• facebook
  • twitter
  • YouTube
  • RSS feed
  • 651-201-6000
  • 800-967-2474
  • 711 TTY

NodeFire Save Document
Home > Protecting Our Lands & Waters > Clean Water Fund > Clean Water Research Program > Evaluation of Alternatives

Evaluation of Alternative Surface-Water Monitoring Protocols for Use in Agriculture TMDL Load Allocation and BMP Evaluation

Principal Investigator: Dennis Busch
Co-Investigator(s): U.S. Geological Survey
Organization(s): University of Wisconsin-Platteville, Pioneer Farm
Sponsor: Clean Water Legacy Act
Award Amount: $32,300
Start Date: 5/18/09 | End Date: 6/1/2011
Project Manager(s): Adam Birr

Evaluation of Alternative Surface-Water Monitoring Protocols Final Report (PDF: 899 KB / 17 pages)


Edge-of-field runoff driven by rainfall, snowmelt, or a combination of rain and snowmelt, produce discharge events that vary greatly in total volume. It is difficult, if not impossible, to accurately pre-program automated samplers so that both small and large events are sampled adequately. What often occurs is insufficient samples are collected during small runoff events and sampler capacity is exceeded during large runoff events. To overcome this challenge, Pioneer Farm, in cooperation with the United States Geological Survey (USGS), has monitored runoff events in real-time via remote connections and adjusted the time between samples so that a much larger range of events can be monitored. Unfortunately, this method increases costs due to increased labor and increased sample handling and preparation.

This project is evaluating low-cost alternatives to the current monitoring procedure used at the Pioneer Farm. This work would be very informative for establishing sampling methodology for both Total Maximum Daily Load (TMDL) source assessment and implementation activities.


This research project compared the current monitoring protocol used by UW-Platteville Pioneer Farm and UW-Extension Discovery Farms that meet EPA guidelines (EPA) with two alternate, low-cost methods: 2-part flow-weighted automated sampling (2-Part FWC) and single-stage passive samplers (SS Siphon).

Current Monitoring (EPA)

Pioneer Farm monitoring methods (referred to as the EPA Method for this project) utilize a pre-calibrated H-flume with a pressure transducer stage recorder to determine flow. Samples are triggered based on time interval that is adjustable real-time remotely via radio telemetry and internet communication. After collection, samples are cooled in the refrigerated sampler until retrieved by a technician.

Alternative Method: 2- Part flow weighted automated sampling (2- Part FWC)

This method used an automated sampler capable of collecting samples based on two flow intervals simultaneously.

For this method, one sampler with a 24 bottle carrousel was set to two flow-weight compositing intervals: bottles 1-12 were set to 0.01 mm interval for small events, and sample bottles 13-24 were set to a 0.4 mm interval for large events.Two replicates (referred to as A and B) were installed at the Pioneer Farm.

  • Flow is determined using the H-flume instrumented with an ultrasonic flow meter
  • Samples collected were not refrigerated
  • This method did not have remote access

Alternative Method: Single-stage passive sampler (SS Siphon)

The method uses a passive sampler, which means it collects a sample without any external control or activation. The sample begins to collect when surface-water height in the flume exceeds the maximum height of the intake tube. This initiates a siphon and the sample bottle fills rapidly. Three siphon samplers are located along side each flume with their intake tube sample heights fixed inside the so that samples are collected at 0.2’, 0.5’, and 1.0’ stages. Two passive sampler replicates were installed at the Pioneer Farm.

Data Collection

All monitoring stations were managed and operated in conjunction with the United States Geological Survey for the duration of this study as described in the USGS Open File report 2008-1015 Methods for data collection, sample processing, and data analysis for edge-of-field, streamgaging, subsurface-tile, and meteorological stations at Discovery Farms and Pioneer Farm in Wisconsin, 2001-7.

It is important to note that not all runoff events that occurred during the experiment are included in this analysis. There were occasions when events were missed due to equipment failure or operator error. Failures occurred on all systems - including the USGS-PF operated sites. For example, large runoff events washed out the flume on one site, sampler lines froze on another, batteries failed on another, and aquarod data was overwritten during runoff events of long duration. While failures were more frequent with the FWC and SS-Siphon samplers when compared to the USGS-PF samplers, PF technicians have been operating these systems for 8 years and are much more familiar with these systems. For this reason, the comparisons were based on events for which data was available from both sampling systems. If all events were included, the observed differences may be more a result of operator experience than inherent capabilities of sampling equipment.

Evaluating Current and Alternative Methods

Three metrics were used to compare the two alternative methods to the EPA method: 

Relative error of event loads (nitrogen, phosphorus, and sediment) calculated by 2-part flow-weight composite and single-stage sampling strategies. The relative error will indicate how closely the alternative method compares to the EPA  method. For example, a relative error of 90% would indicate that the alternative method calculated a load that was 10% less than the actual load.

