Selecting Datasets and Candidate Predictor Variables Sample Clauses

Selecting Datasets and Candidate Predictor Variables. ‌ The emissions and farm and lagoon data used to develop the NH3 EEMs for lagoons were collected under the NAEMS from six swine sites (IA3A, IN4A, NC3A, NC4A, OK3A and OK4A) and three dairy sites (IN5A, WA5A and WI5A). The final reports submitted to the EPA contained daily lagoon emissions calculated based on measurements obtained using both the RPM model and the bLS model. In developing the NH3 EEMs, the EPA used the measurements obtained using the RPM model. The EPA used the RPM data because these measurements were obtained using instrumentation and procedures that were similar to EPA’s developmental test method OTM-10 (Optical Remote Sensing for Emission Characterization from Non-Point Sources). The EPA did not use the bLS emissions measurements because these data were collected under the NAEMS to conduct a validation study of the bLS model performance relative to the RPM model. Furthermore, because the RPM emissions dataset is much larger than the bLS dataset, including the bLS measurements in the EEM development dataset would not provide any additional information on lagoon emissions. Therefore, the EPA chose to only use the RPM emissions data as it represented the largest unique NH3 emissions dataset available from the NAEMS. Due to the very limited number of daily NH3 emissions values reported, the EPA prepared a database of half-hour values by combining the 30-minute emissions data and 5-minute meteorological data provided by the NAEMS researchers. To obtain values for each meteorological variable for a given half-hour, the EPA averaged the 5-minute data values within that half hour, as long as at least one of the values was not missing. Because it is unlikely that the value of any of these meteorological variables would vary enough within the consecutive 5-minute periods that comprise a half hour to significantly change the mean value of the variable during the half-hour, applying a completeness criterion to this aggregation would unnecessarily reduce an already limited data set. In the data aggregation, the EPA did not include any 30- or 5-minute data that had been flagged by the NAEMS researchers as invalid. The resulting dataset contained a small number of negative NH3 emissions (approximately 0.6 percent of the total values). Because such a small number of observations relative to the full dataset would be unlikely to provide information about variability in NH3 emissions measurements, the EPA omitted these values from the EEM development data...