Using a remotely sensed optimized Disturbance Index to detect insect defoliation in the Apostle Islands, Wisconsin, USA

The Disturbance Index (DI) proposed by Healey et al. (2005) has been used successfully to locate and monitor disturbances by various researchers and in many locations. Here, a modification of the DI is presented that adapts it to local conditions by weighting each of its input components to maximize the difference between disturbed and undisturbed forest canopy. The weights reduce the effects of background variation while emphasizing the variation caused by disturbance. The weighted DI was applied to the severe insect defoliation of 2006 in the Apostle Island National Lakeshore with promising results. A supervised maximum likelihood thematic map based on the unweighted DI was 59% accurate and one based on the weighted DI was 62%. A classification with only two classes, defoliated and not defoliated, was also more accurate when based upon the weighted DI (77% verses 69%). In nearly all instances, the user's and producer's accuracies were significantly higher for the weighted DI image, suggesting that weighting the DI to fit local conditions is a superior approach, at least for small-area studies.