Laserfiche WebLink
Relative bias for three species is presented in Figure 7. Kentucky bluegrass is <br />presented to represent an abundant species (the most abundant; 27% relative plant cover), <br />black medick (Medicago lupulina) represents a moderately abundant species (2% relative <br />plant cover), and a sedge (Carex emoryi) represents a species in low abundance (0.6 % <br />relative plant cover). Relative bias of the estimates for each of these species was typical <br />of the bias of relative percent plant cover for species of similar abundance. In general, as <br />the number of sample points increased, the relative bias of the estimates tended toward <br />zero. <br />Proportion confidence interval coverage for the same three species is presented in <br />Figure 8. Proportion confidence interval coverage of the estimates for each of these <br />species was typical of the coverage of relative percent plant cover for species of similar <br />abundance. In general, as the number of sample points increased, the proportion <br />confidence interval coverage of the estimates increased. <br />Relative precision for the same three species is presented in Figure 9. Relative <br />precision of the estimates for each of these species was typical of the precision of relative <br />percent plant cover for species of similar abundance. In general, as the number of sample <br />points increased, the precision of the estimates increased (variance to mean ratio <br />decreased). <br />Program -level Sampling Design Evaluation <br />Program -level PCQ Level of Effort <br />We evaluated the bias, confidence interval coverage, and precision of density <br />estimates under alternate sampling intensities. The tree and shrub density estimates were <br />recalculated using a dataset where the point- centered quarter data (species and distance) <br />had been randomly re- sampled with replacement from the original dataset. The re- <br />sampled datasets were created to simulate from 5 to 60 transects (in increments of 5) for <br />both Cottonwood Ranch and Jeffrey Island and point- centered quarter spacing of 100, <br />200, and 300m along each transect. The sampling procedure and calculation of the <br />estimates was repeated 1000 times. <br />Relative bias was calculated as the mean (over 1000 iterations) of the difference <br />between the re- sampled estimate and the true parameter value, divided by the true <br />parameter value. In this case, the true parameter value was taken as the estimate obtained <br />with all the data. Confidence interval coverage was defined as the proportion of <br />iterations (out of 1000) where confidence interval estimates contained the true parameter <br />value. Relative precision of density estimates was estimated as the mean (over 1000 <br />iterations) of the ratio of the estimated half -width of the density confidence interval, <br />divided by the estimated density. Low values of this ratio indicate more precision <br />associated with the estimates. <br />Relative bias of the tree density estimates ranged from -0.2% to 2.1% for <br />Cottonwood Ranch and from -0.3% to 3.5% for Jeffrey Island. Relative bias of the <br />shrub /sapling density estimates ranged from 3.2% to 104.1 % for Cottonwood Ranch and <br />from 0.7% to 24.3% for Jeffrey Island. Relative bias decreased for both tree and <br />shrub /sapling density estimates as the number of sample transects increased (Figures 10 <br />through 13). The decreased spacing of points along the transect resulted in lower bias <br />