Factors influencing the precision of soil seed bank estimates

Abstract

The dimension of soil augers needed to sample a seed bank of Chenopodium spp. (lamb's-quarters) was determined by randomly sampling a 1.35-ha area within a cornfield in Oxford County, Ontario. Sampling units of three different auger sizes (1.9, 2.7, and 3.3 cm in diameter) were collected. On a per volume basis, there were no significant differences between the three sizes of auger in estimating the number of lamb's-quarters seeds in the soil. Three sampling methods, systematic, stratified random, and cluster, were compared with random sampling in their capacity to minimize the sampling variance. Soil cores of 1.9 cm diameter and 15 cm deep were taken systematically at 3.5-m intervals to form a 32 × 32 matrix. Repeated sampling within the matrix using Monte Carlo techniques indicated that the estimate of sampling variance decreased with increasing sample size, regardless of the sampling method used. No fewer than 60 sampling units should be collected to quantify the seed bank of an abundant weed such as lamb's-quarters. The estimates of sampling variance of systematic and cluster sampling were clearly influenced by the sampling interval and the cluster's shape, respectively. This was attributed to the underlying seed distribution of lamb's-quarters in the soil that was clustered with patterns of high and low seed density parallel to corn rows. There were no significant differences between the estimate of sampling variance of random and stratified random sampling with a fixed sample size of 64 units.