WORKSHEET 11: DATA QUALITY OBJECTIVES
The USEPA has developed a seven-step process for establishing Data Quality Objectives (DQOs) to help maximize the likelihood that data collected in the field will be sufficient for the purposes intended (EPA 2006). The following sections implement this seven-step procedure for this study.
11.1 Step 1: State the Problem
As discussed in Worksheet 10, one premise of the OU2 risk investigation is that Cl2 and HCl are assumed to be acute constituents of potential concern (COPC), and airborne concentrations are known to occur in the vicinity of the U.S. Magnesium Plant. Concentrations of these contaminants in air are expected to vary substantially as a function of distance and direction from the Site, and also as a function of time, depending on both short-term (minutes-hours) and long-term (seasonal) variations in meteorological conditions, as well as variations in release rates from normal plant operations, maintenance activities, and equipment upsets. Available data collected to date, summarized in Worksheet 10, support the concept that the concentrations of these analytes over time are likely to be characterized by intervals of zero or low concentration with a series of intermittent “spikes” of varying concentration and duration occurring at random times.
However, the available ambient air concentration information for these two acute COPCs is too limited to support a risk assessment. The data do not adequately characterize the spatial or temporal patterns of concentration values in air, so additional data are needed to support the characterization of human and ecological exposures to Cl2 and HCl. In preceding RI activities, monitoring systems have been demonstrated that can obtain continuous and representative ambient air concentrations for these target analytes. These systems are to be deployed in Phase 1B to obtain data of sufficient quality and extent to support a risk assessment.
11.2 Step 2: Identify the Goal of the Study
The goal of the study is to collect sufficient data on the concentrations of Cl2 and HCl in air to reliably characterize the pattern of short-term spikes and long-term average exposure concentrations over space and time in order to support reliable evaluation of human and ecological exposures and risks. In effect, the goal of the Phase 1B monitoring program is to deliver sufficient ambient air information with a relatively small number of selected monitoring locations.
The focus for Phase 1B monitoring methodology is the acute exposure timescale. Therefore, the monitoring program must be designed to obtain a sequence of short-term ambient air concentrations in candidate areas in which the presence of human and ecological receptors is likely. Longer-term exposures may then be quantified through analysis of the short-term concentration data.
11.3 Step 3: Identify Information Inputs
The environmental data that need to be collected to support a reliable characterization of exposure and risk from Cl2 and HCl in air consist of reliable measurements of concentration values over short-term and long-term time scales. Because long-term average values may be derived from an adequate set of reliable and representative short-term values, only short-term measurements are included in this study.
Such monitoring data are required at multiple locations around the US Magnesium Plant site, with special focus on locations and times where different groups of humans or ecological receptors may tend to be present.
Additional information inputs to the development of the DQO and execution of the Phase 1B monitoring program are the data validation and quality assurance measures that will be practiced during and after the field portion of the monitoring program. These methods, as developed in a subsequent Sampling and Analysis Plan (SAP) will follow accepted practices to provide data of adequate quality for risk assessment. Appropriate metrics of magnesium plant production will also be recorded during Phase 1B, to verify that operating levels during Phase 1B monitoring are representative of normal production.
11.4 Step 4: Define the Boundaries of the Study
11.4.1 Spatial Boundary
Based on the expectation that concentrations in air will tend to be highest near the US Magnesium Plant and decrease as a function of distance from the Plant, EPA identified an initial study area described by a circle with a radius of 5 miles, centered on the Plant (EPA 2013a). But within this study area, more detailed spatial boundaries have been identified by USEPA as candidate areas for monitoring to assess exposure risk (USEPA August 2014). Based on qualitative considerations, these sites were viewed by USEPA as representative of the likely locations of human and certain ecological receptors, and that generally represented sufficient spatial coverage within the OU2 overall boundary. Further analysis using a combination of dispersion modeling simulations and ranking tools has shown that five monitoring locations can provide sufficient data for risk-based decisions (ERM May 2015).
