Probabilistic Forecasting Exploratory Information

Forecast Process Job Sheet

This Job Sheet steps you through one possible end-to-end forecast scenario using the prototype software and canned data set. This will give you a basic understanding of the process and act as a launching point for exploring all the various capabilities in the prototype software. More in-depth reference material and cheat sheets for all capabilities are accessible in the Forecast Process pull-down menus.

Visualize and Diagnose Ensemble Data


View Ensembles from ALPS

Legend Overlay and Management

Getting Started
  • Start D2D, if not already started and open the Volume->Browser....
  • From the VB, select the following: Sources Ensemble->SREF-21, Fields Basic->Height, and Planes Pres->500MB. Highlight your selection and click Load.
  • Expand the Legend, if not already expanded, by clicking with Mouse Button-1 (MB1) on a plus (+) sign on the left side, note the organization and color scheme (see below, click on image for full resolution). The key point here is that the colors for individual member overlays are now controlled by the segregation strategy, not the standard AWIPS overlay color set.
Using the Collapse Controls
  • Click on a minus (-) sign on the left side of the Legend with Mouse Button-2 (MB2), note that this collapses a group of similarily colored overlays (see below, click on image for full resolution). To re-expand the collapsed group, click MB2 on the + sign on the left side of the collapsed group.
  • Click on a - sign on the left side of the Legend with MB1 to collapse the Legend.
  • Click on a + sign on the left side of the Legend with MB2, note this shows all groups but leaves the groups collapsed (see below, click on image for full resolution).

    Now click on a + sign on the left side of the Legend with MB2, note this expands that group (see below, click on image for full resolution).
  • Expand the Legend again by clicking on a + sign with MB1 on the left side of the Legend.
Toggling
  • Click twice on any item in the Legend (not on the + or - signs) with Mouse Button-3 (MB3), note that the Legend toggles off and on.
  • Click twice on any item in the Legend with MB2, note that the group containing the item toggles off and on.
  • Click twice on any item in the Legend with MB1, note that the item toggles off and on. Expand the Legend using MB1, if not already expanded.
Hiding Empty Overlay Legends
  • Toggle off an item in the Legend using MB1.
  • Click MB3 over another item in the Legend and change the Density to zero.
  • Click MB3 over the main display, away from the Legend, and select Hide Empty Legends.
    • Note: The toggled off and zero density legends disappear, and the small gray more appears on top of the Legend (see below, click on image for full resolution). This capability is there to de-clutter the Legend space by putting aside the non-viewable or unwanted overlays.
  • Click MB1 on the more text at the top right side of the Legend, note the empty overlay items come back and is now part of the "states of rotation" when using Enter from the Keypad.
  • Clear the screen.

Function Overlays

Basic Concepts
Assumption: D2D is running.
  • From the Volume Browser, reload SREF-21 500MB Height. Use MB1 to expand the Legend.
  • Select Volume->Function Overlays from the D2D main toolbar. Tear away the Function Overlays menu and keep to the side.
  • Click on the D2D title bar to get keyboard focus, hit the "<" key on the KeyBoard once to get zero density. From the Function Overlays menu, select Mean (see below, click on image for full resolution).
  • Expand the Legend, if not already expanded, and toggle off a couple of groups using MB2 (if necessary refer to the Toggling section above). Collapse the Legend.
  • Select the Mean Function Overlay again from the Function Overlays menu, note the result is slightly different because the toggled off members are excluded from the Mean computation (see below, click on image for full resolution).
Histogram Overlays
  • Clear the screen and open the Volume Browser.
  • From the VB, select the following: Sources Ensemble->SREF-21, Fields Sfc/2D->Precipitation, Planes Misc->Surface (see below, click on image for full resolution).
  • Load the Color Histogram from Volume->Function Overlays. Press and hold MB1 over an area of plentiful precipitation data and sample. The sampling displays a colored histogram and is useful for identifying outliers. For continuous sampling, click MB3 over the display, away from the Legend, and select Sampling (see below, click on image for full resolution).
    • Note: If you expand the Legend, if not already expanded, and position the cursor over an area with plentiful data with Sampling enabled, you then can hit the plus key on the keyboard several times to toggle through the rotation of the individual members. This enables the identification of outliers in a timely manner.
  • Clear the screen.
  • Exit D2D.

