Lagrangian Modeling Approaches and Measurements to Confirm an Improvement in Environmental Quality due to Reductions in SO2 Emissions
Roland R. Draxler
Overview
Can the effects of reducing pollutant emissions be detected sooner through "smart" sampling methodologies?
Smart = Model Results
Variations in air quality and deposition are primarily due to ...
meteorological, chemical, and emissions factors
on diurnal, seasonal, annual, and decadal time scales
If much of the variance can be explained by a model then ...
remaining trends can be attributed to emissions!
LAGRANGIAN MODELING APPROACHES
RECEPTOR ORIENTATION
Receptor approaches use trajectories, without dispersion, to define the upwind advection path from the measurement point.
Dispersion is not reversible.
CLUSTER ANALYSIS - Single Station Sectorial Sampling
Simple - no restrictions on trajectories
Lagrangian - account for upwind variations
GRIDDED TABULATION - frequencies accumulated on a grid
Simple Single Station
or from ... Multiple Stations
Lagrangian variations can be included
SOURCE ORIENTATION
Calculations including both advection and dispersion are made from a predefined source location to determine the contribution at a receptor.
Emissions and chemistry can be included.
SOURCE-RECEPTOR MATRIX
Calculations performed from all source regions to all receptors.
Results can be inverted to define a receptor orientation
Can include non-linear chemistry
PREDICTION DIFFERENCING
Compute trend in residuals between measurements and calculations.
SUMMARY
CLUSTER ANALYSIS
Advantages - Easy to understand
Yields results quickly
Disadvantages - Primarily a spatial analysis
Multiple receptors
GRIDDED TABULATION
Advantages - Lagrangian factors
Multiple stations
Disadvantages - Application to time series measurements
SOURCE-RECEPTOR MATRIX
Advantages - Daily fractional contributions
Lagrangian factors implicit
Chemistry
Disadvantages - Data management of huge files
Chemistry
PREDICTION DIFFERENCING
Advantages - Source integrated version of matrix
Residual distribution
Disadvantages - Linking specific receptor to source
FUTURE DIRECTIONS
STATISTICAL EVALUATIONS
Non-parametric methods for computing confidence limits
Combination of methodologies
Use cluster methods to identify sites and days appropriate
for prediction differencing
MODELING SYSTEM
Improved precipitation predictions
Switch from NGM to ETA or RUC with observed data
Improved emissions inventory
Nested emissions grids
Finest resolution near sampling sites