Fire was historically the dominant disturbance in southern grasslands ecosystems in United States. Prescribed burning is a current management technique utilized in these systems for species and ecosystem function restoration.  However, knowledge of fire behaviors and effects are limited in these systems. Spatially-explicit fire models, such as FARSITE, were originally developed for western grass and forests ecosystems thus predictive accuracy in southern grasslands is relatively unknown. In this research, the spatial scale of FARSITE predicted fire behaviors and effects were assessed based on comparison of field observations and FARSITE simulations utilizing three different spatial resolutions of fuel map data. These fuel maps were derived from remote sensing sources including digital orthographic quadrangle aerial photography (i.e. DOQ), advanced spaceborne thermal emission and reflection radiometer (ASTER), Landsat 7 enhanced thematic mapper plus (Landsat ETM+).  Fine dead fuels were mapped based on a vegetation index (i.e. NDVI) calculated from these remote sensing data to generate sub-classes of general fuel models described by Anderson (1982).   To test predictive accuracy of the model, two prescribed burns were conducted on the grasslands at Camp Swift, near Austin, TX. The burns were conducted on January 10th, 2008, and July 18th, 2008, with the ignition locations, air temperature, humidity and wind, area of burned, time of arrival of fire, and temperature of fire collected for each fire.

Results/Conclusions

The FARSITE simulations show that, for burn site 1, predicted burn areas was 56 ha when using fine-scale resolution fuel map derived from DOQ data, which is the closest one to actual burn area of 42 ha. A similar result was obtained for the second field burned in this study. All results show that fine scale remote sensing data provides best information for accurately simulating fire area. For the time of arrival, FARSITE simulation with the medium-scale resolution fuel map derived from the ASTER data were the most accurate with on average 10 minute difference between predicted and observed values. For predicted heat energy release, derived from surface temperature, point values were not correlated with observed values, though averages were similar for both fires. These results indicate that spatial heterogeneity of fuel may play an important role in fire heat energy release and ecological impacts which were not captured by non-synoptic remote sensing data used in this study.  Prediction of fires in grasslands is limited by our detailed knowledge about mapping fine fuel loading, structure, contiguity, and interannual variability.