Limitation of Current Solids Measurements in Stormwater Runoff
A discussion of sampling, classifying, and quantifying suspended solids
By
Masoud Kayhanian,
Thomas M Young,
Michael K. Stenstrom
Nearly all treatment systems employed today to remove pollutants of concern from stormwater runoff rely on physical removal processes. The bias toward physical unit operations arises from the fact that such methods can operate passively and are often effective in removing the particles that are the targets of the treatment methods, either for direct (i.e., suspended solids removal) or indirect (i.e., removal of particle-bound contaminants) reasons. Consequently, the design of constructed best management practices (BMPs) is driven largely by their effectiveness in removing small particles (less than 20 micrograms) and their associated organic and inorganic pollutants. However, review of current methods for particle size measurement, particle fractionation, and size-resolved analysis of particulate chemical composition revealed some inherent problems that may influence the design and performance evaluation of current and future BMPs. Delineating these problems and outlining potential solutions are the objectives of this article. Specific topics considered for discussion include (1) classification of solids in stormwater runoff, (2) quantification of solids content using either suspended sediment concentration (SSC) or total suspended solids (TSS), (3) allocation of constituents between particulate and dissolved solid phases, (4) need for improved standardized particle-size-distribution measurement methods, and (5) correlations, or lack thereof, between TSS and most inorganic and organic contaminants.
Classification of Solids in Stormwater Runoff
A simple classification of solids collected from stormwater runoff drainage systems is shown in Figure 1. As shown, diverse types of solids may be collected from stormwater runoff drain inlets and outfalls. The solids can initially be categorized as litter (greater than 6.35 millimeters) and non-litter (less than 6.35 millimeters) components. Litter can further be classified as gross, wet, and dry. Furthermore, a fraction of dry litter can float and is either biodegradable or non-biodegradable. The same is true for non-floatable dry litter. The biodegradability dimension is introduced to better evaluate the impact of these materials on receiving waters, particularly with respect to depletion of dissolved oxygen. Measurement and analysis of litter solids is challenging as it may require special equipment and standard methods that are not widely available. For instance, when the California Department of Transportation (Caltrans) initiated its litter characterization studies about five years ago, there were no laboratories equipped or capable of performing the types of analyses needed. For this reason, Caltrans was forced to develop a litter lab as well as a protocol to monitor litter from its facilities. The Caltrans litter guidance manual is available online at www.dot.ca.gov/hq/env/stormwater/special/newsetup/index.htm. Additional information on litter characteristics, including the biodegradability data, can be obtained from Kayhanian et al. (2002) and Kim, Kayhanian, and Stenstrom (2004).
It is important to note that the majority of litter and larger solids can be removed by most conventional litter devices and BMPs. Removal of any contaminants associated with these solids is straightforward; removing solids that are more stable within the water column because of particle-particle repulsion and/or slow-settling velocities is substantially more challenging. Under the proposed classification scheme, the solid fraction within the water column can be divided into four groups: sediment, gravitoidal, colloidal, and dissolved (Figure 2). The focus of this article is mostly on this type of solids because of the difficulty of removing them with conventional BMPs, especially for the colloidal and dissolved solid fractions. Concerns associated with the measurement of these solids, as well as the measurement of contaminants associated with them, are also discussed in this article.
| Figure 2. Proposed Non-Litter Solids Measurements for Stormwater Runoff Samples |
 |
Limitation of Solids Measurement Within Water Column
SSC Versus TSS Measurement
Suspended solids measurements are becoming an important water-quality parameter for managing stormwater runoff. Suspended solids are one of the primary pollutants of concern because of their impact on water clarity. Suspended solids may also be used as an indicator of overall water quality if robust correlations can be developed with other parameters. Monitoring costs can be reduced if some individual analyses can be eliminated through development of suitable correlations between these parameters and suspended solids concentrations. Improving the accuracy of measurements and representativeness of samples will increase the confidence in loading calculations and correlations based on the individual solids concentration measurements.
The accuracy, precision, and representativeness of solids measurements are affected by how the sample is collected in the field and analyzed in the laboratory. Traditionally, suspended solids are quantified using the TSS test, which was originally developed for wastewater treatment plants. In the absence of alternative tests, many agencies use the TSS test to analyze suspended solids in urban stormwater runoff monitoring and characterization studies. The two most common TSS test methods are EPA Method 160.2 and the American Public Health Association (APHA) Method (Standard Methods for the Examination of Water and Wastewater). In both methods, the TSS concentration is measured by withdrawing a well-mixed 100-milliliter aliquot from the sample container using a wide-bore pipette, and passing the aliquot through a tared glass-fiber filter. The filter and retained sediments are then dried and weighed, and TSS is calculated. This 100-milliliter aliquot might not always be representative of the entire sample, however. The major disadvantage of applying either of the two TSS methods is the associated bias that may result from one or more of the following:
- Pipette orifice size limits the size of particles that can be sampled.
