A Combined Machine Learning and Residual Analysis Approach for Improved Retrieval of Shallow Bathymetry from Hyperspectral Imagery and Sparse Ground Truth Data

Evangelos Alevizos

doi:10.3390/rs12213489

Abstract

Mapping shallow bathymetry by means of optical remote sensing has been a challenging task of growing interest in recent years. Particularly, many studies exploit earlier empirical models together with the latest multispectral satellite imagery (e.g., Sentinel 2, Landsat 8). However, in these studies, the accuracy of resulting bathymetry is (a) limited for deeper waters (>15 m) and/or (b) is being influenced by seafloor type albedo. This study explores further the capabilities of hyperspectral satellite imagery (Hyperion), which provides several spectral bands in the visible spectrum, along with existing reference bathymetry. Bathymetry predictors are created by applying the semi-empirical approach of band ratios on hyperspectral imagery. Then, these predictors are fed to machine learning regression algorithms for predicting bathymetry. Algorithm performance is being further compared to bathymetry predictions from multiple linear regression analysis. Following the initial predictions, the residual bathymetry values are interpolated by applying the Ordinary Kriging method. Then, the predicted bathymetry from all three algorithms along with their associated residual grids is used as predictors at a second processing stage. Validation results show that by using a second stage of processing, the root-mean-square error values of predicted bathymetry is being improved by ≈1 m even for deeper water (up to 25 m). It is suggested that this approach is suitable for (a) contributing wide-scale, high-resolution shallow bathymetry toward the goals of the Seabed 2030 program and (b) as a coarse resolution alternative to effort-consuming single-beam sonar or costly airborne bathymetric laser surveying.

Highlights

Studying the bathymetry of shallow seafloor at coastal areas with optically transparent waters is a developing sector in remote sensing that in recent years has received more and more attention
There has been an increasing use of multispectral imagery along with established approaches for bathymetry retrieval. [10] combined multispectral satellite imagery with optical modeling for estimating bathymetry at a wider depth range up to 30 m of water depth with their results showing a root-mean-square error (RMSE) between 1 and 2 m. [11] proposed a method for deriving bathymetry first by clustering multispectral satellite imagery and by applying empirical models [6,7] and a support vector machine (SVM) algorithm to each cluster separately
It has to be noted that the multiple linear regression analysis (MLRA) utilized a small fraction out of the total number of band ratios provided compared to the kNN and RF algorithms

Summary

Introduction

Studying the bathymetry of shallow seafloor at coastal areas with optically transparent waters is a developing sector in remote sensing that in recent years has received more and more attention. [11] proposed a method for deriving bathymetry first by clustering multispectral satellite imagery and by applying empirical models [6,7] and a support vector machine (SVM) algorithm to each cluster separately This approach minimized significantly the influence of bottom variability in deriving bathymetry resulting in mean absolute errors between 0.2 and 0.5 m. [13] introduced a novel method that is based on the Google Earth Engine cloud computing platform for deriving bathymetry using Sentinel imagery They implemented four empirical models [7,14,15,16], and they obtained bathymetry predictions with RMSE values between 1 and 3 m for various nearshore locations in Greece. This limitation is mainly due to the restricted spectral resolution of multispectral sensors [17], which usually provide 3–4 spectral bands in the visible spectrum

Methods

Results

Discussion

Conclusion