As shown in Table 9, the values of Wilks' lambda and the chi-square for each DF varied from 0.168 to 0.638 and from 40.419 to 156.314. All p-values were less than 0.01, indicating that the spatial DA was credible and effective. In the stepwise DA, seven variables (chl-a, DO, TKN, TP, ammonia-N, magnesium, and chloride) were selected as the most important discriminating variables. The two DFs explained 82 and 60.1% of the group differences, respectively. Likewise, in the wet season, the first DF separated the Miami Canal from the Kissimmee River and the Caloosahatchee River (Figure 5), and was significantly correlated with TP, DO, and TKN (Table 10). The second DF established some separation between the Kissimmee River and the Caloosahatchee River, and was correlated with chl-a and magnesium. The CFs and CMs obtained from the two modes are shown in Tables 11 and 12. In the standard mode, all 12 variables were included and the constructed CFs produced 79.2% accuracy in assigning cases. However, in the stepwise mode, DA produced 81.3% correct assignment using only seven discriminating variables.