Phil A. Kammerer, Jr., Water Quality Specialist

Herbert S. Garn, Chief, Hydrologic Studies and Data Collection Section
U. S. Geological Survey, Water Resources Division, Wisconsin District

Paul W. Rasmussen, Statistician

Joseph R. Ball, Water Resources Specialist

Wisconsin Department of Natural Resources

The U.S. Geological Survey (USGS) and the Wisconsin Department of Natural Resources (WDNR) monitor the quality of Wisconsin’s water resources. Sample collection and processing protocols differ between agencies, and samples are analyzed by different laboratories. Data comparability is an issue to data users who may wish to combine or exchange data, but the degree to which differences in protocols affect results of water-quality monitoring is unknown. The primary objective of this study was to evaluate differences in results of water-quality monitoring caused by differences in sample-collection methods. Other data-comparability issues (sample processing and relative laboratory performance) were examined only to the extent necessary to accomplish the primary objective. Two sample-collection methods, flow-integrated sampling (USGS) and grab sampling (WDNR), were compared for three streams over different flow conditions and for one lake. Constituents used for the comparisons were dissolved orthophosphate, total phosphorus, dissolved chloride, chlorophyll a, and suspended sediment/total suspended solids. Each stream was sampled four times: twice at base flow and twice at high flow. The lake was sampled two times. Concurrent samples were collected by each of the two sample-collection methods. The effects of between-agency differences in sample processing and analytical procedures on results of water-quality analyses were removed by splitting all samples between laboratories and evaluating sample-collection methods independently using the results from each laboratory. The split for each laboratory was further split into triplicate samples to evaluate laboratory precision. Laboratories used in the study were the USGS National Water Quality Laboratory (NWQL), USGS Iowa District sediment laboratory, and the Wisconsin State Laboratory of Hygiene (WSLH).

Split plot analysis of variance and paired t-tests were used to test for significant differences (p<0.05) in constituent concentrations between sampling methods and between laboratories for constituents analyzed by comparable laboratory methods. Concentrations of total phosphorus and dissolved chloride did not differ significantly between sampling methods. Concentrations of dissolved orthophosphate were significantly different among methods, which was not expected, and suspended sediment and total suspended solids also differed significantly between sampling methods. Concentrations of suspended sediment and total suspended solids were usually lower in grab samples than in flow-integrated samples. Chlorophyll a concentrations were significantly different between samplers. Samples collected by WDNR had higher concentrations of chlorophyll a than those collected by USGS. Differences in concentrations of dissolved orthophosphate between samples filtered in the field and samples filtered in the laboratory were not significant. The effect of point of filtration on chlorophyll a, however, was highly significant. Lab-filtered samples analyzed at WSLH gave higher concentrations of chlorophyll a than field-filtered samples at all times.

Differences in concentrations of total phosphorus, dissolved orthophosphate, and dissolved chloride between samples analyzed by the NWQL and samples analyzed by the WSLH were consistent and highly significant. Differences in concentrations of chlorophyll a were also significant. Mean concentrations from the WSLH were higher than those from NWQL for all constituents. Lab differences in general appeared to be more significant and important than sample collection methods. Differences in results between sampling methods and labs may limit the combining of such data for certain purposes, such as for determining trends.

The U.S. Geological Survey (USGS), Water Resources Division and the Wisconsin Department of Natural Resources (WDNR) both have programs that monitor ambient water quality at fixed sites on Wisconsin’s streams and lakes. Sample collection and processing protocols used by the programs differ, and samples are analyzed by different laboratories. The degree to which these differences in protocols affect the results of the monitoring programs is unknown. Data comparability is an issue at both agencies and to other present and potential data users who may wish to combine or exchange data. This study was conducted as part of a pilot project in Wisconsin under a larger effort to improve water-quality monitoring nationwide by the Intergovernmental Task Force on Monitoring (ITFM) Water Quality (ITFM, 1995).