- Variation in the pipette sampling point within the sample container affects sampled particle sizes.
- Sample mixing is not sufficient to keep sand and other heavier material in suspension when withdrawing the 100 milliliters.
As a means to overcome the bias found in the TSS test, the SSC analysis is gradually gaining support as a substitute. With the SSC method, the entire sample volume is used to measure the suspended sediment concentration, thereby eliminating any bias introduced when collecting the 100-milliliter sample aliquot. However, no standard SSC test is established for stormwater runoff characterization. The SSC analysis can be performed using one of three versions of the ASTM D3977-97 test method briefly described in Table 1.
The problems associated with the SSC analysis are:
- It is not a widely recognized test method.
- It is not widely performed by commercial laboratories.
- It is more expensive to run than the TSS test.
- The use of a correction factor in Test Method A may introduce method error.
- Filters used for Test Method B are subject to clogging, resulting in the use of multiple filters or a reduction in the applied sample volume.
- The test requires the entire sample volume to be analyzed. An entire sample must be devoted to the SSC analysis. The “devoted” sample can be collected either by grab sampling or multiple bottle setup in automatic samplers.
Another issue with suspended solids measurements is the variability in the results from different laboratories. Laboratories that receive duplicate samples should generate results that are comparable. However, interlaboratory results are often different for both TSS and SSC samples. For example, duplicate grab samples of runoff were collected from various roadway locations and sent to different laboratories for TSS analysis. Table 2 presents the results of these analyses. As shown, without specifying the test methodology, the results from each lab can be significantly different.
Similarly, two laboratories using ASTM Method D3977-97 B and ASTM D3977-97 A performed the SSC analyses on duplicate samples (Table 3). The results demonstrate the importance of using equivalent test methods when comparing SSC results. Lab B used a method that employs filtration of the sample, and Lab C used an evaporation method without a correction factor. The evaporation method determines the combined concentration of suspended and dissolved solids, thus increasing the sample concentration. Using the evaporation method on road runoff samples from areas where salts are applied during the winter season can significantly increase the SSC concentration.
When duplicate samples were sent to the same laboratory for both TSS and SSC analyses, significant variations were also found in the results (Table 4). Measured solids concentrations varied from 19 to 2,169 milligrams per liter. Variability in the results may be attributed to individual lab procedures, the representativeness of the TSS aliquot, or differences in the sequentially collected samples used to produce the duplicate samples.
Differentiating Between Particulate and Dissolved Solids
The simple concept of classifying the solid particles in runoff as particulate and dissolved solids might not be sufficient to fully assess the quality of water and/or use the information for design and evaluation of BMPs. Particles in stormwater span a continuous size range from below 0.1 to over 10,000 micrograms. As particles decrease in size they have higher specific surface areas, and this characteristic gives small particles the greatest capacity to adsorb constituents that accumulate at surfaces such as metals and non-polar organics (Sansalone and Buchberger 1997; Krein and Schorer 2000). Unfortunately, removing particles using conventional BMPs such as a sedimentation tank or sand filter without a chemical aid to promote aggregation becomes increasingly difficult as the particle size decreases. Runoff containing higher concentrations of fine particles is most likely to have toxicity associated with it.
Using current measurement techniques, particles larger than a specific size, or those removed by a specific process, are considered particulate. Those smaller than a specific size, or not removed by a specific process, are considered dissolved. Using this approach, a specific cutoff size or separation process will delineate the “particulate” and “dissolved” fractions. For instance, the dissolved fraction in Standard Methods for the Examination of Water and Wastewater (1998) is defined as the solids that pass through a specific filter paper, Whatman 934AH (nominal pore size = 1.5 micrograms) or its equivalent. The problem is that some portion of the solids with sizes smaller than 1.5 micrograms might be trapped on the filter paper and hence be counted as particulate solids, and the magnitude of this fraction will depend on the type and concentration of particles and the volume of sample filtered. Errors associated with the results will have significant implications for estimates of dissolved pollutant mass loads.
In an ongoing study of the size and composition of highway runoff particles, we noticed that some of the smaller particles aggregated naturally, producing a distinctive change in the particle size distribution (Li et al. 2005a). To better characterize these smaller particle sizes and any contaminants associated with them, a new approach can be used that considers a number of parameters including size distribution, stability relative to coagulation and fragmentation, and settling rates. Under this approach, the non-litter solids can be classified into four fractions: dissolved, colloidal, gravitoidal, and sediment (Gustafsson and Gschwend 1997). The definition of these four solid fractions is presented in Table 5. A proposed method of measurement for these four solid fractions for samples collected from stormwater runoff is shown in Figure 1. Under this proposed method, 1 gallon of representative runoff sample is placed in a specialized graduated cylinder with valves and opening for each fraction. The sample will be well mixed initially and allowed to settle for up to six hours (exact timing will be determined experimentally). Without disturbing the graduated cylinder, subsamples from each fraction will be collected and analyzed for total solids using standard methods. The captured solids from each fraction can then be used to determine the concentrations of associated contaminants. To date, no studies have investigated contaminant distributions with respect to these four solids classifications.
