Northern Prairie Wildlife Research Center
Prairie vegetation in North America extends from southern Canada into northern Mexico, and occurs throughout much of Texas (Risser et al. 1981). We selected study sites comprised of prairie vegetation in the Gulf Prairies and Marshes and Cross Timbers and Prairies vegetation areas in Texas (Correll and Johnston 1970), and refer to these areas as coastal prairie and central prairie, respectively. Vegetation is typical tallgrass prairie underlain by acid sands, sandy loam, and clay soils along the coast and brown, neutral to slightly acid, sandy or clay loams in the central area. The coastal prairie is characterized by level topography dissected by numerous streams, swamps, and fresh-water marshes, whereas the central prairie is rolling to hilly, and deeply dissected with rapid surface drainage. Average rainfall is 760 to 1,300 mm along the coast and 625 to 1,000 mm in the central prairie.
The coastal prairie occurs along the Gulf coast from southwestern Texas to western Louisiana (Diamond and Smeins 1984, Smeins et al. 1991). Our study area was 13 km south of Sealy, Texas, near the junction of the Gulf Prairies and Marshes and Post Oak Savannah vegetation areas, approximately 100 km from the coast (Correll and Johnston 1970). Eight sites were selected to represent the diversity of habitats:
Our study site in the central prairie was located 10 km north of Glen Rose, Texas, within the Cross Timbers and Prairies vegetation area of central Texas, 300 km to the north of the coastal prairie locality (Correll and Johnston 1970). Three study sites encompassed the diversity of habitat types found here:
Sweep sampling was standardized by having the same person sample the same vegetational association in each site. McFadden (1978) constructed a species-area curve for coastal prairie near Houston, Texas, by plotting the number of sweeps against the number of insect species collected. He found that the asymptote of the species-area curve was at 600 sweeps with 80% of the entomofauna collected. We adopted this level of sampling efficiency for our study, and used 600 sweeps as our minimal standard. We used net diameter and coverage of the net opening by vegetation to calibrate the number of sweeps required to sample an equal amount of plant biomass at each study site. For the coastal prairie, the number of sweeps was set at 600 for Sites 1 and 3800 for Site 2, 1200 for Sites 4, 5, 7, and 8, and 2400 for Site 6. In the central prairie, the number of sweeps taken was 500 for Site 1, 750 for Site 2, and 600 for Site 3. For woody vegetation in both areas, 200 sweeps were taken between the ground and 3 m (details in Tables 1 and 2).
Insects were separated from vegetation within 36 h and were identified to family and morphospecies (Oliver and Beattie [1996b] validated the adequacy of using morphospecies to study biodiversity). Each taxon was classified trophically as herbivore, predator, parasite, or detritivore by inspection of mouth parts and using known food habits of insect families (Chu 1949, Imms 1957, Cole 1969, Borror et al. 1989). Larval food habits were used for trophic classification of species whose feeding habits changed during development. For species where both young and adult stages feed, we scored the species for both habits.
Alpha diversity was computed for each study site with the Shannon-Wiener index (H'; Krebs 1999), and equitability was computed as J = H'/Hmax. Beta diversity was expressed by computing average similarity among sites (Magurran 1988). Similarity in species composition between any two sites was assessed with Sørensen's Coefficient of Similarity [S = (a + b)/2c, where a = species at one site, b = species at a second site, c = species at both sites; Krebs 1999]. Cluster analysis provided closest-pair associations among sites. The unweighted arithmetic averages method joined areas, which were mutually closest and successively recomputed a reduced similarity matrix among sites by considering the newly joined pair as one (Sokal and Michener 1958, Farris 1969, Sneath and Sokal 1973). Gamma diversity was expressed as the total number of species collected among all sites at each geographic locality. Log10 was used for diversity calculations.
Taxonomic comparisons between the coastal prairie and central prairie study areas were tested by summing total number of species across habitats within insect orders. Differences in taxonomic and trophic structure among sites within the coastal prairie and central prairie study area were tested with log-likelihood ratio tests (G-tests; Sokal and Rohlf 1995). We used a posteriori tests to identify statistically homogeneous subsets of sweeps/areas, taxa, or trophic groups (Sokal and Rohlf 1995). Any group, e.g., taxa, trophic level, not in the subset would add significant heterogeneity to yield a G-statistic greater than the critical value. This procedure generated a plethora of nonsignificant subsets resulting from combinations of groups and their subsets. To minimize the total number of comparisons, only those that encompassed the entire set of sites, taxa, or trophic groups were included. Results are reported for analyses that only included those orders with the highest number of species (Homoptera through Hymenoptera on Tables 1 and 2) because the G-test is sensitive to low sample sizes (Sokal and Rohlf 1995). Hence, our significance tests were statistically conservative.