| Hydrological stations | Period | Months missing | Years missing | Missing periods |
|---|---|---|---|---|
| Ubangi / Bangui | 1938–2015 (77 years) | None | None | None |
| Ubangi / Mobaye | 1938–2015 (77 years, 924 months) | 475 months (51%) | 21 years + 20 non complete years (53%) | 1938, 1950, 1958–1969 1971, 1975–2014 |
| Kotto / Kembe | 1948–1995 (47 years, 564 months) | 256 months (45%) | 14 years + 16 non complete years (64%) | 1949–1952, 1956, 1965–1969, 1976–1994 |
| Kotto / Bria | 1948–1995 (47 years, 564 months) | 242 months (43%) | 17 years + 5 non complete years (47%) | 1948–1953, 1964, 1967, 1968, 1977–1989 |
| Mbomu / Bangassou | 1951–1995 (44 years, 528 months) | 274 months (52%) | 14 years + 20 non complete years (77%) | 1951–1954, 1956–1962 1964–1973, 1976–1986, 1992, 1994 |
| Mbomu / Zemio | 1951–1995 (44 years, 528 months) | 100 months (19%) | 4 years + 13 non complete years (39%) | 1951–1953, 1957–1961, 1969, 1976, 1979–1982 1984, 1985, 1993 |
| Uele + Bili | Estimated | No interest | No interest |
Rainfall data are from the Agence pour la Sécurité de la Navigation Aérienne en Afrique et Madagascar (ASECNA) in Bangui for CAR and from the IRD database for Congo. Annual rainfall data were collected from eight stations in CAR and three stations in Congo, spread over the entire Ubangi basin at Mobaye (Figure 6.1), based on the work of Orange et al (1994). From west to east (Figure 6.2), there are: Ippy, Mobaye, Bria, Ouadda Djalle, Yalinga, Bangassou, Rafaï, and Obo in CAR; Bondo, Niangara, and Watsa in Congo. The two rainfall stations at Yalinga and Obo are complete from 1951 to 1995; most of the other stations have common data from 1972 to 1992 (Table 6.2). Reconstruction of the missing data was done by analysis of annual correlations between stations, according to the work of Orange et al. (1994).
The rainfall (P) and hydrology (Q) data have been arranged according to the hydrological year beginning April 1. The annual rainfall for each sub‐basin was calculated using the Thiessen method (Orange et al., 1994). The annual rainfall and hydrological indices (respectively, rainfall index and flow index) are standardized indices. They are calculated at the outlet of each watershed studied according to the following formula:
(6.2)
where Xi is the value of the variable X for year i; Xm and Si are, respectively, the mean and the standard deviation of the variable X observed for the duration of the series studied. The variables considered are the annual precipitated water depth P (expressed in mm) and the annual flow Q (expressed in m3/s). This index makes it possible to define the severity of the drought. Negative annual values indicate drought in relation to the chosen reference period and positive values indicate a wet situation (Ardoin et al., 2003).
Table 6.2 Rainfall data gaps.
| Rainfall stations | Period | Missing years | Missing periods |
|---|---|---|---|
| Yppi(1) | 1951–1995 (44 years) | 11 years (25%) | 1951–1959, 1993, 1994 |
| Mobaye | 1951–1995 (44 years) | 23 years (48%) | 1951–1971, 1993, 1994 |
| Bria | 1951–1995 (44 years) | 11 years (25%) | 1951–1959, 1993, 1994 |
| Ouadda Djalle(1) | 1951–1995 (44 years) | 11 years (25%) | 1951–1959, 1993, 1994 |
| Yalinga | 1951–1995 (44 years) | None (0%) | |
| Bangassou | 1951–1995 (44 years) | 11 years (25%) | 1951–1959, 1993, 1994 |
| Rafai | 1951–1995 (44 years) | 27 years (61%) | 1951–1971, 1989–1994 |
| Obo | 1951–1995 (44 years) | None (0%) | |
| Bondo | 1951–1995 (44 years) | 27 years (61%) | 1951–1971, 1989–1994 |
| Niangara | 1951–1995 (44 years) | 27 years (61%) | 1951–1971, 1989–1994 |
| Watsa | 1951–1995 (44 years) | 33 years (75%) | 1951–1971, 1983–1994 |
(1) Yppi et Bambari are assimilated; Ouadda Djalle et Birao are assimilated.
6.3.2. Break Tests for Hydropluviometric Data
The break tests on the rainfall P and hydrometric Q series were performed with the Khronostat 1.01 software (IRD, 1998) to characterize the stability periods of these two long time sequences, establishing if possible homogeneous flow subsequences, or hydroclimatic periods. Pettitt, Lee and Heghinian rank tests, and Hubert’s segmentation were applied (Hubert & Carbonnel, 1987). The combination of these various methods is justified by the fact that the first two