The Establishment And Application Of Semi-analysis Model On Suspended Sediment Remote Sensing Inversion In Caofeidian Offshore Water

Coastal water is a vital place that combines the exchanging of material and energybetween seas and continents. Marine ecological environment problems are increasinglyclear, such as water degradation, marine ecosystem imbalance, frequency of red tidedisasters, etc. Problems of coastal water resource and environment are receiving more andmore attention. Suspended sediment is an important index of water environment evaluation.Recently, remote sensing has become a significant approach of water suspended sedimentmonitoring. The remote sensing data model of suspended sediment monitoring mainlycontains empirical statistical model, semi-experience and semi-analysis model, andtheoretical analysis model. The semi-experience and semi-analysis model which have acertain theoretical basis is between the other two models, and it has being fully studied andapplied in recent years. Compared with traditional experience statistical pattern, the halfanalysis pattern has improved in both methods and precision of suspended sedimentinversion. It has realized the transformation of black box model to white box model inremote sensing inversion of suspended sediment.This paper takes the ALOS remote sensing data of2009as the source, constructsexperience pattern-based remote sensing models based on the analysis of measured spectraldata in Caofeidian area; The inherent optical quantity of water was calculated with theradiative transfer theory;Combined with the water biological optical theory model, the paperconstructed half analysis-based models of remote sensing inversion by the correlationanalysis of suspended sediment inherent optical index and concentration. The applicationand comparison results of experience model and half-analysis model show: for thesuspended sediment remote sensing inversion of Caofeidian coastal water, the half–analysismodel is superior than the experience model; the average relative error is30.83%, and theRMSE is4.22mg/l.Main research contents:(1)Study of Spectral characteristic of Caofeidian coastal water; measure the spectral data of each site; calculate remote sensing reflectance of waterappearance; make suspended sediment spectral reflectance graphs which changes with theconcentration and analysis its characteristics.(2) Study of empirical statistical inversion model of coastal water suspended sediment inCaofeidian;According to the measured spectrum data and surface suspended sedimentconcentration, this paper constructed remote sensing empirical statistical arithmetic withstatistical regression method on the basis of ALOS RS data.(3)Research of half-experience and half analysis RS inversion model of suspendedsediment concentration in Caofeidian coastal water;Impute the inherent spectrum index of water suspended particles with fieldmeasurement data, particle size data, and surface suspended sediment concentrationmeasurement data. Set up RS inversion half-analysis model on the basis of absorptioncoefficient and back scattering coefficient.(4) Contrastive analysis of empirical model and half-analysis model of suspendedsediment concentration in Caofeidian coastal water.With the inversion precision comparison of half-analysis model and empirical model,some errors were avoided effectively and the inversion precision RS were dramaticallyimproved.Achievement and conclusion:1) The spectral reflection curve is different with the change of suspended sedimentconcentration. The spectral reflection curve of suspended sediment concentration water hastwo obvious peaks, the first peak at the band of550-600nm, the second peak at about800nm. According the spectral characteristic of surface suspended sediment concentration,and the relativity analysis of observed spectral reflectance and observed suspendedsediment concentration, band4, band3/band1, Band3/Band2and Band4/Band1are thesensitive bands to the suspended sediment concentration;2) With measured data, this paper built empirical statistical models of suspended matterconcentration and reflectance of the corresponding remote sensing band. Results show that the correlation coefficient R2of quadratic polynomial of band3/band2is the highest, whichis0.977. Evaluate the inversion precision according to the inversion results of empiricalmodel, the average absolute error is5.34mg/l, the average relative error is36.77%, theRMSE is6.02mg/l.3) Build empirical statistical model for band3of0.998with the correlation coefficient R2of inherent optical quantity and suspended sediment concentration. Combine the twomodels and build suspended sediment inversion half-analysis model: SSC=30.829-494.891*Rrs+2487.318*Rrs2. Evaluate the inversion precision according to the inversionresults, the average absolute error is3.61mg/l, the average relative error is30.83%, theRMSE is4.22mg/l.4) With the half analysis model, the RMSE decreased from6.02mg/l to4.22mg/l, someerrors were avoided effectively. Relatively speaking, errors of the half-analysis model aresmaller and the precision of inversion were dramatically improved.