Forest Disturbance Detection

Submitted by atripp on

Forested animal habitat is impacted by disturbance agents such as wind, insect outbreaks, fire, silvicultural treatments, and climate change. This proof-of-concept study combines disturbance detection with forest management, annual songbird point counts, and continuous forest inventory (CFI) plots from Caddo Lake National Wildlife Refuge (NWR) to improve the understanding of species habitat relationships over time. To monitor forest disturbance, readily available 16-day composite Normalized Difference Vegetation Index (NDVI; MOD13Q1) from the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite sensor and pixel reliability information are used. NDVI is sensitive to changes in plant biomass and leaf area associated with annual and seasonal growing cycles as well as infrequent disturbance events that cause abrupt forest change. Time series NDVI and seasonal trend decomposition is an approach for identifying disturbance events and subsequent recovery periods that may also correspond with animal occurrence and levels of abundance.

Management approaches at Caddo Lake NWR include tree thinning, regeneration treatments, and prescribed burning to develop a diversity of habitat conditions and forest successional stage habitat for songbirds. Forest on Caddo Lake NWR can also experience intense wind and small-scale blow-down events in addition to light intensity silvicultural treatments. Time series decomposition approaches such as breaks for additive season and trend (bfast) and seasonal and trend decomposition using Loess (STL) provide a versatile and robust means of analyzing time series components and nonlinear relationships in time series data. NDVI trends monitored using seasonality harmonics can reveal substantial departures from stable periods that are common in forest systems. In general, major disturbance events were sufficiently detected at the coarse spatial resolution of 250 m MODIS pixels but may fail to detect lower impact understory tree thinning or prescribed burning. Further NDVI cross correlation analyses with monthly MODIS daytime land surface temperature (LST; MOD11B3) and climate hazards group infrared precipitation with station data (CHIRPS v. 2.0) will be used to understand possible climate and drought effects on songbird densities.

Moderate Resolution Imaging Spectroradiometer (MODIS)  time series data for Caddo Lake National Wildlife Refuge (NWR) indicating, a) Normalized Difference Vegetation Index (NDVI) time series information from a single continuous forest inventory (CFI) plot location and grid cell showing disturbance and recovery, b) ‘bfast’ monitoring period between 2007 and 2020 with a significant break detected at the beginning of 2011, c) mapped year of time series breaks related to decreases or increases in plant biomass during the monitoring period, and d) mapped magnitude and direction of change detected since 2007 just prior to the beginning of refuge forest treatments. Mapped changes (c, d) show CFI plot numbers (black) and forest treatment unit polygons (black dashed line).

Author Name
Steven Sesnie
Author Email