Authors / CoAuthors
Huang, Z. | Siwabessy, P.J.W.
Abstract
In the past two decades, multibeam sonar systems have become the preferred seabed mapping tool. Many users have assumed that multibeam bathymetry data is highly accurate in spatial position. In reality, both vertical and horizontal uncertainties exist in every data point. These uncertainties are often represented as one single measure of Total Propagated Uncertainty (TPU). TPU is important to understand because it affects the quality of products generated from multibeam bathymetry data. To account for the magnitude and spatial distribution of this influence, an objective uncertainty analysis is required. Randomisation is the key process in such an uncertainty analysis. This study compared two randomisation methods, restricted spatial randomness (RSR) and complete spatial randomness (CSR), in an uncertainty analysis of a slope gradient dataset derived from multibeam bathymetry data. CSR regards data error in every grid cell as independent and assumes that the data error varies within a known statistical distribution without any neighbourhood effect. RSR assumes spatial structure and thus spatial auto-correlation in the data. We present a case study from a survey of the Oceanic Shoals Commonwealth Marine Reserve in the Timor Sea, conducted in 2012 by the Marine Biodiversity Hub through the Australian Government National Environmental Research Program. The survey area is characterised by steep-sided carbonate banks and terraces with abrupt breaks in slope of limited spatial extent. As habitats, the carbonate banks and terraces are important because they provide hardground for diverse epibenthic assemblages of sponges and corals, with their steep sides marking the environmental transition to deeper water, soft sediment habitats. In this analysis, the data errors in the multibeam bathymetry data were assumed to follow a Gaussian distribution with a mean of zero and a standard deviation represented by the TPU. The CSR and RSR methods were each implemented using a Monte Carlo procedure with 500 iterations. After about 300 iterations, the Monte Carlo procedure converged for both methods. Results for the study area are compared against pre-processed slope data (Figure 1a). The averaged slope gradient from the CSR method is 4.5 degree greater than the original slope layer, whereas for the RSR method this value is 0.03 degree. Moreover, the slope layer from the CSR method resolves noticeably less detail than the original slope layer and is an over-simplification of the true bathymetry (Figure 1b). In contrast, the RSR method maintained the spatial pattern and detail observed in the original slope layer (Figure 1c). This study demonstrates that although the uncertainty in multibeam bathymetry data should not be ignored, its impact on the subsequent derivative analysis may be limited. The selection of appropriate randomisation method is important for the uncertainty analysis. When the data errors exhibit spatial structure, we recommend using the RSR method.
Product Type
nonGeographicDataset
eCat Id
79105
Contact for the resource
Custodian
Owner
Custodian
Cnr Jerrabomberra Ave and Hindmarsh Dr GPO Box 378
Canberra
ACT
2601
Australia
Keywords
-
- External PublicationAbstract
- ( Theme )
-
- bathymetry
- ( Theme )
-
- marine environmental baselines
- ( Theme )
-
- marine
- Australian and New Zealand Standard Research Classification (ANZSRC)
-
- Marine Geoscience
-
- Published_Internal
Publication Date
2014-01-01T00:00:00
Creation Date
Security Constraints
Legal Constraints
Status
Purpose
Maintenance Information
notPlanned
Topic Category
oceans
Series Information
Lineage
This abstract is part of NERP Marine Biodiversity Hub research.
Parent Information
Extents
Reference System
Spatial Resolution
Service Information
Associations
Downloads and Links
Source Information
NERP Marine Biodiversity Hub