H41E-1379
Water turbidity estimation from airborne hyperspectral imagery and full waveform bathymetric LiDAR

Thursday, 17 December 2015
Poster Hall (Moscone South)
Zhigang Pan1, Craig L Glennie2,3, Juan Carlos Fernandez-Diaz2 and National Center for Airborne Laser Mapping, (1)University of Houston, Houston, TX, United States, (2)National Center for Airborne Laser Mapping, Houston, TX, United States, (3)University of Houston, Geosensing System Engineering and Science, Houston, TX, United States
Abstract:
The spatial and temporal variations in water turbidity are of great interest for the study of fluvial and coastal environments; and for predicting the performance of remote sensing systems that are used to map these. Conventional water turbidity estimates from remote sensing observations have normally been derived using near infrared reflectance. We have investigated the potential of determining water turbidity from additional remote sensing sources, namely airborne hyperspectral imagery and single wavelength bathymetric LiDAR (Light Detection and Ranging).

The confluence area of the Blue and Colorado River, CO was utilized as a study area to investigate the capabilities of both airborne bathymetric LiDAR and hyperspectral imagery for water turbidity estimation. Discrete and full waveform bathymetric data were collected using Optech’s Gemini (1064 nm) and Aquarius (532 nm) LiDAR sensors. Hyperspectral imagery (1.2 m pixel resolution and 72 spectral bands) was acquired using an ITRES CASI-1500 imaging system. As an independent reference, measurements of turbidity were collected concurrent with the airborne remote sensing acquisitions, using a WET Labs EcoTriplet deployed from a kayak and turbidity was then derived from the measured backscatter.

The bathymetric full waveform dataset contains a discretized sample of the full backscatter of water column and benthic layer. Therefore, the full waveform records encapsulate the water column characteristics of turbidity. A nonparametric support vector regression method is utilized to estimate water turbidity from both hyperspectral imagery and voxelized full waveform LiDAR returns, both individually and as a fused dataset. Results of all the evaluations will be presented, showing an initial turbidity prediction accuracy of approximately 1.0 NTU. We will also discuss our future strategy for enhanced fusion of the full waveform LiDAR and hyperspectral imagery for improved turbidity estimation.

<< Previous Abstract | Next Abstract >>