Solar radiation transfer in the atmosphere : the effect of aerosols and clouds on solar resource and nowcasting

• Κύριος συγγραφέας: Λογοθέτης, Σταύρος-Ανδρέας
• Άλλοι συγγραφείς: Logothetis, Stavros-Andreas
• Γλώσσα: English
• Έκδοση: 2023
• Θέματα: Aerosols; Solar radiation; Machine learning; Dust aerosol trends; Energy balance; Aerosol optical properties; Solar nowcasting; All-sky imagers; Aerosol classification; Αιωρούμενα σωματίδια; Ηλιακή ακτινοβολία; Τεχνικές μηχανικής μάθησης; Χωροχρονική μεταβολή σωματιδίων σκόνης; Ενεργειακό ισοζύγιο; Οπτικές ιδιότητες αιωρούμενων σωματιδίων; Κατηγοριοποίηση αιωρούμενων σωματιδίων; Πρόγνωση ηλιακής ακτινοβολίας

The aim of this Ph.D. dissertation is to investigate the effects of aerosols and clouds on solar resource and nowcasting in the atmosphere. Aerosols and clouds are the substances with the highest uncertainty in the estimations of the energy budget of the Earth’s−Atmosphere system that configures the Earth’s climate. In Chapter 1, a brief introduction of solar radiation transfer in the atmosphere is given, focusing on the role of aerosols and clouds in the Earth’s climate and solar resource studies. Fundamental concepts and definitions of solar radiation are presented, like Planck’s and Stefan-Boltzmann laws, spectral irradiance, and actinic flux. The role of the absorption and scattering effects in the total extinction within the atmosphere by various substances is also examined, expressing the Rayleigh and Mie scattering theories as well as the initial form and the parts of the radiative transfer equation. Finally, the effect of the most influential atmospheric constituents on the energy budget of the Earth−Atmosphere system is discussed, quantitating their impact along with their uncertainties on Earth’s climate change. In Chapter 2, the various datasets that were used for this Ph.D. are described briefly. The datasets are divided into two primary sources based on the measurement’s origin. The first category of the datasets is derived from ground-based measurements and satellite instruments, and the second category is derived from models. The aerosol information is derived from AERONET, MODIS, MERRA-2, CAMS, and MIDAS datasets, whereas information about solar radiation is derived either directly from BSRN and McClear clear-sky model or indirectly from all-sky imager-based systems. Aerosol studies often require accurate information about aerosol type due to different aerosol origins leading to different aerosol sizes, which have different chemical compositions, optical properties, and removal processes. These differences should be considered by climate models in order to assess accurately the direct aerosol effect on the climate. In Chapter 3, the aerosol optical properties from AERONET Version 3 are used to classify the aerosol types in Europe, Middle East North Africa (MENA), and the Arabian Peninsula, during 2008–2017. Quality-assured data of SSA, FMF, and AE from 39 stations is used to classify the aerosol types based on the threshold limits of these optical properties. The aerosol type depends on the location and the sources of each region of study; for example, in the Atlantic, Arabian Peninsula, and MENA, the dominant aerosol type is coarse absorbing due to dust from the Sahara and Arabian deserts. However, in the Arabian Peninsula, fine particles are also observed in autumn and winter. In addition, the lower percentages of coarse absorbing particles across MENA are observed in the East because of increased fine particle emissions from human activities. In southern Europe, the stations of Group A (southern stations), a bimodal size distribution is revealed, and the dominant aerosol types are fine slightly absorbing and non-absorbing, followed by coarse absorbing due to Sahara dust outbreaks. In the stations of Group B in South Europe (northern stations), fine slightly absorbing and non-absorbing particles are primarily observed since the stations are located in urban/industrial regions. In Central and East Europe, the prevailing aerosol type is fine-non-absorbing, which is followed by the fine, slightly absorbing aerosols from urban/industrial sites. In Chapter 4, the results from the aerosol classification scheme in Chapter 3 are used to quantify the influence of aerosol type on the radiative balance of the Earth’s climate within the same study period. The impacts of SZA, SA, AOD and single SSA on DARF and DARFeff at the BOA and TOA are investigated. Fine slightly absorbing particles show the highest positive gradient of DARFeffBOA with SA and the highest negative gradient of DARFeffTOA with SZA. The mixed absorbing particles provide the highest alteration for DARF at the BOA with the increase of AOD. The analysis of aerosol absorptivity is performed by dividing SSA into six-subgroups. Coarse absorbing particles provide the highest (in magnitude) DARFeff values at the TOA under absorbing aerosol conditions (SSA < 0.89), whereas similar behavior is revealed by the fine absorbing particles for DARFeff values at the BOA. Furthermore, the long-term averages of radiative forcing metrics are analyzed for all aerosol types among the regions of study. At the TOA, fine non-absorbing particles show the highest absolute values for DARFeff (from −75 W m−2 to −79 W m−2) and DARF (from −38 W m−2 to −48 W m−2). At the BOA, coarse and mixed absorbing clusters indicate the highest absolute values for DARF (from −66 W m−2 to −79 W m−2) and DARFeff (from −135 W m−2 to −149 W m−2), respectively. Aeolian dust aerosols affect the Earth’s energy budget through interactions with solar radiation, atmospheric chemistry, and clouds. Erroneous estimations about the historical, present, and future dust aerosol burden in the atmosphere reflect to wrongly estimated aerosol radiative forcing and projection of future