# Kriging observations using additional knowledge from geophysical model

Here is a simplified explanation of the problem I am trying to solve : I have (1) a 100x100 grid with air pollutant concentrations in a city obtained from an air quality geophysical model, and (2) a set of observations of these air pollutant concentrations at 10 stations within the city (i.e. in 10 different cells of my domain).

The map coming from my air quality geophysical model shows a realistic (physically-consistent) spatial distribution of the air pollutant concentrations but is prone to biases/errors (like any other geophysical model). So I want to correct this spatial distribution using the observations.

I read various posts but I am a beginner in geo-statistics and I am still quite confused. I understood that kriging methods could be used to estimate an air pollution map solely based on the information provided by the 10 observations but in my case, I want to use the extra information provided by the air quality model (since I want the output to respect more or less the spatial distribution predicted by the air quality model, showing roads, parks...).

How can I do that? From what I read (but poorly understood...), it seems that universal kriging might be the solution but I am still quite confused on where the information from the air quality model might be taken into account. Should this extra information be included as what we call the external drift?

NOTA : I think that the Python function UniversalKriging from the PyKrig package might be the solution for me (https://geostat-framework.readthedocs.io/projects/pykrige/en/stable/generated/pykrige.uk.UniversalKriging.html) but I am still not sure how to properly use this universal kriging method and this function specifically. In any case, I am open to other recommendations of packages.