Critical Review pubs.acs.org/est
A Review of Selected Inorganic Surface Water Quality-Monitoring Practices: Are We Really Measuring What We Think, and If So, Are We Doing It Right? Arthur J. Horowitz* Georgia Water Science Center, U.S. Geological Survey, Suite 500 1770 Corporate Drive Norcross Georgia 30093, United States Department of Geosciences, Georgia State University, 22 Peachtree Center Avenue Atlanta, Georgia 30303, United States S Supporting Information *
ABSTRACT: Successful environmental/water quality-monitoring programs usually require a balance between analytical capabilities, the collection and preservation of representative samples, and available financial/personnel resources. Due to current economic conditions, monitoring programs are under increasing pressure to do more with less. Hence, a review of current sampling and analytical methodologies, and some of the underlying assumptions that form the bases for these programs seems appropriate, to see if they are achieving their intended objectives within acceptable error limits and/or measurement uncertainty, in a cost-effective manner. That evaluation appears to indicate that several common sampling/processing/analytical procedures (e.g., dip (point) samples/measurements, nitrogen determinations, total recoverable analytical procedures) are generating biased or nonrepresentative data, and that some of the underlying assumptions relative to current programs, such as calendar-based sampling and stationarity are no longer defensible. The extensive use of statistical models as well as surrogates (e.g., turbidity) also needs to be re-examined because the hydrologic interrelationships that support their use tend to be dynamic rather than static. As a result, a number of monitoring programs may need redesigning, some sampling and analytical procedures may need to be updated, and model/surrogate interrelationships may require recalibration.
1. INTRODUCTION Successful environmental/water quality-monitoring programs usually represent a balance between analytical capacity; the collection, processing, and maintenance of uncontaminated and representative samples; and available resources. Historically, the process has been driven by the development of more sophisticated analytical equipment that provided lower detection limits, greater precision, and/or new constituent analyses. In turn, this increased capacity had to be balanced against the limitations of field personnel to collect uncontaminated/representative samples. Long-term environmental databases display significant shifts that exceed natural variability, and which are contemporaneous with methodological changes. In addition to general inflation, almost every monitoring “improvement” has increased program costs resulting from a need for more sensitive analytical instrumentation, cleaner noncontaminating equipment, and bettertrained/educated field personnel. In the current economic climate, resources are a significant limiting factor in maintaining effective monitoring programs. For example, the U.S. Geological Survey (USGS) runs a Mississippi River Basin (MRB) monitoring network to address Gulf Coast land loss, as well as hypoxia. Currently, it costs between US$4,000 and $6,000 to collect a single sample at a mainstem MRB site. Analyses for various physical/chemical This article not subject to U.S. Copyright. Published 2013 by the American Chemical Society
parameters add US$1,500 to 2,000 more per sample. [The cited costs come from four sources: Richard Hooper, NASQAN coordinator from 1992 through 2000; from Charles Demas (2000), former Director for the Louisiana Water Science Center which is responsible for sampling at three sites in Louisiana; Charles Crawford (2007), current Coordinator for USGS Surface Water Sampling; and John R. Gray (2010), USGS Office of Surface Water.] Further funds are expended on program management, database maintenance, data analyses, and publications. These costs limit sample numbers/sites that can be monitored, and have led to questions regarding program effectiveness.1−3 Hence, a review of the current underlying assumptions, as well as the field/laboratory methods in use seems appropriate, to determine if changes can provide better and/or more cost-effective programs within current resource limits. This review is not intended as a critique, per se, of current programs, but hopefully will lead to a re-evaluation of program designs, and sampling, processing, and analytical procedures. Received: Revised: Accepted: Published: 2471
