Interannual to Decadal Rainfall Variability over Greater Horn of Africa: Non-Stationary Teleconnections, Physical Mechanisms, and Predictability

The main objectives of this thesis are to understand the drivers of interannual-to-decadal rainfall variability over the Greater Horn of Africa (GHA), to detect non-stationary teleconnections, to investigate the physical mechanisms, and to assess the skill of the MiKlip decadal forecast systems. To address these objectives, interannual to decadal teleconnections influencing the three GHA major rainy seasons are examined using observational datasets for 1901–2013. Sea Surface Temperature (SST)-based climate indices known to influence Short, Kiremt and Long Rains are used in a comprehensive statistical analysis to detect non-stationary behaviour in teleconnections and to split them into interannual and decadal times scales. Interannual variability in the October–December Short Rains is predominantly influenced by the Indian Ocean Dipole (IOD) with percent variance explained (PVE) of up to 80% in recent years. However, abrupt shifts in this teleconnection are found around 1918, 1951, 1987, and 1994. The Short Rains also correlate strongly with El Niño-Southern Oscillation (ENSO). However, the ENSO influence on Short Rains is mediated by an in-phase fluctuation of IOD. ... mehrDecadal variations in Short Rains are more directly explained by low-frequency variability in the Pacific Ocean (PO). A hitherto undocumented non-stationary relation was found between Atlantic Niño 3 and June–September (JJAS) Kiremt Rains. The non-stationarity seems to be related to a decadal regime shift of the West African monsoon in the late 1960s. The variability of Kiremt Rains is also strongly associated with ENSO, though the recently increased correlation was not a multidecadal change. Consistent with recent studies, the post-1998 March–May Long Rains decline is strongly associated with decadal variability in the PO. The PVEs in the stable correlations with the Pacific Decadal Oscillation (PDO) and the Interdecadal Pacific Oscillation (IPO) indices range from 25–64%, mostly due to low-frequency variability (>8 years).

The recent increase in ENSO influence on Kiremt Rains raises questions about the physical mechanisms. This motivated a modeling study, and the Kiremt Rains interannual variability analyzed using observational data and higher-resolution SST-forced Icosahedral Nonhydrostatic (ICON) experiments for the period of 1981–2017. Such fine-grid global and two-way nests over the GHA were carried out here for the first time. The physical mechanisms that link ENSO influence on the Kiremt Rains in the model and the ERA-Interim reanalyses were also investigated. It is found that the model reasonably simulates the main features of the JJAS rainfall climatology over GHA and also reproduces horizontal wind intensity and patterns at 150, 600, 850, 925 − hPa levels over Africa. It is shown that there is a substantial skill in reproducing the leading modes of Kiremt Rains interannual variability (r = 0.64) given the SSTs are known. The result suggests that the majority (> 50%) of Kiremt Rains anomalies are driven by equatorial Pacific SST variability, while the SST effects
from other regions counteract the ENSO impact in the model. Consistent with previous studies, it is found that the El Niño phase of the ENSO drives a corresponding large-scale circulation anomaly, which weakens the monsoon trough over the Arabian Peninsula, enhances descending motion, and upper-level convergence right over Ethiopia. The subsidence over the GHA region induces upper (lower) level westerly (easterly) wind anomalies over North Africa leading to weakeninging the Tropical Easterly Jet, Somali Low-Level Jet, and reducing the moist westerlies from the Atlantic and Congo basin, and thus lead to a reduction of Kiremt Rains over Ethiopia. The opposite pattern is observed during La Niña events with enhanced surface westerlies leading to a wetter Kiremt Rains. This mechanism represents an anomalous Walker-type circulation for the ENSO-Kiremt Rains teleconnection.

Finally, the skill of MiKlip decadal forecast systems was assessed over GHA. The predictability of climate drivers at seasonal to decadal time-scale and observed teleconnection patterns were evaluated. It is shown that Atlantic Multidecadal Oscillation (AMO) is predictable within a time-scale that depends strongly on the particular MiKlip experiments, ranging from 2 up to 10 years, while the SST interannual variability in the North Atlantic basin appears to be skillfully reproduced at 1-year lead. Nonetheless, the overall performance of the MiKlip prediction skill for PDO is found to be lower than that for AMO. The predictive skill for ENSO is dependent on the lead time. It is shown that the January–March El Niño 3.4 (N34) index has considerable prediction skill (r > 0.85) for 1-year lead, but the skill for the JJAS N34 index is substantially decreased. Moreover, the MiKlip decadal hindcasts failed to reproduce ENSO-Kiremt Rains teleconnection patterns during the JJAS season, and in general, the prediction skill for SST and rainfall is weak over most parts of the tropical regions.

The results of this PhD thesis, therefore, have provided useful insights on the drivers of interannual to decadal rainfall variability, non-stationary teleconnections, physical mechanisms, and predictability over the GHA. The ramifications of the results of this thesis go beyond seasonal-to-decadal forecasting. They are also useful when analyzing the persistent of seasonal-to-decadal teleconnections under anthropogenic climate change.