Surface Solar Spectral Irradiance Retrieval And Evalutaion Using Satellite-based And Ground-based Observation Data

Surface solar radiation drives processes such as the water cycle and carbon cycle on the Earth’s surface through controlling the exchange of energy and matter between the Earth and atmosphere.The spectral distribution of solar radiation is of great importance for climate modeling,photovoltaic power generation,crop yield estimation,and other applications.However,most current algorithms primarily focus on observing and estimating broadband irradiance,with limmited research on its spectral distribution.Additionally,existing ground-based measurements rely on spectrometers,which are expensive and difficult to maintain calibration.The satellite-based estimation suffer from low spatial and spectral resolution,which cannot meet the current demand for high-resolution surface solar spectral irradiance(SSSI).This study firstly proposes a retrivel algorithm for SSSI using ground-based allsky images.Combined with radiative transfer models,the algorithm attempts to obtain surface solar irradiance with higher spectral resolution from the broadband radiation information of the all-sky image.To address the lack of physical mechanism in the retrivel algorithm for all-sky images,a method for calibrating the all-sky image based on the sun’s trajectory is proposed.By fitting the imaging parameters with sun position in sky and actual projection position,the distribution of zenith and azimuth angles of the all-sky image is obtained.Then,a convolutional neural network model is proposed for effectively cloud detection of all-sky images.The all-sky images under clear and cloudy conditions are processed separately based on the cloud detection results.For clear-sky all-sky images,the algorithm establishes the relationship between grayscale value at zenith point and irradiance through radiation calibration algorithms and then estimate the aerosol optical thickness(AOD).The reconstructed SSSI under clear-sky conditions is obtained by combining the SBDART model with the estimated AOD.The comparison with ground-based measurement confirms the accuracy of the algorithm.To address the difficulty of estimating SSSI under cloudy conditions,this study analyzes the efect of three-dimensional clouds on surface irradiance using all-sky images,which provides a foundation for future work.This study then proposes two LUT-based methods for SSSI estimation from MODIS atmospheric products and top of atmosphere(TOA)radiance,respectively.Factors affecting SSSI are inputs of LUT,and surface spectral irradiance data are obtained by interpolating the LUT with real-time atmospheric products.In cases where atmospheric products are incomplete,SSSI is obtained by firstly inverse atmospheric conditions from satellite TOA radiance.To improve lookup efficiency,the effects of water vapor,ozone,and Earth-Sun distance on SSSI are corrected by post-correction.The two algorithms are validated using ground observation data from the Dunhuang station and the Southern Great Plains station of the US Atmospheric Radiation Measurement Plan.The results show that the reconstruction accuracy of the two algorithms is high under clear-sky conditions(around 5%-10%).However,due to the presence of three-dimensional cloud effects and representativeness errors between satellite and ground observations,the root mean square error between satellite and ground observations is about 30%under cloudy conditions.This study achieves SSSI retrieval with a spectral resolution of 5 nm,which has significant implications for research into climate change and the carbon cycle.It also provides important data support for fields such as renewable energy,agriculture,and water resource management.