Remote Sensing Algorithm And Spatiotemporal Distribution Of Phytoplankton Size Classes In Coastal Waters

Phytoplankton size class(PSC),a measure of different phytoplankton functional and structural groups,is a key parameter to the understanding of many marine ecological and biogeochemical processes.Accurate knowledge of phytoplankton functional and structural classes and their distributions is thus critical to the understanding of many marine ecological and biogeochemical processes.The key problem is that the remote sensing inversion of PSC in coastal areas is scarce and pre-existing models still need more verification in turbid waters of China,where optical properties may be influenced by terrigenous discharge and nonphytoplankton water constituents.Here based on measurements from five cruises in turbid waters of Bohai Sea(BS),Yellow Sea(YS)and East China Sea(ECS)during 2015 and 2016,an empirical model and a semi-analytical model are developed and validated to estimate PSC in those seas.Finally,application of the PSC model to satellite data leads to spatial and temporal distribution product of phytoplankton size classes.The main conclusion is listed as follows:(1)Exponential functions are tuned to model the size-specific chlorophyll a concentrations with best remote-sensing reflectance band,Rrs(680),and total chlorophyll a as the model inputs based on the power function between size classes concentration and total chlorophyll a concentration.The R2 of verification samples are 0.957,0.914 and 0.529,respectively.The semianalytical model firstly separate the Cm from total chlorophyll a concentration with R =0.859,based on a two-component classification model.And then,an empirical model based on the relationship between Cp and Cn,p is built to retrieve Cn and Cp with R2=0.895 and 0.835,respectively.The semianalytical model has a better performance than empirical model where MAPE is 117.6%,32.34%,57.47%for empirical model,and 38.40%,34.37%,67.00%for semianalytical model,respectively,when they are applied to the same independent verification samples.(2)Long-term spatial and temporal distribution of phytoplankton size classes was completed using semi-analytical model.The high proportion of microphytoplankton mainly distributed along the seacoast where nano and picophytoplankton concentration is low.The center of BS,YS and ECS is mostly suitable for nano and picophytoplankton to grow.PSC in BS shows little change over time.However,the north YS has a strong microphytoplankton peak value in spring and a weaker peak in autumn.The center of the South YS and ECS only show a peak in spring which are relatively lower than other three areas.The highest peak is located in the Yangtze Estuary with a spring strong peak and a summer strong peak.The peak value distribution of nano and picophytoplankton are opposite to microphytoplankton and the high concentration usually appears in summer and autumn.(3)According to the correlation coefficient of stepwise regression,all areas can be divided into 3 different kinds.Here takes microphytoplankton for example.The first one is the areas where sea surface temperature(SST)shows negative and Kd shows positive relationship with microphytoplankton,controlled by SST and nutrients,like the center of BS,North YS and South YS.The second area shows positive relationship with SST,controlled by vertical mixing,wind and nutrients,like areas approaching Yangtze Estuary.The last kind is different where PSCs do not show a regular relationship for low correlation coefficient between nano and picophytoplankton and is controlled mainly by SST,like the center of ECS.