Wind task 51 and PVPS task 16: how large-scale weather pattern influence short-term solar forecast error?

Overview

This study examines the relationship between large-scale North Atlantic weather regimes and the accuracy of satellite-based intraday solar forecasts. Despite satellite methods generally outperforming numerical weather prediction models at short-term horizons, their performance degrades substantially under adverse atmospheric conditions including convection, fog, and deep depressions. The research quantifies how four distinct weather regimes—Atlantic Ridge, Scandinavian Blocking, NAO+, and NAO——modulate forecast error magnitudes across seasonal and temporal dimensions.

Methods and approach

An 8-year hindcast validation was conducted using satellite-based solar forecasts generated four hours ahead with 15-minute temporal resolution. Forecast accuracy was assessed against pyranometer measurements. The analysis stratified results according to four dominant North Atlantic weather regimes classified through atmospheric pattern recognition. Relative root mean square error (RMSE) was computed across regimes and compared between seasonal periods (winter and summer) and temporal intervals (2016-2020 versus post-2020).

Results

Forecast error magnitude demonstrated substantial regime-dependence. Summer relative RMSE differences between Scandinavian Blocking and Atlantic Ridge conditions ranged from 10-12% during 2016-2020, declining slightly to approximately 10% thereafter. Winter regime-dependent variation was more pronounced, with differences of approximately 20% before 2020 and 15% after 2020. These findings demonstrate that prevailing large-scale atmospheric patterns constitute a significant source of variability in satellite forecast reliability, with measurable seasonal and multiyear variations in regime-dependent error structure.

Implications

The documented relationship between weather regime persistence and forecast error provides a predictability constraint applicable to solar integration planning. Given that weather regime transitions and frequencies are predictable multiple days in advance using atmospheric models, regime-conditional forecast error estimates can inform anticipatory system management decisions. This capacity enables conditional adjustment of forecast reliability expectations prior to the operative forecast window, supporting more robust electricity trading strategies and microgrid dispatch protocols.