Monitoring forest disturbance using time-series MODIS NDVI in Michoacán, Mexico

Abstract

MODIS-based NDVI time series (2000–2016) was applied to monitor sub-annual forest disturbance in the Mexican state of Michoacán, with an algorithm that decomposes the time-series data into a harmonic function and a trend. To detect change, a moving sum of residuals between the observed and predicted NDVI values was compared with that from the reference period. Magnitude of change was computed by subtracting the predicted NDVI from the observed one. By comparing the detected changes with reference data through visual interpretation, a threshold of |0.05| was established as the magnitude of change for forest disturbance detection. The method detected more forest gain than loss for 2013–2016, a result which is supported by recent findings from the national forest inventory. Forest loss decreases yearly for 2013–2016, and forest gain peaks at 2014 and 2015. We verified the findings with data from the global forest cover change project.