Precision of 2-part flow-weight composite and SS Siphon methods. The precision, or coefficient of variation (CV), will indicate how well the alternative methods are able reproduce load estimates. If sampling methods are imprecise (have a large CV) they may not be useful in evaluating BMPs. The CV is a ratio of the standard deviation to the sample mean; therefore events with large differences in sample means can be directly compared.

Costs of alternative methods including: equipment, operation, and maintenance.  


  • Total nitrogen (TN), total phosphorus (TP), and suspended sediment event loads for each sampling protocol
  • Relative error values for TN, TP, and sediment under each sampling protocol
  • Precision estimates for TN, TP, and sediment for each sampling protocol
  • Total cost summary for equipment purchase and installation
  • Discussion of research findings
  • Recommendations for future research


2-part FWC sampler method

  • This method adequately sampled both small and large events.
  • Regression analysis indicates good correlation for discharge estimates between the EPA and FWC methods. The coefficient of variation for discharge measurements, with the exception of the 1/24/2011 event, was less than 0.10 indicating a high precision of discharge estimate.
  • Estimates of total nitrogen concentration generated by the FWC method were generally higher the EPA estimates, with a relative percent error as large as 44.5%, but were not statistically different.
  • Total nitrogen loads for the FWC method were significantly different exceeding the EPA estimates by about 9 kg for all events analyzed. The CV data indicates that the two FWC samplers produced comparable total nitrogen load estimates.
    • Results for all analysis related to total nitrogen (low CV, high r2, and slope of 0.78) may indicate a systematic sampler, field method, or laboratory method error.
  • Event total phosphorus concentrations and loads estimated by the FWC method were both precise and very similar to loads determined by the EPA method.
    • FWC total phosphorus were consistently higher than EPA concentrations and resulted in significantly higher loads for FWC sampler A (observed p value = 0.03), even though the difference between the sum of event loads was only 0.8 kg or 7.8%.
  • Suspended sediment concentrations varied dramatically from EPA method estimates and among FWC samplers A and B but were not significantly different than EPA method concentrations (observed p = 0.48 and 0.47).
    • It is not too surprising that suspended sediment data exhibits the greatest error and least precision, since the within event concentration of suspended sediment varies much more than concentration of total nitrogen or total phosphorus.
  • The cost of equipment, installation, operation, and maintenance are lower for the 2-part FWC method. In the long-run most cost savings will be realized by eliminating the requirement for real-time remote monitoring of sites during events. The total cost per site for a three-year monitoring program would cost $75,120 for the EPA method and $42,320 for the 2-part FWC method.

SS Siphon passive sampler

  • Total nitrogen concentrations and loads were not significantly different from EPA method derived concentrations (concentration observed p values = 0.99 and 0.56, load observed p values = 0.61 and 0.22); however, relative percent error values for event concentrations and loads were large.  The error was not systematic. The sum of event loads estimated by siphon samplers was within 13.2% (sampler A) and -10.6% (sampler B).
  • Total nitrogen concentration estimates generated by single-stage samplers were not as precise as those estimated by the FWC samplers.
  • SS siphon samplers produced total phosphorus concentrations and loads that were not significantly different from EPA method estimates (observed p > 0.12), and were strongly correlated to EPA method values (r2 > 0.96).
  • The single-stage sampler suspended sediment concentrations differed significantly from EPA methods at the 10% significance level and relative error values were large- up to 339%. The sum of event load suspended sediment from the SS Siphon method was 685 kg and 729 kg compared to 419 kg suspended sediment load determined by the EPA method.
  • The cost of equipment, installation, operation, and maintenance are lower for the single-stage sampler method. Due to the simplicity of the equipment, initial investment and installation cost is low for the SS siphon method ($5,100) compared to the EPA methods ($21,120).
  • Technician time required for operation and maintenance is similar to the 2-part FWC method, therefore the cost of a multi-year monitoring project will be similar for the two methods.

Suggestions for Future Research 

  1. It would be of interest to compare the systems using technicians that had no previous experience with any of the monitoring protocols. This would provide information regarding the learning curve associated with each of the methods. Missed events could then be included in the data set and the effect of missed events could be determined on annual yield and load estimates.
  2. This experiment was conducted in one location for a relatively short period of time. A more reliable comparison would include a larger geographic area and longer period of record.
  3. It is assumed that the EPA method is determining the “actual” concentration and load. However, there is variability in all measurement. Future studies should include multiple EPA protocols within the same site so that precision of this system can be quantified.
  4. Lab studies that simulate runoff events using water with known sediment concentrations should be conducted and concentrations estimated with replicate EPA, 2-Part FWC, and SS Siphon sampling methods. With this data, accuracy and precision of all methods could be determined.

For more information please refer to the Final Report


MDA Contact

Margaret Wagner
Supervisor, Clean Water Technical Unit