Detecting Stand-Replacing Disturbance using RapidEye Imagery: a Tasseled Cap Transformation and Modified Disturbance Index

Abstract

The development of robust approaches for monitoring forest cover is a fundamental forest management objective. Timely and detailed mapping of forest disturbance can be used to identify threatened habitat, estimate standing forest carbon stocks, and aid in the sustainable management of forests. Remote sensing technologies have provided new opportunities for mapping forest cover; however, various spatial and temporal resolution limitations have resulted in challenges for gathering up-to-date inventory data. A new generation of satellite constellations may provide solutions to better meet monitoring needs. In this article, we demonstrate a quick and accurate method to locate abrupt, fine-scale forest disturbances using RapidEye 5 m, 5-band imagery. To identify disturbance, we first derived a set of tasseled-cap indices from 50 cloud-free RapidEye images. We then adapted a previously developed disturbance index using Landsat Thematic Mapper data and estimated a simple threshold to differentiate disturbed and unchanged regions. To test the accuracy of the disturbance index, 7 image pairs depicting change were selected and compared with forest inventory data and manually interpreted sample locations. Overall accuracy estimates ranged from 83.65 ± 0.77% to 97.65 ± 0.25% for individual tiles. Across the 7 tiles, overall accuracy was 92.64 ± 0.15%, using forest inventory data, and 96.91 ± 1.16%, using the manually interpreted sample locations. An assessment of producers’ and users’ accuracies for disturbed and undisturbed regions suggested inaccuracies in the forest inventory reference data. In several instances, the modified disturbance index appeared to have provided a more accurate depiction of real change than the forest inventory reference data. Results suggest that the RapidEye disturbance index was successful in identifying stand-replacing disturbances across the landscape and could be implemented to detect disturbances in near real time.