Accuracy

We assessed the accuracy of our classification of the different datasets’ by calculating the error matrix and kappa parameter using the r.kappa module in GRASS GIS. Included in the accuracy assessment are the percent of commission and omission error, kappa values for each class and their variances, overall kappa,  the observed correct classified result (in number of pixels), and percentage of correctly classified pixels. The omission error is one where pixels that belong to a particular class are omitted from that class and commission errors are those where pixels are included in classes they should not have been. The accuracy assessment of our classification of the datasets is listed in table 3. Our results show very strong agreement of the classification with kappa values greater than 0.98 in each case. Note that strong agreement occurs if kappa is greater then 0.80,  moderate agreement occurs when kappa values are 0.40-0.80 and poor agreement occurs when kappa values are less than 0.40 (Jensen 2005).