I am working with the collection 3.1 zonal statistics dataset at the municipality level, which is provided under Downloads. In a sample of 525 Amazon municipalities, I find the following pattern: For each of these municipalities, for at least one year, say t, the area under “natural forests” (land cover id 2 as per collection 3.1) increases in t+1 and then decreases in t+2. I am confused as to why this is happening. As per my understanding, this should not happen - especially because, as per the methodology documentation, such transitions were corrected for in post-classification. Any explanations for why this is happening would be very useful.