Ensemble Forecast Processor

The Ensemble Forecast Processor (EFP) is a plugin designed to compute statistical metrics from ensemble flow time series at specified watershed locations and use those metrics to inform other model alternatives within a forecast. For instance, an HEC-ResSim alternative can incorporate a metric from the EFP—such as a percentile-based flow volume—into a reservoir rule to guide release decisions. The EFP supports several statistical computations over a defined time window, including minimum, maximum, average, and percentile volumes.

This plugin is highly flexible and can be inserted at various points within a forecast's program order, enabling a wide range of applications. One common use case is Forecast-Informed Reservoir Operations (FIRO). In this scenario, forecasted inflows from the ensemble are processed by the EFP to generate a metric—such as the 75th percentile three-day forecasted volume—which can then influence decision rules in the HEC-ResSim alternative to guide reservoir operations.

Before running a forecast, you must:

1. Create a new program order that includes the EFP.
2. Develop an EFP alternative.
3. Assign it a model alternative key.
4. Set up a new ensemble forecast run.

The sections below outline the input and output settings that can be configured when setting up a new EFP alternative.

Input Settings

The Input Settings for an EFP alternative are organized into two tabs: LookBack and Forecast. These tabs allow you to specify locations and enter time series data used by the EFP during its computations. Data from both tabs can be linked within the Model Linking Editor.

LookBack Tab

Use the LookBack Tab to add locations that correspond to time series data within the Lookback Period of the Forecast Time Window. You only need to enter data in this table if:

• You have selected the Ensemble Member with Closest Volume to Metric Compute option, and
• You want to merge observed time series data from the Lookback period with the ensemble member selected for the Forecast period.

For more details, refer to the documentation on Ensemble Member that Contains the Volume Closest to Each Metric.

Forecast Tab

The Forecast Tab is required for any EFP compute. You can use it to define ensemble collection pathnames used in calculating statistical metrics. Each entry must point to an ensemble collection in the HEC-DSS file, whether from an extract list or generated by a previous program in the forecast sequence.

Output Settings

The Output Settings tab defines the outputs generated by the EFP.

Locations to Compute Metrics Tab

This tab lets you define where ensemble metrics should be calculated and what input data to use during the Lookback and Forecast periods. Add a row for each location where metrics should be computed, as specified in the Cumulative Volume Scalar Decision Metrics tab.

Cumulative Volume Scalar Decision Metrics Tab

This tab allows you to define the specific ensemble metrics to calculate for each forecast. Multiple metrics can be computed in a single alternative. 

In the example below a total of 14 metrics are calculated at each computation location. The Cumulative Volume for 1 day and 5 days will be computed, along with the following metric for each duration:

• The minimum value
• The maximum value
• The 20%, 50%, 75%, and 99% percentiles
• The average value

Press Enter to save your entries. Use comma-separated days for Cumulative Volume and decimals for Percentiles.

Store Output as DSS Collection

This optional step, enabled via the checkbox on the Cumulative Volume Scalar Decision Metrics tab, saves your selected output metrics as a single DSS collection for each location.

Example Scenario

You have selected three metrics for the Cloverdale location: 5-Day Cumulative Volume Max, Min, and 90%.

• Standard Output Format (Checkbox Unchecked):
  
• DSS Collection Output Format (Checkbox Checked):
  

To see which collection member is associated with which metric, check the lookup table provided in the Notes section under Location Info/Supplemental Information for each location.

Ensemble Member Compute

Ensemble Member with the Closest Volume to Metric Compute

This optional compute step, enabled via the checkbox on the Ensemble Member Compute tab, finds the ensemble member whose cumulative volume is closest to each computed metric.

For each compute location and metric:

• The EFP identifies the ensemble member with the closest cumulative volume to the metric. Only the cumulative volume during the N-day period is considered.
• If the Lookback record is different from the Forecast record, the returned ensemble member is merged with the Lookback record.
• If the Lookback record is null or the same as the Forecast record, the entire ensemble time series from Lookback Start to Forecast End is returned.

Controlling Output Location

The Controlling Location ensures that the exact same ensemble member is used for the Ensemble Member with Closest Volume to Metric Compute across all locations throughout the watershed (per metric). 

Example Scenario

If Hopland is selected as the Controlling Location in the watershed, the specific ensemble member identified for Hopland will automatically be applied to all other selected compute locations in the watershed, ensuring spatial consistency.

The Store Output as DSS Collection checkbox impacts how this data is saved.

• Standard Output Format (Checkbox Unchecked):
  
• DSS Collection Output Format (Checkbox Checked):
  

The Supplemental Information for each location will store the specific ensemble member that was returned during the compute and merged with the lookback record.

Downstream Integration

This saved DSS collection record can be used directly as an input for model applications that support ensemble computes, such as HEC-ResSim.

Additional Resources

For step-by-step instructions on how to set up a new EFP alternative, refer to this guide.