Spatial micro structure and its change over time is recorded for Norwegian farm firms. Relative strong correlations between geographically close neighbors are expected, either because growing farms swallow the smaller ones, or because they are affected by some spatially related unobserved factors. Strong correlations over time are also expected because of prevalent family farming. The paper proposes a state-of-the-art Markov chain model in order to predict the spatial and temporal micro structure taking account of both non-stationarity and spatio/temporal correlations by means of techniques from non-linear state space modeling and Gaussian Markov random fields. The model and the complete data set is then a device with which one can investigate the consequences of ignoring spatial and/or temporal correlations, both with complete data and with more sparsely sampled data, like FADN panels or USDA's repeated cross-sections (ARMS).