The primary objective of this study was to evaluate between-agency differences in water-quality monitoring results caused by differences in sample collection methods. A secondary objective was to evaluate comparability of data obtained from the two laboratories used in the study. Laboratory precision, laboratory comparability, and sample processing and preservation methods were evaluated only to the extent necessary to accomplish the primary objective.

A sampling program was designed to test for differences in water-quality data resulting from the two sampling methods and use of two laboratories. Sample-collection methods were compared at sites on three rivers and one lake for selected constituents over different flow conditions during the summer of 1993 to summer 1994. The intent of the study was to evaluate sample-collection methods rather than water-quality conditions at particular sites.

For rivers, monitoring programs and protocols included in the comparison were the USGS’ National Stream Quality Accounting Network (NASQAN) and National Water-Quality Assessment Program (NAWQA) and the WDNR’s Ambient Monitoring Network. Two principal sample-collection methods were compared: cross-sectionally integrated, flow-weighted ("integrated") sampling and grab sampling. Both USGS programs collect flow-integrated samples, and the WDNR collects grab samples.

For lakes, monitoring protocols used for the WDNR and the USGS Wisconsin District lake monitoring programs were compared. Both monitoring programs use similar sampling methods; grab samples were collected at discrete depths in the lake. Depth profiles of water temperature, pH, specific conductance, and dissolved oxygen concentration were measured at the time of sample collection.

Many factors can contribute to apparent differences in water-quality measurements. In order to identify differences due specifically to sample-collection methods, other factors must be identified and either measured or eliminated. USGS and WDNR monitoring programs use different analytical laboratories and follow different protocols for field processing, preservation, and sample shipment. The USGS preserves samples, and filters samples for dissolved constituents in the field. For streams, all samples, except those for suspended-sediment concentration analysis, are shipped to the USGS National Water Quality Laboratory (NWQL) in Denver, Colorado. Samples for suspended-sediment concentration analysis are shipped to the USGS Iowa District sediment laboratory in Iowa City, Iowa. For lakes, samples are shipped to either NWQL or the Wisconsin State Laboratory of Hygiene (WSLH) in Madison, Wisconsin, depending on the lake being sampled. The DNR preserves samples in the field and either ships or hand-delivers them to WSLH where samples for dissolved constituents are filtered prior to analysis. Protocols for preservation of nutrient samples also differ between agencies. Analytical methods are "comparable" between agencies for most constituents measured, but differ for some. Within-laboratory precision and accuracy could be expected to differ between laboratories.

The effect of between-agency differences in sample processing and analytical procedures on monitoring results was removed for the purposes of this study by splitting all samples between laboratories. Samples for each laboratory were processed according to the respective standard protocols used by each agency for routine samples (Fishman, 1993;Wisconsin State Laboratory of Hygiene, 1993). The effects of sample-collection method on monitoring results was evaluated independently using the results from each laboratory. Within-lab analytical precision was evaluated by splitting the samples submitted to each laboratory into triplicates (see figure 1). Laboratory accuracy was not evaluated in this study, thus any differences between laboratories reported herein are relative differences rather than absolute differences from a true value. The effect of the point of sample filtration (field versus laboratory) was evaluated for selected constituents only for samples analyzed by WSLH.

The constituents evaluated in this study were selected based on their inclusion in agency monitoring programs and their usefulness as surrogates to represent the behavior of other broader groups of constituents. The constituents chosen for samples from rivers were total phosphorus, dissolved orthophosphate, dissolved chloride, and suspended sediment/total suspended solids. The constituents chosen for samples from the lake were total phosphorus, dissolved orthophosphate, and chlorophyll a. In rivers, suspended material may be unevenly distributed through the stream cross section. Because of this, the greatest differences in monitoring results caused by differences in sampling methods were expected for samples containing large amounts of suspended material. Total phosphorus and suspended sediment/total suspended solids are surrogates for constituents associated with suspended material in rivers. Dissolved chloride is a surrogate for conservative dissolved constituents. Dissolved orthophosphate is of common interest because of its importance as a plant nutrient. For lakes, chlorophyll a concentration is used as a measure of algal production. The effect of point of filtration on monitoring results was tested for dissolved orthophosphate and chlorophyll a.

