New Ensemble Modeling Capability to Support Reservoir Operations

By: Simon Evans, P.E., C.F.M. and Matt Fleming, P.E., Aff.M.ASCE

Introduction

Typically, flood forecasts are created using a single “best” estimate of current and future conditions. However, estimation of flood forecasts are inherently uncertain, due to both variability in natural processes and our imperfect ability to represent those processes in our analyses. The use of ensemble forecasts is one way to integrate and communicate the uncertainty of hydrologic forecasts and improve reservoir operations. An ensemble forecast is a set of forecasts, made up of ensemble members, that include a range of possible outcomes. Uncertainty in initial states for numerical weather and hydrology models, uncertainty in the model structure and algorithms used to represent physical processes, and uncertainty in process-based parameters can be included when generating an ensemble forecast dataset. Figure 1 shows an ensemble flow forecast issued by the California Nevada River Forecast Center (CNRFC) on March 22, 2023. 

An initial ensemble forecasting capability was added to the Corps Water Management System (CWMS), and the publicly available Real Time Simulation (HEC-RTS) software. This initial capability relies on existing flow ensembles which would be applied directly in HEC's Reservoir System Simulation (HEC-ResSim) software. Research into Forecast Informed Reservoir Operations (FIRO) techniques has renewed interest and investment in ensemble forecasts and their application to USACE reservoir operations. FIRO is a reservoir-operations strategy that better informs decisions to retain or release water by integrating additional flexibility in operation policies and rules with enhanced monitoring and improved weather and water forecasts (American Meteorological Society, 2020). One example application of FIRO might be to allow a reservoir to fill above the normal operating zone when no future precipitation is forecasted and evacuate excess water when large, impactful storms are forecasted. This type of operation has shown promise in Western United States watersheds, for improving both flood risk and water supply conditions. 

Additional ensemble modeling capability has been added across the HEC software suite. The new ensemble modeling capabilities provide forecasters and water managers different ways to create meteorologic and flow ensembles and apply them to operate reservoirs.

Cumulus

Cumulus is a USACE web service that provides gridded meteorological and hydrologic data across the United States. Currently, the web service is only accessible to USACE staff and is used primarily by USACE water management staff for creating flood forecasts, as it is tightly integrated with CWMS. Historical datasets within Cumulus can also be used for model calibration in support of other USACE engineering analyses. The Cumulus service clips national gridded products to watersheds and formats the data for application in HEC-MetVue, HEC-HMS, and HEC-RAS. You can access Cumulus from: https://cumulus.corps.cloud/. 

Some of the currently available gridded precipitation and temperature datasets are included in Table 1. Over time, additional deterministic forecast datasets will be added to Cumulus, as well as ensemble datasets. The available products can be explored now to create a limited ensemble forecast in CWMS.

HEC-HMS

The HEC-HMS ensemble compute option was released in Version 4.10. Modelers can add existing simulation runs or forecast alternatives to the ensemble compute while controlling the simulation time window, time step, and level of output. Through the ensemble compute, modelers are able to explore different model configurations (delineation and process methods), as well as parameter sets, initial conditions, and boundary conditions. Example applications in HEC-HMS include:

• Applying different future meteorologic scenarios to the same basin model; i.e., short term forecasts or long term climate predictions.
• Applying the same meteorologic forcing data and varying methods and parameters; for example, exploring post-fire debris predictions by varying the HEC-HMS debris flow model and parameters across multiple basin models.
• Exploring the runoff response by varying the initial conditions defined in the basin model.
• Exploring the runoff response by configuring separate basin models with reasonable parameter sets (similar process models could be used or not).

Currently, the HEC-HMS and CWMS development teams are integrating the HEC-HMS ensemble compute within CWMS. 

HEC-ResSim

HEC-ResSim also provides an ensemble compute option that is currently available through the ensemble alternative type. This alternative type was created to facilitate reservoir operations modeling using ensemble forecasted inflows. A standard HEC-ResSim alternative represents only one correlated set of inflow hydrographs for the basin and produces one set of outflow hydrographs. However, an ensemble alternative represents multiple correlated sets of inflow hydrographs and produces the associated multiple correlated outflow hydrographs. Ensemble inflows and the associated ensemble results utilize HEC-DSS time-series collections.

An ensemble alternative is the same as a standard HEC-ResSim alternative except that instead of representing a single run and computing through a single set of inflow time-series as a standard alternative does, an ensemble alternative represents several runs, one for each selected member of the ensemble inflow datasets.  

CWMS Ensemble Forecast Processor and Possible Use Cases

The Ensemble Forecast Processor is a new plugin available in the upcoming CWMS Version 3.4 software release. The Ensemble Forecast Processor (EFP) allows the user to process various metrics from ensemble flows at specified locations in the watershed, and then use those metrics to influence other model alternatives in the forecast. The EFP plugin is flexible and can be placed throughout the program order of the CAVI forecast to allow for a variety of uses. One possible use case of the EFP is in support of Forecast Informed Reservoir Operations (FIRO). For this example application, ensemble forecasted inflows can be processed by the EFP plugin, and the resulting metrics (i.e., the 75th percentile three-day forecasted volume) can be used to influence the decision rules in the HEC-ResSim alternative to determine potential reservoir operations.

Looking Forward

Future development of ensemble modeling within CWMS will allow for a complete ensemble forecast at each step of the modeling process. A potential workflow would have the modeler select the desired ensemble forecast product (such as the Global Ensemble Forecast System (GEFS) ensemble precipitation forecast) from Cumulus, process the ensemble gridded data within HEC-MetVue, and then use the ensemble gridded data as input to an ensemble HEC-HMS compute. The modeler could then run the Ensemble Forecast Processor (if interested in metrics from the ensemble forecasted inflows computed from HEC-HMS), or run an ensemble HEC-ResSim compute using all the ensemble members. To achieve this potential workflow, additional development beyond what has already been completed is required and includes the addition of ensemble datasets to Cumulus, improvements to HEC-MetVue to process and compute gridded ensemble datasets, and build out of an ensemble compute workflow in CWMS.