Hierarchical Deep Learning for Consistent Multi‐Timescale Hydrological Forecasting
Bibliographic record
Abstract
Abstract This research introduces a novel method for accurate and consistent hydrological forecasting at multiple timescales. Deep learning (DL) models are increasingly being used for hydrological forecasting across various timescales (hourly, daily, etc.). However, one of the main challenges with multi‐timescale DL‐based hydrological forecasting is the potential inconsistency (discrepancy) between forecasts across different timescales. Inconsistent multi‐timescale forecasts can be problematic especially when decision‐making relies on forecasts across different timescales. This paper introduces a hierarchical DL (HDL) model incorporating temporal hierarchical reconciliation (THR) with DL models for consistent multi‐timescale streamflow forecasting. HDL is developed, deployed, and tested for multi‐step (seven days ahead), multi‐timescale (daily and weekly) streamflow forecasting using over 400 catchments across the contiguous United States. HDL is based on long short‐term memory (LSTM) networks and implements THR through a differentiable output layer. HDL consistently improved the median Nash Sutcliffe efficiency (NSE) of daily streamflow forecasts (e.g., by 3.01% for lead time one) compared to a multi‐timescale LSTM benchmark and resulted in significantly more accurate forecasts in more than 65% of the catchments at the weekly scale than daily forecast aggregation. HDL performance is influenced by both the THR formulation and the accuracy of the forecasts at different timescales. For instance, at the weekly timescale, HDL yielded a notable median improvement of 9.45% in NSE for the lowest‐performing decile of catchments (NSE < 0.37), which are typically the most challenging to forecast. The proposed HDL framework offers a novel, generalizable, and promising approach for multi‐timescale forecasting across diverse water resources forecasting tasks.
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 | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| 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) | 0.002 | 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".