Remote Sensing of Breaks and Trends in Vegetation Time Series Data Due to Fire and Drought

Abstract:

Ecosystem disturbances come in multiple forms, impact various extents and last different amounts of time. Areas can be affected by one or multiple disturbances at any given time. Vegetation time series provide the means to examine conditions leading up to a disturbance event, during an incident and how a region recovers after the disruption. Using the Breaks For Additive Season and Trend (BFAST) package for R, 250m 16-day composite Moderate Resolution Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) time series were analyzed to examine drought and fire events. Case studies were conducted to focus in on different disturbances in parts of the western United States. The 2002 drought in Arizona and the subsequent Rodeo-Chediski fire were used to test how BFAST handles drought and post-fire recovery trends. Edwards Plateau was used to focus in on the 2011 drought in Texas. We used BFAST to examine extreme breaks and trends in the vegetation time series along with more subtle changes. We also used pre-disturbance, post-disturbance and temporal anomaly analysis to show when and where landscape changes occurred. This was done in concert with short term and longer multi-year trend analysis to determine the magnitude of these changes and how recovery differed based on disturbance regimes and agents. Some issues arose when dealing with areas impacted by snow. Gradual upward and downward trends in vegetation were distinguished compared to more extreme trends characterized by landscape response to disturbances. This analysis creates the foundation to assess the impact current droughts could have on vegetation measurements. We are also able to examine past fire events and determine how long recovery is taking in order to forecast how long it will take similar regions to recover after more recent wildfires.