XGBoost-Based Corrections for Road Surface Temperature Forecasts.
Saved in:
| Title: | XGBoost-Based Corrections for Road Surface Temperature Forecasts. |
|---|---|
| Authors: | Karsisto, Virve1 (AUTHOR) virve.karsisto@fmi.fi, Hieta, Leila1 (AUTHOR), Kämäräinen, Matti1 (AUTHOR) |
| Source: | Journal of Applied Meteorology & Climatology. Jun2026, Vol. 65 Issue 6, p1-17. 17p. |
| Subjects: | Boosting algorithms, Freezing points, Temperature, Weather forecasting, Machine learning, Road safety measures, Forecasting |
| Abstract: | Accurate road surface temperature forecasts are essential for planning road salting and keeping roads free of ice. However, these forecasts often contain systematic errors due to various factors. This study presents a post-processing methodology aimed at improving forecast accuracy, especially for temperatures near 0 °C, where even small errors can be critical for road safety. The methodology was tested to correct systematic errors in the Finnish Meteorological Institute's road surface temperature forecasts. An XGBoost model was trained using data from several years and approximately 400 road weather stations to learn patterns between forecasted weather variables and road surface temperature errors. A distinctive approach in this study was to apply sample-specific weights during model training, which gave more emphasis to cases where the surface temperature was near 0 °C. Another XGBoost model without weights was trained to perform well across all temperatures. The models were evaluated using data from September 2024 to May 2025. Based on scores calculated over the entire period, the unweighted model reduced the mean absolute error (MAE) by 3–23% compared to original road weather model results, depending on the forecast lead time, that extended up to 62 hours. For cases where the observed surface temperature was between −3 °C and 3 °C and the lead time was 12 hours or less, the weighted model achieved a MAE reduction of 17–34%. In conclusion, both models performed well, with the weighted model showing particular effectiveness in improving forecast accuracy at near zero-degree temperatures. [ABSTRACT FROM AUTHOR] |
| Copyright of Journal of Applied Meteorology & Climatology is the property of American Meteorological Society and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
Be the first to leave a comment!
