| dc.description.abstract | Wildlife management is essentially the balance between maintenance of habitat and control of population density. To demonstrate the application of multivariate techniques for habitat assessment, I evaluated 4 contemporary classification schemes for use as experimental units for mourning dove (Zenaida macroura) research in Texas. I conducted a generalized canonical discriminant analysis (CDA) for each classification scheme using 25 habitat variables obtained adjacent to each of the 133 U.S. Fish and Wildlife Services call-count survey routes within Texas. Classification results from each CDA were used to generate a confusion matrix for each classification scheme (i.e., overall accuracy, average accuracy, and expected agreement). Because classification schemes differed in the number of categories, the Kappa Coefficient of Agreement was used to account for the proportion of agreement due to chance. The Kappa estimates were higher for the Gould (0.760) and Omernik (0.700) classification schemes, than for the Fenneman (0.618) or George (0.673) classification schemes, indicating the newer classification schemes provide a more accurate partitioning of multidimensional habitat space, and are therefore better suited for use as experimental units for mourning dove research in Texas. To demonstrate the impact of human land use on wildlife habitat, I evaluated the spatial-temporal effects of habitat loss and anthropogenic land use on grassland birds from 1993–2012. I used 8 habitat metrics corresponding to the U.S. Census of Agriculture data for Texas during this period, and northern bobwhite (Colinus virginianus) abundance estimates from the Breeding Bird Survey and Texas Parks and Wildlife Department as the proxy grassland bird species. The redundancy analysis indicated that economic, agricultural, and land use metrics accounted for 74.5% of the total variance in bobwhite relative abundance during the period (Radj ² = 60.8%, P < 0.0016), and most anthropogenic land trend variables (e.g., Population Density, Market Value, Production Value) were inversely proportional to quail relative abundance. The canonical discriminant analysis indicated that economic, agricultural, and land use metrics explained 88.6% of the variability among ecoregions (P < 0.0002) and 99.5% of the variability among years (P < 0.0167). These results indicate that land values (market value and production value per hectare) and human population density may signal the onset of anthropogenic land conversion, and might be used to predict future changes that will impact grassland bird species and other natural resources. Finally, to demonstrate the feasibility of combining scientific and citizen-science data to obtain a regional estimate of grassland bird abundance, I obtained congruent estimates of northern bobwhite (Colinus virginianus) abundance using a double-sampling paradigm. Spring cock call-counts were conducted on 12 ranches within the Rolling Plains of Texas during 2012–2014. This sampling effort collected calls and distances at each point, yielding 1,022 total counts, detected 36,415 calls, 4,647 birds, and obtained 4,627 distances. Data were analyzed using program DISTANCE to generate local and regional estimates of quail density for each year, and to calibrate density estimates with birds heard using a double-sampling paradigm. My results demonstrated that it is economically feasible and logistically pragmatic to calibrate metrics obtained through citizen-science efforts (call-counts; relative abundance) with results obtained by more intensive scientific methods (distance sampling; density estimates). Collectively, these results illustrate that it is within the microcosm of single-species management that we test the limits of our ecological knowledge and understanding. | en |
