~~ model still under development ~~
The AP3 model is a model that estimates marginal damages from pollutant emissions that contribute to the formation of fine particulate matter in the contiguous United States. This type of model is called “integrated assessment model” because it combines several steps that are required to derive monetary damages for location-specific pollutant emissions:
- Emissions need to be fed into the model based on their location and type of emission (ground level such as transit is dealt with differently than emissions through smokestacks of power plants) and sorted by pollutant (we deal with five major precursors to particulate matter)
- The next step is a very simplified air quality model to estimate PM2.5 concentrations in all these counties
- The model is then iteratively run with a perturbation of emissions for each location, emission type and precursor pollutant and the differential concentrations from the emission perturbations across all counties are recorded
- Lastly, we compare concentrations after emission perturbation with the baseline in step 2 to compute concentration differentials, store them and then assess the number of people exposed to quantify health impacts
Inputs The inputs for the model are also computed by us, but this step is not part of the model I’m creating for this project. Emissions:
- Area Sources: 3,109 counties x 5 pollutants
- Low Stacks: 3,109 counties x 5 pollutants
- Med Stacks: 3,109 counties x 5 pollutants
- Tall Stacks: 565 facilities x 5 pollutants
- New Tall Stacks: 91 facilities x 5 pollutants Population files are also preprocessed elsewhere and the data stems from post-censal and intercensal estimates. Mortality data is taken from the Centers for Disease Control and Prevention (CDC).
- Population: 3,109 counties x 19 age groups
- Baseline Mortality: 3,109 counties x 19 age groups
These matrices are the core of the model as they describe the transport of pollutant mass from source counties to receptor counties.
- Area: 3,109 counties x 3,109 counties (4 different matrices)
- Low 3,109 counties x 3,109 counties (4 different matrices)
- Med 3,109 counties x 3,109 counties (4 different matrices)
- Tall Stacks: 565 facilities x 3,109 counties (4 different matrices)
- More Tall Stacks: 91 facilities x 3,109 counties (4 different matrices) S/R matrices undergo a change depending on the year in a secondary calibration step.
Emissions, Population and Mortality depending on year (2008, 2011, 2014 and 2017 so far) A number of other parameters are required for the model to run: a dose-response function (or damage function) describes the relationship between exposure to particulate matter and increase in mortality risk — we have two of them: one for infants and one for people aged 30+; age groups 2-29 are assumed to be impervious to adverse health effects from PM exposure. Further, a value of reduced mortality risk (VRMR) monetizes the increased mortality risk.
Potential intermediate results: estimated air pollution deaths by the model in baseline configuration Average estimated concentration Marginal Damages:
- Area Sources: 3,109 counties x 5 pollutants
- Low Stacks: 3,109 counties x 5 pollutants
- Med Stacks: 3,109 counties x 5 pollutants
- Tall Stacks: 565 facilities x 5 pollutants
- New Tall Stacks: 91 facilities x 5 pollutants
To predict ambient PM2.5 concentrations, AP3 augments the Gaussian dispersion modeling of emissions and resolved, speciated pollution with a rudimentary representation of atmospheric chemistry processes that represent the equilibrium between ammonium, sulfates, and nitrates. Emitted SO2 and NOx convert into sulfate and gaseous nitrate. In each county, ambient ammonium, determined by NH3 emissions, reacts with sulfate first to form ammonium sulfate. Then what ammonium remains is free to react with gaseous nitrate to form ammonium nitrate. The two resulting regimes, nitrate-limited and ammonium-limited, critically affect the efficiency with which emissions of NOx form ammonium nitrate, which is a subspecies of secondary PM2.5 whereas gaseous nitrate is not. AP3 then aggregates all subspecies of ambient PM2.5, including sulfate, ammonium sulfate, ammonium nitrate, ammonium aerosols, organic PM2.5, and primary PM2.5 to determine total concentrations in each county.
AP2's approach to the atmospheric chemistry module was to compute the formation of ammonium nitrate only after the formation of ammonium sulfate, a situation leading to no NOx-driven PM2.5 in many counties. However, emissions and resolved concentrations occur throughout the year, not all at once. Therefore, statistical methods were adopted to predict concentrations of ammonium nitrate in counties depending on whether there is or is not free ammonium available based on the annual calculation. Simply put, the modeling went from all or nothing to more or less ammonium nitrate formation from resolved NOx.
AP3 is calibrated using EPA Air Quality System (AQS) monitoring data. The model uses EPA's reported monitor data from PM monitors for calibration, where available it is calibrated to speciated monitor data. The primary calibration step is an iterative process to reduce mean fractional error and mean fractional bias in an iterative process, although the monitors do not include all components of primary particulate matter. A secondary calibration step is therefore conducted to correct the PM2.5-primary receivable portion of the 2.5th percentile of counties in terms of absolute difference between model and monitor output, and subsequently neighboring counties are considered for further adjustments if monitor data and PM model estimates support such decisions.