Elsevier, International Journal of Applied Earth Observation and Geoinformation, (45), p. 42-54
DOI: 10.1016/j.jag.2015.10.009
Full text: Download
In this study we present, evaluate and validate an approach to monitor forest and grassland disturbance. We apply the approach to both Landsat and MODIS imagery for the North Island of New Zealand and validate the results based on high resolution OrbView and Ikonos imagery. We found an overall accuracy of the disturbance index of 98% for the two studied land cover types. The kappa value was 0.770 indicating a 77% better agreement than what would have occurred by chance. We found that there is a difference between the accuracy received for grassland areas compared to the accuracy received for forest areas, with the grassland areas outperforming the forest areas (Kappa of 0.855 vs. 0.656). We split the validation results by soil type and also evaluate the effect of different soil types with respect to grazing pressures. The disturbance index behaved consistently for all available soil orders.