Abstract

Key Words

Introduction

Land Monitor is a coordinated initiative between five Western Australian state government organisations: Department of Agriculture and Food Western Australia (DAFWA), Department of Water (DoW), Department of Infrastructure and Planning (DPI), Department of Environment and Conservation (DEC), Landgate – Western Australian Land Information Authority (WALIA) the Water Corporation and the federal government organisation CSIRO’s Centre for Mathematical and Information Sciences (CMIS); which began in the mid 1990’s as a project focussed on delivering baseline salinity data, areas-at-risk of becoming saline, vegetation data for monitoring changes over time and elevation data. The project area covers 18 million hectares in the agricultural region of the south west of Western Australia.

Land Monitor products

Following each summer the catalogue of Landsat images that were captured over the south west region are reviewed and the most appropriate cloud free images are selected for processing. The CSIRO’s CMIS has developed techniques for the production of the vegetation products (Cacetta et al 2003).

Calibrated Landsat 5 Thematic Mapper (TM) Images

Landsat TM images are rectified to a base image and then radiometrically calibrated to like values using invariant targets to remove seasonal and atmospheric differences and sensor degradation. The spatial resolution of the Landsat data is 25 metres on the ground. Summer imagery provides the input into the processing of the vegetation products while imagery acquired in Spring forms the input for salinity products (Caccetta et al 2000).

The vegetation data products are produced from a time series of Landsat thematic mapper imagery. Currently the project has data for the years 1988, 1990, 1992, 1994, 1996, 1998, 2000, 2002 - 2008 inclusive. 2009 satellite images are currently being processed by CSIRO-CMIS and could be expected to be ready by late August.

Maps of perennial /woody vegetation cover and trends through time

Satellite images are analysed and processed to produce maps that show the current extent of perennial vegetation and changes in vegetation extent (e.g., unchanged, decrease - indicating removal of vegetation, increase - indicating regeneration/growth of vegetation) (Furby, et al 2008a). An accuracy assessment of the 1998 vegetation cover layer that was conducted in the Latham area of the wheatbelt found overall accuracy mapping to be 98.77% when omissions of sparse bush were included (Bryant and Wallace 2001).

Maps indicating vegetation history

The ‘Vegetation Index’ product is multi-channel image file; each band of the file contains the same spectral index derived from a series of dates (Furby et al 2008c). The index values are (band3+band5)/2, calculated from summer dates of calibrated TM data, starting from 1988. This data is the input used to create the ‘Trend Summary’. This data product can be used to produce colour image displays of change in perennial vegetation from selected dates, – similar to the ‘Vegetation History’ product. The data can also be used to plot time traces of index values for pixels or patches of remnant vegetation. These can be used in association with the various change map products to reveal and compare the patterns of response for areas of interest or concern.

The ‘Vegetation History’ product is produced from three dates of calibrated TM data (Furby et al 2008c). It provides visual displays which highlight areas of change in vegetation density over time. The product is a composite 3-band file comprised of a Vegetation Index (TM band 3 + TM band 5) / 2 from three summer dates. The Vegetation Index has been used as it provides a consistent contrast between perennial vegetation (dark) and most other cover types such as soil and dry annual pastures or stubble (bright).

This dataset provides a tool for the visual interpretation of perennial vegetation changes over a 21 year period, 1988 to 2009. No ground validation has been carried out and no attribution of the data is implied.

Vegetation Trends

The ‘Vegetation Trend’ product is multi-channel image file which contains summaries of change in vegetation over time (Furby et al 2008b). It is produced from twenty one summer dates of calibrated TM data, at approximately 2-yearly intervals from 1988 and annual intervals from 2002. The index used is the sum of TM band3 and band5; the summaries of change which are produced for each pixel are recorded as 6 bands, scaled to fit the 1-byte range of 0-255.

