Satellite Derived Bathymetry Of Reef Water Based On Zoning By Depth Invariant Index

As an important topographic element,water depth is the basic datum for ship navigation,port and marine engineering construction,and planning for islands and coastal zones.Water depth information is of wide range of collection and many application requirements.In-situ bathymetry methods are deficient in synchronization,economy,periodicity,flexibility and macroscopicity.In addition,as for areas with harsh environmental conditions or difficult to access due to disputes over rights,on-site water depth measurements are ineffective.Although the water depth derived by remote sensing has a relatively lower accuracy,it has the advantages of large coverage,low cost and short cycle.It is suitable for the comprehensive investigation and dynamic monitoring of topography in shallow waters,and makes up for the lack of field surveys to a certain extent.Relevant scholars have also conducted a lot of research and proposed various water depth inversion models.Sediment is one of the main factors affecting the accuracy of satellite derived bathymetry.How to effectively perform water depth inversion under different sediment conditions is an issue to be explored.At present,most of the relevant domestic and international water depth inversion algorithms are based on clear waters,and most of them have not fully considered the influence of complex substrates on optical signals.In fact,the ocean floor is complex and changeable.Partitioning different substrates and further conducting water depth inversion in different types of areas may reduce the influence of substrates on the inversion and improve the accuracy of inversion.At the same time,remote sensing images used for water depth inversion often have noises such as solar flares,white hats,and thin clouds.These noises will obviously affect the inversion results.In this paper,different noise filtering methods are used to remove image noise,and the water depth inversion ability after denoising is compared and analyzed.1.A comparative analysis of three different water depth models is carried out,which include log-linear model,log-transformation ratio model proposed by Stumpf and improved log-transformation ratio model.Ping Zhou Island in Hong Kong is taken as a research area.The results of inversion experiments show that the log-linear model of B1,B2,B3,and B4(blue,green,red,and near infrared)4-band combination has the highest water depth inversion accuracy.The mean absolute error(MAE)and mean relative error(MRE)of validation points are 1.63m and 12.67%respectively,and the determination coefficient R2 is 0.80.Results of different water depth intervals are analyzed.The results of the three models are good at a water depth of 10-15m with the smallest MAE and MRE of 1.09m and 8.99%,respectively.The inversion error of less than 10m and more than 15m is greater than that of 10-15m,indicating that the above model is more suitable for medium shallow water.2.A depth-invariant index model based on dual-band radiance is proposed and compared with the depth-invariant index proposed by Lyzenga.The results of East Island and North Island of Xisha Islands show that both models obtain good classification results and reduce water depth inversion errors.The proposed model in this paper is applied in the classification of North Island and East Island,with the accuracy improved by 2.9%and 0.38%respectively while compared with the Lyzenga model.The MAE of inversion based on classification by proposed model is 0.10 m lower than that of the unclassified.For area where the substrate type is coral reef,the inversion accuracy is significantly improved compared to that of the unclassified type with the maximum error reduced by 0.20m.While the inversion accuracy of the substrate type of coral sand is slightly higher than that of the unclassified type.3.Different noise filtering methods are applied to the images for water depth inversion.The study areas are Wuzhizhou Island of Hainan and Kachel Island in Andaman Sea.Through comparison and analysis,it is found that the water depth inversion results are different under different noise filtering methods.Compared with Retinex algorithm,Dct algorithm,median filter and Gaussian filter,the water depth inversion result obtained by Sym4 wavelet algorithm is better.Among them,for depth inversion result with the filtering of Sym4 of Wuzhizhou Island,the mean absolute error is 1.49m,and the mean relative error is 16.32%.While for that of Kachel Island,the mean absolute error of 1.52m and mean relative error of 19.03%is achieved.