View Ensembles from the GFE

Logging In
  • Log in to the GFE as GFETEST with TestConfig as the config file.
Loading Ensemble Members
  • Choose GFE-->Weather Element-->Weather Element Groups --> SREFTemps. The temperature ensemble members will appear in the Grid Manager (see below, click on image for full resolution).
View Ensemble Statistics
  • Select a "t" grid in the Grid Manager (MB1 over the grid)
  • Choose ProbTools-->CalcEnsembleStats. In the dialog that pops up:
    • Choose "t"
    • Choose "Mean" and "StdDev"
    • Click Run/Dismiss
  • Grids for the Mean and StdDev will be created in the Grid Manager. Click on each to view them in the Spatial Editor. You will want to change the color table for each: Click MB3 over the color bar and choose "Fit to Data."
  • Use the Sample Tool to view the values (see below, click on images for full resolution).
        
View Ensemble Relative Frequencies in a Histogram
  • Draw an Edit Area in the Spatial Editor: Click the "Draw Edit Area Tool" and draw area.
  • Choose ProbTools-->EditHistogram. In the dialog that pops up:
    • Choose "t"
    • Click Run/Dismiss
  • A Histogram appears depicting the ensemble relative frequency values over the selected Edit Area. Colors indicate the contribution of each ensemble member to each histogram bin. Change the Edit Area on the GFE and re-run the EditHistogram tool. Note that the selected area can make a big difference in the resulting histogram (see below, click on image for full resolution).
View Ensemble Relative Frequencies in a Table
  • Select a time range by clicking on one of the ensemble grids in the Grid Manager
  • Draw an Edit Area in the Spatial Editor (preferably use the same Edit Area for which you generated the Histogram in the previous step): Click the "Draw Edit Area Tool" and draw area.
  • Bring up the Formatter Launcher: Products --> Formatter Launcher
  • In the Formatter Launcher, choose Products --> EnsembleTable
  • Click the "Run Formatter" button
  • When the dialog comes up,
    • you can choose to produce a table for t, tp, and/or wind by checking the buttons,
    • select "Draw" to use the edit area you have drawn,
    • you can select "Test Mode" to use just a few ensembles to speed performance.
    • Click "RUN".
  • Tables will be generated for your edit area showing the ensemble relative frequencies for the chosen elements. For temperature and wind speed, each entry represents the percentage of the ensemble members that selected the given value. For precipitation, each entry represents the percentage of the ensemble members that selected the given value or greater.
  • If you are running the GFE as a WFO (rather than on the US domain), you have the option of choosing areas in the Zone Combiner. There will be a row for each area selected (regardless of combinations) in the resulting table (see below, click on image for full resolution).

Weight the Ensemble Members

After viewing and diagnosing the ensemble data, you may want to weight the ensembles, for example, to de-emphasize outliers. The weights database is automatically started up with the ifpServer.

