TS-CRC Student project - Spatial patterning of resources for graminivores: Developing a model for habitat management

Charles Darwin University

Yue Zhang

Background | Objectives | Research outcomes | More Information |

Prime site for fire: understorey grass in the Yinberrie Hills

Prime site for fire: understorey grass in the savannas of Yinberrie Hills, 50 km north-west of Katherine in the NT
Photo: Zhang Yue

Background

Many savanna birds dependent on grass seeds have declined in distribution and abundance. These declines often appear to have preceded extensive disturbance such as land clearing, suggesting that changes in the condition of grasslands began with extensive pastoralism. Possible sources of change include the direct impacts of grazing, as well as alteration of fire regimes to favour pastoral objectives.

Other work supported by the TS-CRC indicates that the early wet season is a period of resource shortage, when many graminivorous birds are dependent on a small suite of perennial grasses that are patchily distributed in the landscape. Important species include Chrysopogon fallax and Alloteropsis semialata . Within seasons, disturbances such as fire affect both the size of seed crops and the timing of their production, and over the longer-term influence the distribution and abundance of the plants.

This project was designed to describe the configuration and quality of perennial grass patches needed to sustain regional graminivore populations, to identify the landscape and management determinants of that patterning, and to develop a model for grassland management to maintain graminivore populations. It included sites with varying degrees of grazing pressure.

Objectives

Specific objectives were to:

  1. Employ existing data to characterise grasslands used by graminivores in terms of resource density, patch configuration, and position in the landscape.
  2. Develop a GIS-based model of the correlates of distribution and abundance of favourable grassland patches using available themes including vegetation/land system/land units, soils, geology, DEM, and other (Auslig) topographic data.
  3. Relate ground-based measures of grassland pattern to remotely-sensed images of landscape pattern. Integrate classifications with GIS model.
  4. Assemble fire history of study sites used for characterisation of favourable patches, using methodologies developed by the NT Bushfires Council Fire Ecology group.
  5. Predict location of favourable grassland patterning in the landscape with and without reference to fire history.
  6. Conduct ground surveys of vegetation and grassland patterning to verify model predictions and explore implications of different fire histories.
  7. Conduct ground surveys of graminivorous birds using the site during the wet season.
  8. Relate results of avifaunal surveys to fire history and landscape context. Identify additional landscape elements with which different taxa are strongly associated.
  9. Develop GIS-based models to identify additional sites at which particular taxa are expected to occur. Verify against existing records of avifaunal distribution and design and perform surveys to determine presence/absence.
  10. Develop models for management of graminivore habitat through the appropriate use of fire at a landscape scale.
  11. Integrate outputs with models of landscape function developed in pastoral landscapes.

Research Outcomes

An operational method to map bush fire history with these Landsat TM data was developed. This interpretation approach is practicable for the use of large numbers of images by which a multi-yearly fire history wildfire mapping and spatial analysis can be implemented.

With the visible red, NIR and MIR bands, an unsupervised digital image classification was carried out to delineate the burnt patches. These patches were labelled by using field knowledge as well as by on-screen assessment of the raw data and signature files for previously confirmed fire scars.

Spectral overlap between fire scars, water bodies, shadows and miscellaneous geological features was observed and was eliminated by using of a binary spatial mask and the Digital Elevation Model (DEM).

The validity of the fire mapping was assessed with the help of field data as well as using a high spatial resolution IKONOS image acquired at the time of the TM data recording in 2000. Preliminary analyses showed accuracy rates of fire mapping in the image of 2000 were 91 per cent and 94 per cent, according to field data and IKONOS data, thus the validation of the method was proved to be feasible.

Thirty-two Landsat TM images were offered by the Parks & Wildlife Commission of the NT through the TS-CRC in August 1999 and a number of rectification procedures for the images were carried out. This data set crosses 11 years but unfortunately in some years there are just one scene available so that this mapping result still need to be enhanced by acquiring required images in certain years.

Vegetation mapping was also carried out using a new approach. One Landsat TM image acquired in March 1997 was used to implement vegetation interpretation with red, NIF and MIF bands. To simplify the complexity of tropical savanna landscape characteristics in the study area—such as very small areas of vegetation or geological distribution that differs from the types around them—a statistical filter was used to merge those pixel assemblages with the pixels around them.

A supervised classification with maximum likelihood algorithm was carried out using training data that included ground data collected within the study area, hard-copy vegetation maps and digitised vegetation maps carried out by the NT Department of Lands, Planning & Environment (NTDLPE).

Seven classes of vegetation types were mapped with the resolution of Landsat data, but this should be significant enough for a fire history analysis to quantify fire-sensitive vegetation cover both temporally and spatially.

The digital vegetation data of study area in the format of ArcView shape file was provided by NTDLPE, which divides the vegetation of the study area into 17 types of communities. Except for training signatures for supervised classification, the spatial modelling method included in ArcView Spatial Analyst can be implemented directly to delineate the fire influences on the different covers of vegetation types.

To provide a preliminarily analysis of the spatial changes to pattern in this landscape during the fire season, binary grid fire maps were generated in a GIS and these were used to compute spatial pattern indices: number of patches, mean patch size, mean shape index, and mean patch fractal dimension. The fluctuations of these indices illustrate the changes in spatial patterning of vegetation in a savanna landscape that is significantly affected by dry-season burning.

The implementation of tropical wildfire mapping, vegetation mapping and spatial pattern analysis will be used as the basic data stream for habitat analysis and modelling of graminivores.