Band 1 : Mean index brightness over all dates
Band 2 : linear trend (slope) in brightness over time (scaled)
Band 3 : Quadratic trend (scaled)
Band 4 : standard deviation (sd) about mean (scaled)
Band 5 : residual sd after fitting linear trend (scaled)
Band 6 : residual sd after fitting linear and quadratic trends (scaled)

Areas which were never classified as having perennial vegetation cover over the period have been masked out and given a zero value in all bands; these will include some areas of very thin native vegetation. The bands can be displayed separately or together to summarise trends and stability of vegetation over time as measured by the index, and in particular to highlight areas with different patterns of change. One recommended display shows positive and negative linear trends in different colours; while other bands can be used to examine deviations from these trends.

The ‘Trendclass’ product is a simplified one-band summary of the linear trend over the whole period broken into five classes chosen from inspection only. The classes have been determined by numeric changes in the reflectance index without ground or ecological interpretation, and may be interpreted tentatively as follows:

• Major positive trend in vegetation density over the entire period
• Positive trend
• No major change
• Negative trend
• Major negative trend

Salinity Monitoring Product

The Land Monitor project has produced maps showing the changes in salt-affected land over the south-west agricultural area of Western Australia. The product indicates the extent of salinity in the years 1987-1992 and then how the extent has changed in 1995 / 2000.  The method involves the processing of historic Spring calibrated Landsat satellite imagery in conjunction with elevation data and ground information to identify areas of land that are of consistently low productivity (McFarlane, George and Caccetta 2004).

Valley Hazard Maps / Areas at Risk of Developing High Water Tables

By integrating Landsat imagery with landform information derived from height data, ground truthing and other existing mapped datasets, areas at risk of developing high water tables can be predicted (Dunne, Caccetta, and Beetson 2001).

Digital Elevation Model (DEM)

Through the Land Monitor project, high quality DEMs have been produced using automated ortho-photo techniques for the south-west of WA's agricultural area (Caccetta et al 2000).  The Land Monitor DEM is based on a 10metre grid and has a vertical accuracy of +/-1.5m.

Data distribution via Web Map Service

“We need a website where anyone interested can access information on the amount of vegetation at a regional and sub-regional scale, because vegetation is a good surrogate for biodiversity.” (O’Neill 2008)

The Land Monitor web map service provides three forms of login, two of which are available to the public. The “Public access” login requires no username or password and allows overview viewing of all of the datasets however it has a zoom restriction. The “Project Subscriber” login requires a username and password, which can be obtained by providing your contact details to the Land Monitor Technical Officer, allows unrestricted viewing and zoom of all data sets As of April 2009 the “Project Subscriber” service provides for non-commercial purposes, free limited (1.5MB) subset and download of all vegetation products (as .jpeg format) upon acceptance of disclaimer and license conditions. The “Project Partner” login is only available to employees of the project partner organizations.

Important features of the web map service include “Other” viewable data sets (e.g., Cadastre, Basins, catchments, pastoral boundaries, roads, water bodies, city/towns, shire boundaries and pipelines) are able to be overlaid the landsat imagery and vegetation products, turned on/off to enhance navigation and viewing.

The web map may be queried by selecting the “Information” button and clicking on a point or layer of interest. This will return information about all the viewable layers that intersect with the point clicked on.

The toolbar also contains print tools (including A4 and A3 portrait and landscape) and a bookmark tool which will allows a location of interest to be rapidly returned to and/or shared with other parties.

Larger subsets or different file formats of Land Monitor information products may be purchased by the public. Contact the Land Monitor Technical Officer for further information.

Land Monitor products applications

The calibrated landsat images and vegetation monitoring products have now been produced annually since 2002 are supplied to the project partners and made available to the public and university researchers. There has been significant use of the data by conservation and local land management and catchment groups with more than one hundred direct enquiries and requests for the data per year since 2006 from these groups, universities and the public.

In 2006 a survey of Land Monitor project partner users found the most used data products were the DEM, vegetation extent and change, calibrated Landsat scenes and salinity data sets.

