Logistic regression modelling of thematic mapper data for rubber (Hevea Brasiliensis) area mapping / Mohd Nazip Suratman ... [et al.]

Logistic regression modelling of Landsat Thematic Mapper (TM) was applied for mapping the area of rubber plantations in the study area ofSelangor, Malaysia. TM bands 2-5 and 7
were included in the final logistic regression model, and all were highly significant at the 0.0001 level. The tf value (23247.9) for the model was highly statistically significant (P<0.0001), which implies the estimated model fitted the model building data. TM bands 4 and 5 were the two most influential variables affecting the odds of rubber area occurrence on the imagery. Using probabilities of > 0.5, the model correctly classified 94.5% of the observations in both the training and validation data sets. This high accuracy suggests that the model is appropriate for predicting the presence of rubber trees in the pixels based on selected spectral bands measured by Landsat TM.