Stormwater | Improving Stormwater Monitoring

A comprehensive system developed in California helps ensure monitoring programs are cost-effective and produce scientifically credible data.

Stormwater monitoring involves a unique set of challenges. Doing it right - loosely defined as conducting the monitoring so as to produce meaningful, representative, and useful data of verifiable quality - requires extensive knowledge and preparation and an even greater commitment. Success requires careful and thorough advance planning, a 24/7 commitment during the targeted wet season, and detailed follow-up through a rigorous series of data validation steps.

Even selecting appropriate sites can be a challenge when you need to provide safe access in all kinds of weather, avoid unwanted influences such as backwater or aerial drift from nearby land uses, supply security for field personnel and any installed equipment, make available a reliable power source for automated monitoring equipment, and ensure with some degree of certainty that flow ranges will be within reasonable limits - not too low, not too high.

You've got to line up an analytical laboratory - or, more often, laboratories - not only to provide sufficiently low detection limits and documented data quality but also often to provide special services, such as off-hours sample acceptance, logging, and perhaps preparation (composite splitting, filtration, preservation). Bottles and sample tubing have to be precleaned, and you'll need to have plenty of sterile gloves and other sampling gear on hand. Be sure to charge the cell phone, and remember to throw your boots and rain gear in the back of the vehicle.

The weather is going to be notoriously difficult to predict, even if you've hired a consultant to help you interpret those maddening radar images and fine-tune the Intellicast and National Weather Service forecasts. But when you need to know not only whether it will rain or not but also how much it will rain, for how long, and when it will start and stop, that difficulty is multiplied.

Then you're out there collecting samples during adverse conditions - it's raining, first of all, and often windy, cold, and dark. Next you've got to get the samples to the labs in time to beat the holding time requirements (a special challenge for coliform samples) and ensure prompt sample filtration and preservation - all of which requires quick action by the field crew and the cooperation of your friendly lab personnel, often at odd hours. You can, of course, split samples, filter, and preserve in the field if you have to - lots of fun on a stormy Saturday night.

When the skies have cleared - and you've gotten some sleep - you'll need to follow up with the lab(s) to be sure everything is being done the way it's supposed to. Then you'll need to carefully check the data against your data-quality objectives when finally you get the lab reports back, often after a harassing phone call or two to your best buddy, the client manager at the lab.

Finally, after you've gotten the fully validated and/or qualified data entered into some kind of electronic format, you'll need to deal with those pesky nondetects - that is, if you're going to make proper sense of the data set you've just gone to all that time, trouble, and expense to produce (this is where the "useful" part of doing it right comes in).

Speaking of doing it right, stormwater programs normally are considered to be successful if they get to first base safely; that is, if they do a decent job of characterizing runoff quality through physical and chemical water-quality measurements. But a fully developed stormwater monitoring program also should incorporate sediment/particle quality, litter characterization, and toxicity studies. (Note that this article focuses on monitoring stormwater [i.e., runoff] discharges; receiving-water monitoring can involve another suite of activities, including stream morphology analysis, habitat evaluation, and bioassessment.) A brief overview of each of the stormwater monitoring areas follows.

Runoff Water Quality. This is the issue one typically associates with stormwater monitoring, with a focus on measuring the chemical and physical properties of runoff, through water-quality sampling and analysis. This is not limited merely to rainfall runoff - the higher elevations of the Sierra Nevada mountain range also experience abundant snowfall (yes, it snows even in California) - and snowmelt runoff (Regenmorter et al., 2002a). Whenever possible, runoff water-quality monitoring programs also include flow measurement so that pollutant loadings may be calculated as well.

Sediment/Particle Quality. Much of the pollutant load associated with stormwater runoff is carried by sediment, and the removal of sediment or prevention of its transport to surface waters is the major aim of most stormwater pollution controls (Regenmorter et al., 2002b). Efforts therefore have been devoted to providing protocols for determining the particle size distribution and chemical composition of the sediment found in its drain inlets and pollution control structures.

Litter. This is not what we typically think of when we think about water quality, and it's often overlooked. But in some areas, particularly those near both large urban centers and coastal waters, litter is a leading water-quality concern. Litter in many cases represents the most visible water-quality impact of human activity. To address this concern, monitoring protocols have been designed to characterize the volume, mass, and composition of litter transported by stormwater runoff.

