Extending Satellite Ocean Color Remote Sensing To The Ultraviolet Bands

Ultraviolet(UV)radiation accounts for about 10%of total solar radiation.UV radiation has profound effects on marine life,which include the phytoplankton growth,vertical movement of zooplankton,decomposition of dissolved organic matter,while the measurements of UV radiance improve aerosol monitoring and the inversion of water constituents.Understanding the distribution of UV radiation in the ocean can contribute to further understanding of marine biogeochemical cycles.Historically the research of UV radiation has been relied on on-site measured data in ocean optics,where the sampling interval and space are significantly limited,which cannot provide adequate and repetitive observations for broad areas.Ocean color satellites have the advantages of repetitive coverage of large areas,which have become an important method of current ocean research.However,historically and even today,most ocean color satellites have the shortest wavelength as 410 nm,thus they cannot provide radiance measurements in the UV,and thus UV penetration,in the global ocean.Based on the importance of UV radiation and the current limitations of UV measurement in ocean color remote sensing,this study carried out related researches on the expansion of ocean color satellites to the UV bands,as well as inversions of optical parameters in the UV band.The main research contents and results are as follows:(1)Developed a remote sensing reflectance(Rrs)prediction model for the UV bands of ocean color satellites.To fill the data gap of UV penetration in the global ocean,especially for measurements after the launch and operation of modern ocean color satellites,a deeplearning system(UVISRdl)using simulated data set is developed to estimate Rrs at the near-blue UV bands(nbUV,specifically at 360,380,and 400 nm in this study)with Rrs in the visible(410-670 nm)as the input.Taking VIIRS as an example,it is found that UVISRdl-estimated Rrs(nbUV)agree very well with those from field measurements,although larger differences are found between VIIRS Rrs(nbUV)and matchup in situ data when measurements were taken in coastal regions.(2)Established a CDOM inversion algorithm using UV band.Due to CDOM strong absorption properties in ultraviolet radiation,it plays an important role in marine research.The quasi-analytical algorithm(QAA)that introduces UV band(QAA_UV)proves that the UV band helps to improve the separation accuracy of CDOM nonalgal particle absorption coefficient(adg)and phytoplankton absorption coefficient(aph).But QAA-UV does not provide a way to separate CDOM absorption coefficient(ag).This study uses the Rrs(nbUV)predicted by the developed deep learning model to improve QAA_UV,and combines with inversion formula of ag(QAA_CDOM)to perform ag inversion.In the study,the simulated data set of IOCCG and the measured data in the China sea area were used to verify the model.The study found that the improved semi-analytical algorithm has higher accuracy,where the inversion result of ag(440)mean absolute relative difference(MARD)=0.48,the coefficient of determination(R2)=0.72.The model was applied to the estuary area,the separation of suspended particles and CDOM was observed,which provided a method for the detection of the movement distance of the dilute water in the estuary area.(3)Evaluation of algorithms for the estimation of the diffuse attenuation coefficient(Kd)in UV band.Six algorithms(both empirical and semi-analytical)developed for the estimation of Kd in the nbUV domain were evaluated from a dataset of 316 stations covering oligotrophic ocean and coastal waters.In particular,the semi-analytical algorithm used Rrs(nbUV)estimated from a recently developed UVISRdi as the input.For Kd(380)in a range of 0.018-2.34 m-1.it is found that the semi-analytical algorithm has the best performance,where the MARD=0.19,and R2=0.94.For a VIIRS and in situ matchup dataset(N=62),the MARD of Kd(380)=0.21(R2=0.94)by the semi-analytical algorithm.These results indicate that a combination of deep learning system and semianalytical algorithm can provide reliable Kd(nbUV)for past and present satellite ocean color missions that have no spectral bands in the UV,and help for subsequent long-term series of UV radiation studies.(4)Long-term data analysis in the UV band.Using the obtained Rrs(nbUV)and Kd(nbUV),combined with chlorophyll(Chi),sea surface temperature(SST),mixed layer depth(MLD),photosynthetically available radiance(PAR)data,long-term data analysis was carried out to obtain the changes of optical parameters.By studying the monthly average data from 1997.09 to 2021.02,and taking the data of the south pacific gyre(SPG)and north pacific gyre(NPG)as an example,the changes of data on a long-term scale were analyzed.The results show that Chl has obvious seasonality,with lower SST,higher MLD,higher Chl,lower Rrs,higher Kd,and lower penetration depth(Z1%)in the SPG.The NPG seasonal characteristics are not very obvious.Combining the 380nm irradiance(Ed(380))data of the OMI satellite and the MLD and Kd data,it is found that the 380nm Ed has obvious spatial variation characteristics at the MLD.Limited by the Ed incident and MLD depth,the 380nm Ed only appears high value in the tropics and subtropical regions.In time scale,the organisms grew further in March and April,resulting Ed in a low value of the MLD in the equatorial upwelling region.The research results provide important methods and data for remote sensing research of UV band ocean color in the global ocean,and are of great significance to the study of long-term biogeochemical cycles at the global scale.