Lack of Standardized Particle Size Measurement Approaches
As discussed previously, a wide variety of particle-sizing techniques have been used to characterize stormwater. A large variety of instruments have been developed for characterizing particles in the water column, and their advantages and disadvantages were discussed in a recent report (Grant et al. 2003). Several researchers have measured the particle size distribution (PSD) in highway runoff or stormwater (Characklis and Wiesner 1997; Sansalone et al. 1998; Legret and Pagotto 1999); however, no consistent experimental method has evolved. In addition, few researchers have systematically measured PSD over smaller particle size ranges (1 to 50 micrograms) or over entire storm events.
In a recent study (Li et al. 2005a), a series of experiments was performed to establish a standard protocol with defined accuracy and repeatable results for PSD. These experiments were performed to understand five key sampling concerns: reproducibility, sample contamination, sample representativeness, impact of sample storage time and temperature, and the issue of sampling technique. These parameters were selected after three months of rigorous experimental evaluation using stored samples. The following conclusions resulted from these experimental results.
1. The reproducibility of the PSD obtained by measuring duplicate stormwater samples was represented by a difference proportion (DP), calculated as follows:
where N1 and N2 are the number of particles in a specific size range for the first and second samples. The difference for duplicate samples was within 10% for particles less than 30 micrograms. The DP increased for larger particles, and the difference was approximately 76% for particles in the range of 200 to 1,000 micrograms. This resulted in part because there were fewer particles in this size range. To decrease the variability of the large particle measurements, larger sample volumes should be collected or the measuring range should be modified to include more particles.
2. A hand-washing procedure for glass bottles was a suitable method of preventing contamination of samples being analyzed for PSD.
3. Gentle inversion (five to six times) of the sample bottle was an appropriate mixing method that prevented sedimentation or particle shearing.
4. Particles showed a natural aggregation, which required analysis as soon as possible but within six hours of sample collection.
5. Particle concentrations in samples collected by automatic samplers were lower than a flow-weighted average of corresponding grab samples. Results suggest that automatic composite samplers should not be used to collect samples for PSD analysis until further development is completed. The issue of sample collection and the associated contaminants is further discussed below.
These results may be used as a basis for a standard protocol for roadway runoff particle size measurement. Additional experiments need to be performed to refine a protocol for standard quantification of roadway runoff PSD.
Sampling Issues Influence the Differentiation of Dissolved and Particulate Contaminants
How the sample is collected in the field can affect its representativeness, especially if an automatic sampler is used. Automatic samplers are designed to conduct sampling from a single fixed location. The sample tube intake is typically anchored to the bottom or side of a conduit. Only the portion of the flow within the immediate vicinity of the intake is collected. The suspended sediment contained in this portion may or may not represent the suspended sediment concentration found within the entire flow cross-section. Intakes located at the bottom of a conduit may collect a higher proportion of heavy sediments, while intakes located downstream of a sediment trap may miss the heavier material completely. This uncertainty is partially due to the spatial variability of sediment concentrations in runoff, which can vary with depth and cross-sectional location. Heavier particles tend to flow along the bottom of a conduit. Variations in velocity along a cross-section create spatial variations in the suspended sediment concentrations. Spatial variation can become more pronounced as the flows deepen and/or become wider. Shallow, narrow flows tend to be more uniformly mixed.
| Figure 3. Two Cases of Non-Representative Solids Captured by an Automatic Sampler |
 |
(a) An automatic sampler strainer under heavy sediment particles
(b) An automatic sampler strainer within the flume after heavier sediment particles have been captured through flow dissipation |
 |
Automatic samplers probably collect the most representative samples when the sampling pipe is not full, the flow is turbulent, and sediments are uniformly suspended through the water column. Of course, this condition rarely exists, and hence the solid measurements are not fully representative with automatic samplers. Two of these non-representative cases are shown in Figure 3. This is especially problematic for large particles that have higher settling velocities and will consequently be depleted near water surfaces and enriched near the bottom. One way to prevent this problem is to redesign the sampling strainer so that it is always at the midpoint of the water column as shown in Figure 4.