climate change. In Chapter 5, the global, regional and seasonal temporal dust changes as well as the effect of dust particles on total aerosol loading using the MIDAS fine-resolution dataset are investigated. MIDAS delivers DOD at fine spatial resolution (0.1° x 0.1°) spanning from 2003 to 2017. Within this study period, the dust burden increased across the central Sahara (up to 0.023 yr−1) and Arabian Peninsula (up to 0.024 yr−1). Both regions observed their highest seasonal trends in summer (up to 0.031 yr−1). On the other hand, declining DOD trends are encountered in the western (down to −0.015 yr−1) and eastern (down to − 0.023 yr−1) Sahara, the Bodélé Depression (down to −0.021 yr−1), the Thar (down to −0.017 yr−1) and Gobi (down to −0.011 yr−1) deserts, and the Mediterranean Basin (down to −0.009 yr−1). In spring, the most negative seasonal trends are recorded in the Bodélé Depression (down to −0.038 yr−1) and Gobi Desert (down to −0.023 yr−1), whereas they are in the western (down to −0.028 yr−1) and the eastern Sahara (down to −0.020 yr−1) and the Thar Desert (down to −0.047 yr−1) in summer. Over the western and eastern sector of the Mediterranean Basin, the most negative seasonal trends are computed at summer (down to −0.010 yr−1) and spring (down to −0.006 yr−1), respectively. The effect of DOD on the AOD change is determined by calculating the DOD-to-AOD trend ratio. Over the Sahara the median ratio values range from 0.83 to 0.95, whereas in other dust-affected areas (Arabian Peninsula, southern Mediterranean, Thar and Gobi deserts) the ratio value is approximately 0.6. In addition, a comprehensive analysis of the factors affecting the sign, the magnitude and the statistical significance of the calculated trends is conducted. Firstly, the implications of the implementation of the geometric mean instead of the arithmetic mean for trend calculations are discussed, revealing that the arithmetic-based trends are overestimated when compared to the geometric-based trends over both land and ocean. Secondly, an analysis interpreting the differences in trend calculations under different spatial resolutions (fine and coarse) and time intervals is conducted. AOD constitutes a key parameter of aerosols, providing vital information for quantifying the aerosol burden and air quality at global and regional levels. In Chapter 6, a machine learning strategy for retrieving AOD under cloud-free conditions, based on the synergy of machine learning algorithms (MLAs) and solar irradiance data is presented. The retrieval methodology is applied twice, using a) ground-based solar irradiances and b) clear-sky model simulations. The performance of MLAs is investigated by applying different components of solar radiation. The use of direct instead of global irradiance as a model feature led to better performance. Regardless of the underlying climate and aerosol environments, the MLA-based AODs encompassed RMSE between 0.01 and 0.15. Among the MLAs, artificial neural networks outperformed the other algorithms in terms of RMSE at 54% of the measurement sites. Compared to MERRA-2 and CAMS, the ML-AOD retrievals revealed the highest accuracy. The performance of MLAs was also assessed by replacing the ground-based solar irradiance measurements with estimations from McClear model. The MLA performance depends on the underlying climate and aerosol environments with higher deviations in equatorial and arid climate areas as well as in regions dominated by biomass burning and mineral dust particles. Fluctuations of the incoming solar irradiance impact the power generation from photovoltaic and concentrating solar thermal power plants. Accurate solar nowcasting becomes necessary to detect these sudden changes of generated power and to provide the desired information for optimal exploitation of solar systems. In Chapter 7, a benchmarking exercise has been conducted relying on a bouquet of solar nowcasting methodologies by all-sky imagers (ASI). The work of Chapter 7 is conducted in the framework of the International Energy Agency’s Photovoltaic Power Systems Program Task 16, where four ASI systems nowcast the GHI with a time forecast ranging from 1 to 20 minutes during 28 days with variable cloud conditions spanning from September to November 2019 in southern Spain. All ASIs demonstrated their ability to accurately nowcast GHI, with RMSE ranging from 6.9% to 18.1%. Under cloudy conditions, all ASI nowcasts outperform the persistence models. Under clear skies, three ASIs are better than persistence. Discrepancies in the observed nowcasting performance become larger at increasing forecast horizons. The findings of Chapter 7 highlighted the feasibility of ASIs to reliably nowcast GHI at different sky conditions, time intervals, and horizons. Such nowcasts can be used either to estimate solar power at distant times or to detect sudden GHI fluctuations. In Chapter 8, the results from Chapter 7 are used to detect sudden GHI fluctuations based on ASIs' GHI nowcasts. ASIs 1–2 and ASIs 3–5 can capture the true ramp events by 26.0–51.0% and 49.0–92.0% of the cases, respectively. ASIs 1–2 provided the lowest (< 10.0%) falsely documented ramp events while ASIs 3–5 recorded false ramp events up to 85.0%. On the other hand, ASIs 3–5 revealed the lowest falsely documented no ramp events (8.0–51.0%). ASIs 1–2 are developed to provide spatial solar irradiance forecasts and have been delimited only to a small area for the purposes of this benchmark, which penalizes these approaches. These findings show that ASI-based nowcasts could be considered as a valuable tool for predicting solar irradiance ramp events for a variety of solar energy technologies. The combination of physical and deep learning-based methods is identified as a potential approach to further improve the ramp event forecasts.