The requirement of annotated data as a source of ground truth is often the bottleneck in machine learning applications. The process of manual annotation is expensive and labor intensive. Simulation has been investigated as a possible source of ground truth annotated data, but in many cases, simulated datasets fail to accurately capture enough physical characteristics of machine learning classes to be useful. For satellite image target recognition, an alternative to manual annotation and simulation is to use image data from one sensor as a source of ground truth for training the other sensor. In this study, we investigate the correlation between multispectral time series image data of Amazon deforestation which were manually annotated and processed by AI, and SAR data collected over the same region during overlapping time intervals. For a set of geographical points which exceeded some threshold of change in time series of multispectral data, the change detection seen in an overlapping time series of SAR data had positive Pearson correlation coefficients ranging up to 0.7. This result suggests the possible utility of using SAR change detection data as a source of ground truth for machine learning for multispectral imagery of deforestation change.