Toxicity. It is well known that urban runoff often is toxic to aquatic life, and toxicity studies represent a means of directly measuring the effects of the pollutants present in stormwater runoff. To help determine whether runoff produces toxic effects, protocols have also been developed for toxicity studies in accordance with standard EPA procedures. Toxicity testing typically is undertaken in association with runoff water-quality monitoring.

To meet these challenges, a group of environmental consultants, university researchers, lab geeks, public agency folks, and assorted critics pooled their knowledge and conspired to develop a comprehensive system for stormwater monitoring in California. This advanced monitoring system is designed to ensure that stormwater monitoring programs will be cost-effective, will produce data that have scientific credibility, and will result in information that is useful in managing runoff.

This intrepid and loosely affiliated group has developed - and tested and revised - a comprehensive set of protocols and tools for stormwater monitoring that, in combination, is considered to be unique in the field. This system is designed to cover all aspects of stormwater monitoring, from project planning to data management, and includes the following components:

This set of tools and protocols provides monitoring personnel with the means to plan and implement sound monitoring programs and to verify and interpret the monitoring data. Description of the major features of this advanced monitoring system, as well as their practical application to stormwater management, is the principal focus of this article.

A description of each of the major components in this advanced monitoring system follows.

The use of standardized procedures is essential if monitoring data are to be compared - not just from one monitoring event to another, but also from project to project, region to region, year to year. If consistent methods are not applied, it might be difficult to perform data comparisons, and the value of the data might be diminished.

And it is not enough just to standardize the procedures - they also must be the right methods for the job. The sample collection regimen (composites, grabs), sample-handling techniques, and analytical methods all must be designed to adequately characterize the subject flow. Analytical detection/reporting limits must be established at levels low enough to provide quantifiable results more often than not for the particular type of sample matrix being monitored. (If this issue is not given adequate attention during the planning of the monitoring program, lots of money can be wasted in the production of lab reports reading "ND.") If flow-proportioned composites are to be collected, the flow rate and volume must be regularly tabulated; again, during project planning the flow measurement technique and equipment must be designed to suit the characteristics of the particular monitoring site.

We therefore developed monitoring protocols guidance manuals to ensure that the data produced by stormwater monitoring projects use consistent methods, produce scientifically defensible and credible data, and are cost-effective. A technical committee was established to guide production of each of the individual subject area guidance manuals, led by experts experienced in the field. The principal aims of the manuals are to:

The guidance manuals cover all activities related to planning and implementing monitoring programs; there is a manual for each of the four major data categories (water quality, particle/sediment quality, litter, toxicity). The manuals provide guidance in project planning, from establishing reasonable, achievable study objectives, to selecting appropriate sample collection and analytical methods. The manuals also include specific step-by-step procedures and easy-to-use checklists for field crews. Some of the key aspects of these protocols are briefly described below.

Automated Composite Sampling. Flow can vary significantly throughout a runoff event, and runoff quality is known to vary as well (Kayhanian, 2002; Kayhanian et al., 2003). Flow-proportioned composite samples therefore are considered to be the most representative sampling regimen for runoff monitoring. The preferred setup for stormwater monitoring projects is to employ automated monitoring equipment to collect an equal sample volume (aliquot) for every increment of a preset runoff flow volume. The key elements of the standard automated setup include an automated composite sampler, a flow meter, a rain gauge, and a programmable data logger/controller (Figure 1). The runoff volume increment is set in advance based on the quantity of precipitation forecast, such that an adequate number of aliquots will be collected to provide sufficient composite sample volume for all planned analyses. The composite sample then covers the full event hydrograph - and accounts for variation in flow and/or runoff quality throughout the runoff event.

Grab sampling, or even time-proportioned compositing, by contrast, generally will produce data that are less representative of any given runoff event. Depending on the sample timing and the nature of the events monitored, this can lead to a bias in the reported results. For constituents such as bacteria, for which grab sampling is required, an effort should be made to collect samples near midstorm, and ideally at the peak of the hydrograph, so as to provide the most representative characterization.

