A Fast Cloud Geometrical Thickness Retrieval Algorithm For Single-layer Liquid Clouds Using OCO-2 Oxygen A-band Measurements

Clouds are common but important,widely distributed,and changeable.Knowledge of cloud geometrical thickness is of importance for the study of radiative balance and cloud microphysics.Previous studies have shown that it is feasible to retrieve multiple cloud properties simultaneously based on the space-borne hyperspectral observation in the oxygen A-band,such as cloud optical depth,cloud-top height,and cloud geometrical thickness.However,it is still challenging to design a fast retrieval algorithm for cloud geometric thickness,which requires accurate estimation of land surface albedo,rapid calculation of cloud reflectance,evaluation of retrieval accuracy,and error source analysis.1)Land surface albedo estimation in the oxygen A-bandExcept the cloud reflection,the land surface reflection is the strongest,a strong disturbance in the cloud geometrical thickness retrieval,and is closely related to the land surface albedo.We propose a method for estimating the land surface albedo in oxygen A-band based on the multi-channel black/white albedos of MCD43C3 products and emphasize the importance of land cover type in the estimation.We verified the reliability of the multi-channel model based on the tests in a different time and different space,where the coefficients of determination were all over 0.9 and the root-meansquared error was 0.026.In addition,we verified the multi-channel model for different land cover types.The multi-channel model was always superior to the single-channel linear model,whether applied to the best performing type of the ‘barren or sparsely vegetated',or to the worse performing type of the ‘snow and ice'.The quality of the seven-channel albedo data is the most important factor affecting the accuracy of the land surface albedo estimation in the oxygen A-band.The root-mean-squared error of the estimation with the best input quality was slightly better than 0.02 and increased to more than 0.05 as the input quality decreased.Besides,inaccurate references also led to errors in the comparison.2)A rapid parameterization of hyperspectral reflectance in the oxygen A-bandHyperspectral remote sensing is time-consuming if based on the precise radiative transfer solution that counting multiple scattering of light.To speed up the radiation transfer solution in cloud scenarios for nadir space-borne observations,we developed a rapid parameterization of hyperspectral reflectance in the oxygen A-band for singlelayer water clouds.The parameterization takes into account the influences of cloud droplets forward-scattering and nonlinear oxygen absorption on hyperspectral reflectance,which are improvements over the previous studies.The performance of the parameterization is estimated through comparison with DISORT(Discrete Ordinates Radiative Transfer Program Multi-Layered Plane-Parallel Medium)on the cases with solar zenith angle ,the cloud optical depth ,and the single-scattering albedo in the range of 0° ? ? 75°,5 ? ? 50,0.5 ? ? 1.The relative error of the cloud reflectance is within 5% for most cases,even for clouds with optical depths around five or at strong absorption wavelengths.Applying in the hyperspectral simulation,the relative error of the out-of-band radiance is less than 4%,and the relative error of the intra-band radiance ratio is less than 4%.The radiance ratio is the ratio of simulated observations with and without in-cloud absorption and is used to assess the accuracy of the parameterization in quantifying the in-cloud absorption.The parameterization is preparation for rapid hyperspectral remote sensing in the oxygen A-band.It would help improve retrieval efficiency and provide cloud geometric thickness products.3)A fast cloud geometrical thickness retrieval algorithmThe retrieval of cloud geometrical thickness(H)remains challenging,especially for passive instruments.In this work,we derive a semi-analytical algorithm for retrieving H of single-layer liquid cloud based on oxygen A-band hyperspectral measurements from NASA's Orbiting Carbon Observatory-2(OCO-2).The algorithm can retrieve H in real-time as it does not require the use of the time-consuming radiative transfer model for radiation calculation during each retrieval.In addition,the algorithm currently requires cloud optical depth,cloud top height,and aerosol properties measured by other instruments as input.In idealized simulations using ten thousand Aband spectra spanning a range of cloud cases,the root-mean-squared error(RMSE)of the retrieval is approximately 2.0 h Pa(for low clouds,1h Pa is about 10 m).We also retrieve H based on millions of real OCO-2 observations and compare the retrieval results with the cloud product from Cloud Sat/CALIPSO(Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations).After abnormal samples are removed,the correlation coefficient is 0.716,the average bias is-15.6 h Pa,and the RMSE is 27.4h Pa.The statistical results show that the absolute bias increases systematically with the reference cloud geometrical thickness,which may be caused by the unrealistic vertical homogeneous cloud assumption.The phenomenon was also found in comparison with OCO2CLD-LIDAR-AUX,a retrieval product based on an accurate radiative transfer model.