Effects of Large-scale Wildfires on Rodents in Southern California Denise R. Clark*, Cheryl S. Brehme, Carlton J. Rochester, and Robert N. Fisher U.S. Geological Survey, Western Ecological Research Center, San Diego Field Station, 4165 Spruance Road, Suite 200, San Diego, California, U.S.A. * Phone: (619) Fax: (619) Introduction: In October 2003, wildfires burned large areas of San Diego County. After the fires, we surveyed rodents in both burned and unburned control plots across four sites and four vegetation types using a combination of pitfall and Sherman live-traps. Prior to the 2003 wildfires, we conducted studies at one site that included rodent surveys, habitat quality assessments and vegetation characterization using established methods (Table 1). Based on previous studies, we expected to see shifts in rodent community composition after fire and to see changes in the density and distribution of individual rodent species. The objectives of this study were to determine the effects of fire on rodent species diversity, community composition and relative abundances of individual rodent species. Trapping Methods: Unbaited Pitfall Trap Array (Figure 2) Three 15-m arms of drift fence Seven 19-L pitfall traps Three hardware-cloth funnel traps Baited Sherman Traps (Figure 2) 9 small traps (8 x 9 x 30.5 cm) 9 large traps (10 x 11.5 x 38 cm) Data Collection: All captured animals were processed by recording species, sex, reproductive condition, age class, weight, length and hair clipping for identification of recapture. Figure 2. Study plot design. Study Area: Four study sites (Figure 1) Each site was impacted by 2003 wildfires in part or in whole Four vegetation types Chaparral (CHAP) Coastal sage scrub (CSS) Grassland (GRASS) Woodland/riparian (WR) Figure 1. The four study sites in San Diego, CA, were Little Cedar Ridge (CED), Elliott Chaparral Reserve (ELL), Rancho Jamul Ecological Reserve (RAJ), and Santa Ysabel Open Space Preserve (SYR). Prefire sampling Postfire sampling Unburned plots Burned Plots Elliot--010 Cedar--09 Rancho Jamul2010 Santa Ysabel01014 Total20 43 Discussion: There were no significant changes in overall rodent species diversity two to three years after the 2003 fires, but there were significant changes in rodent community composition in coastal sage scrub, grassland and chaparral vegetation types. Postfire changes found in these rodent communities were likely caused by both direct and indirect effects of the fires. Direct effects include differences in species mortality, while indirect effects include changes in rodent competitive interactions and postfire habitat suitability. Individual rodent species largely responded in a manner consistent with their life history characteristics. Generalist and open habitat specialists generally increased in relative abundance, while closed habitat specialists decreased. Among habitats, significant increases in relative abundance were found for the harvest mouse (Reithrodontomys megalotis), Dulzura kangaroo rat (Dipodomys simulans) and deer mouse (Peromyscus maniculatus). In contrast, significant decreases in relative abundance were found for the California mouse (Peromyscus californicus), California pocket mouse (Chaetodipus californicus), desert woodrat (Neotoma lepida) and the dusky footed woodrat (Neotoma macrotis) which prefer shrubs, trees and closed habitat. Figure 4. Changes of relative abundance of rodent species in coastal sage scrub, grassland and chaparral communities, shown with standard errors. Change (postfire minus prefire values) estimated separately for impact and reference plots in CSS (A.) and GRASS (B.) communities. C) Relative abundance of rodent species from postfire surveys of impact and reference plots in chaparral communities. Levels of significance is indicated as follows: *** p < 0.001, * p < 0.05, + p < Species codes are Chaetodipus californicus (CHCA), Chaetodipus fallax (CHFA), Dipodomys simulans (DISI), Microtus californicus (MICA), Neotoma lepida (NELE), Neotoma macrotis (NEMA), Peromyscus eremicus (PEER), Peromyscus maniculatus (PEMA), Peromyscus californicus (PECA) and Reithrodontomys megalotis (REME). Statistical Analyses: To reduce seasonal and daily variability, we averaged the number of captures for each species per sample