As an example, Figure 5 shows the dissolved fractions of lead, nickel, copper, and zinc determined using composite and grab samples collected as part of a first flush highway runoff characterization study (Lee et al. 2004). Figure 5a includes the results of dissolved metals for 237 flow-weighted composite samples collected from more than 50 highway sites during the 2000–2003 wet seasons. Figure 5b includes 729 grab samples per metal from three highway sites for the 1999–2003 wet seasons. The composite samples show approximately equal portions in the dissolved and particulate form, except for lead, which is less than 30% (median) in the dissolved form. The grab samples show larger dissolved fractions than the autosampler derived values, except for lead, which is only 11% (median) in the dissolved form.
Another problem that may arise from automatic sampling relates to changes in metal partitioning with increasing holding time. It is well known that the soluble and particulate metals concentrations in water-quality samples may not be at equilibrium at the time of sample collection (Sansalone and Buchberger 1997). For this reason, Standard Methods and EPA methods require that samples for particulate phase metal concentrations be filtered within 24 hours of collection. Composite samples collected by automatic sampler for a storm event that is longer than 24 hours may therefore differ from grab samples in their dissolved particulate metal distribution because of changes in partitioning. In a recent study (Kayhanian and Stenstrom 2005), first flush samples were collected as grab samples during the rainfall event, which usually lasted eight to 12 hours. The samples collected in the first four hours were transported to the laboratory and filtered, which means that sample holding time was shorter than for flow-weighted composite samples. The nature of composite samples means that equilibrium processes can occur for at least the length of the storm (e.g., six to 24 hours) and the allowable holding time before filtration (less than 24 hours). Therefore, if the equilibrium of metals is shifting toward the particulate phase, composite samples should show a higher fraction of metals in the particulate phase than observed in grab samples. Sansalone and Buchberger (1997) showed modest increases in the particulate fraction over time (0 to 24 hours). The difference in dissolved fractions between the two highway studies mentioned above may be partially related to equilibration processes occurring during sample collection or analysis. The shift toward greater particulate-bound metal fractions with increasing storage time may also be a result of the natural aggregation processes, which may cause sufficient growth in fine particles to move their associated metals from the “colloidal” to the “particulate” size classes.
The time required to attain an equilibrium distribution of constituents between the dissolved and particulate fractions in stormwater merits further research investigation. If there is a significant trend toward increased particulate phase metals during storage, then this trend could be useful in BMP selection. BMPs that can store the runoff can provide time for increased particulate phase concentrations and therefore greater metals removal, because metal removal by BMPs is primarily due to suspended solids removal. The hydraulic detention time for many BMPs, such as detention basins, is quite small compared to the time typically allowed for equilibrium to be reached in most samples collected as part of monitoring programs.
Advertisement
Utility of Correlations Between TSS and Constituents
Total suspended solids have often been used as a surrogate for other pollutants in stormwater runoff (Driscoll, Shelley, and Strecker 1990). This approach has interesting implications for the prediction of related contaminant concentrations and offers the prospect for tremendous analytical cost savings. It will also have a profound impact on first flush pollutant mass loading, because the particle dynamics may have major impacts on emission rates of other pollutants (Li et al. 2005b). Recently the existence of a seasonal first flush for many pollutants has been noted, but suspended solids generally showed less seasonal first flush magnitude than other pollutants (Lee et al. 2004). Two other studies have concluded that there is no significant correlation between TSS and most organic and inorganic constituents (Caltrans 2003; Han et al. 2004). This may be in part attributed to (1) rainfall intensity effects on TSS emission rates, (2) non-representativeness of the captured solids, and (3) error associated with TSS measurement. However, a relatively good correlation was observed between particulate metals and TSS except for cadmium. The result of this correlation analysis is shown in Table 6. In Table 6, numbers above the diagonal are Pearson’s coefficients, r, and the numbers below the diagonal are probability, or P, values. Improving sampling-method as well as sample-measurement techniques outlined above may enhance the correlation coefficient.
Conclusions
Clearly, there are problems associated with current methods of sampling, classifying, and quantifying solids found in stormwater runoff. Organic and inorganic contaminants dynamically transfer between particle fractions by sedimentation, erosion, coagulation, fragmentation, adsorption, and desorption. These dynamic phenomena should be considered and ideally exploited in BMP design and performance evaluation. Widely accepted methods for particle characterization and a database of particle composition and aggregation characteristics developed over a wide range of geography and climate are prerequisites for fully exploiting these processes in BMP design. Classifying collected solids into four fractions (dissolved, colloidal, gravitoidal, and sediment) and measuring the concentrations of associated solids represent a first step along this research path.
Author's Bio: Masoud Kayhanian, Ph.D., is a research professor and associate director of the Center for Environmental and Water Resources Engineering in the Department of Civil and Environmental Engineering at UCD.
