Decision-tree and rule-induction approach to integration of remotely sensed and GIS data in mapping vegetation in disturbed or hilly environments

Abstract

/ The integration of Landsat TM and environmental GIS data sets using artificial intelligence rule-induction and decision-tree analysis is shown to facilitate the production of vegetation maps with both floristic and structural infermation. This technique is particularly suited to vegetation mapping in disturbed or hilly environments that are unsuited to either conventional remote sensing methods or GIS modeling using environmental data bases.Current concerns worldwide with the management of native forests have led to a renewed interest in the development of rapid, cost-effective methods ibr forest mapping. Before any assessment can be made of the nature of vegetation in a given area, it is necessary to know what species are predominant, how they are distributed, and their relative abundance. Basic to the description of vegetation is a knowledge of floristics, lifeform, and structure. Traditional methods of mapping these attributes using a combination of field survey and photointerpretation are both labor-intensive and slow. Hopes that digital remote sensing would reduce time and costs, yet routinely produce acceptable thematic maps of vegetation and forest type, have not yet been met. In southeast Australia, where this study was carried out, remote sensing can often successfully distinguish structural communities (Walker and others 1986, Jupp and others 1986). The identification of floristic associations and life-forms is much less successful using only remote sensing. Although the difference between the reflectance of eucalypt leaves, in particular, appears to offer promise that some species discrimination is possible using remote sensing (Figure 1), leaf reflectance varies greatly within a single tree where there are mature and juvenile leaves present along with old, insectdamaged ones. Laboratory measurements show that the spectral variability of leaves within a eucalypt species

Spatial modelling of biodiversity at the community level

“…In this strategy, also known as 'predict first, classify later' and 'classification-then-modelling' Davis & Goetz (1990), Franklin & Wilson (1991), Lees & Ritman (1991), Moore et al (1991), Fitzgerald & Lees (1992), Brzeziecki et al (1993), Brown (1994), Brzeziecki et al (1995), Lewis (1998), Zimmermann & Kienast (1999), Keith & Bedward (1999), Hilbert & Ostendorf (2001), Ferrier et al (2002) 1b. Modelling of pre-derived species groups McKenzie et al (1989), Bojórquez-Tapia et al (1995), Ferrier et al (2002) 1c.…”

Convergence of LBS and AI

“…As discussed in [2] the two main properties required by almost all GIS based application is classification and feature extraction. In [2] discussion also has been made for implementing these properties using the techniques of Artificial Neural Networks.…”

“…In [2] discussion also has been made for implementing these properties using the techniques of Artificial Neural Networks. In [1] the discussion has been made for demonstrating the usefulness of ANNs in modeling spatial interaction and classifying spatial data.…”

The application of GIS based decision-tree models for generating the spatial distribution of hydromorphic organic landscapes in relation to digital terrain data

“…The trees created are usually very large with multiple terminal nodes, meaning that the models are intimately fitted on the training data (Lagacherie et al, 1995). Each terminal node is assigned to the label of the majority class (Lees and Ritman, 1991). Splits or rules defining how to partition the data are selected based on information statistics that measure how well the split decreases impurity (heterogeneity or variance) within the resulting subsets (Clarke and Pregibon, 1992).…”