Staff Research Highlight - Improved process based streamflow simulation through ensemble and stochastic data driven approaches

Aquanty staff recently had the pleasure to meet the newest member of our modelling team over an informal ‘Lunch ‘n’ Learn’ presentation. David Hah, MSc. did his undergrad at the University of Waterloo for environmental engineering where he completed two co-op terms with Aquanty. He subsequently received a Masters degree in civil engineering to improve hydrological simulations using data-driven approaches. David recently joined the Aquanty team as a data scientist, and is helping us to continue developing and improving our near real-time hydrologic forecasting platforms like HGSRT. David’s research focused on the use of data-driven models to improve the results of hydrologic simulations.

Hydrologists have developed process-based hydrological models (HMs) and data-driven models (DDMs) to better understand the complexities of water movement on Earth, both of which have been applied to a variety of water resources applications (e.g., flood forecasting, reservoir operations, drought monitoring, hydraulic design). HMs attempt to simplify hydrological processes of interest (e.g., snowmelt, subsurface flow), whereas DDMs use historical data to estimate statistical relationships between explanatory/input and response/target variables. Although these two models have traditionally been improved independently, there is growing interest in using DDMs to improve HM simulations. Here, a new method is introduced to correct HMs probabilistically to enable reliable and conservative simulations for various water resource applications.

Keep your eye on the Aquanty blog for updates on this research, we hope to see this research published soon.

Abstract:

To better understand the complexities of water movement on earth, hydrologists have developed process-based hydrological models (HMs) and data-driven models (DDMs), both of which have been applied to a host of water resources applications (e.g., flood forecasting, reservoir operations, drought monitoring, hydraulic design). HMs attempt to simplify hydrological processes of interest (e.g., snowmelt, sub- surface flow), while DDMs estimate statistical relationships between explanatory/input and response/target variables using historical data.

Traditionally, HMs and DDMs have been developed independently, however, there has been growing interest in using DDMs to improve HM simulations. Among various approaches for combining process-based theory with DDMs, the conceptual data-driven approach (CDDA) was recently proposed, where DDMs are used to correct the residuals (errors) stemming from ensemble HMs followed by the stochastic CDDA (SCDDA) that used HM simulations as input to DDMs within a stochastic framework - both approaches improved ensemble HMs' simulations. Here, a new SCDDA is introduced where CDDA uncertainty is estimated (instead of DDM uncertainty in the original SCDDA). Using nine HM-DDM combinations for daily streamflow simulation in three Swiss catchments, the new SCDDA improved CDDA's mean continuous ranked probability score up to 15% and performed similarly without a snow-routine in a snowy catchment, suggesting that SCDDA may account for missing processes in HMs. The stochastic framework can convert unreliable ensemble models into more reliable (stochastic) models at the cost of simulation sharpness. The coverage probability plot is proposed as a diagnostic tool, predicting SCDDA's out-of-sample reliability using validation set data (CDDA simulations and observations).