11.4.2 Temporal Boundaries
Concentrations of contaminants in air at any specified location within the study area are expected to vary substantially as a function of meteorological conditions, plant production patterns and operational events, both over the short-term and also seasonally. Therefore, it is important that sampling be of sufficient duration to capture the range of values that occur over both the short-term (minutes-hours) and the long-term (seasonally). For purposes of designing the Phase 1B monitoring program, this DQO includes 1-minute instantaneous air concentration readings as the short-term unit of the airborne dataset. This monitoring interval can be reliably performed in field measurements and is a suitable metric for assessing acute exposure. To capture the effects of a full year of seasonal conditions and the potential emission rate variability, it is proposed that the duration of the Phase 1B sampling program be one year.
11.5 Step 5: Develop the Analytic Approach
Risks from exposures to short-term spikes of Cl2 or HCl will be characterized by estimating either the duration (minutes per year) or the frequency (events per year) that a specified receptor will experience. Repeated exposure to short-term spikes is also an aspect of long-term potential risk. Chronic risks will be evaluated by comparing the long-term effective mean exposure concentration to an appropriate chronic reference concentration. A summary of how both acute and chronic exposure will be estimated from 1-minute instantaneous air concentration readings is provided below. The specific methodologies of how these data will be used to assess potential human or ecological risks will be described at a later time in the baseline human health and ecological risk assessment technical memoranda for OU2.
11.5.1 Characterization of Hazard for Human Receptors
11.5.1.1 Risk-Assessment Data for Short-Term Exposures
Available continuous monitoring data surrounding the Plant is limited. But air data obtained during the DMA field work for OU2 does demonstrate the general pattern of concentration events. The variability of local winds was found to result in substantial periods with no measurable acute toxicants concentration present at a monitor with relatively brief, well-spaced events of detectable concentration. In effect, the acute toxicant concentrations pattern at a typical location can be represented as a time-series of short term concentration “spikes.” An example of this concentration data pattern is illustrated in the time-series of 10-minute average chlorine concentrations obtained during the Air Quality DMA in June 2013.
E
Figure 11.1 – Example DMA Field Data for Acute Toxicant Gas
The continuous contaminant data collected during the Phase 1B program at OU2 is expected to support a risk assessment that assesses human exposure to repeated, infrequent, acute exposures. On this basis, the monitoring data will support evaluation of concentration events that exceed a recognized threshold for reversible acute effects. Acute toxicity thresholds protective of human health and ecological receptors will be presented in the baseline human health and ecological risk assessment technical memoranda for OU2.
11.5.1.2 Characterization of Risks to Humans from Chronic Exposure
To quantify the potential chronic exposure to human receptors, longer-term average concentrations will be calculated using the 1-minute data collected during Phase 1B. The availability of a full-year of 1-minute data at several locations will allow mean average concentrations to be calculated for different timeframes and locations of interest. In constructing this average, an appropriate treatment of the prevalence of 1-minute data that are effectively non-detects is necessary. The bounding cases would be to either treat all non-detects as true zero concentrations, which has some basis in the observation that the plume transport is the only transient mechanism that conveys the gases to a given receptor. A more conservative bounding case would be to assign a level to the non-detects that reflects a background levels that are below the detection limit. Assessment of the treatment of non-detects and the comparison of long-term averages to toxicity thresholds will be addressed in the Baseline Human Health Technical Memorandum.
11.6 Step 6: Specify Performance or Acceptance Criteria
11.6.1 Decision Error Goals
In evaluating exposures from Cl2 and HCl in air, two types of decision errors are possible:
Type I error: In this case, it is concluded that exposure is within acceptable limits, when in fact the true exposure exceeds acceptable limits.
Type II error: In this case, it is concluded that exposure is above acceptable limits, when in fact the true exposure is within acceptable limits.
EPA is primarily concerned with minimization of the chances for a Type I error, since an error of this type could result in a failure to address exposures that are of potential human or ecological health concern. In general, EPA has a goal that the probability of making a Type I error should not exceed 5%.
