Design of stream sampling networks and a GIS method for assessing spatial bias

Bias in the design of stream-sampling networks is a major cause of inaccurate characterization of water quality at state and national levels. The subject of sampling design is difficult because it includes spatial and temporal dimensions and is dispersed in scientific literature. This study identifies sources of bias that commonly occur in water-quality assessments and develops methods for recognizing and reducing bias.; 'Design bias' associated with location and frequency of sampling is distinguished from 'analytical bias'. Locational design bias is emphasized since other sources of bias are inconsequential if spatial design is inappropriate. Recent adoption of watershed-based approaches to water-quality management by government agencies has increased the importance of addressing spatial-design issues.; Environmental indices, which are compatible with the approach to water-quality assessment advocated by EPA for the 305(b) program, are developed to help water-resource managers recognize and reduce spatial bias in watershed sampling networks. The indices are also useful for prioritizing subsheds for sampling and for aggregating water-quality data into basinwide measures that provide a consistent means for comparing water-quality condition of watersheds within a region.; Indices reflect the relative influence that each subshed is likely to exert on basinwide water quality. 'Influence' is assessed by differentiating subsheds according to natural landscape and anthropogenic features that affect variability of water quantity and quality. Subsheds with more influence have higher index values, indicating greater landscape complexity or anthropogenic stress, than subsheds with lower index values. Spatial bias is reduced by ensuring that each subshed is sampled in proportion to its influence.; Use of the indices is demonstrated in a New Hampshire watershed. Indices are developed for six subsheds and include a 'landscape' subindex, which incorporates measures of topographic slope, elevation, and subshed size, and a 'stress' subindex, which incorporates measures of population and of point and nonpoint pollution sources. Indices are used to assess spatial bias in an existing sampling network and reveal a disproportionate number of stations in the subshed containing most point sources. Index values are used as weights to reapportion stations throughout the watershed. The weights are also applied to sampling data to produce basinwide water-quality measures.