The survey found that the 128 primary projects that used the Land Monitor data included salinity management (13%), catchment management (12%), vegetation change monitoring (11%), planning (11%), landscape management (9%), farm management (8%), conservation (6%), biodiversity (5%), ecological (4%), regulation enforcement (4%), re-vegetation (3%), bushland management (3%), fire management (3%) and translocation (2%) however a number of survey respondents reported that they distributed the data to an additional 55 other projects. The purposes that the data was most commonly used were: as background image, for further spatial analysis, to determine the current status of vegetation or salinity extent, as a baseline reference, for communication, to interpret temporal vegetation or salinity changes or trends, to produce a value-added product or to assess vegetation health and density. The types of value-adding that were reported as being performed on Land monitor data sets were primarily area analysis, vegetation condition assessment, vegetation land use models and vegetation classification types but also included value-adding that ranged from habitat distribution models, fragmentation index, intersection with soil-landscape mapping, vegetation decline statistics, interpretation of hazard or risk and resource protection.

Indeed vegetation monitoring products produced in map form have high communication value to land managers. Maps of vegetation change can prompt ground reviews and management actions as well as can help to quantify and communicate the effects of implemented management strategies.

Conclusion

The Land Monitor project delivers accurate, spatially detailed, timely and accessible vegetation monitoring information through time (1988-2009) about the extent and condition of perennial vegetation in Western Australia’s south west. The vegetation products are annually delivered to the partner agencies and made available to the public via the web map service and by direct request to the Land Monitor Technical officer.

References

Bryant GS and Wallace JF (2001).Accuracy assessment of the Land Monitor vegetation cover layer, July 2001. www.landmonitor.wa.gov.au. Accessed 17 July 2009.

Dunne R, Caccetta P, and Beetson B (2001). Prediction of Areas at risk of salinity: The agricultural area of Western Australia Report Number: CMIS 01/183. http://www.landmonitor.wa.gov.au/reports/salinity_risk/summary_report.pdf Accessed 17 July 2009.

Caccetta PA, Bryant G, Campbell NA, Chia J, Furby SL, Kiiveri HT, Richards G, Wallace JF and Wu X (2003). Notes on Mapping and Monitoring Forest Change in Australia Using Remote Sensing and Other Data, 30th ISRSE Honolulu, Hawaii, November 10-14. www.cmis.csiro.au/rsm/publications.htm Accessed 17 July 2009.

Caccetta PA, Campbell NA, Evans FH, Furby SL, Kiiveri HT and Wallace, J. F. (2000), Mapping and monitoring land use and condition change in the South-West of Western Australia using remote sensing and other data, (2000), Proceedings of the Europa 2000 Conference, Barcelona. www.cmis.csiro.au/rsm/publications.htm Accessed 17 July 2009.

Furby S, Zhu M, Wu X and Wallace JF (2008a). Vegetation Extent and Change 1988-2008 South West Agricultural region of Western Australia. http://www.landmonitor.wa.gov.au/reports/landmon_II/LM2008_VegMask_1988_2008.pdf Accessed 17 July 2009.

Furby S, Zhu M, Wu X and Wallace JF (2008b).Vegetation trend 1990-2008 South West Agricultural region of Western Australia. http://www.landmonitor.wa.gov.au/reports/landmon_II/LM2008_VegTrend_1990_2008.pdf Accessed 17 July 2009.

Furby S, Zhu M, Wu X and Wallace JF (2008c).Vegetation History and Index Sequence 1988-2008 South West Agricultural region of Western Australia. http://www.landmonitor.wa.gov.au/reports/landmon_II/LM2008_IndexSeq_History_1988_2008.pdf Accessed 17 July 2009.

McFarlane DJ, Georg, RJ and Caccetta PA (2004), The Extent and Potential Area of Salt-affected Land in Western Australia Estimated Using Remote Sensing and Digital Terrain Models, Engineering Salinity Solutions, 9-12 November, Perth, Western Australia. www.cmis.csiro.au/rsm/publications.htm Accessed 17 July 2009.

O’Neill, G. (2008) Jump starting environmental monitoring. ECOS magazine. http://www.ecosmagazine.com/nid/206/issue/4132.htm. Accessed 17 July 2009.