Hyperspectral Remote Sensing of the Coastal Ocean: Adaptive Sampling and Forecasting of In situ Optical Properties

Phytoplankton species composition research is key to understanding phytoplankton ecological and biogeochemical functions. Hyperspectral optical sensor technology allows us to obtain detailed information about phytoplankton species composition. In the present study, a transfer learning method to inverse phytoplankton species composition using in situ hyperspectral remote sensing reflectance and hyperspectral satellite imagery was presented. By transferring the general knowledge learned from the first few layers of a deep neural network (DNN) trained by a general simulation dataset, and updating the last few layers with an in situ dataset, the requirement for large numbers of in situ samples for training the DNN to predict phytoplankton species composition in natural waters was lowered. This method was established from in situ datasets and validated with datasets collected in different ocean regions in China with considerable accuracy (R2 = 0.88, mean absolute percentage error (MAPE) = 26.08%). Application of the method to Hyperspectral Imager for the Coastal Ocean (HICO) imagery showed that spatial distributions of dominant phytoplankton species and associated compositions could be derived. These results indicated the feasibility of species composition inversion from hyperspectral remote sensing, highlighting the advantages of transfer learning algorithms, which can bring broader application prospects for phytoplankton species composition and phytoplankton functional type research.

Download Full-text

An Algorithm to Estimate Suspended Particulate Matter Concentrations and Associated Uncertainties from Remote Sensing Reflectance in Coastal Environments

Remote Sensing ◽

10.3390/rs12132172 ◽

2020 ◽

Vol 12 (13) ◽

pp. 2172 ◽

Cited By ~ 1

Author(s):

Juliana Tavora ◽

Emmanuel Boss ◽

David Doxaran ◽

Paul Hill

Keyword(s):

Remote Sensing ◽

Optical Properties ◽

Particulate Matter ◽

Suspended Particulate Matter ◽

Ocean Color ◽

Major Constituent ◽

Coastal Environments ◽

Suspended Particulate ◽

Spm Concentration

Suspended Particulate Matter (SPM) is a major constituent in coastal waters, involved in processes such as light attenuation, pollutant propagation, and waterways blockage. The spatial distribution of SPM is an indicator of deposition and erosion patterns in estuaries and coastal zones and a necessary input to estimate the material fluxes from the land through rivers to the sea. In-situ methods to estimate SPM provide limited spatial data in comparison to the coverage that can be obtained remotely. Ocean color remote sensing complements field measurements by providing estimates of the spatial distributions of surface SPM concentration in natural waters, with high spatial and temporal resolution. Existing methods to obtain SPM from remote sensing vary between purely empirical ones to those that are based on radiative transfer theory together with empirical inputs regarding the optical properties of SPM. Most algorithms use a single satellite band that is switched to other bands for different ranges of turbidity. The necessity to switch bands is due to the saturation of reflectance as SPM concentration increases. Here we propose a multi-band approach for SPM retrievals that also provides an estimate of uncertainty, where the latter is based on both uncertainties in reflectance and in the assumed optical properties of SPM. The approach proposed is general and can be applied to any ocean color sensor or in-situ radiometer system with red and near-infra-red bands. We apply it to six globally distributed in-situ datasets of spectral water reflectance and SPM measurements over a wide range of SPM concentrations collected in estuaries and coastal environments (the focus regions of our study). Results show good performance for SPM retrieval at all ranges of concentration. As with all algorithms, better performance may be achieved by constraining empirical assumptions to specific environments. To demonstrate the flexibility of the algorithm we apply it to a remote sensing scene from an environment with highly variable sediment concentrations.

Download Full-text