The role of North Atlantic Oscillation teleconnections in solar irradiance nowcasting error variability

The movement and dynamics of clouds significantly impact solar radiation and energy production from photovoltaic (PV) systems. Short-term solar irradiance forecasts, ranging from hours to days, are essential for reliable energy supply through PV plants. Forecasts using geostationary satellites outperform numerical weather prediction models for intraday forecasts. However, forecast accuracy depends heavily on prevailing weather conditions.

The North Atlantic Oscillation (NAO), a key teleconnection over the Euro-Atlantic region, significantly shapes weather patterns in western Europe and impacts the accuracy of satellite-based solar irradiance forecasts. The present study analyzes eight years (2016-2024) of Global Horizontal Irradiance (GHI) forecasts at the SIRTA Observatory in Palaiseau, near Paris (France). The forecasts are generated four hours ahead with 15-minute time step using a cloud motion vector (CMV) computation to extrapolate the cloud over. These forecasts are validated against pyranometer observations. GHI forecast errors are analyzed for two periods (2016-2020 and 2020-2024), focusing on seasonal variations and the impact of NAO teleconnection indices provided by the Climate Prediction Center of the National Centers for Environmental Prediction (NCEP CPC).

The GHI forecast error values were averaged across all forecast horizons (0 to 240 minutes). The results indicated that the relative root mean square mean error (RRMSE) is 32.7% for spring and autumn seasons from 2016 to 2024. NAO+ and NAO- teleconnection indices are respectively associated with lower (29.5%) and higher (36.2%) RRMSE values across spring and autumn seasons and both time periods (2016-2020 and 2020-2024). NAO+ events are characterized by anticyclonic circulations over the Atlantic Ocean, bring reduced precipitation and stable weather across Europe, resulting in clearer skies and lower forecast errors. Conversely, NAO- events lead to higher errors due to less stable conditions. These findings are particularly significant as North Atlantic weather regimes, typically reliable predictors of forecast errors, appear less effective during transitional seasons like spring and autumn.

In winter and summer seasons, distinct patterns in GHI forecast errors were observed. During the winter of 2016-2020, NAO+ and NAO- events yielded higher (44.5%) and lower RRMSE in GHI forecast (31%), respectively. This trend reversed during the winter of 2020-2024, with NAO+ and NAO- events respectively, showed lower (43%) and higher (49%) RRMSE values. These seasonal variations during winter align with changes in the frequency of NAO events from 2020-2024, when NAO- occurrences increased while NAO+ occurrences decreased. During summer, similar seasonal trends were observed, though with reversed magnitudes during both NAO+ and NAO- regimes for 2016-2020 and 2020-2024.

Changes in GHI forecast errors emphasize the importance of understanding large-scale atmospheric patterns for a better interpretation of GHI forecasts. Errors linked to NAO indices in winter and summer should be further studied, as they may also be influenced by other teleconnections and weather regimes. As a dominant teleconnection over Europe and the Atlantic, advanced knowledge of NAO indices and their interaction with other weather systems helps in anticipating forecast errors, offering critical insights for energy traders and grid operators to enhance smart grid management.