Monitoring insect outbreaks following a drought event in southwest USA

Forests act as carbon sinks and have a large share in the terrestrial sink term of the global carbon cycle. Yet the sink behaviour can be highly dynamic. A major source of short-term variability and possible long-term trends are disturbances. In order to fully assess carbon fluxes during disturbances, it is important to evaluate the impact of combined disturbances. After extreme weather conditions, such as droughts, a significant increase of insect outbreaks can be observed within forests. There are currently no detection methods for insect outbreaks to generate homogeneous global data sets at large scale. The detection of outbreaks during or right after another disturbance can be complicated by the difficulty to disentangle the signals of the difference disturbances. This study aims at analyzing the possibility to detect outbreaks of insects in forests by using Moderate Resolution Imaging Spectroradiometer(MODIS) data,with a specific focus on data from a 500,000 km² area in Western USA in the years 2001 - 2006. A strong drought affected this region until 2003, and in the following year a peak of insect disturbances was identified in aerial detection surveys from the USA Forestry Service. Based on the analysis of MOD13Q1 products, including MIR (2.105-2.155 μm) and the normalized difference vegetation index (NDVI) with a resolution of 250m and 16 days, we tried to identify the particular fingerprint of insect disturbance on forested areas. On a second step, two detection methods based on anomalies of NDVI and MIR or their temporal evolution proposed in the literature were evaluated for this study region. Due to low performance, we tested possible modifications for the particular study region and for the type of insect. In total, overall accuracy of about 77% could be obtained. Finally, possible causes for the low performance were examined, such as contamination from recovery from previous fires and land-cover misclassification.