Quantifying post-fire ponderosa pine snags using GIS techniques on scanned aerial photographs

Snags are an important component of forest ecosystems because of their utility in forest-nutrient cycling and provision of critical wildlife habitat, as well as associated fuel management concerns relating to coarse woody debris (CWD). Knowledge of snag and CWD trajectories are needed for land managers to plan for long-term ecosystem change in post-fire regimes. This need will likely be exacerbated by increasingly warm and dry climatic conditions projected for the U.S. Southwest. One of the best prospects for studying fire-induced landscape change beyond the plot scale, but still at a resolution sufficient to resolve individual snags, is to utilize the available aerial photography record. Previous field-based studies of snag and CWD loads in the Southwest have relied on regional chronosequences to judge the recovery dynamic of ponderosa pine (Pinus ponderosa) burns. This previous research has been spatially and temporally restricted because of field survey extent limitations and uncertainty associated with the chronosequence approach (i.e., space-for-time substitution), which does not consider differences between specific site conditions and histories. This study develops highly automated methods for remotely quantifying and characterizing the spatial and temporal distribution of large snags associated with severe forest fires from very high resolution (VHR) landscape imagery I obtained from scans of aerial photos. Associated algorithms utilize the sharp edges, shape, shadow, and contrast characteristics of snags to enable feature recognition. Additionally, using snag shadow length, image acquisition time, and location information, heights were estimated for each identified snag. Furthermore, a novel solution was developed for extracting individual snags from areas of high snag density by overlaying parallel lines in the direction of the snag shadows and extracting local maxima lines contained by each snag polygon. Field survey data coincident to imagery coverage for post-fire ponderosa pine forests allowed calibration and accuracy assessment of these new tools. These new methods may allow for broader estimation of snag dynamics in post fire landscapes while significantly lowering the human and material costs of conducting such surveys. Outcomes for these methods were mixed. Both the snag count and snag height values were chronically underestimated using a feature extraction method and an edge detection method; so, an adjustment constant was developed for the categories of each method. Average accuracies ranged from 54 to 46% lower than the field-based values for the count attribute and 2-12% over the ground values for the snag height attribute using these two approaches. These methods show much promise, and are less resource intensive than field surveys, but more research is needed to improve overall accuracy.