October 17, 2012 February 4, 2013 February 13, 2013 February 13, 2013 dx.doi.org/10.1021/es304058q | Environ. Sci. Technol. 2013, 47, 2471−2486
Environmental Science & Technology
Critical Review
2. SAMPLE MEDIA In typical environmental/water quality-monitoring programs the most common samples collected and analyzed are filtered and unfiltered (whole) water. The former are used to determine dissolved constituents whereas the latter are used to determine suspended sediment concentrations (SSCs)/ grain-size distributions, and sediment-associated chemical constituents by subtraction. There are unique issues associated with the collection and analyses of both sample types; the success of any monitoring program, regardless of spatial/ temporal scale, requires addressing them. 2.1. Filtered Water. Filtered water samples are analyzed for dissolved constituents. The two most important issues associated with this sample media are contamination4−13 and processing (filtration) artifacts.14−25 The contamination issue became clear in the late 1980s and early 1990s. Until then, for example, typical reported trace element levels were in the mg/L range. However, many of these concentrations resulted from contamination added during field sampling/processing, rather than reflecting ambient levels. This problem was addressed by implementing “clean” sampling/processing protocols that attempted to eliminate most systemic contamination, but they could not completely eliminate random contamination.7−11,13 Hence, these protocols include quality assurance/ control (QA/QC) procedures. All these changes substantially increased the cost of sample collection/processing, but brought reported trace element levels into the μg/L/sub μg/L range. Filtered water should be a solution that is defined as a mixture of one substance(s) dissolved in another (water) so that the properties are homogeneous.26 A solution consists of a solute (e.g., trace elements, nutrients) and a solvent (water). The solute is the substance being dissolved; the solvent does the dissolving. Typically, the solute is of molecular size and in an ionic (charged) form. This is a chemical definition. However, the environmental/water quality-monitoring community, as well as many standards organizations (e.g., ASTM),27,28 almost always use an operational physical definition for a dissolved constituent; it is one that passes a 0.45 μm filter. However, the actual operational definition may vary depending on the constituent(s) of concern. For example, (1) a 0.45 μm capsule filter (trace elements/nutrients); (2) a 0.70 μm prefired glass-fiber plate filter (organic compounds); and (3) a 1.5 μm plate filter (suspended sediment). Despite these differences, most organizations concatenate all filtered water (“dissolved”) data as if they were consistent/ equivalent.13,29 Filtered water actually is a dispersion consisting of a solute, a solvent (water), and finely divided micrometer to submicrometer solids (colloids) that may or may not be homogeneous throughout the fluid.26 As a result of current preservation procedures (e.g., nitric acid to pH ≤2 for trace elements), as well as the use of high temperature analytical equipment (e.g., inductively coupled plasma (ICP) systems), colloidally associated constituents often are at least partially quantified contemporaneously with truly dissolved constituents. Hence, a substantial portion of what currently is termed dissolved actually is colloidally associated.14−25 While it may be intuitively obvious that different filter pore sizes can produce different results for “dissolved” concentrations, depending on the mass of colloidal material passing the filter, what is not so obvious is that even using consistent pore-sized filters can produce different dissolved concentrations
(Figure. 1; Supporting Information (SI) Figure 1, Table 1). Hence, anything affecting the amount of colloidal material
Figure 1. The concentration of Fe in sequential 100 mL aliquots from four different filters for a sample collected in Keg Creek, GA. The cellulose acetate filters are 0.45 μm and the polycarbonate filters are 0.40 μm. All the concentrations on the graph meet the current operational definition of a dissolved constituent. The numbers to the right of each filter series represent the concentration in 1000 mL of filtrate.
