Implementation methodology for spatial data collation and analysis
following describes process that will be used for developing the GIS system, transferring
skills and providing information to inform management planing.
The development of the spatial data system involves the following
The first task is to access pre-existing
spatial data for inclusion in an ARC-View based GIS.
Satellite Image analysis.
processing for land cover classification and fire mapping will provide the primary derived
data for analysis.
collection and ground truthing
GPS based surveys will provide additional spatial data and accuracy assessment
for the derived satellite data.
Spatial analysis will be conducted examining the relationship of derived
fire histories to various biophysical and cultural landscape layers.
Derived map products
will be produced as a primary means of communication and field verification.
On-going training in Satellite image processing, GIS development, analysis
and field survey technique
2.1 Data Collation.
following data sets are being collated for this project.
Topographic map sheets (1:50000) have been acquired
as both hard copy and digital coverages. The hard copy
map sheets have been scanned and will be registered to the rectified digital coverage.
These mapped data are the most accurate available and so will form the base layer
for co-registration of all other spatial data layers. These map data
where developed from recent aerial photography (~1995) and so provide reliable
infrastructure and land-cover mapping.
Ariel Photography will be sourced from the appropriate
Indonesian government archive. Photography used for the topographic mapping over
the study sites will be used to add to and verify the topographic mapping. Historic
aerial photography will be obtained, where available, for the study sites to be
used in land cover change analysis. All photography will be scanned and rectified
to the topographic map base.
Landsat-ETM imagery has been obtained as ninth scenes (60x60km).
Two of these sub-scenes are required to cover the study sites of Dorameli and Dhersa in Flores and
Kiritana and Lukuwingir in
Sumba respectively. Two images will be purchased for each year to cover early
dry season (May to early August) and the late
dry season (August to November). This imagery will be used for fire mapping, landcover
mapping and landcover change detection.
Early Landsat-MSS imagery will be used for landcover change analysis. We currently have a full MSS scene from 1973 for the Sumba study
area and a 1973 scene has been ordered for the Flores study area.
MODIS and NOAA satellites imagery will be used for hot-spot
mapping. Current mapping being undertaken by the CSIRO, using MODIS imagery, covers
NTT. It is proposed that this mapping be combined with NOAA derived hot-spot data
by the Western Australian Dept of Land Administration (DOLA) to produce comprehensive
mapping that would be posted daily through DOLA’s established
web based hot-spot distribution system.
Satellite Image analysis
scar mapping will be conducted on the ninth scene ETM images. Using ER-Mapper the two yearly images will be combined to produce one
multi-date image. From this a difference image will be produced for each date
by subtracting one image date from following date to highlight change between
dates and thus highlight fire-scars. A Principle Components Analysis image will
then be produced from the difference image to further highlight this change. Training
sites will be selected from areas of significant change and a supervised classification
run. The resulting imagery will be filtered to reduce speckle misclassification.
The classification will then exported to ARC-View as a vector file for manual
cleaning. The PCA-difference and raw imagery will be used as an underlay to guide
MODIS data will be examined
for is ability to augment the fire scar mapping being undertaken with Landsat.
A broad scale land cove map will be produced from the ETM imagery
for the study areas on Sumba and Flores. Using field surveys and Ariel photography,
classification training sites will be selected and a supervised classification
conducted. Further ground truthing will be used to refine the map product. This
vegetation map will be used as a base layer for analysis of the fire derived fire
It is of interest to this project to develop an understanding
of recent and historic landscape change. By comparing Landsat-MSS imagery going
back to 1972 to current imagery we will quantify land cover change for the overlapping
scene. Landsat MSS imagery has a lower spectral and spatial resolution (80x80m
than current ETM imagery. This will limit the accuracy of cover change mapping.
Once co-registered a difference image will be created from
the MSS and ETM imagery highlighting changes in spectral response in the overlapping scene area. This difference
image will then be classified into positive and negative change, spectral response.
The classified \image will then be compared to the two image dates for manually
interpretation of the change response. Classification will be altered or annotated
according pattern response and local knowledge. Ground truthing of change areas
will be conducted through site visits and discussion with locals.
Hot Spot Detection
Using thermal sensors on the MODIS
and NOAA satellites hot spots are detected and mapped on a daily basis through
out Australia using an automated detection algorithm.
Due to the coarse resolution of NOAA
imagery the location accuracy of the hotspot data is, at best, ±1–2 km. Typical
daily fluctuations of fire intensity and the timing of the satellite data collection
effect hotspot detection. The inclusion of the higher resolution MODIS data (±250m)
increases both the spatial and temporal resolution of the mapping.
Training will be provided in the interpretation of hot-spot
data. A ground truthing program will be developed to ascertain the accuracy and
effectiveness of hot-spot data for management.
2.3 Field collection
and ground truthing
GPS surveys will map the following:
-Traditional and government tenure boundaries
will be mapped for relation to current burning regimes and management options.
sites need to be mapped as integral features in developing fire management plans.
This is also true for…
- Areas of particular ecological significance.
ground based mapping of land cover features such as forest edges and cultivated
land a base layer for developing a strategic burning approach. It will be base
data for accessing land cover change through comparison to historic data and ongoing
truthing of satellite imagery and derived data is important for classification
attribution and accuracy assessment.
is particularly important for the determining the accuracy of hot-spot and fire
scar mapping. It is assumed that image
processing techniques developed in Australia will readily transfer to the rugged
savanna landscapes of eastern Indonesia. This assumption
needs to be tested.
2.4 GIS processing and analysis
Digital elevation models will be created from the
1:50000 topographic map contours. These will be used for landscape 3D visualisation,
Land-cover mapping and derived management maps ie. erosion potential.
scar mapping will be will be aggregated
into two seasonal coverages: fire maps for May to early August will be combined
to produce an Early Dry Season (EDS) coverage, and fire maps for late August to
November will be combined to produce a Late Dry Season (LDS) coverage. Over time these data will be combined to produce
fire history parameters: total fire frequency – the number of times
burnt over the study period; frequency of EDS fires; frequency of LDS fires. These fire history
parameters will be intersected with landcover, tenure,
management strategy and land cover change data and relationships quantified.
2.5 Derived map products
It will be an important and on-going part of the project to
provide hard copy and digital map products to land managers, local and provincial
government administrators for verification, analysis and discussion.
Satellite image maps of
the study areas, with and without infrastructure vector over lays, will form a
base layer for discussion and explanation. Image maps will be provided at the
start of the project to each project site village head-men (Kepala-Desa).
Land cover maps will be produced for the study areas
and provided to villages and field officers to assist in developing and implementing
Hot spot maps will be provided in near real time
over the internet.
history maps will
be produced as the imagery is processed. They will form a basis for ground truthing
and field discussion.
Management maps, such as erosion potential or burning
sensitivity, derived from GIS analysis of combined spatial layers could be produced
to aid land management decisions.
Training will be provided in combination with development of
the GIS and subsequent analysis primarily to the two GIS project officers. Software
training will specifically focus on the use of ER-Mapper
for satellite image analysis and ARC-View for GIS analysis. It is envisaged that
simple hand books will be created in English and Indonesian describing technical
methodology and processing pathways developed for this project. Field training
will focus on the use of GPS for mapping and accuracy assessment.
Training in the interpretation of satellite imagery and the
derived products, will be provided to all team members and relevant government
administrators., This will create an understanding of management opportunities
and limitations of these tools. This process
will be particularly important for assessing the utility of hot spot mapping
CIFOR will provide training in the integration of social cultural
data into a spatial context for analysi