It does not look like you have any spatial structure in your data. As such, there is little you can say about the spatial process and interpolation is not supported. Perhaps your sampling distance is too coarse and you missed the spatial process you were after. This is why it is so important to apply a sample design before you start collecting data. You may want to consider collecting more data and sampling in a different grid pattern. I have found that hexagonal sampling works quite well for things like soil compaction. Based on the semivariogram, your data looks somewhat categorical. Is this the case? If so you need to read up on methods that support discrete data (not kriging).
Given your current sample you can model the expected semivariance and design a secondary sample, that does capture the spatial structure, to augment your current data. There are a some very good papers on dual-phase sampling that may allow you to salvage your study. A good starting point is:
Delmelle E. and P. Goovaerts (2009). Second-Phase Sampling Designs for Non- Stationary Spatial Variables. Geoderma, vol. 153: 205-216.