passing a 0.45 μm filter will affect dissolved concentrations.14−25 Field and laboratory studies demonstrate that dissolved concentrations can be affected by such factors as filter type, filter diameter, filtration method, sample volumes processed, SSC and grain-size distributions, and the concentrations of colloids and organic matter.12,16 Using consistent protocols can eliminate some differences, but others are a function of environmental conditions beyond the control of field personnel. Hence, dissolved chemical data should be accompanied by descriptions of how samples are collected, processed, and analyzed, and comparing and/or concatenating data generated using different procedures should be undertaken with caution. The issue of filtration/processing artifacts likely will continue until the environmental/water quality-monitoring community adopts a chemical rather than a physical definition for separating/analyzing dissolved constituents. 2.2. Whole Water. While dissolved constituents are presumed to be homogeneous in fluvial cross sections, suspended sediment is vertically and horizontally heterogeneous because suspended sediment consists of different-sized and -density materials, and fluvial velocities vary cross sectionally (Figures 2 and 3).30−35 Generally clay- and siltsized particles (≤63 μm) are distributed homogeneously, whereas sand-sized particles (≥63 μm) are not. These distributions have consequences for sampling/processing whole-water samples, as well as for the determination of ambient SSCs, grain-size distributions, and suspended sediment-associated chemical concentrations (Figures 2 and 3). The chemical variations are caused by the “grain-size effect” that leads to increasing concentrations of most sedimentassociated constituents with decreasing grain size.30−33 Consequently, collecting whole-water samples by dipping a device at one location and at one depth in a river is unlikely to collect representative samples (SI Figure 2). Sampling protocols intended to generate representative whole-water samples for subsequent physical/chemical analyses are available. Collectively, these methods are termed equalwidth increment (EWI) or equal-discharge increment (EDI) 2472
dx.doi.org/10.1021/es304058q | Environ. Sci. Technol. 2013, 47, 2471−2486
Environmental Science & Technology
Critical Review
Figure 2. The upper graphs display changing suspended sediment concentrations in depth-integrated verticals across two different rivers. Note that the changing concentrations result from increasing concentrations of sand-sized particles as the sampling points move toward higher velocities near the centroid of flow. The lower graphs display changing concentrations of sediment-associated constituents; the decreasing concentrations toward the centroid of flow are due to dilution from increaseing amounts of sand-sized particles.
depth-and-width-integrated isokinetic sampling.35 These procedures entail sampling an entire fluvial cross section by using devices that fill with whole water as they pass between the surface and the river bed and return, along a series of verticals that are either equally spaced (equal-width) or display equal velocities (equal-discharge). No external force (e.g., pumping), which can bias the sample, is used to fill the sampler, only river velocity does. The marked heterogeneity of suspended sediment distributions in fluvial cross sections has implications for autosamplers and for various surrogates (e.g., turbidity) that have been introduced to provide whole-water samples, SSC/grain-size distribution data, and/or chemical concentrations in the absence of manual samples because almost all these devices function at a single point and depth in fluvial cross sections.32,36−42 Calibration against depth-and-width-integrated isokinetic samples is needed to ensure that measurements and/ or autosample-derived data are cross-sectionally representative. Also, varying hydrologic conditions will change the crosssectional position of these devices; hence, calibration is required over a range of flows. Further, as many fluvial systems are subject to changing anthropogenic effects that may alter
upstream/instream conditions, calibration checks must be maintained while monitoring continues. Typically, there are three exceptions to collecting wholewater samples without using depth-and-width-integrated procedures.35,36,42 Under low-flow conditions (≤0.6 m/s), current equipment cannot sample isokinetically. In such cases, nonisokinetic (open bottle) depth- and-width-integrated sampling is used. Also under low-flow conditions, when SSCs are below 20 mg/L, a single grab sample is likely to collect a representative whole-water sample because at low velocities, suspended sediment usually consists only of clay- and silt-sized material which normally are homogeneously distributed in most fluvial cross sections. Lastly, under high flow conditions (e.g., storms), there may be sufficient turbulence to homogeneously distribute suspended sediment, regardless of particle size. However, in all cases, calibration is required to ensure a representative sample is collected. 2.3. Suspended Sediment. Most U.S. environmental/ water quality-monitoring programs are regulatory-driven in response to the 305(b)/303(d) requirements of the Clean Water Act of 1972.43 Section 305(b)44 requires stream quality reports every 2 years, and Section 303(d)45 requires a listing of 2473
dx.doi.org/10.1021/es304058q | Environ. Sci. Technol. 2013, 47, 2471−2486
Environmental Science & Technology
Critical Review
Figure 4. The partitioning of the annual fluxes of selected chemical constituents between dissolved and sediment-associated phases in small urban streams in the City of Atlanta, GA. Note that with the exception of total nitrogen, greater than 80% of the constituent fluxes are associated with the solid phase.
