Unravelling the power of neural networks for flood prediction across complex hydrological systems
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The increasing frequency and intensity of flooding events, driven by climate change and land-use modifications, call for the development of more advanced prediction tools to support early warning systems and disaster mitigation strategies. This chapter explores the use of multiple neural networks for flood prediction, focusing on their application in two complex and hydrologically challenging gauging stations in the USA: the Satilla River near Waycross, Georgia, and the coastal area of Socastee, South Carolina. We examined three state-of-the-art neural networks with different algorithmic structures: N-HiTS (Neural Hierarchical Interpolation for Time Series Forecasting), a residual-based model; LSTM (Long Short-Term Memory), a recurrent neural network optimized for capturing sequential dependencies; and PatchTST, a transformer-based architecture utilizing self-attention and patch embedding strategies. All models were trained using multi-year hydrometeorological time-series data (2007–22) and evaluated on an independent testing set (2022–24) across multiple prediction horizons (1, 3, 6 and 12 h). Among multiple models, N-HiTS consistently outperformed LSTM and PatchTST in both flood-prone settings. N-HiTS demonstrated superior accuracy, especially under complex tidal conditions, due to its hierarchical structure and multi-scale feature representation. PatchTST performed competitively in stable hydrological regimes, while LSTM struggled with long-term dependencies and dynamical shift in hydrological behaviours. These results emphasize the effectiveness of N-HiTS in capturing flood dynamics across multiple horizons, enhancing flood prediction reliability across multiple temporal and spatial scales.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 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 it