Priority in site selection was given to sites that were included in the routine monitoring programs of the participants and where the water was expected to contain measurable (detectable) concentrations of the constituents of interest under all sampling conditions. Additional site-selection criteria for rivers were that they represent a variety of hydrologic settings and that they were capable of producing substantial amounts of suspended material at high flow.

The three rivers sampled during this study were the Milwaukee River at Milwaukee (USGS station number 04087000), the Manitowoc River at Manitowoc (USGS station number 04085427), and the Wolf River at New London (USGS station number 04079000). Lake samples were collected from Little Green Lake near Markesan, Wisconsin.

Each of the rivers was sampled four times: twice under stable low-flow conditions and twice at high flow. For the purpose of this study, high flow was defined as a condition when surface runoff was entering the river, and concentrations of suspended material in the water appeared to be higher than at low flow. Low-flow samples were collected in August and October, 1993, and high-flow samples were collected in April–July, 1994.

During each low-flow sampling, field teams representing each of the three monitoring programs collected a series of three consecutive samples at about the same time using their standard sample-collection protocols. Each of the resulting nine samples was then split between laboratories and the split for each lab was further split into triplicate samples (figure 1). All samples were processed according to the standard protocols for the laboratory that would receive the sample. The concurrent samples were collected to measure the variability in results between teams caused by differences in sampling methods. The consecutive samples were collected to measure the variability in results for a team repeating the same procedure under assumed stable water-quality conditions in the river.

During each high-flow sampling, each of the three field teams collected a single concurrent sample which was split and processed in the same manner as the low-flow samples. Consecutive samples were not collected because the assumption of stable water-quality conditions through the sample-collection period was not valid under high-flow conditions, when water-quality conditions may change rapidly.

Lake samples were collected in August 1993 and April 1994. Each of the two field teams collected three consecutive concurrent samples from adjacent boats. The concurrent samples were collected to measure the variability in results between teams using similar sampling methods. The consecutive samples were collected to measure the combined variability due to a single team repeating the same procedure and temporal and spatial changes in lake water quality that took place during the sample-collection period. Sample splitting and processing were done according to the protocols employed for river samples.

Cross-sectionally integrated, flow-weighted (flow-integrated) sampling is used to collect a composite water sample in a stream cross-section such that the dissolved and suspended material in the sample is in proportion to water flow in the cross-section. Variation of water quality in a stream cross-section is often significant and is most likely to occur because of incomplete mixing of upstream tributary inflows, point-source discharges, or variations in velocity and channel geometry. The flow-integrated sampling technique employed by USGS in this study is known as the equal width increment/equal transit rate (EWI) method (Edwards and Glysson, 1988; Ward and Harr, 1990). In this method, an isokinetic sampling device (a sampler that allows water to enter without changing its velocity relative to the stream) is lowered and raised at a uniform transit rate through equally-spaced verticals in the stream cross-section. Samples were collected either by wading with hand-held samplers or from a bridge using a crane-mounted sampler, depending on river stage and flow conditions. The number of verticals employed differed between sites and between sampling teams.

Grab sampling involves dipping a sample from one or more points in a stream cross section. Grab sampling techniques employed by WDNR in this study were consistent at each site, but differed between sites, and ranged from wading and dipping samples at from one to three points in the cross section to sampling from a bridge at a single point using a Van Dorn-type sampler.

Lake-water samples were collected by both agencies at a depth of 1.5 feet below the lake surface using Van Dorn-type samplers.

Sample splitting was done using a cone splitter routinely employed by the NAWQA program. Splits are accomplished by pouring a sample through the splitter, which splits the sample into 10 equal volumes. Successive splits of water from the initial split were done as needed to fill sample bottles directly or provide water for field filtration. Sample bottles for suspended sediment/total suspended solids, total phosphorus, dissolved orthophosphate (WSLH only), dissolved chloride (WSLH only), and chlorophyll a (WSLH only) analyses were filled directly from the splitter.