determination/identification of the sources/sinks of suspended sediment (e.g., sediment fingerprinting), or identifying waterquality trends, or for evaluating land-use impacts, or in dealing with eutrophication issues, then sediment-associated chemical constituents should be included.32,33,42 Inorganic constituents that primarily are sediment-associated are called hydrophobes, and normally are bound to sediment particle surfaces.57 This group includes heavy metals/trace elements (e.g., Pb, Zn, Hg) and nutrients (e.g., N, P). In countries where sediment-associated chemistry is regulated, elevated levels of these constituents are viewed as contaminants. Also, numerous sediment-associated constituents are markers for various land-use effects.57−63 The annual fluxes of sediment-associated nutrients can be substantial. For example, in the lower MRB more than 75% of the P, 50% of the C and 30% of the N are sediment bound.64 While sediment-bound nutrients may not be bioavailable when the material is in suspension, they may become so after deposition in lakes/ impoundments due to postdepositional chemical remobilization and/or to bacterial action. This is particularly true for P, the limiting nutrient for freshwater eutrophication.65−67 Lastly, although not surface bound, and typically not viewed as contaminants, suspended sediment is a source for matrix-bound constituents like Al, Fe, and Ti. The concentrations/fluxes of matrix-bound constituents usually reflect local geology/ pedology.68,69 2.4. Bed Sediment. Although bed sediments have been used in large-scale (e.g., national) environmental surveys, as well as for geochemical reconnaissance (e.g., mineral exploration), they rarely are used in traditional monitoring programs.68−72 However, of all the potential sample media, bed sediments probably are the easiest to collect and process, and are least likely to suffer from contamination or insufficient sample mass/analytical detection issues. Hence, their lack of use is surprising, but results from several factors. First, since the publication of the Hawkes and Webb73 treatise on geochemical exploration, as well as the publication of several national
Figure 3. The upper graphs display changing suspended sediment concentrations in isokinetic point samples taken at 20, 40, 60, and 80% of river depth. Again, the changing concentrations are due to increasing concentrations of sand-sized particles as the sampling points move deeper in the water column. The lower graphs display changing concentrations of sediment-associated constituents; the decreasing concentrations are due to increasing amounts of sandsized particles.
impaired river segments that do not meet water quality requirements for designated-use categories. As such, the majority of sample analyses are for dissolved constituent concentrations. Turbidity, which results from variations in SSC and color, is responsible for most current sediment-associated total maximum daily load (TMDLs) calculations.45,46 Other than SSCs and/or turbidity, sediment-associated chemical concentrations largely are unregulated because there are few if any established limits for sediment-associated chemical constituents. The reasons for this appear to relate to disagreements over how to estimate sediment-associated constituent bioavailability.47−57 Currently, only five countries regulate sediment-associated chemical constituents: Canada, New Zealand, Australia, The Netherlands, and Germany; however, many have established guidelines.54,57 Contrariwise, when environmental/water-quality monitoring program objectives are nonregulatory, for example, estimating annual fluxes for various chemical constituents (Figure 4; SI Figure 3), or the 2474
dx.doi.org/10.1021/es304058q | Environ. Sci. Technol. 2013, 47, 2471−2486
Environmental Science & Technology
Critical Review
geochemical/environmental atlases,68−72 bed sediments have been used to detect geochemical/environmental spatial differences, but rarely are viewed as sensitive to short-term (geo)chemical variations; whereas the latter are the usual goal of most monitoring programs. Second, bed sediment surveys typically employ grain-size limited aliquots (e.g.,
