Semi-Automated Sample-Based Forest Degradation Monitoring with Photointerpretation of High-Resolution Imagery

Abstract

Forest fragmentation and degradation are a problem in many areas of the world and are a cause for concern to land managers. Similarly, countries interested in curtailing climate change have a keen interest in monitoring forest degradation. Traditional methods for measuring forested landscape pattern dynamics with maps made from classified satellite imagery fall short with respect to the compatibility of their forest definitions with information needs. In addition, they are not easily amenable to interpretation using tools like confidence intervals derived from survey sampling theory. In this paper, we described a novel landscape monitoring approach that helps fill these gaps. In it, a grid of photo plots is efficiently created and overlaid on high-resolution imagery, points are labeled with respect to their land-use by a human interpreter, and mean values and their variance are calculated for a suite of point-based fragmentation metrics related to forest degradation. We presented three case studies employing this approach from the US states of Maryland and Pennsylvania, highlighted different survey sampling paradigms, and discussed the strengths and weaknesses of the method relative to traditional, satellite imagery-based approaches. Results indicate that the scale of forest fragmentation in Maryland is between 250 and 1000 m, and this agrees with compatible estimates derived from raster analytical methods. There is a positive relationship between an index of housing construction and change in forest aggregation as measured by our metrics, and strong agreement between metric values collected by human interpretation of imagery and those obtained from a land cover map from the same period. We showed how the metrics respond to simulated degradation, and offered suggestions for practitioners interested in leveraging rapid photointerpretation for forest degradation monitoring.