analiza aiqbench
This study evaluates the impact of LoRA adaptation strengths on an atmospheric foundation model within a WRF coupling framework, specifically targeting heavy rainfall forecasting. Results reveal a critical trade-off where unadapted models maximize detection success (CSI), while intermediate adaptation levels significantly reduce error metrics but sacrifice some detection capability. The findings underscore the necessity of balancing multiple downstream criteria when fine-tuning foundation models for specific regional weather applications.
LoRA Adaptation Strength in Aurora-WRF: Trade-offs in Regional Heavy-Rainfall Forecasting
arXiv:2610.03747v1 Announce Type: cross Abstract: We investigate how parameter-efficient adaptation of an atmospheric foundation model affects downstream regional precipitation in a controlled Aurora-WRF coupling experiment over the Beijing-Tianjin-Hebei region. A single-step, precipitation-weighted LoRA adaptation of AuroraPretrained is trained on May-September 2020-2021 and selected using 2022 validation data. Five inference-time adaptation strengths are coupled to an unchanged 9-km WRF configuration and compared with a GFS-driven control under common ERA5 initialization and auxiliary-field rules. Evaluation uses 17 initializations in 2023, hourly ERA5 precipitation, and conservative remapping to a fixed 356-cell, 0.25-degree verification grid. Heavy rainfall is defined exclusively as at least 50 mm in a continuous 24-hour window, advanced hourly. For window starts from 0 to 48 hours, unadapted Aurora attains the highest mean critical success index (CSI), 0.216, compared with 0.144 for the GFS control. An intermediate LoRA strength of 0.75 reduces the Aurora-driven 24-hour precipitation mean squared error by 30.7% and brings pooled frequency Bias from 1.510 to 1.024, but lowers CSI to 0.182. GFS retains the lowest whole-domain error. These results characterize a trade-off between rainfall detection, frequency calibration, and intensity error, showing the importance of evaluating adaptation strength against multiple downstream criteria within a common coupling framework.
Abstract:We investigate how parameter-efficient adaptation of an atmospheric foundation model affects downstream regional precipitation in a controlled Aurora-WRF coupling experiment over the Beijing-Tianjin-Hebei region. A single-step, precipitation-weighted LoRA adaptation of AuroraPretrained is trained on May-September 2020-2021 and selected using 2022 validation data. Five inference-time adaptation strengths are coupled to an unchanged 9-km WRF configuration and compared with a GFS-driven control under common ERA5 initialization and auxiliary-field rules. Evaluation uses 17 initializations in 2023, hourly ERA5 precipitation, and conservative remapping to a fixed 356-cell, 0.25-degree verification grid. Heavy rainfall is defined exclusively as at least 50 mm in a continuous 24-hour window, advanced hourly. For window starts from 0 to 48 hours, unadapted Aurora attains the highest mean critical success index (CSI), 0.216, compared with 0.144 for the GFS control. An intermediate LoRA strength of 0.75 reduces the Aurora-driven 24-hour precipitation mean squared error by 30.7% and brings pooled frequency Bias from 1.510 to 1.024, but lowers CSI to 0.182. GFS retains the lowest whole-domain error. These results characterize a trade-off between rainfall detection, frequency calibration, and intensity error, showing the importance of evaluating adaptation strength against multiple downstream criteria within a common coupling framework.
Atmospheric and Oceanic Physics (physics.ao-ph); Artificial Intelligence (cs.AI)
arXiv:2610.03747 [physics.ao-ph]
(or arXiv:2610.03747v1 [physics.ao-ph] for this version)
https://doi.org/10.48550/arXiv.2610.03747 Focus to learn more arXiv-issued DOI via DataCite
REPRODUCIBILITY.md
analysis/case_dispersion.csv
analysis/case_metrics.csv
analysis/event_support.csv
analysis/leave_one_start_out.csv
analysis/map_selection.json
analysis/summary.csv
analysis/window_metrics.csv
build_analysis.py
build_lead_uncertainty.py
inputs/archive_hashes.json
inputs/dates.npy
inputs/drivers.npy
inputs/dual_scope_summary.csv
inputs/export_validation.json
inputs/latitude.npy
inputs/longitude.npy
inputs/truth.npy
uncertainty/make_preview.py
uncertainty/uncertainty_statistics.json
verify_numeric.py
(21 additional files not shown) You must enabled JavaScript to view entire file list.
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 326 cuvinte
