24-10-2012, 03:29 PM
Objective Reduction of the Space-Time Domain Dimensionality for Evaluating Model Performance
ABSTRACT
In the USA, photochemical air quality models are the principal tools used by governmental agencies to develop emission
reduction strategies aimed at achieving National Ambient Air Quality Standards (NAAQS). Before they can be applied with
confidence in a regulatory setting, models’ ability to simulate key features embedded in the air quality observations at an
acceptable level must be assessed. With this concern in mind, the U.S. Environmental Protection Agency (EPA) has recently
completed several executions of the Community Multiscale Air Quality model (CMAQ) and the Regional Modeling System
for Aerosols and Deposition model (REMSAD) to simulate air quality over the contiguous USA during year 2001 with a
horizontal cell size of 36 km x 36 km. The meteorological model MM5 and the emission processor SMOKE were used to
generate the input fields necessary for CMAQ and REMSAD. Since these annual model simulation generate a huge amount
of information, failure to properly organize the results may lead to confusion and hamper the evaluation procedure. The
challenge is therefore to identify a technique that would make use of all pertinent observations over a large region and clearly
indicate which spatial and temporal features are reproduced by the model. To address this challenge, we propose a procedure
to objectively condense the spatial and temporal observational domain into a limited number of homogeneous categories.