Comparison of sequential and double sampling designs for estimating point density within seedling populations

• Published in: Canadian Journal of Forest Research
• Main Author: Newton, P.F.
• Format: Article in Journal/Newspaper
• Language: English
• Published: Canadian Science Publishing 1994
• Subjects: Ecology; Forestry; Global and Planetary Change; Newfoundland
• Online Access: http://dx.doi.org/10.1139/x94-190; http://www.nrcresearchpress.com/doi/pdf/10.1139/x94-190

Sequential and double sampling plans calibrated for estimating point density (mean number of seedlings per quadrat) at a specified level of precision (ratio of the standard error of the mean to the mean) within mixed black spruce (Piceamariana (Mill.) B.S.R)–balsam fir (Abiesbalsamea (L.) Mill.) seedling populations were compared using sampling simulations. Specifically, plans employing a precision level of 0.25 and 10.1-m 2 quadrat size were assessed on the following criteria: (i) total number of samples required; (ii) bias in the resultant point density estimates; and (iii) bias in precision levels attained (i.e., difference between the precision level specified in the plans and the actual levels obtained during sampling). The data base consisted of three population clusters derived from forty-three 0.121-ha sample plots situated on recently disturbed sites throughout central insular Newfoundland. Variation in stocking level and spatial pattern heterogeneity characterized differences between clusters: low stocking levels and highly aggregated spatial patterns (cluster A), moderate stocking levels and aggregated spatial patterns (cluster B), and high stocking levels and slightly aggregated spatial patterns (cluster C). Results indicated that the sequential plan required fewer samples than the double sampling plan irrespective of cluster conditions. The percentage of point density estimates within ±25% of their population cluster means obtained from sequential sampling was 63, 71, and 68% for clusters A, B, and C, respectively. Corresponding values for double sampling were 68, 75, and 73% for clusters A, B, and C, respectively. The percentage of sequential simulations in which the specified precision was attained or exceeded was 34, 64, and 61% for clusters A, B, and C, respectively. Corresponding values for double sampling were 51, 77, and 71% for clusters A, B, and C, respectively. Although these results suggest that double sampling may be a practical alternative to sequential sampling, further research into ...