3.2. Statistical Analysis
Table 3 shows the results of correlation analysis among the six variables in terms of the Pearson’s correlation coefficients and the Spearman’s rank correlation coefficients. There are moderate correlations among the duration and the total rainfall depths. The maximum discharge is correlated with the total runoff volume and with the bulk runoff coefficient, respectively. Correlation cannot be found in the other combinations of two variables. The difference between the two total rainfall depths in each event implicates spatial variability of rainfall distribution within the catchment area. As the absolute values of the Spearman’s rank correlation coefficients are small in general, the six variables are considered independent and separately fitted to probability distributions. The weak correlation between the total rainfall depth at the observation system and the total runoff volume also implies a significant limitation of the conventional SCS runoff curve number method
where with the curve number represents the potential maximum soil moisture retention, and the parameter determines the initial abstraction (Zakaria et al. 2012). Figure 4 compares the observed total runoff depth () for each flash flood event with the estimations by of different and . The sum of the absolute errors, the sum of the square errors, and the maximum absolute error are minimized with = , , and , respectively. With the commonly used value of , those are minimized with = 76.8, 77.1, and 77.1, respectively. The incapability of the SCS runoff method to reproduce the total runoff volumes from the total rainfall depths is evident. To sum, we recommend direct monitoring of runoff discharges rather than rainfall depths.
Table 3. Results of correlation analysis among the six variables