INVESTIGATING SPATIAL AND TEMPORAL DYNAMICS OF THE LEEUWIN CURRENT FROM MODIS SEA SURFACE TEMPERATURE IMAGES USING OBJECT-BASED TECHNIQUES  

Authors / CoAuthors

Huang, Z.

Abstract

The dominant ocean current off the Western Australian (WA) coast is the Leeuwin Current (LC) [1]. It is a warm, poleward flowing surface current up to 300 metres in depth and exhibits significant seasonal differences in intensity, ranging from strong during the austral winter and weak during the austral summer [2-3]. At a regional scale, the LC is significant because it directly influences the temperature and nutrient dynamics of the WA ocean ecosystem [3]. As a result, it has been shown that the LC affects the production of phytoplankton [4-5], the recruitment of western rock lobster [6], and the distribution of fishes and algae [7]. The LC can be observed from Sea Surface Temperature (SST) satellite-derived images. However, delineating an accurate map showing the extent and spatial structure of the LC from a SST image remains a challenge. And given the large area covered by the LC, an automatic approach is desirable. This study aims to test an object-based image processing technique from time-series MODIS SST data for the above purpose. SeaDas image processing software was used to process MODIS images from daily raw data to Level 3 products. The monthly SST4 layers between June 2009 and May 2010 were the inputs for this study. The SST layer shown in Figure 1a clearly indicates a warmer (than off-shore) southward flowing current (LC) that extends from Exmouth, passing Cape Leeuwin, into the Great Australian Bight. Previously, relative temperature differences have been employed to identify LC structure from SST images [8]. An off-shore SST profile crossing the LC (Figures 1a & b) shows that the LC, indicated by warmer temperatures, occupies a zone approximately between 10 km and 90 km from the coast, with a core current between 35 km and 70 km. This study utilised two characteristics of MODIS-derived sea surface temperatures to identify the extents of the LC. The first characteristic is that the LC is warmer than surrounding waters. The second characteristic is the connectivity between the core LC current and the eddies. According to the first characteristic, the SST images were treated as elevation surfaces where the LC occupies slope and ridge positions. Topographic Position Index (TPI) was then derived from these SST layers to identify topographic positions [9]. As shown in Figure 1c, the LC approximately corresponds with areas of large positive TPI values. In the next step, the multi-resolution algorithm in eCognition Developer was employed to segment the SST and TPI layers of each month into objects. The objects were classified into a pseudo LC class if their mean TPI values are greater than 0.25 of the global standard deviation value. The second characteristic of the LC was then used to remove false positive objects. To do that, a small number of objects at known LC locations were selected as 'seeds'. In a looping process, any objects that connect with these seeds were classified as true LC class. The extents of the LC for the 12 months analysed here areshown in Figure 2. The LC during austral winter is clearly stronger (e.g., larger in extent) than during austral summer, which confirms the findings of other studies [2-3]. The LC during the summer time is patchier, which required more seeds (8-14) than during the winter time (less than 5 seeds). The core summer current is also slightly further away from the coast. In addition, eddies are clearly visible in most months. In summary, the proposed object-based approach was semi-automatic and effective in delineating the extents of the LC although there is a degree of subjectivity in the selection of accurate seeds. The weak summer current, however poses some difficulty for the approach and future work is aimed at improving the modelling accuracy.

Product Type

nonGeographicDataset

eCat Id

75159

Contact for the resource

Keywords

• External PublicationAbstract

( Theme )

• remote sensing

( Theme )

• marine

( Theme )

• NERP

Australian and New Zealand Standard Research Classification (ANZSRC)

• Oceanography

• Published_Internal

Publication Date

2013-01-01T00:00:00

Creation Date

Security Constraints

Australian Government Security ClassificationSystem    

Classification - unclassified

Legal Constraints

Creative Commons Attribution 4.0 International Licence    

Access - restricted

Use - license

Status

Purpose

Maintenance Information

notPlanned

Topic Category

oceans

Series Information

Lineage

The abstract was written by Zhi Huang for the IGARSS conference to be held in Melbourne, July 2013.

Parent Information

Extents

Reference System

Spatial Resolution

Service Information

Associations

Downloads and Links

Source Information

NERP Marine Biodiversity Hub