3.1. Redundancy index analysis
A total of 315 hydrological indices of magnitude, frequency, duration, timing, and rate of change have been calculated which characterize the hydrological series in NPRS for different periods of the year (Table A3). The PCA showed an inertia ranging between 36.2 and 38.0% for the first axis and between 16.3 and 16.9% for the second (Fig. 2), reaching approximately 70% when grouping the first five components for the thresholds of 0, 1, 2, and 5 l/s. Results revealed multiple linear correlations between indices. It is also noted that the indices of frequency and duration of zero flow had a positive correlation with the first axis, while flow magnitude had a positive correlation association with the second (Fig. 2).
(Here Fig. 2)
Additionally, the first indices with the highest loading coefficient associated with the first, second and third dimensions were frequently the same, only changing for the fourth and fifth dimensions for the different thresholds (Table C1). Consequently, only 171 indices were selected for the redundancy index analysis based on clusters with correlation distance between hydrological indices within the same attribute (magnitude, frequency, duration, timing and rate of change) in different thresholds (Table C1 and Fig. C2 to Fig. C7). As a result of the dimensionality reduction and the redundancy analysis, five hydrologic indices were selected: number of days per year without flow (DU035), annual percentage of months without flow (FE052), mean daily annual flows (MA001), CV of Julian date of the annual start of zero flow (TI002) and annual rise rate (RC001) (Table 1).
(Here Table 1)