Published online by Cambridge University Press: 20 January 2017
Weed management could be more efficient and require less herbicide if growers could afford to estimate the composition, density, and distribution of weed seed banks. Spatial distribution of a weed seed bank will affect the accuracy of both mean estimates and interpolated maps of density. Consequently, information about the general characteristics of spatial distributions of seeds in a seed bank is needed to identify the most efficient strategies for sampling. Seed banks were sampled on 8.4-m square grids in eight irrigated corn fields to identify the common features of distributions of seed banks of annual weeds. Spatial dependence was described with correlograms for four to eight species in each field. Spatial dependence was detected for 36 of 45 distributions, and seed counts were correlated to an average distance of 25 to 150 m for a distribution. Seed banks of different species and fields had common features of spatial correlation: spatial pattern accounted for less than half of the total variability of seed counts, spatial correlation decreased rapidly over short distances, and ranges of spatial dependence varied with direction. For half of the distributions, the maximum range of spatial dependence was at least twice as long as the minimum range. Seed counts were correlated for the longest distances in the direction of the crop row for 16 distributions, and the distance was longer in the direction of the crop row than across rows for 26 of the 36 samples. Researchers should be able to design more efficient sampling plans for growers if the common features of spatial dependence are considered. For seed banks like these, the accuracy of maps and estimates of seed bank density may be improved by collecting multiple cores around each sampling location to mitigate the effect of short-scale spatial variability. In addition, sampling may be more efficient with grids and interpolation methods that account for ranges that are 1.5 to 2 times longer in the direction of the crop row than perpendicular to the row. With a 55- by 30-m sampling grid, adjacent observations would be correlated, and maps could be made for 80% of these seed banks. More closely spaced observations would be needed to describe the rapid decline in spatial correlation with distance for a more accurate or finer-scale map. Whether sampling seed banks for making management decisions will be cost-effective is not clear. However, potential methods to sample and map seed bank distributions more efficiently have not been exhausted.