Gridded precipitation post-processing for Ensemble Streamflow Forecasts to improve the prediction of anomalously high precipitation and inflow
Research Team Members
Objective:
This project aims to develop an enhanced ensemble forecast system in HEFS that can be potentially implemented in the NextGen framework by developing an improved gridded ensemble postprocessor to reduce spatial displacement error and creating enhanced MEFP FEWS adapter for easy adaptation for RFC operation.
Approach:
The project team will develop an enhanced MEFP Flood Early Warning System (FEWS) adapter with the enhanced precipitation grid postprocessor and implement it to an ongoing CIROH HEFS project. Tasks are: (1) develop an improved gridded ensemble postprocessor, (2) generate enhances ensemble streamflow forecast for four Texas FIRO pilot reservoirs in the central Texas, (3) implement TEEHR for hindcast evaluation, and (4) develop an enhanced MEFP FEWS adapter for RFC operation.
Impact:
Improved uncertainty representation in precipitation ensemble will better capture possible occurrences of extreme events and it will benefit forecast informed reservoir operation. This will directly benefit RFCs and Army Corp of Engineers for better reservoir operation.Abstract:
The existing precipitation postprocessor for the Hydrologic Ensemble Forecast System (HEFS) has been identified to have limitations in capturing extreme storm events when these events do not occur directly over the watershed of interest. This results in a tendency to underrepresent the probability of extreme inflows to reservoirs. The project team aims to develop an enhanced MEFP Flood Early Warning System (FEWS) adapter that incorporates an improved precipitation grid postprocessor and facilitate forecast evaluation through the TEEHR system currently being developed by another CIROH project.
Development and testing will primarily focus on the Brazos River basin but will be sufficiently generalized to allow configuration for other locations. Gridded precipitation pre-processing will be designed to produce inputs in a TEEHR-compliant format (i.e., Parquet), providing a platform for testing and comparing standard MEFP processed forcings and improved gridded forcings. Standard verification metrics for horizons up to 7 days will be utilized, with an emphasis on comparing streamflow forecast skill for higher QPF events. The results will be directly coordinated with WGRFC and USACE to assess the magnitude of forecast improvement and its impact on their operations and decision-making processes.
The current status of the project is explained below:
- The project team has evaluated three precipitation postprocessors: Mixed Meta-Gaussian Distribution (MMGD), Conditional Bias Penalizing Regression (CBPR), and Censored-Shifted Gamma Distribution (CSGD). Preliminary results indicate CSGD outperforms the other methods at test locations.
- The team will continue model evaluation to finalize the method for implementation. They have also started setting up a standalone WGRFC ensemble forecasting for FEWS adapter development and hindcast generation, showing significant progress towards developing an enhanced precipitation postprocessor.