Hello! I am working to expand subseasonal to seasonal hydrologic forecasting capabilities for large navigable rivers, with a focus on the Upper Mississippi River in the US, using AI/ML hydrologic models. I am based at the University of Minnesota in the US. Currently I have a proposal to do this in development. I would love to incorporate some of these state-of-the-art S2S AI/ML-based forecasts being generated for the AI Weather Quest into an eventual project focused on this. I will also be using some traditional and currently operational models, but this would be an interesting outlet to develop a ready-made application of this work beyond meteorology. I am wondering if any teams would be interested in discussing this application of your work. Let me know in this forum and I am happy to get in touch and discuss further.
1 Like
In hydrologic models, would you be considering the surrogate calibration such as on Wflow.jl https://deltares.github.io/Wflow.jl/v0.3/ with deep learning https://www.mdpi.com/2072-4292/17/11/1916
Hi! I have not used those methods but I am open to suggestions. The basic method we are looking at is an encoder-decoder LSTM framework. The fact that the approaches you link to can be integrated into Deltares products such as FEWS which is already used by the US National Weather Service is certainly an advantage. The method we are hoping to expand is shown here - https://doi.org/10.22541/essoar.172900696.63551165/v2 - McEachran et al., In Press. Water Resources Research. 10.1029/2024WR039064