Not logged in - Login
< back

IWDW Analysis Products

The objective of this page is to provide guidance on the types of data and analysis products that need to be available from the 3SDW. Past discussions on analysis products used the concept of the five most relevant plots for air quality modeling analysis:

  • scatter plots - can compare model-obs and model-model by plotting concentrations paired in space and time along each axis; statistics summarizing the model performance are also nice additions to these plots. Configuration options for this plot should include the species, the monitoring network(s), the model simulation(s), the time period, the spatial subset of monitors (i.e. all monitors in the domain or CO monitors only), and a cutoff concentration (i.e. only values that are above a certain concentration) to display in the scatter and the metrics to include in the table of performance stats
  • time series plots - model-obs and sometimes multiple model-obs results displaying concentrations paired in space and time with concentrations on the y-axis and time on the x-axis; most useful for single site analysis. Configuration options for this plot should include the species, the monitoring network, spatial subset (state ID), the model simulation(s), the time period, and the monitor to display; populate a drop down list of monitors based on the selected network and spatial subset
  • spatial stats - spatial display of statistics of model performance at monitor locations; colors indicate the value of the statistic. Configuration options for this plot should include the species, the monitoring network(s), the model simulation, the time period, and the statistical metric to display. Relevant metrics are correlation coefficient, MB, ME, NMB, NME, FB, FE, RMSE, index of agreement, R^2
  • tile plots - spatial plot of the model results; often displayed as daily/monthly/annual maximums, 4th highest daily maximum 8-hour average O3, or another relevant metric; Configuration options for this plot should include the species, the model simulation, the time period; may also have options for creating a difference plot between two simulation and observation overlays
  • soccer plots - summary of statistics that plots model error vs. bias with lines representing performance goals superimposed on the plot. Configuration options for this plot should include the species, the monitoring network, spatial subset (state ID), and the model simulation

Task 1: Time Series Plots (11Nov2014)

The first plot to work on will be the time series plots to display the model vs. obs results at specific monitors in the modeling domain. A database export from the UNC AMET instance for this project that includes paired data for all of the FRM networks is available on Viking:

/mnt/data2/3SDW/3SAQS/2011/Base11a/AQ/CAMx/AMET/CAMx_3SAQS04_B11a

Use these data to prototype the analysis database and time series display tool. After getting the time series display to work we will move on to creating other display products. This dump includes both model and obs data, however you may choose to just use the model results as the 3SDW already include a database of monitoring data. Included here are also lat lon coordinates and monitor IDs, which should make cross-referencing with the existing obs database fairly easy

Note that a slick display for creating the plots will be a map of monitoring locations that a user can click on to select which monitor they would like to create a time series for. This would then kick you into a configuration window to choose the simulation, species, and time period to plot

Task 2: Visualization tool development (Feb 2015)

Links to two dynamic tools for displaying model-to-obs datasets, and emissions data are shown below. These tool provide the bulk of the time series functionality laid out in Task 1, above. Please use this wiki to enter comments about these tools, and to suggest how they might be modified to produce the plot types listed above.

  • model-to-obs tool
  • emissions review tool

Comments (when adding comments, please enter the date you entered the comment and the tool, if the comment applies to a specific tool)

Feb. 3 data ops:

  • In table view below graph, add a column that shows the number of data points used to calculate average if n>1.
  • Model-to-obs: allow for multiple site selection and dynamic averaging.
  • It would be slick to add a map selection feature so users could select groups of sites.
  • Provide other site selection filters, such as high elevation in CO, proximity to C1A, etc.
  • add scatter chart type (Pending upgrade to latest ChartFX version and Analysis add-on)

Feb. 3 (Tom) - general comments:

  • 1) units on the graph
  • 2) help link to DW metadata and reports in the selection box area
  • 3) improved labeling on the graph (legend with descriptive name of each source shown [Source Category name from selection doesn’t carry forward so it appears in legend], descriptive title on X and Y axes, colors, et cetera).

Feb 3 (Zac) - emissions browser:

  • add source categories for biogeneic, WBD, fires
  • (Done) Capitalize the source category names: e.g. “air, locomotive, marine" should be "Aircraft, Locomotive, Marine” (Done)
  • roll up chemical species modes to base species (e.g. “CO in Exhaust” should be CO or Carbon Monoxide and thrown in the bin with all of the other sources of CO.)
  • Removeremove the Non-SESARM prefixes in the descriptions of the sectors
  • We also talked about drilling down to the SCC level in these charts. So when you click in a county you would get a list of the emissions by SCC in that county. You could only get county level SCC summaries for one county.
  • Tabulated data below the chart. Show the data displayed in the chart as a table

Feb 3 (Tom/Zac) re. non-SESARM prefixes

  • Tom - I understand and generally agree with Zac’s comments. I don’t understand the comment about “…Non-SESARM prefixes…”. I think it is likely that we will need to be able to show that different groups originate emissions estimates and it may make sense to bring that info along to the selection tool and onto the graph and associated data table.
  • Zac - Tom, the non-SESARM tag is an artifact of the SEMAP project. SESARM developed their own inventories for their states for that project, so we needed to distinguish between the SESARM and the non-SESARM inventories.

For the 3SAQS project all of the inventories with the exception of the O&G basins in the three-states came from EPA. We do distinguish between the 3SAQS O&G and non-3SAQS O&G data, where we added the descriptor “(Across the USA)” for the non-3SAQS inventories. I think we’ll need to change that a bit, but we do have the capacity to identify that these are two distinct sources of O&G data. For the rest of the sectors, which all came from EPA, we may be able to add an ID to the sector name like “(NEI2011v1)” if you think that it would add value to these plots.

  • Tom - Distinction also applies to fire, biogenics, mobile is not the same either.

  • Zac - You’re right, and we also have these documented in the modeling protocols and elsewhere. How much metadata about these emissions results do you want to display on the tool? My inclination is to keep the tool simple and then defer to the documentation of the runs if people want details about what “Biogenic” means in terms of simulation 3SAQS Base08b. If you would like for this tool to display more detail about what went into the different sectors, let me know and I can try to provide concise inventory IDs to display along with the sector names.