Interpreting Deep Machine Learning for Streamflow Modeling Across Glacial, Nival, and Pluvial Regimes in Southwestern Canada
Bibliographic record
Abstract
The interpretation of deep learning (DL) hydrological models is a key challenge in data-driven modeling of streamflow, as the DL models are often seen as “black box” models despite often outperforming process-based models in streamflow prediction. Here we explore the interpretability of a convolutional long short-term memory network (CNN-LSTM) previously trained to successfully predict streamflow at 226 stream gauge stations across southwestern Canada. To this end, we develop a set of sensitivity experiments to characterize how the CNN-LSTM model learns to map spatiotemporal fields of temperature and precipitation to streamflow across three streamflow regimes (glacial, nival, and pluvial) in the region, and we uncover key spatiotemporal patterns of model learning. The results reveal that the model has learned basic physically-consistent principles behind runoff generation for each streamflow regime, without being given any information other than temperature, precipitation, and streamflow data. In particular, during periods of dynamic streamflow, the model is more sensitive to perturbations within/nearby the basin where streamflow is being modeled, than to perturbations far away from the basins. The sensitivity of modeled streamflow to the magnitude and timing of the perturbations, as well as the sensitivity of day-to-day increases in streamflow to daily weather anomalies, are found to be specific for each streamflow regime. For example, during summer months in the glacial regime, modeled daily streamflow is increasingly generated by warm daily temperature anomalies in basins with a larger fraction of glacier coverage. This model's learning of “glacier runoff” contributions to streamflow, without any explicit information given about glacier coverage, is enabled by a set of cell states that learned to strongly map temperature to streamflow only in glacierized basins in summer. Our results demonstrate that the model's decision making, when mapping temperature and precipitation to streamflow, is consistent with a basic physical understanding of the system.
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 | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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".