Water Chemistry Field Sampling
30
sampling sites can be stratified according to accessibility. This can help ensure
that few sites will be included that are remote and difficult to access for sampling. However, use of this approach can result in lower precision in quantifying conditions of remote sites.
Selection of streams or lakes for sampling may be nonrandom if information is needed on specific waters or watersheds rather than on a population
of waters or watersheds. Nonrandom sampling may therefore be perfectly
acceptable for certain studies. Nevertheless, one should carefully consider
that the gains realized in ease of sampling or availability of data collected previously in a nonrandom study must be weighed against the loss of the ability to
quantitatively extrapolate directly from the sampling sites to the population
of interest.
For nonrandom sampling, especially if the lake or stream is intended to be
part of a long-term monitoring (LTM) effort, it may be desirable to select water
bodies that exhibit particular characteristics. For example, it makes little
sense to spend many years monitoring a body of water that is not acid sensitive if the objective is to evaluate acidification response. Thus, one might purposely select one or more highly sensitive sites for monitoring or for detailed
study. Similarly, it would not be logical to focus a study of atmospheric nutrient N enrichment on a surface water that is P limited. For certain studies, it
is logical to select a site (or sites) highly sensitive to the stressor in question.
Nevertheless, it can be difficult to interpret the results of such studies without
first determining where the studied sites fall within the distribution of site
sensitivities across the study area or across the region. Such distributions of
regional site characteristics can sometimes be provided by various statistically based large synoptic sampling programs, such as EPA’s WSS, National
Stream Survey, National Lake Survey, or EMAP. Statistically based survey data
can be used to aid in selection of sites for LTM. LTM sites may be chosen at
random from among randomly selected survey sites so that the resulting monitoring data will be representative of the entire population of interest. This
approach was taken in EPA’s Temporally Integrated Monitoring of Ecosystems
(TIME) project (Kahl et al. 2004).
Thus, connection of survey or monitoring sites to the broader regional
population of lakes or streams is almost always highly desired. This connection allows one to extrapolate (whether statistically or merely semiquantitatively) results to more bodies of water than just the ones sampled. Ideally,
study sites should be statistically selected. If this is not possible, it may be
possible to express the results for a given study site relative to the broader
population by quantifying its chemistry relative to the population of lake or
stream chemistry determined in one of the larger regional or national surveys, such as those conducted by EPA. Alternatively, the feasibility of conducting a synoptic survey targeted to the specific forest or region should
30
sampling sites can be stratified according to accessibility. This can help ensure
that few sites will be included that are remote and difficult to access for sampling. However, use of this approach can result in lower precision in quantifying conditions of remote sites.
Selection of streams or lakes for sampling may be nonrandom if information is needed on specific waters or watersheds rather than on a population
of waters or watersheds. Nonrandom sampling may therefore be perfectly
acceptable for certain studies. Nevertheless, one should carefully consider
that the gains realized in ease of sampling or availability of data collected previously in a nonrandom study must be weighed against the loss of the ability to
quantitatively extrapolate directly from the sampling sites to the population
of interest.
For nonrandom sampling, especially if the lake or stream is intended to be
part of a long-term monitoring (LTM) effort, it may be desirable to select water
bodies that exhibit particular characteristics. For example, it makes little
sense to spend many years monitoring a body of water that is not acid sensitive if the objective is to evaluate acidification response. Thus, one might purposely select one or more highly sensitive sites for monitoring or for detailed
study. Similarly, it would not be logical to focus a study of atmospheric nutrient N enrichment on a surface water that is P limited. For certain studies, it
is logical to select a site (or sites) highly sensitive to the stressor in question.
Nevertheless, it can be difficult to interpret the results of such studies without
first determining where the studied sites fall within the distribution of site
sensitivities across the study area or across the region. Such distributions of
regional site characteristics can sometimes be provided by various statistically based large synoptic sampling programs, such as EPA’s WSS, National
Stream Survey, National Lake Survey, or EMAP. Statistically based survey data
can be used to aid in selection of sites for LTM. LTM sites may be chosen at
random from among randomly selected survey sites so that the resulting monitoring data will be representative of the entire population of interest. This
approach was taken in EPA’s Temporally Integrated Monitoring of Ecosystems
(TIME) project (Kahl et al. 2004).
Thus, connection of survey or monitoring sites to the broader regional
population of lakes or streams is almost always highly desired. This connection allows one to extrapolate (whether statistically or merely semiquantitatively) results to more bodies of water than just the ones sampled. Ideally,
study sites should be statistically selected. If this is not possible, it may be
possible to express the results for a given study site relative to the broader
population by quantifying its chemistry relative to the population of lake or
stream chemistry determined in one of the larger regional or national surveys, such as those conducted by EPA. Alternatively, the feasibility of conducting a synoptic survey targeted to the specific forest or region should