Human/Environment Interactions, Remote Sensing, and Artificial Neural Networks: Modeling Longleaf Pine Sandhill Leaf Area and Burn History in North-Central Florida

摘要

Longleaf pine sandhills are intensely managed pyric (fire-controlled) ecosystems that occupy 2% of their previous area. The ability to map and monitor sandhill biophysical characteristics, such as leaf area index (LAI), could be a useful tool to stave off succession. This paper examines the use of remote sensing data and artificial neural networks (ANN) to estimate sandhill LAI as a function of time since last burn (0-6 years; 7-9 years; > 9 years) and frequency of burns in the last nine years (0, 1, or 2 burns). Significant differences were found between sandhills that last burned within 6 years and those that last burned 10 or more years earlier, and between sandhills that burned between 7 and 9 years earlier and those that burned 10 or more years earlier, indicating that sandhill LAI increases as a function of time since last burn. Significant differences were also found between each of the frequency groups, indicating that sandhill LAI is inversely related to burn frequency and fire frequency may be more important to sandhill maintenance than time since last burn. The paper concludes by considering the real-world implications of the study and its policy relevance.