Developing process-based and spatially consistent approaches for correcting streamflow biases in watershed hydrology simulations
The goal of the proposed research is to develop a new streamflow bias-correction method that can be used by Reclamation to create input time series for water resources models at a daily time step. The proposed method will avoid existing artifacts in the bias-corrected streamflow, such as discontinuities on month boundaries and inconsistent local inflows. The methodological advances are that the bias-correction method will develop corrections based on the dominant hydrologic process (e.g. snow melt) rather than on time-of-year (e.g. a correction based on month) and that the method will account for the connectivity between successive downstream locations to result in realistic incremental flows.
Need and Benefit
In an ideal world, our models would be perfect and streamflow bias-correction would not be necessary. Unfortunately, our model simulations are not perfect, due to model structural errors, parameter uncertainty, errors in forcing data and errors in the naturalized streamflow that we use to calibrate and evaluate our models. Streamflow bias-correction is in most cases a necessary step before modeled streamflow time series can be used in water resources and flow regulation studies.
Reclamation currently has access to streamflow bias-correction procedures that operate at a monthly time step. Modeled daily flows are then scaled to match the bias-corrected monthly totals. However, this quantile-mapping based method can introduce severe flow discontinuities on month boundaries and can radically change the shape of the modeled hydrographs for future conditions. The UW has developed another bias-correction method that does not suffer the same problem, but which does not exactly reproduce the monthly hydrograph for the reference period. Because of these shortcomings, neither method fully addresses the needs of Reclamation.
In addition, neither method accounts for the connectivity imposed by the stream network, that is, all sites are corrected independently. This can result in abrupt changes in local inflows, which are simply an artifact of the bias-correction method, especially at shorter time steps (e.g. daily rather than monthly).
Erik Pytlak (Manager, Weather and Streamflow Forecasting, and Climate Change Technical Lead) from BPA noted in a letter of support for the proposed work (based on experience with the RMJOC-II project) that: "Streamflow post-processing (bias correction) is one of the areas which has proved particularly challenging. […] the use of traditional month-averaged bias correction techniques proved to be highly problematic. […] additional work is needed to extend this approach to other parts of the hydrologic modeling chain, including the derivation of useful headwater and incremental flows".
Simulated streamflow using existing methods of bias-correction often lead to modeled river operations that are not believable by stakeholders, potentially leaving them with a lack of confidence in Reclamation modeling studies.
The benefits of this research are: improved bias-correction methods will lead to better projections of future hydrologic conditions and will improve planning and operations for the coming decades; improved bias-correction methods will lead to study results that stakeholders will have confidence in to inform their river operations, thereby increasing the likelihood of improved water conservation and improved reliability of water supplies.
Existing bias-correction methods do not perform well enough for Reclamation to have much confidence in their application in studies evaluating potential future hydrology. In some cases, the bias-correction procedure distorts projected changes in hydrology to such an extent that the bias-correction dominates the final time series. Consequently, Reclamation, its partners and stakeholders cannot always use the bias-corrected projections of future hydrology in their regulation studies with any great degree of confidence. If this project is not funded, Reclamation will have to continue its current bias-correction procedures, which have known and potentially severe shortcomings. As a result, Reclamation may not be able to fully achieve its mission.
Contact the Principal Investigator for information about partners.
Please contact email@example.com about research products related to this project.