[R] How to produce rainfall maps

[Micha Silver]

Hi

On 11/23/2017 10:04 AM, Stefano Sofia wrote:
> Thank you Sarah and Mike for your explanations. My final objective is 
> to produce maps (png image or any kind of extension I can import in 
> LaTeX) where rainfall data are interpolated, using the Inverse 
> Distance method or Kriging. My input file (pointfile.csv in the 
> reported example) reports the station code, lat and long of the 
> meteorological station and the rainfall value (which might be the 
> cumulate of a week or ten days or the period I need to investigate). 
> Here a small example: Station_Code, Init_Year, Init_Month, Init_Day, 
> Init_Hour, Init_Minute, Fin_Year, Fin_Month, Fin_Day, Fin_Hour, 
> Fin_Minute, Rainfall_Cumulate, Long, Lat 1056, 2017 , 11 , 1 , 0 , 0 , 
> 2017 , 11 , 11 , 0 , 0 , 28.40, 12.786904, 43.851849 1064, 2017 , 11 , 
> 1 , 0 , 0 , 2017 , 11 , 11 , 0 , 0 , 27.20, 12.967556, 43.762669 1072, 
> 2017 , 11 , 1 , 0 , 0 , 2017 , 11 , 11 , 0 , 0 , 21.80, 12.897710, 
> 43.907555 As far as I can understand (as you can see, I am not an 
> expert on GIS or any spatial topic) - my input file pointfile.csv is 
> in "observation-per-row form"; - I need a grid (file.asc) where I can 
> interpolate my rainfall data (I can get it, a resolution of 1km will 
> be enough for me)
If I understand, in order to interpolate point data to a grid, the steps 
you need to do in R are:

 1. import the CSV of rain data and convert to a SpatialPointsDataFrame
 2. Convert that SPDF to a projected coordinate system, such as UTM, for
    kriging
 3. create an empty grid as the target for kriging, probably based on
    the extent of the rain data
 4. Run kriging using the point rain data and target grid

Here's a basic workflow for the above

#-----------------------------------

# Required libraries

library(gstat)
library(automap)
library(rgdal)

# Read in CSV file and convert to SPDF
rain_data <- read.csv("pointfile.csv")
str(rain_data)

# Check what you have so far

point_coords <- rain_data[c("Long","Lat")]
coordinates(rain_data) <- point_coords
p4str <- CRS("+init=epsg:4326")  # Since the coordinates are in 
Long/Lat, first declare this CRS
proj4string(rain_data) <- p4str


# Now Convert to UTM

p4str_UTM <- CRS("+init=epsg:32633")
rain_data_UTM <- spTransform(rain_data, p4str_UTM)

str(rain_data_UTM)    # Check that this is a SPDF in the UTM coordinate 
system


# Create Grid for kriging output, using the extent of the rain data SPDF
minx <-  rain_data_UTM at bbox[1,1]
maxx <- rain_data_UTM at bbox[1,2]
miny <- rain_data_UTM at bbox[2,1]
maxy <- rain_data_UTM at bbox[2,2]
pixel <- 1000        # Each pixel will be 1000 meters
grd <- expand.grid(x=seq(minx, maxx, by=pixel), y=seq(miny, maxy, by=pixel))
coordinates(grd) <- ~x+y
gridded(grd) <- TRUE
proj4string(grd) <- p4str_UTM

# Kriging, using autoKrige which creates a best guess variogram

# The formula for ordinary kriging is "<data_column> ~ 1"

OK_rain <- autoKrige(Rainfall_Cumulate ~ 1, rain_data_UTM, grd)

#-----------------------------------


The kriging result contains a component "prediction" which you can 
either plot directly, convert to an R raster object for plotting with 
ggplot2, or export to a Geotiff.


HTH,

Micha

> Dear R users,
>> I need to produce rainfall maps using R. I know that this is 
>> possible, I looked though the web, I found the example below reported 
>> (the author is Andrew Tredennick). I would ask you if this is the 
>> most performing way to make rainfall maps; if yes would someone be 
>> able to give me an example of how file.asc and pointfile.csv should 
>> be? If no would somebody please show me another way providing a small 
>> example? Thank you for your help Stefano

-- Micha Silver Ben Gurion Univ. Sde Boker, Remote Sensing Lab cell: 
+972-523-665918