I'm looking to create an interpolated surface from 5 weather stations spread throughout southern Ontario. The stations are roughly equidistant from each other, and have temp & precip data for the exact same times. Before, I typically used the Kriging tool inside 3D analyst. This time when I try it, it simply creates a single-colored square between the outer stations.
This is my working environment in ArcPro, with the 5 stations:
What would I put in the Parameters on the left? Is there perhaps another, better interpolation method in ArcPro?
You've run into a classic problem with interpolating meteorological data, a small number of sampling locations with a lot of measurements across time. None of the kriging methods in ArcGIS Pro deal with time explicitly. I suspect what is happening is that the Kriging tool is getting all of its neighbors from the same location so it sees very little variation in the data. Do you have access to the Geostatistical Analyst extension? If so, I suggest trying the Empirical Bayesian Kriging tool. It has an environment setting (Coincident Points) that will summarize all of your temporal observations down to a mean value. Five points is not a lot of information to build any interpolation model on. You many need to 'help' the model by providing additional explanatory variables. You can do this with either the EBK Regression Prediction tool or Forest Based Classification and Regression tool. This Learn lesson might help: https://learn.arcgis.com/en/projects/analyze-urban-heat-using-kriging/ Cheers.