3. Model validation
The baseline is defined as the scenario representing the current conditions and current processes without soil conservation. Following Schreinemachers et al., 2007 and Schreinemachers et al., 2010, we estimate the goodness-of-fit of the first 2 years of the baseline by regressing the simulated data on the observed data from the farm survey. The regressions were conducted at the level of population clusters and not at the level of individual agents, because the agent population in the MP-MAS was initialized using a cluster-level Monte-Carlo approach (see Section 2.1.5). Table 2 shows the results for three aspects of the validation.