APPLICATION OF A MULTIPURPOSE UNEQUAL PROBABILITY STREAM SURVEY IN THE MID‐ATLANTIC COASTAL PLAIN¹

Abstract

A stratified, spatially balanced sample with unequal probability selection was used to design a multipurpose survey of headwater streams in the Mid‐Atlantic Coastal Plain. Objectives for the survey include unbiased estimates of regional stream conditions, and adequate coverage of unusual but significant environmental settings to support empirical modeling of the factors affecting those conditions. The design and field application of the survey are discussed in light of these multiple objectives. A probability (random) sample of 175 first‐order nontidal streams was selected for synoptic sampling of water chemistry and benthic and riparian ecology during late winter and spring 2000. Twenty‐five streams were selected within each of seven hydrogeologic subre‐gions (strata) that were delineated on the basis of physiography and surficial geology. In each subregion, unequal inclusion probabilities were used to provide an approximately even distribution of streams along a gradient of forested to developed (agricultural or urban) land in the contributing watershed. Alternate streams were also selected. Alternates were included in groups of five in each subregion when field reconnaissance demonstrated that primary streams were inaccessible or otherwise unusable. Despite the rejection and replacement of a considerable number of primary streams during reconnaissance (up to 40 percent in one subregion), the desired land use distribution was maintained within each hydrogeologic subregion without sacrificing the probabilistic design.

Watershed-Based Survey Designs

“…All three distributions are similar for all subregions combined. Similar analyses for separate subregions indicated that bias among sampled streams in terms of watershed size, agricultural area, and urban area was also limited (Ator et al, 2003) Despite the rejection and replacement of a considerable number of primary streams, the desired land use distribution was maintained in the Mid-Atlantic Coastal Plain survey without sacrificing the probabilistic design (Ator et al, 2003). Site rejections during field reconnaissance may be less problematic in areas for which more accurate, regionally consistent data are available for defining the sampling frame.…”

“…A stratified random sample with unequal probability selection (within each stratum) was used to select individual streams for sampling (Ator et al, 2003). The survey was designed to include 175 streams, including 25 within each subregion.…”

“…Results from this study will support the development of landscape-indicator models for benthic ecological conditions and for concentrations of pesticides and nutrients in ground water discharging to small streams. The design for this study is described in more detail in Ator et al (2003).…”

“…First-order streams were chosen to minimize variability among data-collection sites with respect to hydrogeology, land use, or other geographic variables under investigation (Ator et al, 2003). The population of interest of 10 144 first-order streams was constructed from available digital hydrography for the Coastal Plain at a scale of 1:100 000 (Horn and Grayman, 1993;McKay et al, 1994).…”

Population viability analysis of the Endangered shortnose sturgeon

“…The primary purpose of this design was to provide data needed to develop good empirical models (Ator et al 2003). Two problems that often plague modeling efforts can be addressed at the sample selection stage are too little variability among important predictors and multicollinearity among predictors.…”