Hi! First you have to tell us more precisely what kind of data do you have. How big is the area covered by TRMM satellite, how many "pixels"? Then, how many observatory stations? I guess your satellite data is spatially continous, so that may be "primary value" (Z1) in cokrigging. Observatory stations are for sure less dense than satellite pixels, so that may be "co-variable" (Z2). And what is " elevation" you mentioned? Some model of precipitation based on elevation above sea level?

Hello,

The Surfer, a product of Golden software is what you want and can do that.

https://www.goldensoftware.com/products/surfer

Mateusz Lukowski Hello Mr. Mateusz Lukowski! Thank you very much for your responds. About the size of area covered by TRMM Satellite in our study is 3x2 pixels. Amount of observatory stations are 11. And for the elevation, it will be used as a co-variable to get spatial distribution of precipitation based on interpolation results. In additional, precipitation data from observatory stations will be the main variable and secondary variables (co-variables) are TRMM satellite and the elevation of covered area. Specifically, the study case area is in Bandung (6°46’46.92’’ S – 7°11’9.24’’ S and 107°18’52.92’’ E – 107°49’37.92’’ E).

May it's quite detail and you can get it clearly.

Best regards.

Hello Mr. Raoof Mostafazadeh , thank you very much for your responds. I'll save and try your software recommendation.

Hello Mr. Joaquín Duque Lazo Thank you very much for your answers. Hmmm, I've used it before but those softwares have limit capability for some reasons, And I had ever get suggest to use Arcpy in ArcGIS too, but the module is only for the ordinary kriging, Sir..

So to summarize: you have 11 stations and 2x3=6 pixels. Please remember that co-variable must be less numerous than primary variable, so in fact you may use only several elevations. In my opinion your dataset is too poor to use kriging. I would start to think about geostatistics if I had more than 30 spatial points.

Hi, Said! When you have covariates available for every pixel (such as satellite imagery and digital elevation models) it is better to use regression kriging than cokriging. There is a very good book here showing how to perform it in R and Saga GIS (https://download.tuxfamily.org/tuxgis/geodescargas/hengl__A_practical_guide_to_geostatistical_mapping_2009.pdf). We usually use cokriging when your covariates are available just as points, and in different locations than your main variable.

You can use kriging with external drift (ked) or cokriging when you would use auxiliary variable along with the primary variable. You can implement these in ArcGIS geostatistical analyst extension platform. Also, you can do it on the R platform.

Even regression kriging is another option as suggested by Vitor in a reply above. For this one, you can use the R platform.