Upper Midwest Environmental Sciences Center
The LTRMP sampling designs include those that rely on stratified random and on nonrandom sampling (in LTRMP parlance,"SRS" and "fixed-site sampling," respectively). From 1988 through 1991, sampling locations were selected nonprobabilistically. Beginning in 1992 and 1993, however, stratified random or probability sampling was introduced. At present, a predominance of sampling effort and inferences are derived from data collected using probability-based designs. Use of data from nonrandom sites is discussed under Using Data from LTRMP Nonrandom ("Fixed-Site") Locations.
Probabilistic Sampling Designs
The LTRMP samples probabilistically using a stratified random sampling design within five reaches of the Upper Mississippi River (UMR) and one reach of the Illinois River. These six reaches represent a judgment sample of reaches within the Upper Mississippi River System (UMRS), and, as a result, inferences about either the UMR or the UMRS from LTRMP data must rely on models rather than on the design.
The LTRMP has also historically treated the four water quality sampling episodes (winter, spring, summer, fall) as separate populations. The argument for doing so rests more on scientific considerations than on water’s sampling design: The limnological characteristics of most large rivers change seasonally, and often substantially, in response to discharge.
The LTRMP's sampling design, within strata, is not strictly random. Our sampling frame represents the intersections of a square grid (cell areas are defined at http://www.umesc.usgs.gov/ltrmp/stats/population_sizes.pdf), with random sampling from the grid intersections. The use of the grid precludes many extreme (clumped) sampling outcomes. Hence, our sampling variances are expected to be smaller than under full stratified random sampling.
Probabilistic sampling began in 1992, 1993, and 1998 for LTRMP's macroinvertebrate, fish and water quality, and vegetation components, respectively. Within reaches, the list of sampling locations is stratified by broad geomorphic features (Wilcox 1993); sampling locations are selected randomly within these strata. The number of strata per reach is small (roughly four), and strata definitions have been constant (excepting a minor change in the vegetation component in 1999). Sampling intensities have varied by component, reach, and stratum but have remained roughly constant across sampling years within component-reach-stratum combinations. With the exception of vegetation sampling within Pool 8 (years 2001 through 2004), sampling locations were rerandomized between sampling episodes. Sampling units represent locations defined on a grid laid over the aquatic portions of the six reaches in 1992.
Further information about LTRMP's probabilistic sampling designs are provided in reports published for fish, macroinvertebrates, vegetation, and water quality components. Additional information is provided for the fish component as fish metadata.
Wilcox, D. B. 1993. An aquatic habitat classification system for the Upper Mississippi River System. U.S. Fish and Wildlife Service, Environmental Management Technical Center, Onalaska, Wisconsin, May 1993. EMTC 93-T003. 9 pp. + Appendix A. (NTIS PB93-208981)
Contact: Questions or comments may be directed to Brian Gray, LTRMP statistician, Upper Midwest Environmental Sciences Center, La Crosse, Wisconsin, at firstname.lastname@example.org.