session. Sample plots were treated as replicates. Prefire and postfire averaged capture rates were used as the basis for all further analyses. Effect of Burn on Species Diversity [Shannon Index (SI)] : Mixed effects models [Random factor = site, Fixed factors = burn condition (reference/impact plots), vegetation type, % shrub/tree cover, elevation] Models ranked using Akaike’s Information Criterion (AIC) Effect of Burn on Community Composition (Multivariate Analyses) : Non-metric multidimensional scaling (MDS; Primer-E software) Analysis of similarity (ANOSIM) - Factors: Vegetation type, site or both – if significant, test effect of burn on each vegetation type separately controlling for site as fixed effect - Factors: burned (postfire impact plots) vs. unburned (prefire reference plots, postfire reference plots and prefire impact plots) Biological-environmental matching (BIOENV) - Percent shrub/tree cover Single Species Comparisons (Contrast Tests by Vegetation Type and Burn Condition) CSS and GRASS plots compared postfire (impact - reference plots) - prefire (impact - reference plots) CHAP and WR plots compared postfire reference to postfire impact plots Acknowledgements: We thank the field crew for their many hours of hard work. Thanks to Dr. Julie Yee who provided statistical advice. This project benefited greatly from reviews by Dylan Schwilk and Dr. Julie Yee. For site access, we thank the managers of the four study sites (California Department of Fish & Game, San Diego County Parks & Recreation Department, University of California, San Diego - Elliott Chaparral Reserve, Bureau of Land Management). Funding for this project was provided in part by the San Diego MSCP Initiative and the San Diego Association of Governments. Use of trade names does not imply USGS endorsement. Results: The percentage of shrub and tree cover: Reference (unburned) plots Impact (burned) plots CSS 53% (se = 4.30) 13% (se = 3.7) CHAP 67% (se = 4.31) 30% (se = 3.89) GRASS 3.8% (se = 1.81) 2% (se = 1.26) WR 34% (se = 7.55) 24% (se = 6.35) Mixed Effects Models: “No difference in diversity between reference & impact plots:” Top model accounted for 62.4% of model weights and showed both ‘burn condition’ and vegetation type to be predictive of rodent diversity. Rodent diversity was highest in coastal sage scrub and chaparral, with SI of 1.11 (se = 0.11) and 1.02 (se = 0.13), respectively. Woodland was at 0.71 (se = 0.13). Rodent diversity was least in grasslands with SI of 0.46 (se = 0.13). Multivariate Community Analyses: ANOSIM tests showed significant differences in composition of the rodent communities among vegetation types (CSS, GRASS and WR) and sites. Rodent community composition in burned plots was significantly different from the composition in unburned plots in coastal sage scrub (R = 0.23, p = 0.050), chaparral (R = 0.714, p = 0.067) and grassland (R = 0.40, p = 0.021) but not woodland (R = 0.009, p = 0.444). BIOENV showed that the percentage of shrub and tree cover was predictive of the change in rodent community composition for coastal sage scrub, (ρ = 0.195, p = 0.005; Figure 3). Postfire Changes in Relative Abundance of Rodent Species by Vegetation Type: (Figure 4) CSS plots showed statistically significant differences in the proportions of three species in impact plots. GRASS plots produced significant results for four species. Five species had significant differences in CHAP plots. Figure 3. Non-metric multi-dimensional scaling plots (MDS) based on Bray-Curtis similarity matrix using square root transformed capture data show how the rodent communities in burned plots (B) of coastal sage scrub have shifted away from the communities detected in the unburned plots (U). Bubble size represent the relative cover from shrubs and trees. Reduced cover was a significant predictor of the change in rodent community composition in burned plots. Distance between bubbles represents how similar plots are to each other. RAJ array 36 (prefire)RAJ array 36 (1 year postfire) A. B. C. SYR array 11 (prefire) SYR array 11 (4 years postfire) Table 1. Number of unburned and burned plots sampled by study site Reithrodontomys megalotis Dipodomys simulans Chaetodipus fallax Neotoma lepida CSS Rodent Community GRASS Rodent Community CHAP Rodent Community