US Geological Survey


Date of this Version



The Journal of Wildlife Management, Vol. 69, No. 2 (Apr., 2005), pp. 794-799


This article is a U.S. government work, and is not subject to copyright in the United States.


Moist-soil managers manipulate hydrology, soils, and vegetation to provide habitat and foods for waterfowl and other wildlife in seasonally flooded herbaceous wetlands. Increasing seed availability for waterfowl is a priority, but managers also provide resources such as invertebrates, tubers, and browse (Fredrickson and Taylor 1982). An important principle in moist-soil management is maintaining a large component of early-successional plant species whose reproductive strategies include production of abundant seed (Cronk and Fennessy 2001). Low and Bellrose (1944) first referred to the annual species that colonize mudflats as moist-soil plants and documented their potential seed production. Fredrickson and Taylor (1982) developed guidelines for modern moistsoil management in the 1970s and use of moist-soil methods increased rapidly thereafter (Fredrickson 1996). In the Mississippi Alluvial Valley (MAV), state and federal wildlife agencies now manage >8,000 ha in 300 impoundments for moist-soil habitat (U.S. Fish and Wildlife Service 2002). Several methods have been used to quantify seed availability in moist-soil habitats. Harvesting seeds from inflorescences has been the most common method of estimating seed production of individual plant species (Low and Bellrose 1944, Fredrickson and Taylor 1982, Haukos and Smith 1993). Other researchers have tried to simplify estimating seed production by developing species-specific predictive models relating seed yield to plant morphology (Laubhan and Fredrickson 1992; Gray et al. 1999a,b). We believe more effort is needed to develop methods to estimate seed availability for management units rather than for individual species because of the increasing number of impoundments managed and the need to understand the role of moist-soil habitat in meeting food requirements of nonbreeding waterfowl (Reinecke and Loesch 1996, Miller and Newton 1999). Double sampling for stratification (hereafter double sampling; Thompson 1992:143) potentially increases precision of estimates but does not assume that the stratum membership of plots or the sizes of strata are known. Sample units (plots) are assigned to strata during the first sampling period based on predetermined criteria, and stratum sizes are estimated as proportions of plots assigned to strata in the first sample. Successful stratification reduces sampling costs by decreasing the size of the second sample needed to achieve the desired level of precision to inform management decisions. We used double sampling to estimate seed availability in moist-soil impoundments. Our strategy involved 2 sampling steps. We selected a large first sample of plots and used qualitative criteria that we believed were correlated to seed availability to assign plots to different strata (levels of seed availability). Then we selected a second (sub)sample of the first sample, and in these plots we measured seed availability by collecting soil cores and plant inflorescences just before waterfowl arrived. We used double sampling to achieve the increased precision associated with stratified designs, and we measured seed availability by collecting soil cores and inflorescences just before waterfowl arrived to assess the abundance of resources actually available to the birds. Previous studies (Low and Bellrose 1944, Laubhan and Fredrickson 1992, Haukos and Smith 1993) have assumed that no mortality of seeds occurs between the time seeds are harvested by researchers during the growing season and the time waterfowl arrive in fall or winter. Our general objective was to determine if double sampling would provide precise, cost-effective, and unbiased estimates of seed availability in moist-soil impoundments. Specific objectives were to (1) estimate mean seed availability for 3 impoundments in each of 2 years; (2) compare the statistical and cost efficiency of double sampling to that of simple random sampling; and (3) determine if incomplete seed recovery from soil cores leads to biased estimates of seed availability