Lizard community data were collected as part of the National Park Service Inventory and Monitoring Program (science.nature.nps.gov/im/). Recognizing the severe data limitations for ectotherm vertebrates at regional scales we here assume that each park represents one community. Data was accessed through the NPSpecies portal and through requesting reports from individual park units. Inventories were initiated within all national park units identified as containing significant natural resources with goals including documenting at least 90% of the species present and establishing a monitoring baseline. We focused on the data collected within the Southern Colorado Plateau Inventory & Monitoring Network (SCPN) between 2001 and 2003 as the parks were consistently intensively surveyed with robust methods.
We assessed both quantitative and qualitative quality control measures for data from an initial set of 54 parks. We then selected only those parks that had at least 50 person hours of survey efforts, had areas of at least 2 km2, and were qualitatively considered to have utilized consistent, repeatable methods. Further, we excluded parks without representative and complete coverage of all habitats. This resulted in a final set of 18 national parks, recreation areas and monuments (including park codes) as follows (Drost, Persons & Nowak 2001; Persons & Nowak 2004, 2006a,b, 2007; Prival & Goode 2005; Persons, Nowak & Hillard 2006): Amistad (AIMS), Bandelier (BAND), Big Bend (BIBE), Carlsbad Caverns (CAVE), Chaco Culture (CHCU), Death Valley (DEVA), El Malpais (ELMA), Fort Davis (FODA), Glen Canyon (GLCA), Guadalupe Mountains (GUMO), Hovenweep (HOVE), Manzanar (MANZ), Mojave, (MOJA), Petrified Forest (PEFO), Petroglyph (PETR), Salinas Pueblo Missions (SAPU), White Sands (WHSA) and Wupatki (WUPA). The primary sampling method that we included in our analysis was time–area constrained search, where both the time spent searching and area covered are standardized (Crump & Scott 1994). These searches were generally constrained for 1 h within 1 ha plots. For the larger parks (i.e. BAND, CHCU, ELMA, PETR, SUCR, WACA and WUPA), plots were randomly allocated within the park boundaries. The survey effort per park averaged 290 person hours (median: 141; range: 52–940). Species richness averaged 10·2 with a median of 9·5 (range: 5–18). The number of individuals sampled per park averaged 779 with a median of 556 (range: 124–2074). Abundances were normalized by search effort (lizards/person hour) throughout. Our findings are robust to sampling biases induced by differential detectability of species (Appendix S1, Supporting information).
We used park boundary polygons to extract mean environmental data (temperature, precipitation and elevation) for each park (data from 1961 to 1990 with 10′ latitude/longitude resolution; New et al. 2002). To estimate energy availability, we use consensus mean annual NPP estimates compiled from numerous models by the Potsdam institute (g C m−2, 30′ resolution, Cramer et al. 1999).