Site Selection. There are many ways that stormwater samples can be compromised - and therefore many considerations to address in selecting monitoring sites. Adjacent land uses might contribute pollutants to the runoff from the area you are trying to characterize either through aerial drift or via direct run-on onto your site. It might not be feasible to gain safe access to the flow, so avoid those midstreet access holes and freeway breakdown lanes and look for an alternative site if you can't get safe access. You will need to ensure an adequate power supply for your monitoring equipment and security for the equipment and your field personnel as well. If you are doing flow-proportioned compositing, you will need to be sure you can measure flow fairly accurately under all flow conditions that can reasonably be expected. The specific configuration of the site might affect the types of flow measurement options you can choose from. Our guidance manuals provide descriptions of the various factors to consider when selecting monitoring sites, and a site evaluation form is provided for use in the field (yes, you really do need to go out there before you can safely determine whether it is a good site or not).

To provide superior-quality analytical data, low-level analytical detection limits generally are required for trace metals, organic compounds, and some nutrients. The lowest analytical detection limits reasonably achievable for runoff samples (a moderately dirty matrix) must be determined in advance, and labs must be selected principally based on their ability to provide superior analytical performance. Cost - while obviously a consideration - should not be the main driver in lab selection. It is better to spend more dollars per sample and get good data than to spend less and get data rife with nondetects or beset with questionable quality-control issues. Our guidance manuals provide specifications for analytical methods and reporting limits.

In turn, when you are using low-level analytical methods, clean-sample handling techniques are required to reduce the possibilities of sample contamination. The sampling protocols should contain specific clean-sampling instructions.

A comprehensive QA/QC program is necessary to ensure that sample contamination is minimized and to provide data with documentable accuracy and precision. The guidance manuals provide a series of steps to be taken both in the field and in the lab. Each monitoring project must specify a schedule for the monitoring season, listing the events and locations for collection of field blanks, field duplicates, laboratory duplicates, and matrix spike samples (i.e., all the samples for which field crews will need to collect extra water). Labs must be instructed to perform the specific analyses required (according to your methods, with your reporting limits) and should be requested to provide the results of additional internal lab quality-control analyses, such as surrogate spikes and lab control samples.

Then, following receipt of the laboratory data, a thorough data-quality evaluation is performed. In this evaluation, the results of QA/QC sample analyses are compared to the project's data-quality objectives and suspect data are qualified (flagged) as necessary, following guidelines established by EPA for evaluation of inorganic and organic analyses.

Sample Representativeness.Two measures were developed to determine whether a composite sample might be deemed adequately representative of the runoff event from which it was collected. Each composite sample consists of a number of individual sample aliquots collected on a flow-proportioned basis throughout the runoff event; the aliquots then are composited together to form a composite sample that can be analyzed by the laboratory. First, a minimum number of sample aliquots must be collected for the event, based on the overall rainfall amount. Second, a minimum "percent capture" is also specified for each event; essentially this is defined as the percentage of total event runoff flow during which composite sample collection occurred. An example is shown in Table 1.

Table 1. Monitoring Event Representativeness Requirements
Total Event Precipitation	Minimum Acceptable Number of Aliquots	Percent Capture Requirement
0-0.25 in.	6	85
0.25-0.5 in.	8	80
0.5-1 in.	10	80
>1 in.	12	75
Source: Caltrans, 2000 (Table 10-1)

Even if your monitoring teams are all "doing it right" and following the techniques you've carefully established for sample collection and analysis, the product of your hard work still will be a relatively useless pile of data reports if there isn't consistency in, and an established protocol for, data reporting.

For each of our four monitoring data categories (runoff water quality, particle/sediment quality, litter, and toxicity), we have developed a complete set of data-reporting protocols. The specific contents of the data-reporting protocols were derived from the design of the database (see component 5: Data Management Tool). These protocols provide detailed specifications for data fields and instructions for content (how to populate each field). The protocols help ensure that data from all projects will be reported in a consistent format - and that the data records will include sufficient information to permit their full use within a common database.

One of the tricks to having the data-reporting process proceed smoothly is to coerce the labs into reporting the data in your format. Maybe you will have to pay them a little extra for them to build your format into their laboratory information management systems software, but it will be worth it.

Another key is that the field crews (and the monitoring task manager) need to get used to recording all of the site and event information required to fully characterize each monitoring event (in the format you provide, of course). This all requires advance planning (before you develop your database structure), but this front-end effort proves to be absolutely essential in terms of making the data accessible and useful down the road.

