2003 Piru Fire Historical Imagery

The increased occurrences of drought and fire may be contributing to the loss of biodiverse ecosystems in Mediterranean regions. Specifically, the conversion of diverse native shrublands, such as chaparral, to non-native annual grassland by fire is of great conservation concern in California. To avoid or slow the loss of chaparral, it is important to understand the underlying causes of landscape conversion. Studies investigating the interaction of multiple potential drivers are particularly crucial to identification of vulnerable areas of the landscape. Here we used aerial imagery to evaluate vegetation transitions between chaparral, sage scrub, grassland, and tree domination and their potential drivers within Ventura County, California, a strongly Mediterranean climate region. We used random forest algorithms and conditional inference trees to determine the climatic, topographic, and fire-related variables contributing most to vegetation change. Our results support that chaparral conversion to grass (27% of chaparral plots) is a result of landscape position, fire, and drought acting in tandem. In particular, lower elevation, southwest facing slopes that experience a post fire drought are at very high likelihood of conversion to non-native annual grass. Additionally, our results show that these grasslands, once formed, rarely convert to other community types. Therefore, protecting shrub dominated areas that are most likely to convert (low elevation, more southwest facing slopes, less annual precipitation) is crucial to preserving native vegetation diversity. Methods 1930 aerial imagery was obtained from the Map and Image Laboratory at the University of California, Santa Barbara. Imagery is from the 1930 flight C-870 and was mapped on a scale of 1:18,000. The 2009 imagery was obtained from the United States Geological Survey Digital Orthophoto Quarter Quads and is a true color aerial photograph at one-meter spatial resolution. The 1930 imagery was georectified to the 2009 imagery using 300-400 ground control points. Ground control points were selected from temporally stable objects such as trees, rock outcrops, roads, and permanent structures. Each image was then warped using triangulation and pixels were resampled to the nearest neighbor, creating a georectified image with one-meter spatial resolution. Mosaicked images were then validated for spatial accuracy by identifying 40-100 ground control points and ensuring a root mean square error of 10 pixels or less.