Decomposing streamflow for improved river flow prediction accuracy of machine learning models
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
Accurate streamflow prediction can mitigate flood losses, optimize power generation, and reduce drought impacts. Streamflow time series has many inherent natural frequencies driven by both climate cycles and watershed characteristics. Decomposition methods have been used to isolate underlying fluctuations related to influencing variables such as climate oscillations from a single streamflow time series. Variational mode decomposition (VMD) was recently developed to improve upon common decomposition methods. These methods are known to suffer from various limitations, including the time–frequency trade-off, boundary effect (not significant in hindcasting), and predefined bases functions. Extreme gradient descent boosting (XGBoost) is an increasingly popular ML approach that has shown promise in many fields but has not been thoroughly applied to streamflow forecasting. This study develops a VMD-XGBoost model for daily streamflow forecasting. Since XGBoost allows for a customizable loss function, various loss functions are implemented in model training. Specifically, a seldom-recognized forecasting performance measure, horizontal error (HE), is used to improve model susceptibility to imitation error. The VMD-XGBoost model is compared to a standalone XGBoost model. It highlights that VMD significantly improves forecasting, by reducing HE from 0.94 to 0.41 while improving NSE from 0.82 to 0.84, and bias from 1.20 m3/s to 0.20 m3/s.
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