Author's Bio: Thomas M. Young is an associate professor in the Department of Civil and Environmental Engineering at the University of California-Davis.
Author's Bio: Michael K. Stenstrom, Ph.D., P.E., is a professor in the Department of Civil and Environmental Engineering at UCLA.
July-August 2005
Limitation of Current Solids Measurements in Stormwater Runoff
A discussion of sampling, classifying, and quantifying suspended solids
By
Masoud Kayhanian,
Thomas M Young,
Michael K. Stenstrom
Nearly all treatment systems employed today to remove pollutants of concern from stormwater runoff rely on physical removal processes. The bias toward physical unit operations arises from the fact that such methods can operate passively and are often effective in removing the particles that are the targets of the treatment methods, either for direct (i.e., suspended solids removal) or indirect (i.e., removal of particle-bound contaminants) reasons. Consequently, the design of constructed best management practices (BMPs) is driven largely by their effectiveness in removing small particles (less than 20 micrograms) and their associated organic and inorganic pollutants. However, review of current methods for particle size measurement, particle fractionation, and size-resolved analysis of particulate chemical composition revealed some inherent problems that may influence the design and performance evaluation of current and future BMPs. Delineating these problems and outlining potential solutions are the objectives of this article. Specific topics considered for discussion include (1) classification of solids in stormwater runoff, (2) quantification of solids content using either suspended sediment concentration (SSC) or total suspended solids (TSS), (3) allocation of constituents between particulate and dissolved solid phases, (4) need for improved standardized particle-size-distribution measurement methods, and (5) correlations, or lack thereof, between TSS and most inorganic and organic contaminants.
Classification of Solids in Stormwater Runoff
A simple classification of solids collected from stormwater runoff drainage systems is shown in Figure 1. As shown, diverse types of solids may be collected from stormwater runoff drain inlets and outfalls. The solids can initially be categorized as litter (greater than 6.35 millimeters) and non-litter (less than 6.35 millimeters) components. Litter can further be classified as gross, wet, and dry. Furthermore, a fraction of dry litter can float and is either biodegradable or non-biodegradable. The same is true for non-floatable dry litter. The biodegradability dimension is introduced to better evaluate the impact of these materials on receiving waters, particularly with respect to depletion of dissolved oxygen. Measurement and analysis of litter solids is challenging as it may require special equipment and standard methods that are not widely available. For instance, when the California Department of Transportation (Caltrans) initiated its litter characterization studies about five years ago, there were no laboratories equipped or capable of performing the types of analyses needed. For this reason, Caltrans was forced to develop a litter lab as well as a protocol to monitor litter from its facilities. The Caltrans litter guidance manual is available online at www.dot.ca.gov/hq/env/stormwater/special/newsetup/index.htm. Additional information on litter characteristics, including the biodegradability data, can be obtained from Kayhanian et al. (2002) and Kim, Kayhanian, and Stenstrom (2004).
It is important to note that the majority of litter and larger solids can be removed by most conventional litter devices and BMPs. Removal of any contaminants associated with these solids is straightforward; removing solids that are more stable within the water column because of particle-particle repulsion and/or slow-settling velocities is substantially more challenging. Under the proposed classification scheme, the solid fraction within the water column can be divided into four groups: sediment, gravitoidal, colloidal, and dissolved (Figure 2). The focus of this article is mostly on this type of solids because of the difficulty of removing them with conventional BMPs, especially for the colloidal and dissolved solid fractions. Concerns associated with the measurement of these solids, as well as the measurement of contaminants associated with them, are also discussed in this article.
| Figure 2. Proposed Non-Litter Solids Measurements for Stormwater Runoff Samples |
|
Limitation of Solids Measurement Within Water Column
SSC Versus TSS Measurement
Suspended solids measurements are becoming an important water-quality parameter for managing stormwater runoff. Suspended solids are one of the primary pollutants of concern because of their impact on water clarity. Suspended solids may also be used as an indicator of overall water quality if robust correlations can be developed with other parameters. Monitoring costs can be reduced if some individual analyses can be eliminated through development of suitable correlations between these parameters and suspended solids concentrations. Improving the accuracy of measurements and representativeness of samples will increase the confidence in loading calculations and correlations based on the individual solids concentration measurements.
The accuracy, precision, and representativeness of solids measurements are affected by how the sample is collected in the field and analyzed in the laboratory. Traditionally, suspended solids are quantified using the TSS test, which was originally developed for wastewater treatment plants. In the absence of alternative tests, many agencies use the TSS test to analyze suspended solids in urban stormwater runoff monitoring and characterization studies. The two most common TSS test methods are EPA Method 160.2 and the American Public Health Association (APHA) Method (Standard Methods for the Examination of Water and Wastewater). In both methods, the TSS concentration is measured by withdrawing a well-mixed 100-milliliter aliquot from the sample container using a wide-bore pipette, and passing the aliquot through a tared glass-fiber filter. The filter and retained sediments are then dried and weighed, and TSS is calculated. This 100-milliliter aliquot might not always be representative of the entire sample, however. The major disadvantage of applying either of the two TSS methods is the associated bias that may result from one or more of the following:
- Pipette orifice size limits the size of particles that can be sampled.