Edit and publish weights in Simple Editor

  • Try editing the weights from a rudimentary editor. From the bin directory (same directory as runIFPServer and runGFE) type:
    • editWeights
  • A window will appear with the weights and slider bars. Adjust the sliders (if desired). Press "Publish" to set the Weights in the database server. When you publish your model weights other applications can read them and modify the display based on the values of the weights. When you're all done, press "Exit".
To set the weights from the GFE:
Logging In
  • Log into the GFE as GFETEST with TestConfig as the config file.
Running EditHistogram
  • Choose ProbTools-->EditHistogram. In the dialog that pops up:
    • Choose "t"
    • Click Run/Dismiss
Examine Histogram
  • The histogram displayed shows the distribution of temperature over the selected area at the selected time. For temperature the tool displays the data in one degree Fahrenheit intervals or bins. Intervals for wind are one knot and for QPF (tp) they are 0.01inches. The colors reveal the contribution of each model to that bin. The models have been grouped for easier management and were chosen under the assumption that ensemble members with similar names will also perform similarly. Model groupings can be configured differently. The vertical axis "COUNT" indicates the total number of grid points whose values fall within the corresponding bin displayed on the x-axis. For example, if the particular bin shows a count of 1000, all of the models in that bin contributed that total number of grid points for the bin value.
Adjust the Weights
  • The key located in the upper left corner of the window shows the colors assigned to various groups. Individual models within its group are each assigned a different shade. Clicking on any color-coded group label with Mouse Button-1 (MB1) will toggle off or on all the models in that group. This action causes the histogram to be recalculated without the contribution of the selected group. MB1 clicking on any of the colored rectangles will toggle off or on that particular model and remove it from the histogram as well.
  • Models can also be assigned individual weights. To adjust the weight of a model, press and hold MB1 within its corresponding colored rectangle and move the mouse up or down to adjust the weight. As you drag the mouse up and down, the currently assigned weight for that model is displayed nearby.
Publish the weights
  • The weights defined in the interface can be published to the Weights Database so that other applications such as ALPS can incorporate the weights when displaying ensemble data. Select the "Publish/Dismiss" button to publish your weights.

Produce Deterministic Grids

Produce deterministic forecast grids for T, Wind, and QPF for the Today and Tonight period.
  • Log in to the GFE as any user.
  • Copy all grids from a model of your choice (e.g. NAM12) as a starting point.
  • If desired, edit those grids to your liking.

Produce Probabilistic Grids

The deterministic forecast grids will serve as the 50th percentile to generate the 10 and 90 grids. To do so:
  • Log in to the GFE as GFETEST with TestConfig as the config file.
Run GenerateProbGrids
  • Choose ProbTools-->GenerateProbGrids. In the dialog that pops up:
    • Choose "T", "Wind"
    • Click "Run/Dismiss"
  • Grids will be created for T10Pct, T90Pct, Wind10Pct, and Wind90Pct. Note: the generation of QPF 10 and 90 percentile grids has been disabled for now until we develop a high quality algorithm for generating 10 and 90 QPF grids. The algorithm used for T and Wind is relatively simple and works reasonably well. Because QPF grids can be mostly zeros, the current algorithm works poorly.

Generate Products

From your feedback, we can proceed to develop experimental products from the 10-50-90 grids.
  • Log in to the GFE as GFETEST with TestConfig as the config file.
View Prototype Probabilistic Phrases
  • To get an idea of some possible probabilistic phrasing, run the experimental ZFP-style Probabilistic Phrases that are generated directly from the ensemble members (weighted or unweighted).
  • Load ensemble members
    • Choose Weather Element-->Weather Element Groups --> SREFTempsPrecip. The temperature ensemble members will appear in the Grid Manager.
  • Bring up the Formatter Launcher
    • Products --> Formatter Launcher
  • In the Formatter Launcher, choose Products --> ProbabilityPhrases
  • Click the "Run Formatter" button and choose options on the dialog:
    • Choose t, tp, wind
    • Choose Today
    • Choose Weighted or non-weighted ensembles
    • Choose "Test Mode" if you want to use just a few ensembles (It will run faster this way)
    • Click "OK".
  • Phrases describing temperature, wind, and precipitation will appear. This is a very crude beginning, but we hope it will jog your thought process! (see below, click on image for full resolution)
View Prototype Probabilistic Graphic
  • Click here to view a possible graphic representation of probabilistic information. We are working on generating this graphic automatically from 10-50-90 grids. Let us know what you think!