UTILITY OF VHR REMOTE SENSING DATA FOR LANDSCAPE SCALE ASSESSMENT OF THE ENVIRONMENTAL WEED PARA GRASS UROCHLOA MVTICA, (FORSSK), NGUYEN ON A TROPICAL FLOODPLAIN

摘要

The conservation and management of tropical wetlands requires a strategic approach in order to manage multiple threats to multiple assets under budgetary constraints, and such an approach is embedded in ecological risk assessm ent frameworks. Aquatic weeds are currently the gre atest widespread threat to our relatively pristine wetlands in the Northern Territory (NT) of Australia, and a spatially-explicit ecological risk assessm ent framework that incorporates knowledge on the distribution and abundance of aquatic weeds in relation to their ac tual or potential impacts on natural and cultural conservation values, is essential for cost -effective long-termmanagement. However, whilst m apping and mon itoring weed distr ibution is c ritical f or undertaking spatially-based ecol ogical risk assessm ents and developing long-term control strategies, there is incomplete information on the distribution of most aquatic weeds, their rates of invasion and im pacts in speci fie floodplain habitats. Remote sensing has the potential to address information gaps associated with m anaging weeds on typically remote and inaccessible wetlands across northern Australia. W e report init ial results of a pilot study investigating the utility of very high reso lution remote sensing (VHR) to map and m onitor the d istribution and abundance of para grass (Urochloa mutica) on the Magela Creek floodplain in Kakadu National Park, NT, a serious environm ental weed that has a wide distribution aero ss northern Australia. QuickBird? multispectral im agery was captured in conjunction with f ield validation surv ey data for a 64 km~2 area of the central portion of this floodplain containing the most extensive and densest patches of para grass Am aximum likelihood classificati on produced an overal 1 accuracy of 86% and a producer accuracy for para grass ranging from 90 to 97%, indicating that there is potential to monitor para grass using VHR imagery.The expansion rate of para grass on the Ma gela Creek floodplain were estim ated also by comparing current m ap information with historic records since 1983. Results show that para grass currently occurs over 35% of the floodplain, has displaced about 11% of native vegetation, and is expanding on average at 14% p.a. or doubling in extent every 5 years.In summ ary there is a need to d evelopad edicated remote sensing monitoring program that target specific information needs for robust ecologi calrisk assessment of the threat from weeds to Ram sar listed wetlands on World Heritage Kakadu National Park. The inherent tern poral variability of floodplain envir onments, driven by the seasonal hydrological cycle, presents a challenge for the remote sensing of weeds and key native wetland habitats. Acquisition of high quality and reliable Iong-term tim e-series data used to monitor and model weed distribution require protocols for planning the timing of VH R captures and, designing appropriate field surveys to assess the accuracy of spectral classifications.