Once the data reports come in from the labs, there is a crucial step that often is overlooked - the data and lab reports must be reviewed and evaluated both for conformance with the data-reporting protocols and for compliance with the project's data-quality objectives. To provide for efficient evaluation of analytical data, Automated Data Validation software was developed (Amano et al., 2001). This automated program permits quick and efficient evaluation of lab data against data-quality objectives and standard measures of data quality and provides extensive error-checking for a standard set of possible analytical or data transcription errors. The resulting electronic data deliverable (EDD) is then ready for final checking prior to entry into the California Department of Transportation (Caltrans) stormwater-quality database.

The final data validation step involves checking that the EDD conforms to the Caltrans Data Reporting Protocols for the specific data type; corrections are made as necessary to provide information for any missing or improperly populated data fields.

Given that we have established representativeness criteria for composite samples, monitoring personnel need a means with which to evaluate conformance with those criteria. To make this into something that can be done reasonably, even in the field, we developed a hydrologic software utility to provide the information needed to determine whether the sample representativeness criteria have been met for a given monitoring event. This software utility is used to convert flow and sample aliquot data into usable information and allow assessment of sampling representativeness on-site.

The hydrologic utility actually serves multiple purposes: It allows monitoring personnel to assess representativeness, but it also serves to standardize calculation of important storm and sampling parameters, such as total flow volume, total event rain, and estimated percent capture. In addition, the utility generates a hydrograph and a hyetograph in a standardized format from measured hydrologic data.

The hydrologic utility is installed as an "Add In" in Microsoft Excel and is composed of a number of Visual Basic subroutines. The calculations and graphs are created from user-controlled input parameters as well as datalogger exports of rainfall, flow rate, and sample data records. The utility output is in two parts; a new worksheet and a new plot (chart) are added to the workbook. The output worksheet contains the processed input data that act as the source data for the plot and a summary table of important calculations. The output hydrograph and hyetograph plot note the timing of the primary composite sample aliquots, when they are provided, and include a table of important summary data. The plot and the calculated parameters are output to a worksheet page that can be printed and added to event reports. An example of the hydrologic utility output is shown in Figure 2.

Data management - i.e., data storage and retrieval - is the next big challenge. A giant pile of paper data reports is obviously an inefficient and obsolete way to do business. But even data in electronic format (such as Microsoft Excel spreadsheets) can be unwieldy or even nearly impossible to use effectively, without an additional level of organization. A well-designed database that provides a consistent format for data storage and also provides the means for efficient data retrieval is essential. Add a user-friendly interface and you've got a data management tool that someone might actually be able to use! And that is exactly what we did. It took some development, and adding the particle/sediment data, litter data, and toxicity study data has been a real chore, but when someone asks for a certain set of data, we can give it to them easily and efficiently. And this is a benefit that will continue to pay dividends well into the future.

In our system, once the error-checking and data validation process is complete, the data from the validated EDDs (in the form of Excel spreadsheets) are imported into an MS Access database (you can, of course, choose your own favorite database platform). Data are stored in three main tables: sample description, event description, and site description. The fields and content guidelines for each of these tables are described in the Data Reporting Protocols for each category of data (runoff water quality, particle/sediment quality, litter, toxicity).

The database includes a user-friendly interface with a geographic information system based map feature and menu-driven query screen. This interface permits quick and easy retrieval of data based on user-selected parameters. Users can select individual sites or groups of sites from a map or a menu and then step through a logical menu system to refine the query.

This custom-designed user interface and the database together comprise a powerful data management tool. A screen shot from this data management tool is shown in Figure 3.

Even with all of the planning and precautions you've taken to minimize generation of nondetect data, you'll still end up with a few "ND"s here and there. But even though those data are "censored" (the statistician's term for nondetect data) they still provide potentially useful information. That is, it should be worthwhile to know that copper is not present at a level above 1 mg/L. When you instead substitute an arbitrarily selected value, such as zero or half the reporting limit for nondetect results (as we all used to do), you might be artificially biasing the statistics (Kayhanian et al., 2002). Our team evaluated all available science-based statistical methods and selected the most suitable technique to calculate descriptive statistics (i.e., mean, median, standard deviation) when the data set contains nondetects (Shumway et al., 2002).

