AAIR Project: Sub-task 5.4 Regional Classification of Fuel Types

Contact

Tiago Oliveira / Mário Caetano

Centro Nacional de Informação Geográfica

Rua braamcamp, 82 - 1ºdto

1200 Lisboa

Portugal

E-Mail: tmo@rigel.isa.utl.pt

Description

This is a methodology based on linear mixture modelling. A Regional Classification of fuel types will be produced. A linear mixture model is used to convert satellite radiance values into proportions of forest, shrub and soil (Smith et al., 1990).

Input

digital terrain model

satellite image

image endmembers (extracted from the image)

reference endmembers (measured with a spectroradiometer)

Output

The outpu of the model for each endmember is a fraction image, where in each pixel is retained the abundance of the pure component on the ground.

Method

The image radiance is modelled as a linear combination of the radiance of pure components, i.e. endmembers. We apply the model to the raw image and to the data after radiometric and atmospheric corrections. The model that we use is integrated in the IPW software. The output, corresponding to the fraction images regarding shrubs, will be check and validated throught a GPS survey.

Uses

Application of this methodology will produce a map corresponding to the Regional Classification of fuel types.

References

Smith, M.O., S.L. Ustin, J.B. Adams, and A.R. Gillespie. 1990. Vegetation in deserts: I. A regional measure of abundance from multispectral images. Remote Sensing of Environment 31: 1-26.

Contact:

Maria J. Vasconcelos

Centro Nacional de Informação Geográfica

Rua Braamcamp 82 - 1Dto

1250 Lisboa

Email : maria@cnig.pt

Description

This is a knowledge based system that uses rules inferred from maps and from field data to produce regional level fuel maps.

Inputs

The field data consists of the relevant attributes found at the sample sites. The type and level of aggregation of the parameters collected in the field are decided upon completion of the exploratory phase and depend on the scale of analysis. The cartographic information is derived from a spatial data base that includes the results of a functional/structural vegetation classification and of the fuels maps obtained by the spectral mixture models.

Outputs

A regional level fuel map.

Method

The knowledge based system consists of a knowledge base, a set of rules, and and inference engine. The knowledge base contains spatial and attribute data obtained from the cartographic data base and from field work. The rules of inference are obtained by induction of decision trees. Basically, the rules are inferred from the data by analysing the values and relations of the attributes of the sites designated in a stratified random sample of the study region, and by studying the correspondence to the present fuel type (learning by example). This analysis may also have a temporal ascpect.

Uses

Regional level fuel mapping by remote sensing using ancillary data.