Algorithm for continual monitoring of fog life cycles based on geostationary satellite imagery as a basis for solar energy forecasting

Abstract. Detection and monitoring of fog and low stratus (FLS) is particularly important in the context of photovoltaic power production, as FLS, unlike moving clouds, can persist longer and impact larger regions simultaneously, making regional power grid balancing harder. Although photovoltaic power production is limited to daytime hours, its short-term forecasting (especially during the early hours of the day) in the context of high PV penetration systems grid operation, benefits from a complete knowledge of FLS life cycle. As this life cycle usually begins at night and ends during the day, a day-night consistency in the algorithms used for monitoring FLS is required. This study presents an algorithm for detection of FLS over Europe based on the infrared bands of the SEVIRI (Spinning Enhanced Visible and InfraRed Imager) instrument onboard the Meteosat second generation geostationary satellites. As the method operates based on the SEVIRI infrared observations only, it is expected to be stationary in time and thus can provide a coherent and detailed view of FLS development over large areas over the 24 H day cycle. The algorithm is based on a gradient boosted trees machine learning model that is trained with ground truth observations from Meteorological Aviation Routine Weather Reports (METAR) stations and the SEVIRI observations at bands cantered at 8.7, 10.8, 12.0 and 13.4 μm wavelengths. The METAR data used here comprises a total number of 2,544,400 datapoints spread over the winters (i.e., 1st of September to 31st of May) of the years 2016–2022 and 356 locations across Europe. Among them, the datapoints corresponding to 276 stations and the winters of 2016–018 and 2019–2021 (~45 % of all datapoints) were used to train the algorithm. The remaining datapoints comprise four independent datasets which were used to validate the algorithm’s performance and applicability to the time spans and locations in the study area (i.e., Europe) that extend beyond particular locations and time spans covered by the datapoints used for training the algorithm. Additionally, the algorithm’s accuracy at the locations of METAR stations with that of the stablished state-of-the-art daytime FLS detection algorithm Satellite-based Operational Fog Observation Scheme (SOFOS). Validation of the algorithm against the METAR data, showed that the algorithm is well suited for detection of FLS. Specifically, the algorithm is found to detect FLS with probabilities of detection (POD) ranging from 0.70 to 0.82 (for different inter-comparison approaches), and false alarm ratios (FAR) between 0.21 and 0.31. These numbers are very close to those achieved by SOFOS for discriminating the FLS from other sky conditions at the tested locations and time spans. These results also showed that the technique’s applicability in the study region extends beyond the particular locations and time spans covered by the datapoints used for training the algorithm.