Checking for missing content type metadata ...
This resource contains content types with missing metadata required to make it public or discoverable. Show missing content type metadata.
Click on the edit button ( ) below to edit this resource.
Checking for non-preferred file/folder path names (may take a long time depending on the number of files/folders) ...
This resource contains some files/folders that have non-preferred characters in their name. Show non-conforming files/folders.
This resource contains content types with files that need to be updated to match with metadata changes. Show content type files that need updating.
A Deep Learning Post-Processing Method for Improving the Accuracy of NWM Streamflow Forecasts: Design and Validation of a Modified ETSformer
| Authors: |
|
|
|---|---|---|
| Owners: |
|
This resource does not have an owner who is an active HydroShare user. Contact CUAHSI (help@cuahsi.org) to determine if accessing this resource is possible. |
| Type: | Resource | |
| Storage: | The size of this resource is 1.5 GB | |
| Created: | Oct 29, 2026 at 6:21 a.m. (UTC) | |
| Last updated: | Oct 30, 2026 at 2:20 p.m. (UTC) | |
| Citation: | See how to cite this resource |
| Sharing Status: | Discoverable |
|---|---|
| Views: | 29 |
| Downloads: | 0 |
| +1 Votes: | 1 other +1 this |
| Comments: | No comments (yet) |
Abstract
Operational long-range (30-day) streamflow forecasts from NOAA’s National Water Model (NWM) carry a combination of model structural error and forecast-forcing error that grows with lead time, limiting their practical value for water-resources decision-making. This paper presents a learned post-processing framework that improves NWM long-range forecast skill without retraining the underlying physics-based model. The framework adopts the Exponential Smoothing Transformer (ETSformer) as its backbone and augments it with three components: a Conditional Frequency Channel Attention (CFCA) module that incorporates static basin attributes such as dam-regulation status, Reversible Instance Normalization (RevIN) to address non-stationary distribution shift across basins, and a zero-cost Linear Scaling step that removes residual systematic bias. We train and evaluate the framework exclusively on the NWM’s archived operational long-range forecast product, retrieved from the public BigQuery archive for Sep,2018 to Nov, 025 across 20 USGS-gauged river reaches distributed across the contiguous United States. We benchmark the framework against four LSTM variants (LSTM, SAM-LSTM, MHAM-LSTM, BOA-LSTM) and two alternative Transformer architectures (Pyraformer, iTransformer). The framework reduces unbiased root-mean-square error by approximately 77% relative to the strongest LSTM baseline and maintains its skill across the full 30-day forecast window rather than degrading with lead time. Ablation experiments show that CFCA, RevIN, and Linear Scaling each contribute, and that their gains are complementary. A leave-one-basin-out generalization experiment indicates that the model transfers to unseen watersheds with only marginal loss in accuracy. The results demonstrate that a Transformer combined with explicit non-stationarity correction and static-attribute conditioning can deliver stable long-horizon bias correction for the NWM operational forecast product.
Subject Keywords
Content
How to Cite
This resource is shared under the Creative Commons Attribution CC BY.
http://creativecommons.org/licenses/by/4.0/
Comments
There are currently no comments
New Comment