Samples for dissolved chloride, dissolved orthophosphate, and chlorophyll analyses by NWQL were filtered in the field. Extra samples were also filtered in the field and sent to WSLH for chlorophyll a and dissolved orthophosphate analyses to evaluate the effect of point of filtration.

Chlorophyll a samples for WSLH, and all phosphorus samples, were iced following processing. All phosphorus samples sent to NWQL were preserved with mercuric chloride. Total phosphorus samples sent to WSLH were preserved with sulfuric acid, and dissolved orthophosphate samples were not preserved. Filters from field-filtered chlorophyll a samples for NWQL were stored and shipped frozen on dry ice.

Analytical methods used by NWQL and WSLH (table 1) for dissolved orthophosphate and dissolved chloride analyses are considered comparable methods, so analytical results for these constituents can be compared between labs in addition to comparison of sampling methods. Analytical methods for total phosphorus used by both laboratories are comparable for lake samples and nominally comparable for stream samples. Total phosphorus methods used for stream samples differed in the digestion method used prior to analysis. Analytical methods employed by WSLH for chlorophyll a and total suspended solids differ from those employed by NWQL for chlorophyll a and by the USGS Iowa sediment lab for suspended sediment concentration.

Laboratory data were analyzed using statistical techniques to evaluate differences in central values and variability of the sampling results, primarily split plot analysis of variance (ANOVA) that reflected the physical splitting of samples, and paired t-tests. Differences were considered significant at p<0.05. Tests were evaluated for each of the constituent concentrations among monitoring programs and between laboratories (where constituents were analyzed by comparable laboratory methods) and the various interactions. Data from low-flow dates were used for analyses involving consecutive sampling time and data from the first sampling time on each date for analyses combining low and high flow conditions. The statistical comparisons included:

For the stream samples, no significant differences among monitoring programs were detected for concentrations of total phosphorus and chloride. For both stream and lake samples, there was a highly significant difference among samplers for dissolved orthophosphate. Although the analysis indicated significant differences among samplers, the differences were smaller than the standard deviation among lab replicates. In further analysis of these differences by lab, differences were significant only for samples analyzed at NWQL for data from low-flow samples. There were also significant differences among monitoring programs for concentrations of total suspended solids or suspended sediment. Because of differences between labs (methods are not comparable), separate analyses were repeated for each lab. For samples collected at high flow, the means for NAWQA and NASQAN flow-integrated samples were significantly greater than those from WDNR grab samples. The pattern was similar for both labs. Although the pattern was similar at low flow (USGS samples had larger means than WDNR samples), the differences were smaller and not significant.

For lakes, chlorophyll a concentrations were found to be different between samplers. For chlorophyll a samples analyzed at WSLH there were significant differences among samplers, but for samples analyzed at NWQL there were no significant differences. Samples collected by WDNR had higher concentrations of chlorophyll a than those collected by USGS (mean difference of 4.16 mg/L).

There were no significant concentration differences for total phosphorus, orthophosphate and chloride data among sequential samples collected on the same day during low-flow conditions for any of the monitoring programs. There was significant variability associated with sequential samples among monitoring programs for total suspended solids or suspended sediment, but the differences varied among dates. A summary of the results of statistical analyses for each of the comparisons is provided in table 2.

In a similar study in Kentucky, comparing surface-grab and flow-integrated stream sampling methods (Martin et al, 1992) for a larger number of constituents, concentrations of suspended sediment and some sediment-associated constituents such as total phosphorus, total iron, and total manganese were found to be significantly lower in grab samples than in integrated samples. The magnitude of the differences generally increased with streamflow. Median percent differences in concentration were about 20-25 percent. Concentrations of most of the dissolved constituents and common physical properties were not consistently different.

