[R-sig-Geo] Spatio-temporal kriging with missing time periods

[Nima Mehrafshan]

Hello everybody,

I have four years of spatial data with a couple of gaps (missing time
periods of up to one year).

    > summary(sample_data)
          lon              lat              x                month
     Min.   : 8.338   Min.   :54.41   Min.   :-1982.94   Min.   :2013-09-01
     1st Qu.: 9.551   1st Qu.:54.54   1st Qu.: -428.76   1st Qu.:2014-03-24
     Median :13.280   Median :54.63   Median :   86.03   Median :2014-09-01
     Mean   :11.856   Mean   :54.68   Mean   : 2470.73   Mean   :2014-11-08
     3rd Qu.:13.395   3rd Qu.:54.83   3rd Qu.:  997.35   3rd Qu.:2015-06-01
     Max.   :13.664   Max.   :55.02   Max.   :30967.30   Max.   :2017-08-01

(I have rounded the times to the next first of the month)

I would like to do spatio-temporal kriging on the target variable x using
the gstat package. However, when I run...

    > sp <- SpatialPoints(sample_data[,c('lat', 'lon')], CRS('+proj=longlat
+datum=WGS84'))
    > sp <- spTransform(sp, CRS("+proj=utm +zone=32 +ellps=WGS84
+datum=WGS84"))
    > sample_data <-STIDF(sp, time = sample_data$month, data =
sample_data[,'x',drop=F])
    > vv <- variogramST(x~1, data=sample_data, cutoff=1000,
assumeRegular=FALSE)

...the computation stops after 7% with no error message and the resulting
data frame (vv) has NAs in np, dist, and gamma in all but the first lag. I
guess this is because there are no observations for the second lag ?

Can variogramST only be used for data without gaps in the time dimension?

Do I have other options for spatio-temporal interpolation that work with
this kind of data?

Thank you.
Nima Mehrafshan

PS: I've also posted this question here:
https://gis.stackexchange.com/questions/254127

	[[alternative HTML version deleted]]