- Variation in the pipette sampling point within the sample container affects sampled particle sizes.
- Sample mixing is not sufficient to keep sand and other heavier material in suspension when withdrawing the 100 milliliters.
As a means to overcome the bias found in the TSS test, the SSC analysis is gradually gaining support as a substitute. With the SSC method, the entire sample volume is used to measure the suspended sediment concentration, thereby eliminating any bias introduced when collecting the 100-milliliter sample aliquot. However, no standard SSC test is established for stormwater runoff characterization. The SSC analysis can be performed using one of three versions of the ASTM D3977-97 test method briefly described in Table 1.
The problems associated with the SSC analysis are:
- It is not a widely recognized test method.
- It is not widely performed by commercial laboratories.
- It is more expensive to run than the TSS test.
- The use of a correction factor in Test Method A may introduce method error.
- Filters used for Test Method B are subject to clogging, resulting in the use of multiple filters or a reduction in the applied sample volume.
- The test requires the entire sample volume to be analyzed. An entire sample must be devoted to the SSC analysis. The “devoted” sample can be collected either by grab sampling or multiple bottle setup in automatic samplers.
Another issue with suspended solids measurements is the variability in the results from different laboratories. Laboratories that receive duplicate samples should generate results that are comparable. However, interlaboratory results are often different for both TSS and SSC samples. For example, duplicate grab samples of runoff were collected from various roadway locations and sent to different laboratories for TSS analysis. Table 2 presents the results of these analyses. As shown, without specifying the test methodology, the results from each lab can be significantly different.
Similarly, two laboratories using ASTM Method D3977-97 B and ASTM D3977-97 A performed the SSC analyses on duplicate samples (Table 3). The results demonstrate the importance of using equivalent test methods when comparing SSC results. Lab B used a method that employs filtration of the sample, and Lab C used an evaporation method without a correction factor. The evaporation method determines the combined concentration of suspended and dissolved solids, thus increasing the sample concentration. Using the evaporation method on road runoff samples from areas where salts are applied during the winter season can significantly increase the SSC concentration.
When duplicate samples were sent to the same laboratory for both TSS and SSC analyses, significant variations were also found in the results (Table 4). Measured solids concentrations varied from 19 to 2,169 milligrams per liter. Variability in the results may be attributed to individual lab procedures, the representativeness of the TSS aliquot, or differences in the sequentially collected samples used to produce the duplicate samples.
Differentiating Between Particulate and Dissolved Solids
The simple concept of classifying the solid particles in runoff as particulate and dissolved solids might not be sufficient to fully assess the quality of water and/or use the information for design and evaluation of BMPs. Particles in stormwater span a continuous size range from below 0.1 to over 10,000 micrograms. As particles decrease in size they have higher specific surface areas, and this characteristic gives small particles the greatest capacity to adsorb constituents that accumulate at surfaces such as metals and non-polar organics (Sansalone and Buchberger 1997; Krein and Schorer 2000). Unfortunately, removing particles using conventional BMPs such as a sedimentation tank or sand filter without a chemical aid to promote aggregation becomes increasingly difficult as the particle size decreases. Runoff containing higher concentrations of fine particles is most likely to have toxicity associated with it.
Using current measurement techniques, particles larger than a specific size, or those removed by a specific process, are considered particulate. Those smaller than a specific size, or not removed by a specific process, are considered dissolved. Using this approach, a specific cutoff size or separation process will delineate the “particulate” and “dissolved” fractions. For instance, the dissolved fraction in Standard Methods for the Examination of Water and Wastewater (1998) is defined as the solids that pass through a specific filter paper, Whatman 934AH (nominal pore size = 1.5 micrograms) or its equivalent. The problem is that some portion of the solids with sizes smaller than 1.5 micrograms might be trapped on the filter paper and hence be counted as particulate solids, and the magnitude of this fraction will depend on the type and concentration of particles and the volume of sample filtered. Errors associated with the results will have significant implications for estimates of dissolved pollutant mass loads.
