3.2. Rainfall vs river discharge
Similar to rainfall, observed river discharge was also statistically decomposed using the singular spectral analysis. This approach is warranted because unlike rainfall, the discharge time series is a one column vector unit and decomposing it similar to rainfall allows comparison between their leading modes. The temporal patterns of the leading modes of rainfall and discharge were compared with one another (Figs. 8a-b). The relationship between leading modes (PC-1) of rainfall and river discharge show marked fluctuations during the entire period (1901-2010). The strongest linear correlation (r = 0. 71) was observed during 1991-2010 period while the lowest (r = 0. 59) was observed at the start of the century (1903- 1930). The other two climatological periods show moderately strong correlations (r = 0. 68 and 0. 64 for the 1930-1960 and 1961-1990 periods, respectively). Maximum correlation between the second modes of rainfall and discharge is also linear and strong but with a one-month phase lag. For example, the second modes of rainfall and discharge during the 1903- 1930 (r = 0. 71)
and 1991-2010 (r = 0. 66) were found to be well associated at one-month lag. The total variability accounted for by the two leading modes of rainfall and discharge are summarised in Table 1. The annual signal in GPCC explains an average of 42.7% of variability in discharge while the multi-annual signal accounts for about 50% at 1-month lag. During all climatological periods, temporal relationships between the first modes of rainfall and discharge are moderately strong and significant (r = 0. 05). At the start of the century, more of the variability in rainfall were rather consistent with those of discharge (92% and 96.7% total variability for rainfall and discharge, respectively) as opposed to the 1990- 2010 period (91% and 98.3% total variability for rainfall and discharge, respectively) when there was increased variability in discharge (Figs. 8a-b). Similar to rainfall (Figs. 5a-b and 6a-b), the variability in leading modes of discharge fluctuated during the century (between 69% and 76%). This suggests that apart from rainfall, Congo river discharge is driven by other factors, which may include the influence of human activities through deforestation and sand mining. While the interaction and exchange of fluxes within wetlands could disturb the variability of the Congo river, Conway et al. (2009) confirmed the dominant control of inter-annual and decadal rainfall variability in river flows whilst acknowledging the key roles of human interventions. Notably, a significant proportion of changes in the dominant discharge patterns is still not explained by those of rainfall. This information signals the threshold of complex hydrological processes in the region, and perhaps suggest the influence of anthropogenic contributions and strong multi-scale ocean-atmosphere phenomena as key secondary drivers of hydrologic variability.