Recommended Changes STI

Steve Brown and Song Bai have provided the following recommendations for EPA PMF 4.0. Please send me your commnets/suggestions.

Major proposed work for PMF:
• Develop functions to retain inputs and outputs for base model runs, F-peak runs, and constrained model runs; users can retrieve model results for further analyses (e.g., to explore more pulling scenarios without rerunning the associated base run each time). This is something I heard from a couple of AMS PMF users who were also interested in EPA PMF, and would be useful if one is doing an analysis that includes pulling, examining FPeak runs, etc., so you don’t have to do it all in one session. This however is a huge step for EPA PMF.
• Conduct further research regarding DISP functions, which may include, but not limited to, the following: 1) exploring the methodology of DISP and identifying key DISP parameters that users may change through the model interface; 2) analyzing how DISP method may change the Q values and how to specify appropriate dQmax values for base model runs and constrained model runs; and 3) conducting DISP analyses based on real-world data sets such as the G.T. Craig PM2.5 and Baton Rouge VOC data, and comparing results from the DISP and bootstrapping methods.
• Modify the user interface to allow multiple F-peak runs and provide graphs to compare multiple F-peak runs with the base run (e.g., generate graphs showing how Q values change against various F-peak values that the user has specified).
• Develop functions to allow a batch run that includes different number of factors; generate comparison graphs for the user to summarize how key variables (e.g., residues, Q/Qexpecte) may change against the number of factors.
• Implement additional graphic diagnostic features regarding Q/Qexpected and residuals, and also explore results with different datasets so we can give some interpretation of results in user’s guide
-Q/Qexpected overall
-Q/Qexpected and total residual by sample (time series graphs)
-Q/Qexpected and total residual by species (profile type graphs)
-allow dQmax in pulls to be either a value or a % of original Q
• Incorporate “selective factor” switch somehow – via time series graph? Also figure out what it’s actually doing first by exploring it with a dataset.
Minor proposed work for PMF:
• For the xlsx file – allow the installation of PMF package without updating the “Access” software.
• Config file – allow the user to access the config file through the model interface; incorporate notes into the config file or in other formats, so the user can add notes for each major step or module.
• Base model run – allow the user to control seeds that will be used in ME2.
• Constrained model – add a Pie Chart in the constrained model results.
• Deal with the wrapping of the ME-2 output if N species is > ~220.