Sunteți pe pagina 1din 18

www.environment.gov.

au/ssd

Monitoring the wetland weed para grass (Urochloa mutica) using remotely sensed data in Kakadu National Park, Northern Territory
Dave Walden, James Boyden & Renee Bartolo
Environmental Research Institute of the Supervising Scientist

Supervising Scientist - working to protect the environment from the impacts of uranium mining

Project Overview

July 2004 a multispectral QuickBird satellite capture of 64 km2 of the central region (highest density of para grass) of the Magela floodplain in KNP Airboat and helicopter field validation surveys in March 2003 and June 2004 Able to estimate the cover of para grass within the image boundary and the rate of spread using historical data Data also incorporated into cost-of-control and Bayesian habitat suitability models
Supervising Scientist Division 2

Para grass as a weed

A trailing perennial grass with long, robust creeping culms (stolons or stems) that sprout new roots at the nodes wherever they contact the ground Also colonises floating vegetation mats that can break apart and float downstream, thus increasing spread Fast growing with no pests/diseases and a very broad environmental niche tolerant of inundation at a wide range of water depth from moist ground to 200 cm Widespread distribution as it is valued by pastoralists as ideal buffalo/cattle fodder and actively planted in the region as early as 1922 and right up to the KNP declaration in 1979
Supervising Scientist Division 3

Key impacts of para grass

Can modify landscapes by forming monocultures over large areas thus reducing biodiversity Para grass readily outcompetes (amongst other flora) wild rice (Oryza spp.) and sedges (Eleocharis spp.). Species that are essential food and nesting resources for magpie geese and some other waterbirds The denser structure and higher biomass reduces water flows thus increasing sediment deposition. Greater fuel loads increase intensity and extent of floodplain fires which can kill fringing paperbarks and other species
4

Supervising Scientist Division

Kakadu National Park, the Magela floodplain & the extent of the QuickBird image

Supervising Scientist Division

Woodland Raw 2004 QuickBird image Wetland

Supervising Scientist Division

Image analyses vegetation map

Supervised classification using six training classes para grass (2 forms), dominant native vegetation types, and open water Training areas for classification were selected using the spatially referenced ground and low-level helicopter survey data. Wherever possible, training sites were selected from within discreet homogeneous patches of a particular class
Map class Para grass (high greenness) Para grass (low greenness) Overall (all classes) Producer accuracy 90 % 96 % 86 % User accuracy 96 % 92 %

Supervising Scientist Division

Vegetation class map from supervised classification

Supervising Scientist Division

Image analyses para grass cover

The vegetation class map was resampled to 5 m pixels. From this map a raster layer was produced for para grass only (other map classes removed) Using this data (and a 250 m zone-grid overlay), the percentage cover of para grass within each 250 m grid cell was estimated i.e. dividing the total number of 5 m para grass classed pixels falling within each 250 m grid cell, by the total area of each grid cell

Supervising Scientist Division

Para grass cover map derived from the vegetation map

Supervising Scientist Division

10

Early para grass spread on the Magela

Small patches observed amongst the native Hymenachne acutigluma on the Western Plains of the Magela floodplain KNP in 1982-83 An eriss project (Knerr 1998) based on aerial photography and ground mapping of the same region showed that para grass cover in 1991 was 132 ha and by 1996 had spread to 422 ha (more than tripled in 5-6 years) This initial spread increased rapidly as the grazing and trampling impacts of buffalo were reduced during the 1980s early 1990s
11

Supervising Scientist Division

More recent para grass spread

In 2004 the total area of para grass cover in this central floodplain region as derived from the cover map was 1250 ha Para grass was distributed over 35% of the floodplain with 10% displacement of the native vegetation being largely wild rice mixed with sedges. Many satellite infestations particularly to the north (ie downstream) Average spread rate calculated at 14% p.a. or a doubling in extent every 5 years Comprehensive helicopter surveys (Parks, NT Weeds Branch and CDU) in 2009 calculated 3513 ha of para grass over the central region
12

Supervising Scientist Division

Para grass spread (not including 2009 data)

Linear regression between Loge extent (km2) of para grass and time (yrs) (R2=69%, n=5, P<0.04) derived from: early observations in the 1980s: the study by Knerr (1998) who measured spread between 1991 and 1996; and the 2004 cover
Supervising Scientist Division

map

13

Future work

The Spatial Sciences and Data Integration group at eriss now acquires an annual (VHSR) WorldView-2 image capture of the Magela floodplain and Ranger minesite Using the GEOBIA approach it should be possible to continue monitoring and detect more subtle changes (including any control efforts) in para grass cover compared to previous mapping efforts In addition, SSDI will soon acquire a LiDAR 30cm DEM of the Magela floodplain. This will yield valuable hydrological/inundation data, factors which are critical in determining the distribution of floodplain flora Such information should further our knowledge of spread rates and habitat suitability, information that can be used to develop and improve management strategies
14

Supervising Scientist Division

WV-2 image May 2010

Para grass

Supervising Scientist Division

15

QuickBird July 2004

WV-2 image May 2010

Supervising Scientist Division

16

Bayesian Habitat Suitability Model by Keith Ferdinands

Supervising Scientist Division

17

Thank you

WV-2 image May 2010

Gratuitous crocodile photo

Supervising Scientist Division

18

S-ar putea să vă placă și