13.5, 1, 16, 1, 1)), .Names = c("Col0", "Col1", "Col2", "Col3", 
"Col4", "Col5", "Col6", "Col7", "Col8", "Col9", "Col10", "Col11", 
"Col12", "Col13", "Col14", "Col15", "Col16", "Col17", "Col18", 
"Col19", "Col20", "Col21", "Col22", "Col23", "Col24", "Col25", 
"Col26", "Col27", "Col28", "Col29", "Col30", "Col31", "Col32", 
"Col33", "Col34", "Col35", "Col36", "Col37", "Col38", "Col39", 
"Col40", "Col41", "Col42", "Col43", "Col44", "Col45", "Col46", 
"Col47", "Col48", "Col49"), row.names = 143:148, class = "data.frame")


# Import the required libraries
library(rgdal)
library(maptools)
library(gstat)
library(sp)
library(automap)
library(XLConnect) 

# Read the station data from xls file
stnrain = readWorksheetFromFile(path_fileName,"Sheet1", region = "D1:BA187", header = FALSE)
N = nrow(stnrain)
rain = stnrain[4:N,]
lat = as.numeric(t(stnrain[2,]))
lon = as.numeric(t(stnrain[3,]))
lonlat = cbind(lon,lat)

#Transform from GCS to UTM protection
sp = SpatialPoints(lonlat,proj4string = CRS("+proj=longlat")) 
sp_utm = spTransform(sp, CRS("+proj=utm +zone=43N +datum=WGS84"))

krige_value = list() #prepare a list for storing the autokrige output
krige_stderr    = list()

nRows = nrow(rain)
for (i in 1:nRows) 
    {   
      irain = rain[i,]
      miss_indx = (irain == "NaN")
      irain = irain[!miss_indx]
      irain = as.numeric(irain)
      isallZeros = (max(irain) == 0) # To take care of the cases of dry day(irain =0) 
      irain = as.data.frame(irain)
      M = nrow(irain)

      if ((M > 5) & (!isallZeros)) # To avoid cases of NaN across many stations
       { 
         print(i)
         foo_utm = sp_utm[!indx]# Removing the locations with NaN values
         data = data.frame(foo_utm,irain)
         names(data) = c("Easting","Northing","rain")
         coordinates(data) = c("Easting","Northing") 

         variogram = autofitVariogram(rain~1,data,model = "Sph",fix.values=c(0,NA,NA)) 
         p = plot(variogram, main="Semi-variogram (Spherical Model)",xlab="Distance(m)",ylab="Semi-Variance(mm2)", sub=paste("Range: ",variogram$var_model$range[2], "Day",i))
         print(p)
         png(p)
         dev.off()
        }
       else
        {
            krige_value[[i]] = list(rep(0, L))
            krige_stderr[[i]]   = list(rep(0, L))
         }
     }
  }

Q2) How can i save the variogram fit png file in a loop. I understand that dev.off() should be used after each saving the figure, which i had done, but I am not able to save the the figure. Any help would be appreciated.

Thanks,

Any suggestions would be appreciated?

Answer 1

In regard to your first question, the sample variogram is built using points up to a maximum distance of around 1/3 of the diagonal of the area of interest. The assumption here is that points farther away form that are not related, and because they are not in the sample variogram or variogram model they are plotted. This is just a choice, and might not be the correct choice, but when I wrote autofitVariogram it seemed to work well for my data. The variogram model you show confirms this, the range is smaller than 60 km.

For saving your png's I have two suggestions. First, call the plot command inside the png() dev.off pair, so not:

 print(p)
 png()
 dev.off()

but:

 png()
 print(p)
 dev.off()

In addition, I would create meaningful names for the png files.

To create sets of variogram plots, I would use ggplot2. This uses geom_line and facet_wrap. ggplot2 cannot deal directly with gstat/automap variogram models, luckily you can create distance semivariance data using the function variogramLine from gstat. See for example figure 3.1, and the plots in appendix A of this report I wrote. This answer I wrote earlier does also include an example of using ggplot2 for spatial data, this time to plot a grid map.