Evaluation of ML techniques for downscaling hydrodynamic simulations
Bibliographic record
Abstract
The accurate representation of fluvial hydrodynamic characteristics, including water depth, and velocity field, is crucial for many hydro-environmental, geomorphological, and engineering applications. Traditionally, the hydrodynamic characteristics are determined using a 2D physics-based model that solves the shallow-water equations. However, employing these physics-based models for fine resolution simulations on large domains is often impractical due to the high computational cost. Consequently, spatial resolution often needs to be reduced to enable simulations. In this paper, we introduce and compare three Machine Learning (ML) approaches for downscaling: a linear dense model, as well as two deep neural networks (DNN) with convolution layers and fully connected layers. The objective is to enhance the spatial resolution of hydrodynamic models while maintaining reasonable computational efficiency and high precision. To evaluate and compare the different neural networks, a synthetic test case, the confluence of two synthetic channels, will be used. The results showcase that a DNN, the U-Net style, gives the best results due to the powerful convolutional layers, the non-linear activation function and skip connections. Then, we used a similar U-Net model to enhance the spatial resolution of a real-world riverine system in the St. Lawrence River, Canada. In conclusion, we will identify the strengths and weaknesses of existing architectures in different hydrodynamic contexts. This can pave the way for the development of an architecture specific to hydrodynamic phenomena.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".