Step 1 in developing numeric nutrient criteria (NNC) using the reference condition approach is to establish your reference condition (e.g., least disturbed condition). If using the spatial reference approach, base your selection of reference sites on those that meet designated uses and a combination of landscape, local habitat, and other factors. Reference water bodies typically have upstream land uses comprised predominantly of natural habitat such as forests or wetlands and water body geomorphology that is undisturbed such as in channel or lake beds that do not have excessive sedimentation. If you are using the temporal reference approach, identify waters with a known historical period of time predating human disturbance that had land uses comprised predominantly of natural habitat and supported designated use(s) as in the spatial approach. In step 1, you establish data screening requirements to identify appropriate reference sites. You will use the resulting data sets later, in steps 2 and 3, to classify sites and create statistical distributions of each variable of interest and, based on a selected percentile, determining numeric nutrient criteria (NNC).

For lakes, EPA recommends that you develop data sets for total phosphorus (TP), total nitrogen (TN), chlorophyll a, and Secchi depth. Other parameters, such as dissolved oxygen (DO), macrophyte growth or speciation, and organic carbon also are sometimes useful. For rivers and streams, EPA recommends using TP, TN, and chlorophyll a as primary variables. Secondary variables might include DO and pH, conductivity, and dissolved organic carbon. For estuarine and coastal waters, EPA recommends you use TN, TP, and silica as primary variables and chlorophyll a, macroalgal biomass, measures of water clarity (e.g., light attenuation coefficient, Secchi depth), DO, carbon compounds, and benthic macroinfauna as secondary variables.

To identify reference sites, you will first establish a set of screening requirements or rules designed to ensure final data sets containing high-quality data that truly represent sites that satisfy the a priori definition for the reference condition. You will assess your reference database based on those rules and eliminate any data that do not meet or conform to them. Your final product of this step will be sets of high-quality data for each class that represent the best range of reference conditions that can be expected of similar waterbodies in the class and, therefore, reflect the desired ecological condition that supports designated uses.

Developing Rules

In general, focus on creating screening rules for three main areas:

• Data quality
• Data quantity
• Representativeness of reference conditions

Data quality

The following elements of data quality are important to developing defensible criteria and are similar to requirements for any analytical approach:

• Ensuring sample integrity is maintained, evidenced by complete and correct sample transmittal documentation as well as records of adherence to approved sample preservation methods, transportation, and sample-handling protocols.
• Having associated metadata, so data can be traced to a sampling site, date, and time.
• Using approved (EPA/state) sample collection and laboratory analysis methods. Sufficient use of quality control measures in the laboratory to establish the precision, accuracy, sensitivity, selectivity, and potential bias associated with the analytical system and associated results.
• Documenting all data generated so that all field sampling, handling, and laboratory processing can be reconstructed and ensure that data are verified and validated.
• Recording instrument calibration and verification of performance.
• Carefully documenting and communicating adherence to data quality objectives.

The quality of older historical data sets is a recurrent problem because the data quality is often unknown. This is especially true of long-term repositories of data and academic databases, where objectives, methods, and investigators could have changed many times over the years and are frequently undocumented. The most reliable data tend to have been collected by a single agency using the same protocol for a number of years. Always examine supporting documentation to determine the consistency of sampling and analysis protocols.

When selecting which data to use, filter data for acceptance considering a number of variables, including the following:

• Location
• Analytical methods
• Laboratory quality control
• Data collecting agencies
• Time period: Long-term records are critical to establishing trends

Data quantity

To properly characterize reference conditions, adequate data associated with the reference waters must be available. If existing data are deemed to be insufficient for characterization purposes, then an adequate number of reference sites must be sampled to obtain the additional data necessary to characterize reference conditions. Confidence in estimated percentiles depends on the number of samples (i.e., reference sites) and the percentile being estimated.

• The smaller the sample size, the lower the confidence level (greater the uncertainty) around a percentile estimate.
• Percentiles distant from the central tendency are estimated with less confidence than percentiles close to the central tendency (i.e., 90^th percentile estimates are more uncertain than 75^th percentile estimates, all else being equal).

Other important elements to consider in data quantity are both homogeneity and heterogeneity of data in space and time.

There is no definitive rule for establishing a minimum number of reference sites appropriate to the development of a reference condition, and the greater number, the better. A general rule of thumb for reference sample size should be about 20 to 40 water bodies per water body class in order to estimate upper percentiles confidently.

The quantity of data should enable:

• Capturing of variability across space/time (ideal case)
• Spatial or temporal representativeness
  • Site-specific—Need considerable representation over time
  • Regional—Need considerable representation over space

Representativeness of reference conditions

A set of screens should be developed and used to identify least disturbed reference sites in space and during the timeframe of interest, including removing any locations with state 303(d) listings. Ensure that the final sites selected accurately reflect the desired ecological condition and designated use. Data screening criteria should also be applied to ensure the representation of reference conditions over the period of record. Locations or time periods that do not meet the reference site requirements or are impaired (e.g., data from time periods on the 303[d] list) should be excluded from the reference data set.

Variables to consider for screening sites include the following:

• Impairment status: EPA has found that 303(d) listing and biological condition screens can be helpful and applicable in identifying minimally or least disturbed sites
• Land cover/landscape development intensity
• Habitat condition
• Water quality
• Biological indicators, such as a biological condition index
• Best management practices
• Dischargers and the presence of point source dischargers
• Management
• Riparian zone condition

In characterizing reference conditions for nutrient criteria, it also is important to determine if trends exist in the reference site database. For example, since passage of the Clean Water Act and elimination of most discharges, many water bodies have improved markedly. Other waters, subject to increased nonpoint source runoff, might have declined in overall quality.

Examples of temporal screens include the following:

• Demonstrated attainment of other criteria during that time period (i.e., not impaired during the designated time period)
• No evidence of adverse nutrient impacts
• No evidence of adverse trends in nutrient concentrations or nutrient related responses over time
• Pre-discharger conditions