Type II errors are of lesser concern, since a Type II error does not result in unacceptable exposures. However, Type II errors may result in the unnecessary expenditure of resources to address exposures that are actually within acceptable limits. Consequently, EPA typically seeks to limit the probability of Type II errors to within a reasonable tolerance. Although there is no standard rule for Type II errors, a value of 20-30% is often identified as a goal.
11.6.2 Characterizing Uncertainty In Short-Term Observations
Using Phase 1B continuous data, both short-term concentration event duration and frequency are to be calculated based on counts of the number of concentrations above an accepted threshold that are observed during some sampling duration. Type I and Type II errors may occur because of random statistical variation in the observed counts of concentration events. That is, the observed count may be either higher or lower than the true long-term average count.
Sampling at the representative monitoring locations as informed by air dispersion modeling provides reasonable assurance that Type I errors will not occur. These locations can be shown by data adequacy evaluations and possibly by additional modeling during the execution of Phase 1B to show that the selected locations have experienced concentrations that are representative of previous levels as predicted by long-term modeling. Therefore, if during Phase 1B the selected monitoring locations do not exhibit actual concentrations above the Cl2 or HCl exposure risk thresholds, then there is reduced probability that other locations could have experienced such exposures.
In order to help minimize the probability of Type II errors, it is desirable to sample for a sufficient duration such that if the monitored frequency of elevated concentration events is below a health-based threshold, then it can be concluded that the upper bound of the possible frequency of such events will also be below those thresholds. Because uncertainty decreases as a function of the total number of observations, longer sampling durations result in decreased statistical uncertainty. Because suitable health-based thresholds are not yet established, quantitative estimates of sampling duration needed to limit Type II errors are not possible. However, once the data are collected, the probability of a Type I and Type II decision error associated with a decision based on either the best estimate or the upper bound may be calculated and presented as part of the uncertainty discussion.
11.6.2 Characterizing Uncertainty in Long-Term Average Exposure
As noted above, the long-term average exposure concentration will be derived by first calculating the mean averages for each day that monitoring has occurred, and then using that data set to derive estimates of the longer-term mean and the Upper Confidence Limit (UCL) of the mean. Use of the UCL to estimate exposure concentration automatically limits the probability of a Type I error to 5%. Because the distributional pattern and variance of the data are not known, it is not possible to perform a power calculation to estimate the number of one-day values needed to achieve target tolerances for Type II decision errors. However, because each one-day concentration is derived from many short-term measurements (e.g., 1/minute * 60 min/hr * 24 hrs = 1440), it is expected that the high variance that is likely to be present in the short-term measurements will be greatly reduced in the set of one-day values. Consequently, it is expected that uncertainty in the estimate of long-term average exposure will be small.
11.7 Step 7: Develop the Plan for Obtaining the Data
Several different types of continuous sensors that have the ability to measure Cl2 and HCl in air have been evaluated at the Site (ERM 2014, 2015a). Based on field trials of the instruments, monitors available from Gastronics have been identified as the most reliable, and these monitors will be used to collect data at selected monitoring locations within the 5-mile study area. Instantaneous values of both Cl2 and HCl will be measured and recorded once per minute.
Details on instrument sensitivity, calibration, and maintenance as well as protocols for data collection, storage and management are provided in this section, and in the attached Gastronics specifications and Operating Manual.
11.7.1 Location of Sampling Stations
As noted above, the values and frequency of observed concentrations of Cl2 and HCl are expected to vary as a function of both direction and distance from the US Magnesium Plant. In order to ensure the data are adequate to reliably characterize spatial variability, an initial set of 12 candidate sampling station locations were proposed by establishing six equal pie-shaped sectors radiating outward from the plant. The orientation of the sectors was selected so that the first sector (designated “A”) lies with its center line oriented in the predominant downwind direction (southeast) from the Plant. Two monitoring locations were placed within each sector, focusing on areas where human and/or ecological receptors are expected to be present. The 12 candidate locations are shown in Figure 11-1 (Panels A and B).