Interpolation techniques generally and specifically the ArcGIS Geostatistical Wizard produce raster maps (which later can be converted to contour maps, but what for).
If you have enough and quality data (spatially representative etc.), which more or less fulfil the basic geostatistical requirements, you can rely on the tools provided by Geostatistical Analyst, semivariogram can be also modelled on advanced level.
What would be very interesting, once you have a raster map of CO2 in soil, would be to measure the rate of change from different times (Delta CO2/time). This will allow you to answer the following question: Where is the soil gaining CO2 and where is it loosing it? And possibly infer why or do some spatial regression analysis to try to find an explanatory variable. That is much easier done with raster maps than contour maps.
In our Lab we use three different platforms for spatial analysis including geostatistical interpolation: ArcGIS, SAGA and R. All of them has its own advantages and disadvanteges. But the process is always carried out within one environment. If we krige in ArcGIS, we also model the semivariogram with ArcGIS Geostatistical wizard.