In an ongoing study of the size and composition of highway runoff particles, we noticed that some of the smaller particles aggregated naturally, producing a distinctive change in the particle size distribution (Li et al. 2005a). To better characterize these smaller particle sizes and any contaminants associated with them, a new approach can be used that considers a number of parameters including size distribution, stability relative to coagulation and fragmentation, and settling rates. Under this approach, the non-litter solids can be classified into four fractions: dissolved, colloidal, gravitoidal, and sediment (Gustafsson and Gschwend 1997). The definition of these four solid fractions is presented in Table 5. A proposed method of measurement for these four solid fractions for samples collected from stormwater runoff is shown in Figure 1. Under this proposed method, 1 gallon of representative runoff sample is placed in a specialized graduated cylinder with valves and opening for each fraction. The sample will be well mixed initially and allowed to settle for up to six hours (exact timing will be determined experimentally). Without disturbing the graduated cylinder, subsamples from each fraction will be collected and analyzed for total solids using standard methods. The captured solids from each fraction can then be used to determine the concentrations of associated contaminants. To date, no studies have investigated contaminant distributions with respect to these four solids classifications.
Lack of Standardized Particle Size Measurement Approaches
As discussed previously, a wide variety of particle-sizing techniques have been used to characterize stormwater. A large variety of instruments have been developed for characterizing particles in the water column, and their advantages and disadvantages were discussed in a recent report (Grant et al. 2003). Several researchers have measured the particle size distribution (PSD) in highway runoff or stormwater (Characklis and Wiesner 1997; Sansalone et al. 1998; Legret and Pagotto 1999); however, no consistent experimental method has evolved. In addition, few researchers have systematically measured PSD over smaller particle size ranges (1 to 50 micrograms) or over entire storm events.
In a recent study (Li et al. 2005a), a series of experiments was performed to establish a standard protocol with defined accuracy and repeatable results for PSD. These experiments were performed to understand five key sampling concerns: reproducibility, sample contamination, sample representativeness, impact of sample storage time and temperature, and the issue of sampling technique. These parameters were selected after three months of rigorous experimental evaluation using stored samples. The following conclusions resulted from these experimental results.
1. The reproducibility of the PSD obtained by measuring duplicate stormwater samples was represented by a difference proportion (DP), calculated as follows:
where N1 and N2 are the number of particles in a specific size range for the first and second samples. The difference for duplicate samples was within 10% for particles less than 30 micrograms. The DP increased for larger particles, and the difference was approximately 76% for particles in the range of 200 to 1,000 micrograms. This resulted in part because there were fewer particles in this size range. To decrease the variability of the large particle measurements, larger sample volumes should be collected or the measuring range should be modified to include more particles.
2. A hand-washing procedure for glass bottles was a suitable method of preventing contamination of samples being analyzed for PSD.
3. Gentle inversion (five to six times) of the sample bottle was an appropriate mixing method that prevented sedimentation or particle shearing.
4. Particles showed a natural aggregation, which required analysis as soon as possible but within six hours of sample collection.
5. Particle concentrations in samples collected by automatic samplers were lower than a flow-weighted average of corresponding grab samples. Results suggest that automatic composite samplers should not be used to collect samples for PSD analysis until further development is completed. The issue of sample collection and the associated contaminants is further discussed below.
These results may be used as a basis for a standard protocol for roadway runoff particle size measurement. Additional experiments need to be performed to refine a protocol for standard quantification of roadway runoff PSD.
Sampling Issues Influence the Differentiation of Dissolved and Particulate Contaminants
How the sample is collected in the field can affect its representativeness, especially if an automatic sampler is used. Automatic samplers are designed to conduct sampling from a single fixed location. The sample tube intake is typically anchored to the bottom or side of a conduit. Only the portion of the flow within the immediate vicinity of the intake is collected. The suspended sediment contained in this portion may or may not represent the suspended sediment concentration found within the entire flow cross-section. Intakes located at the bottom of a conduit may collect a higher proportion of heavy sediments, while intakes located downstream of a sediment trap may miss the heavier material completely. This uncertainty is partially due to the spatial variability of sediment concentrations in runoff, which can vary with depth and cross-sectional location. Heavier particles tend to flow along the bottom of a conduit. Variations in velocity along a cross-section create spatial variations in the suspended sediment concentrations. Spatial variation can become more pronounced as the flows deepen and/or become wider. Shallow, narrow flows tend to be more uniformly mixed.
| Figure 3. Two Cases of Non-Representative Solids Captured by an Automatic Sampler |
|
(a) An automatic sampler strainer under heavy sediment particles
(b) An automatic sampler strainer within the flume after heavier sediment particles have been captured through flow dissipation |
|
Automatic samplers probably collect the most representative samples when the sampling pipe is not full, the flow is turbulent, and sediments are uniformly suspended through the water column. Of course, this condition rarely exists, and hence the solid measurements are not fully representative with automatic samplers. Two of these non-representative cases are shown in Figure 3. This is especially problematic for large particles that have higher settling velocities and will consequently be depleted near water surfaces and enriched near the bottom. One way to prevent this problem is to redesign the sampling strainer so that it is always at the midpoint of the water column as shown in Figure 4.
