In a precision agriculture context, the spatialization of existing crop models by downscaling processes to simulate agronomic variables at a within-field scale is of interest to better adapt technical decisions at this scale. The evaluation of spatial crop models needs to be based on both aspatial and spatial pattern error. However, current aspatial model metrics and existing spatial metrics have known limitations to evaluate the performances of spatial crop models. To address these limitations, a new metric, the spatial balanced accuracy (SBA), is proposed. The SBA is a novel metric, based on connectivity analysis that incorporates both aspatial and spatial aspects of model performance. The theory behind the metric development is presented here along with a comparison with existing model metrics applied to synthetic simulated data that covers a range of potential conditions.