You are here:
Development and Application of a Simple Model Assessment R Tool (SMART)
Citation:
Kang, D. Development and Application of a Simple Model Assessment R Tool (SMART). 2018 CMAS Conference, Chapel Hill, NC, October 22 - 24, 2018.
Impact/Purpose:
It is always desirable to have a simple, portable, and easy-to-use tool to perform model evaluations using observational data sets. This R-based model evaluation tool can be used to perform quick model assessment for model development, evaluation, and applications.
Description:
As an environment for statistical computing and graphics, the programming language R is widely used and freely available. The EPA’s Atmospheric Model Evaluation Tool (AMET) is developed using R to produce statistical analysis and generate various graphics. Complementary to AMET, SMART provides an alternative tool to perform model evaluation for simple projects independent of databases and other languages for pre-processing. With the design of unique data structure and common interfaces to process model outputs and observations, SMART can easily pair any model outputs with observations from common observation networks and special field studies. The large memory requirement and slow processing speed associated with the R platform (the language and running environment) when dealing with large datasets are resolved by using SQLDF packages and parallel computing packages. Saved R objects serve as pseudo database tables to perform statistical analysis and generate graphics for various model scenarios. With functionality to subset and merge for the designed data structure, it is easy and fast to generate statistics and graphics for any subsets of the model-observation pairs in time and space. Since R is the only language used for site-compare and statistical analysis, it is completely portable to any computing environment with R installed. To run SMART, except for the shell script that is used to define all the environmental variables (such as input and output directories and variable names), users are not required to have any knowledge of R.
