Estimation And Modeling Forecast Of Surface Chlorophyll A Concentration In The Seas Around Tropical Marine Ranching

Marine ranching is an effective means of protecting fishery resources and ecosystem,its construction and management activities and natural process mutually influence each other.To better manage and evaluate the effects of marine ranching,requires monitoring of necessary indicators.Chlorophyll a(Chl a),an important parameter of marine ecosystem,represents marine primary productivity and biomass adequately.Therefore,it is of great significance to study the temporal and spatial distribution characteristics and driving factors of Chl a and concentration forecast works.Ocean color remote sensing is among the most commonly used ways of estimating the concentration of sea surface Chl a,it enables efficient and large-area continuous spatiotemporal observation of Chl a.In order to develop a local empirical algorithm and to explore the spatiotemporal distribution characteristics and driving factors of Chl a and concentration forecast,in this work,we did the following:(1)A local band-ratio algorithm was developed based on in situ and satellite data in the seas around marine ranching of Yazhou bay,coefficient of determination reached0.74,and applied to produce Chl a product further needed.It was found that Chl a was highly spatial autocorrelation and high-value clustering by spatial analysis.Chl a peaked in winter at high concentration but in summer at low concentration on local seasonal patterns,high in Dongluo Bay but low in Yazhou Bay.(2)The correlation coefficient of DN values of HY-1C/D CZI and Landsat 8 OLI is more than 0.6.Compared to MODIS Aqua standard Chl a product,the root mean square error between sensors is about 0.1 ug/L.Hot spot analysis showed that the spatial trends of them are basically the same,CZI is locally closer to the OLI.We present a metrics Synchronized Trend Rate to evaluate the consistency of time series trend.In 2020,the STR of OLI and CZI Chl a time series is 0.55.(3)In order to better describe spatial distribution characteristics of Chl a,this paper uses three percentiles to create a combined mean,which is more stationary than the mean.After the stationarity and pure randomness testing of the combined mean series of Chl a concentration passed,the ARIMA model was chosen.The ARIMA(2,0,2)(1,1,1)[4] model fit the combined mean time series of Chl a data well,the point forecast error was less than 0.1 ug/L in four quarters of 2021.