Ensemble Forecasts

Forecast Runs

An Ensemble Forecast Run can be added through the Forecast Run Editor. The Forecast Run Editor supports creating, editing, and deleting of an Ensemble Forecast Run type.

Ensemble forecast run names are displayed in italic text throughout the HEC-RTS interface. This visual distinction appears in the Setup tab’s watershed tree, the Modeling tab’s forecast run tree, the Model Linking Editor, and any generated reports or plots, helping differentiate ensemble runs from deterministic ones. 

Model Linking and Extracts

In the Extract Editor, any input for a model alternative marked as coming from the Extract List in the Model Linking Editor appears as a Required Input. Each input is labeled with the specific model alternative and forecast run that requires it. If the same alternative is used in multiple forecast runs—all referencing the Extract List—the same pathname will appear for each run. When the Hide Duplicate Pathnames option is enabled, these repeated pathnames are shown as a single entry. This setup means that a given input (HEC-DSS pathname) can only have one source. However, since a model alternative can be part of both standard and ensemble forecast runs, the system includes a way to distinguish inputs for each run type. For ensemble forecast runs, you can specify whether the input is a single time series or an ensemble dataset.

To support this separation, Required Input pathnames must be distinguishable by forecast run type. Inputs used in ensemble forecast runs are shown in italic text, allowing them to be treated as unique (not hidden as duplicates). This distinction enables you to assign either an ensemble or single time series dataset to the same pathname, depending on the forecast run type.

Model Linking Editor

The figure below displays the Model Linking Editor for the ResSim-Ensemble Forecast Run.

Extract Editor

All input sources are set to Extract List, which means their corresponding pathnames will appear in the Required Input section of the Extract Editor for the HEC-ResSim model alternative. Since this is an ensemble forecast run, the pathnames (see figure below) are shown in italic to indicate that they are associated with an ensemble input.

Forecasts

Once you’ve set up an ensemble forecast run, configured the new extract lists, and verified model linking, you’re ready to create the ensemble forecast. Navigate to the Modeling tab and create the new forecast using the ensemble Forecast Run.

File Management

HEC-RTS ensemble computations place each iteration of the compute member in a separate folder within the directory structure of the current forecast. For ensemble forecasts, a new folder named EnsembleRuns is created. Inside this folder, subfolders are labeled with numbers, representing the forecast watershed copies for each selected ensemble member that is computed (in the example below, the numbers are 1, 3, and 5-8). As the compute progresses through each ensemble member, the File Manager creates these ensemble watershed directories. If the directories already exist, the File Manager only copies the edited or missing files from the forecast watershed to the ensemble watersheds.

Each ensemble member folder contains its own forecast.dss file. Within this file, each model alternative of a single ensemble trace reads its input and writes its output. The File Manager creates this file and transfers the ensemble member’s HEC-DSS input into it from the parent forecast.dss file. During this process, the collection ID is removed from the HEC-DSS pathnames stored in the ensemble member’s forecast.dss file.

The original (parent) forecast.dss file is used to collect output from the ensemble members, utilizing HEC-DSS collections to store individual traces in ensemble datasets. After completing the computation of all selected ensemble traces, the File Manager copies the output from each ensemble member’s forecast.dss file into the parent forecast.dss file, prepending a collection ID string that includes the ensemble member’s ID to the F Part of each HEC-DSS pathname.

The other outputs for the ensemble members remain in their respective subfolders, and the File Manager will retrieve them upon request.

Any edits to the model alternatives of an ensemble forecast should be made to the alternative in the original (parent) forecast watershed. This ensures that all ensemble members compute from the same version of the edited data. Edited files are copied to the ensemble directories each time a compute is launched.

Compute Log

A compute log is generated for each ensemble member during an ensemble forecast run, as well as an additional log for the Compute Manager itself. The compute log viewer for an ensemble forecast provides access to the logs for each ensemble member that was computed, along with the Compute Manager’s log.

You can access the log from the Reports tab.

Results

After the forecast compute is complete, you can choose the desired member from the Select Active Ensemble Member dialog to view results in the map window.

Ensemble Plots

Unlike standard plots, ensemble plots display results from all computed ensemble members simultaneously. Because ensemble alternatives may include dozens of members, viewing them all on a single plot can quickly become overwhelming and make interpretation difficult.

To improve clarity and support better analysis, all ensemble members (traces) are shown in the same color, using thinner lines with moderate transparency. This creates a visual effect where overlapping traces appear darker, highlighting areas of convergence among the results.

In addition, five summary curves are automatically generated for each parameter: minimum, maximum, median, 25th percentile, and 75th percentile. These curves are displayed with thicker, fully opaque lines to clearly stand out from the individual traces and provide statistical context at each time step.

To view these ensemble collection plots, you must use HEC-DSS collection records displayed in the Condensed-Group Collection format within CWMS-Vue. Please note: these plots are not accessible from time series icons in HEC-RTS and must be viewed directly in CWMS-Vue.

An example collection plot is shown in the figure below.