Our team developed a software tool that incorporates probability-based techniques for handling nondetects. The Data Analysis Tool (DAT) can be run on user-selected data sets to efficiently generate descriptive statistics for monitoring data. The tool runs directly from the database interface screen (with the click of a button), or it can be used as a stand-alone Excel Add-In. The DAT employs a Regression on Order Statistics technique for appropriate statistical treatment, including handling of nondetect data. This tool uses the detected values and a combination of regression and probability analysis to determine a "fill-in" concentration value to assign to all data points below the reporting limit (nondetects) based on an assumed log-normal probability distribution. The filled-in values are then used in statistical analysis. An example of the output from the DAT is shown in Figure 4.

Summary

Stormwater monitoring is not an easy job. And that fact is partially responsible for the very wide variation in methods used to produce stormwater monitoring data - and the corresponding variation in overall quality (and usefulness) of that data. Recognizing that it is a tough job is perhaps a good first step in learning how to "do it right." Developing a comprehensive set of protocols that provides for a consistent approach to monitoring - covering the whole sequence of events needed to successfully plan and implement the program - is a good second step.

It is essential to understand that stormwater monitoring is a complex activity that requires specialized expertise and training, extensive advance planning, and real dedication to follow-through. Much of the knowledge needed to plan and implement a successful stormwater monitoring program cannot be found in EPA guidance because the required knowledge is the result of practical experience. That experience is derived from lots of effort invested in trying to figure out how to make programs really work, from establishing achievable objectives to providing effective data management. As this article demonstrates, there are resources and tools available to assist those who have had a stormwater monitoring program dropped into their laps (from those of us who have been there).

A successful program can result from the proper blend of committed staff, technical wherewithal, and of course funding resources. All that and a willingness to brave the elements on a Saturday night - in the name of getting it done right.

Amano, R., L. Flynn, M. Kayhanian, and E. Othmer. "Automated Verification and Validation of Caltrans Storm Water Analytical Results." Proceedings of Annual Waste Testing and Quality Assurance Conference. Arlington, VA. August 2001.

Kayhanian, M.D. "Establishment of an Effective Total Maximum Daily Load Through Reliable Water Quality Data." Proceedings of StormCon 2002 Conference. August 2002.

Kayhanian, M., A. Singh, and S. Meyer. "Impact of non-detects in water quality data on estimation of constituent mass loading." Water Sci. and Technol., Vol. 45, No. 9, pp. 219-225. 2002.

Kayhanian, M., A. Singh, C. Suverkropp, S. Borroum. "The Impact of Annual Average Daily Traffic on Highway Runoff Pollutant Concentration." ASCE Journal of Environ. Engineering. 2003.

Regenmorter, L., M. Kayhanian, and K. Tsay. "Stormwater Runoff Water Quality Characteristics from Highways in Lake Tahoe, California." Presentation given at StormCon 2002 Conference, Marco Island, FL. August 2002a.

Regenmorter, L., M. Kayhanian, R. Chappel, T. Burgessor, and K. Tsay. "Particles and the Associated Pollutant Concentrations in Highway Runoff in Lake Tahoe, California." Proceedings of StormCon 2002 Conference. August 2002b.

Shumway, R., R. Azari, and M. Kayhanian. "Statistical Approaches to Estimating Mean Water Quality Concentrations with Detection Limits." Environ. Sci. Technol., Vol. 36, No. 15, pp. 3345-3353. 2002.

Acknowledgements

Funding for the development of the guidance manuals and other tools described in this paper was provided principally by the Caltrans Division of Environmental Analysis Stormwater Management Program. Caltrans conducts studies involving monitoring of runoff from transportation facilities, including highways, construction sites, park-and-ride lots, and maintenance yards. Since 1998, Caltrans has monitored runoff from numerous storm events at sites throughout California. The current statewide database of Caltrans stormwater monitoring data includes more than 120,000 data records.

Many people contributed to the development of the monitoring guidance and various tools described in this paper. We especially thank Kuen Tsay of Caltrans for his input.

Armand Ruby is with Larry Walker Associates in Davis, CA. Masoud Kayhanian, Ph.D., is with the University of California, Davis.