As an example, Figure 5 shows the dissolved fractions of lead, nickel, copper, and zinc determined using composite and grab samples collected as part of a first flush highway runoff characterization study (Lee et al. 2004). Figure 5a includes the results of dissolved metals for 237 flow-weighted composite samples collected from more than 50 highway sites during the 2000–2003 wet seasons. Figure 5b includes 729 grab samples per metal from three highway sites for the 1999–2003 wet seasons. The composite samples show approximately equal portions in the dissolved and particulate form, except for lead, which is less than 30% (median) in the dissolved form. The grab samples show larger dissolved fractions than the autosampler derived values, except for lead, which is only 11% (median) in the dissolved form.
Another problem that may arise from automatic sampling relates to changes in metal partitioning with increasing holding time. It is well known that the soluble and particulate metals concentrations in water-quality samples may not be at equilibrium at the time of sample collection (Sansalone and Buchberger 1997). For this reason, Standard Methods and EPA methods require that samples for particulate phase metal concentrations be filtered within 24 hours of collection. Composite samples collected by automatic sampler for a storm event that is longer than 24 hours may therefore differ from grab samples in their dissolved particulate metal distribution because of changes in partitioning. In a recent study (Kayhanian and Stenstrom 2005), first flush samples were collected as grab samples during the rainfall event, which usually lasted eight to 12 hours. The samples collected in the first four hours were transported to the laboratory and filtered, which means that sample holding time was shorter than for flow-weighted composite samples. The nature of composite samples means that equilibrium processes can occur for at least the length of the storm (e.g., six to 24 hours) and the allowable holding time before filtration (less than 24 hours). Therefore, if the equilibrium of metals is shifting toward the particulate phase, composite samples should show a higher fraction of metals in the particulate phase than observed in grab samples. Sansalone and Buchberger (1997) showed modest increases in the particulate fraction over time (0 to 24 hours). The difference in dissolved fractions between the two highway studies mentioned above may be partially related to equilibration processes occurring during sample collection or analysis. The shift toward greater particulate-bound metal fractions with increasing storage time may also be a result of the natural aggregation processes, which may cause sufficient growth in fine particles to move their associated metals from the “colloidal” to the “particulate” size classes.
The time required to attain an equilibrium distribution of constituents between the dissolved and particulate fractions in stormwater merits further research investigation. If there is a significant trend toward increased particulate phase metals during storage, then this trend could be useful in BMP selection. BMPs that can store the runoff can provide time for increased particulate phase concentrations and therefore greater metals removal, because metal removal by BMPs is primarily due to suspended solids removal. The hydraulic detention time for many BMPs, such as detention basins, is quite small compared to the time typically allowed for equilibrium to be reached in most samples collected as part of monitoring programs.
Utility of Correlations Between TSS and Constituents
Total suspended solids have often been used as a surrogate for other pollutants in stormwater runoff (Driscoll, Shelley, and Strecker 1990). This approach has interesting implications for the prediction of related contaminant concentrations and offers the prospect for tremendous analytical cost savings. It will also have a profound impact on first flush pollutant mass loading, because the particle dynamics may have major impacts on emission rates of other pollutants (Li et al. 2005b). Recently the existence of a seasonal first flush for many pollutants has been noted, but suspended solids generally showed less seasonal first flush magnitude than other pollutants (Lee et al. 2004). Two other studies have concluded that there is no significant correlation between TSS and most organic and inorganic constituents (Caltrans 2003; Han et al. 2004). This may be in part attributed to (1) rainfall intensity effects on TSS emission rates, (2) non-representativeness of the captured solids, and (3) error associated with TSS measurement. However, a relatively good correlation was observed between particulate metals and TSS except for cadmium. The result of this correlation analysis is shown in Table 6. In Table 6, numbers above the diagonal are Pearson’s coefficients, r, and the numbers below the diagonal are probability, or P, values. Improving sampling-method as well as sample-measurement techniques outlined above may enhance the correlation coefficient.
Conclusions
Clearly, there are problems associated with current methods of sampling, classifying, and quantifying solids found in stormwater runoff. Organic and inorganic contaminants dynamically transfer between particle fractions by sedimentation, erosion, coagulation, fragmentation, adsorption, and desorption. These dynamic phenomena should be considered and ideally exploited in BMP design and performance evaluation. Widely accepted methods for particle characterization and a database of particle composition and aggregation characteristics developed over a wide range of geography and climate are prerequisites for fully exploiting these processes in BMP design. Classifying collected solids into four fractions (dissolved, colloidal, gravitoidal, and sediment) and measuring the concentrations of associated solids represent a first step along this research path.