2.4 Statistical analysis
The geomorphic parameters (HL , A , Pe , andHDC ) and indices (CC , FF , RR , LR , andHI ) of the catchments were grouped according to the Rift margin and Valley floor derivations. Normality of each geomorphic parameter and index in each group was tested with the Kolmogorov-Smirnov test (α = 0.05) and the Shapiro-Wilk test (α = 0.05), and homogeneity of variance was tested with the Levene test (α = 0.05). For the parameters and indices that were tested for normality and homogeneity of variance, a t-test was applied to detect differences in the mean values between the two groups; otherwise, a non-parametric test was applied.
Least square regression analysis was applied for log-transformeds (slope of the gully head) and a (upslope drainage area) to draw the threshold line that represented the gully topographical thresholds at each of the 12 main study gullies on a double logarithmic chart. Based on the appearance of the 12 threshold lines, hypotheses were made, and the topographical threshold of gully heads were categorised into different subgroups, e.g., periods of gully head incision or land use items around gully heads. Analysis of covariance (ANCOVA) was made to find there were significant differences (α = 0.05) in the mean values of the dependent variable (s or a ) between the different subgroups (controlled independent variable, or factor) while taking into account the influence of the uncontrolled independent variable (a or s as a covariate). ANCOVA assumes that: (i) the dependent variables are normally distributed, and those variances are equal overall subpopulations; (ii) the interaction between the factor and covariate are negligible; (iii) a linear relationship between the covariate and dependent variable. In the analysis, these assumptions were tested, and post hoc tests were made using the Bonferroni correction method to find the subgroups that had significant differences in the mean values. SPSS ver. 20 (IBM) was used for the statistical analyses.
RESULTS