There was no significant difference in concentrations of dissolved orthophosphate in paired samples that differed only in point of filtration (field versus lab). The effect of point of filtration on chlorophyll a, however, was highly significant. Lab-filtered samples analyzed at WSLH gave higher concentrations than field-filtered samples at all times and were relatively less variable. The mean difference between methods was 3.3 mg/L.

A highly-significant difference between laboratories was found for total phosphorus (for both stream and lake samples), dissolved orthophosphate, and dissolved chloride concentrations. Mean concentrations from the WSLH were higher than those from NWQL for all three constituents in stream samples, but the magnitude of the differences varied among dates. For lake samples, total phosphorus concentrations from NWQL were greater than those from WSLH. The mean differences for stream data were approximately 0.025 mg/L for total phosphorus concentrations up to about 0.30 mg/L, 0.005 mg/L for dissolved orthophosphate concentrations up to about 0.14 mg/L, and no consistent difference for dissolved chloride concentrations up to about 60 mg/L. The mean difference between labs was 0.017mg/L total P on high-flow dates and 0.037 on low-flow dates.

The analytical techniques employed by WSLH for chlorophyll a and total suspended solids are different from those employed by NWQL for chlorophyll a and by the USGS Iowa sediment lab for suspended sediment concentration, so results from the two labs were expected to be different. Because of the different lab methods used, analytical results for these constituents were used primarily to emphasize comparisons of factors within labs, and those between labs are discussed briefly. In general, the sediment value from the USGS lab was larger than the suspended solids value from WSLH, and the difference was greater at low flow than at high flow. In regard to chlorophyll a, the values from WSLH were larger than those from NWQL for every sampling time on each date, although there were no consistent differences among sampling times.

Within-laboratory variation for each constituent was estimated from the variance among the three replicates sent to each lab for each sample. For stream samples, the variance in concentrations for total phosphorus and dissolved chloride was significantly greater for NWQL than for WSLH. For lake samples, total P variability among lab replicates was not significantly different for the two laboratories. Chlorophyll a sample replicates, however, analyzed at the NWQL were more variable than those analyzed at WSLH. Lab variability was much greater for suspended sediment concentrations from the USGS Iowa District sediment laboratory than for total suspended solids concentrations from WSLH.

A study comparing results between the USGS and Illinois Environmental Protection Agency laboratories (Melching and Coupe, 1995) also found that differences for some constituents were statistically significant and large enough to concern water-quality planners and engineers. Findings from this study also implied that lab differences were more important than sample collection differences from concurrent sampling, and that data from different laboratories should not be mixed when doing statistical analyses.

Major findings of this study may be summarized as follows:

For the flow conditions encountered in this study, sample collection method affects monitoring results for some constituents. Sample collection method (flow-integrated versus grab sample) does not appear to affect monitoring results for some dissolved constituents as represented by chloride or for constituents associated to some degree with suspended material as represented by total phosphorus. Sample collection method does appear to affect monitoring results for direct measures of suspended material as represented by total suspended solids concentration and suspended sediment concentration.

For the constituents measured and the range of concentrations encountered in this study, there were statistically significant differences in monitoring results for samples split between laboratories and analyzed by comparable analytical methods. Differences include both effects of laboratory performance and effects of sample processing between the time of collection and receipt by the laboratory. Sample processing and analytical methods for total suspended solids and suspended sediment concentration measurements are not comparable and do not yield comparable results.

Lab differences in general appeared to be more significant and important than sample collection methods. Generally, interaction terms of lab methods with dates or flow conditions were highly significant components in the analyses, indicating that differences are not consistent from sampling time to time to allow the application of simple correction factors. Differences in results between sampling methods and labs may limit the combining of some data for certain purposes, such as for determining trends.

* USGS Iowa Sediment Laboratory.

[NS = not significant, p>0.05; * = significant, p<0.05; ** = highly significant, p<0.01; - = not tested; NWQL = National Water Quality Lab, USGS; WSLH = Wisconsin State Lab of Hygiene.]