Northern Prairie Wildlife Research Center
During the spring of 1994, seven oak savannas were selected to be treated with prescribed fire between fall 1994 and spring 1996 and fit our definition of an established oak savanna (Figure 1). In this study, an established savanna was defined as an area with an overstory crown closure between 10 and 80%, a herbaceous layer that contained both prairie and forest species, and a current prescribed fire management regime (Nuzzo 1986). In addition, a shrub layer may or may not be present. The seven study sites encompassed a range of savanna conditions from almost completely open to nearly closed canopy, as reflected in the basal areas for each site (Table 2).
The climates of the study sites are similar, averaging approximately 40 in. of precipitation per year with the wettest months being April, May, and June (May receiving the most precipitation: 4.78 in.) while the driest months are December, January, and February (January receiving the least precipitation: 1.70 in.). Annual high and low temperatures average 65.2°F and 44.5°F, respectively, with July being the warmest month on average (78.3°F) and January the coldest month on average (28°F) (Pers. comm. Adnan Akyüz, Missouri State Climatologist).
|Figure 1. Study site locations within Missouri.|
Fire behavior data were collected along fire behavior transects randomly placed within each savanna and allocated proportionally based on the size of the study sites (Table 3). Rate-of-spread (ROS) data were collected using easily made, inexpensive, reusable rate-of-spread clocks (Blank and Simard 1983, Grabner et al. 1997, Simard et al. 1984). In cases where ROS data was not available, it was usually because of clock failure (Table 3).
BEHAVE predictions were analyzed using simple linear regression techniques. Predictions were considered reliable if the slope of the regression line was not different from one, and the intercept was not different from zero (Sneeuwjagt and Frandsen 1977, Andrews 1980, Smith and Rose 1995). Residual error between observed and predicted ROS was calculated as observed minus predicted, where a negative residual error indicated that ROS was underpredicted (Reynolds 1984). The percent prediction error was also calculated as the absolute difference in predicted and observed ROS values divided by the larger of the two values (Andrews 1980). Data was analyzed by comparing each predicted and observed ROS (individual ROS) and predicted and observed ROS summarized by study site (mean ROS). Comparing each predicted and observed ROS, the reliability of BEHAVE predictions were tested for specific locations within a burn unit. Comparing ROS data summarized by study site is a more appropriate test of BEHAVE predictions because BEHAVE predictions are based on average rather than specific fuel conditions (Bushey 1985).