Application of distance sampling for Pima pineapple cactus (Coryphantha scheeri var. robustispina) population estimation and monitoring

Efficient and accurate surveys methods are essential for understanding abundance and habitat relationships of plants and wildlife, especially species of conservation concern. The Pima Pineapple Cactus (Coryphantha scheeri var. robustispina) is a federally-listed endangered species in southern Arizona that is commonly surveyed for compliance with federal law and for research and monitoring. We tested a new survey method for this species based on distance sampling (DS), which involves measuring distances to focal objects from lines or points and modeling a detection function that adjusts estimates of abundance for variation in detection probability. We compared estimates of density and population size from DS with values obtained with the recommended survey protocol (Roller method) that focuses on a complete census of all individuals in a focal area. Additionally, we assessed factors that influenced detection probability and detection distance, associations between local estimates of density from DS and various environmental factors, and summarized recommendations for future applications of DS in this system. We recorded 105 live and 15 dead Pima Pineapple Cacti while DS along 36.9 km of line transects at 11 focal sites in the Brawley and Santa Cruz watersheds. Density averaged 1.47 live individuals/ha with an estimated total of 294 individuals overall based on DS, and precision was high (CV= 0.139). Based on what we presumed to be known values from census efforts, both density and abundance were fairly well estimated by DS with bias equaling -11.4% across the population. Moreover, estimates from DS were highly correlated with values from censuses on both the untransformed (r = 0.82, p = 0.002), and especially, logarithmic scales (r = 0.92, p < 0.001). Estimates of detection probability averaged 0.49 (CV = 0.076) overall, and 0.96 and 0.92 at 2 and 3 m from lines, respectively, suggesting results based on the Roller method fail to detect 4-8% of individuals on average. Detection probability declined as soils became increasingly dominated by larger rocky particles compared to those dominated by sand or silt. Detection probability also seemed to decline with increasing understory vegetation volume, grass cover, and decreasing cactus height, although additional effort is needed to confirm these patterns. Local densities increased with increasing slope and soil substrate size, and decreased with increasing understory vegetation volume (p ≤ 0.022). There was some evidence that bias of estimates among focal sites increased as grass cover increased, suggesting more focused survey effort in areas of dense grass near lines will improve accuracy of DS. Although our results indicate DS is an effective method for surveying this species for research and monitoring applications, reducing the spacing of neighboring line transects from 50 to 40 m, surveying perpendicular to elevation contours, and other recommendations we present should improve
accuracy.