Rapid Eutrophication Assessment Based On Support Vector Machine(SVM)

The phenomenon of global eutrophication occurs frequently, shelf marginal seanear our country is the world’s largest,and eutrophication is serious as result ofdevelopment of extensive economic.Our country becomes one of the most seriouseutrophication countries. The first generation of evaluation model is widelyused.However, parameters of the first generation of eutrophication evaluation modelhas long test period and big workload.At the same time,it needs professionalequipment and personnel.So,it cannot make the real-time monitoring and evaluationof coastal waters realize. The second generation of eutrophication evaluation model isnot suitable for offshore eutrophication evaluation in our country.And it needs a largenumber of long-term survey data. In view of this situation, a real-time, rapidevaluation of eutrophication model is needed.Evaluation parameters inclues Excitation-emission matrix spectroscopy andUltraviolet absorption of CDOM with quick monitoring eutrophication factor on thesite(DO,Chla, turbidity et.al). Fast evaluation model of offshore eutrophication basedon support vector machine (SVM) is built.It is successfully applied in East China Sea.The main research conclusions are as follows.(1)The model of NQI is used to evaluate the eutrophication of Bo hai,The YellowSea,The East China Sea.The result shows eutrophication of East China Sea is serious,followed by Bo hai,The Yellow Sea. The main contribution factor of eutrophication isanalyzed.It shows DIN,COD is the main contribution factor of Bo hai and East ChinaSea, The Yellow Sea is DIN,COD,DIP.Relationship between the main contributionfactors and Excitation-emission matrix spectroscopy and Ultravioletabsorption(220nm,254nm,270nm,350nm) of CDOM is analyzed for the built of fastevaluation model.(2)Based on support vector machine, Excitation-emission matrix spectroscopy(C1,C2,C3,C4) and Ultraviolet absorption(220nm,254nm,270nm,350nm)of CDOM instead of DIN,DIP,COD of NQI. Fast evaluation model of offshoreeutrophication is realized by LIBSVM in the MATLAB platform.Compareed withresult of NQI,E, fuzzy evaluation, classification accuracy of Cross validation isapproximately90%(NQI, fuzzy evaluation),70%(E). Classification accuracy of Crossvalidation doesn’t increase with the increase of the train set. Fast evaluation model ofoffshore eutrophication is built by NQI of163training samples. Discriminant functionis established.(3) Fast evaluation model is successfully applied in East China Sea.Theevaluation results is similar with result of fuzzy evaluation and literature.Andabnormal station is analyzed. From what has been discussed above,the evaluationresults of Fast evaluation model is reasonable.