Rainfall interpolation by kriging

A continuous rainfall surface built from point station data with ordinary kriging in R, with the variogram and prediction error mapped.

Question

How can scattered rain gauge readings be turned into a reliable rainfall map, and where is that map least certain?

Tools

R

Data

EDIT: rainfall station dataset and study area

Context

GIS module assignment, University of Galway

Problem

Rainfall is measured at a limited number of stations, but planning and environmental work need values everywhere. Kriging predicts values between stations and, unlike simpler methods, also gives an error estimate for each location.

Data

Add: the station dataset, number of stations, time period and study area.

Method

• Explored the station data and checked for trend and outliers
• Fitted an empirical variogram and a variogram model to describe how rainfall similarity changes with distance
• Ran ordinary kriging over a prediction grid
• Mapped both the predicted rainfall surface and the kriging variance
• Checked accuracy with cross-validation

Result

Predicted rainfall surface (left) and kriging variance (right).

Add: the prediction and variance maps, the fitted variogram plot, and the cross-validation error (e.g. RMSE). If your LISA analysis has the stronger visual, swap this project for that one.

Code

Add: link to the GitHub repo with the R script.