Kakadu National Park Landscape Symposia Series 2007–2009

səhifə	10/16
tarix	25.06.2016
ölçüsü	5.42 Mb.

1 ... 6 7 8 9 10 11 12 13 ... 16

References

Boyden J, Bartolo R, Walden D& Bayliss P 2010. The role of remote sensing for monitoring wetland weeds in Kakadu National Park. In Kakadu National Park Landscape Symposia Series 2007–2009. Symposium 2: Weeds management, ed Winderlich S, 27–28 November 2007, Jabiru Field Station, Supervising Scientist Division, Kakadu National Park. Internal Report 565, January, Supervising Scientist, Darwin, 32–39.

Introductory Weed Management Manual 2008. Cooperative Research Centre for weed management, Department of the Environment and Heritage, Australian Government Canberra. http://www.weedscrc.org.au/documents/manual.pdf

McNaught I, Thackway R, Brown L & Parsons M 2006. A field manual for surveying and mapping nationally significant weeds. Bureau of Rural Sciences, Canberra.

Walden & Gardener 2008. Invasive species: Weed management in Kakadu National Park. In Kakadu National Park Landscape Symposia Series 2007–2009. Symposium 1: Landscape Change Overview, 17–18 April 2007, South Alligator Inn, Kakadu National Park. Internal Report 532, April, Supervising Scientist, Darwin, 66–83.

Weed Management Branch. NT Department of Natural Resources, Environment and the Arts 2007. Guidelines for weed data collection in the Northern Territory. Northern Territory Government of Australia, Darwin.

6 The role of remote sensing for the monitoring and control of wetland weeds in Kakadu National Park

J Boyden^⁹, R Bartolo¹, D Walden¹& P Bayliss^¹⁰

6.1 Introduction

Cost-effective and strategic weed management is achieved though timely detection of new incursions and monitoring and controlling the extent of known weed infestations. Management priorities are defined by the distribution of weeds in context to the native vegetation (assets) vulnerable to invasion in conservation areas, where native vegetation communities (or the habitats for biodiversity conservation) can be considered as the mapping units for landscape management. Implicit is the need for accurate current (and historical) maps of weeds in context to the distribution of these assets. However, in remote and inaccessible landscapes such as Kakadu National Park, vegetation maps have tended to be incomplete or out of date, resulting in uncertainty in setting management priorities in remote conservation areas (Harvey & Hill 2001). We contend that remotely sensed data have the potential to address these information gaps and to assist in developing an integrated, landscape-scale, weed monitoring and control policy.

A general framework for the spatial risk assessment of weeds and the role of remote sensing (RS) in monitoring weeds on wetlands is provided in Figure 1. Captured over appropriate spatial extents and spatio-temporal scales, RS allows repeated mapping of vegetation, and allows managers to monitor changes in the distribution of weeds in relation to the current and historical extent of native vegetation. Repeated mapping of weeds in relation to native vegetation is also useful for developing spatial models to predict the likely occurrence of future weed invasions. Predictive maps are among the decision support tools that may be used in operational planning to prioritise control of weeds in key areas. A monitoring program that incorporates remotely sensed data has potential also for assessing the effectiveness of weed control strategies in preserving the integrity of key habitats.

RS has been applied to map weeds and native vegetation condition on wetlands, ranging in scales from aerial photography to coarser scale satellite products such as Landsat (Menges et al 1996, Phinn 1996, Phinn et al 1999, Tuxen et al 2007). In general finer spatial scale (<2.5 m) imagery produce higher accuracy and are better able to detect small weed patches, while coarser scale (>25 m pixel resolution) products can resolve only large weed infestations reliably (Catt & Thirarongnarong 1992). This paper focuses on the application of Very High Resolution (VHR) satellite imagery such as Quickbird, which offer a more cost effective solution to traditional aerial photography techniques.

Figure 1 General framework for monitoring and weed risk assessment, indicating the role of RS in providing timely information on weed distribution and native habitat displacement
(adapted from USEPA 1998)

6.2 Very high resolution (VHR) satellite imagery for mapping weeds

Satellite sensors are now capable of characterising vegetation distribution at comparatively very high spatial resolutions ranging from 0.6 m to 2.5 m pixels with several advantages over conventional aerial photography. These include: lower cost with increased spatial accuracy; a greater field of view, producing a image with reduced spectral noise^¹¹ (Tuxen et al 2007); and greater spectral range (visible to near infrared). Revisit time for VHR QuickBird imagery is 10 days in the NT Top End region, allowing flexibility in capture time.^¹²

Users of VHR maps can navigate to locations confidently using GPS technologies as spatial accuracy of VHR imagery is high (in the order of ±15m or ± 2.5m if surveyed ground control targets are also deployed). Importantly, VHR makes the link between ground validation information (including photos) and satellite imagery easier by allowing for clear recognition of small objects in the landscape, such as weed patches or individual trees (Phinn et al 1999, Tuxen et al 2007).

In general terms VHR can delineate fine-scale detail of habitat distribution in highly variable wetland ecosystems, and is complementary to field sampling by providing continuous coverage in expansive and often inaccessible wetland environments (Phinn et al 1999, Everitt et al 2004, Everitt et al 2007). Other authors have shown that VHR can also provide a good surrogate to ground-based measurements of native vegetation condition and is generally more cost-effective and a less invasive solution for assessing condition across extensive areas (Phinn et al 1999, Johansen et al 2007). VHR has advantages over systematic visual surveys of weeds by air or ground. Maps produced by aerial surveys are of coarse resolution and do not identify the specific locations of weed patches. Grounds surveys tend not to be practical over extensive and often inaccessible areas and may in fact exacerbate the dispersal of weeds into more pristine areas.

In Kakadu National Park (KNP) pilot studies conducted on the Magela floodplain and at Boggy Plain (lower South Alligator Floodplain) demonstrated the utility of VHR to map para grass (Urochloa mutica) and key wetland habitats as illustrated in Figures 2a & b (Boyden et al 2007, Boyden & Bartolo 2008). Accuracy of para grass detection using computer based map classification was in the order of 95% on the Magela floodplain.

Figure 2b Major native habitats mapped in May 2003 at Boggy Plain, South Alligator River, Kakadu National Park (Digital Globe® all rights reserved). For best interpretation of Figures 2a&b, please see online version of this report on the SSD website.

The utility of VHR is improved through proactive use of imagery by KNP field staff. Weed managers trained in aerial and ground survey techniques are an important resource as they can also be adept at VHR image interpretation. In this context it is also worth noting that KNP rangers have successfully used VHR imagery sourced through Google Earth to locate new patches of para grass in the South Alligator region of KNP (Atkins pers comm 2008).