Abstract
Heavy rainfall in East Africa between late 2019 and mid 2020 caused devastating floods and landslides throughout the region. These rains drove the level of Lake Victoria to a record-breaking maximum in the second half of May 2020, when the lake reached its highest level since measurements began in the late 1800s. The combination of high lake levels, consequent shoreline flooding and flooding of tributary rivers, caused loss of lives and severe damage to housing, agriculture and public infrastructure in Uganda, Tanzania and Kenya. The floods triggered international attention and various media sources proposed a causal link with climate change. However, a formal attribution study identifying the possible role of anthropogenic climate change in increasing the likelihood of such record-breaking water levels has not been carried out so far. The main objective of this thesis is therefore to determine whether climate change influenced the likelihood of the flooding observed along the Lake Victoria shoreline in mid to late 2020.To this end, first, the flooding event is characterised by estimating the spatial extent and number of people impacted by the flooding of the lake shoreline and tributary rivers. Second, the record-high lake levels are reconstructed with an observational water balance model for Lake Victoria, which is extended to 2020 using a state-of-the-art satellite-derived precipitation product and is evaluated with remotely sensed lake level measurements. Drivers of the flooding are then investigated. Finally, the water balance model is used as an impact model within a probabilistic event attribution framework to estimate the influence of anthropogenic climate change on the likelihood of the observed flooding. To this end, the water balance model is forced with historical and natural forcing only bias-adjusted precipitation data from six Earth system models from the \gls{CMIP6} ensemble, made available through the \gls{ISIMIP} Phase 3b (ISIMIP3b).
Based on state-of-the-art remote sensing tools for flood detection, a total area of over 640 km\si{^2} located within 50 km of the lake shores is estimated to have been flooded between April and July 2020, affecting over 29'000 people. Precipitation over the lake and its basin in 2020 was the highest recorded in the last four decades, and was consistently above average between May 2019 and May 2020. The lake level rise witnessed between late 2019 and mid 2020 was approximately five times above climatology. Both lake precipitation and inflow drove the anomalous lake level rise in 2020, contributing respectively to approximately 70\% and 30\% of the anomaly, and were not balanced by above-average outflow.
The flood event is defined for attribution as a rate of change in lake levels over 180 days as extreme as the one observed in 2020, when lake levels rose by 1.21 m between November 2019 and May 2020. This event was the third most extreme in the historical record, after 1998 and 1962. Based on observations, the 2020 rise in lake levels is approximately a 48-year event in the current climate (CI 18 - 612 years) and would have been slightly less likely in a pre-industrial climate, with a return period of 65 years (CI 26 - 1606 years). The best estimate from observations is that the event is approximately 1.3 times more likely in the current climate (CI 0.6 - 4.5), but this change is found to be statistically not significant. Based on a synthesis of observations and climate model simulations, the best estimate is that the event is approximately 1.4 times more likely in the current climate (CI 0.9 - 2.6), and that in a pre-industrial climate the increase in lake levels would have been approximately 5 cm less than observed (CI -4 - 10 cm). Nonetheless, the slight increase in likelihood and intensity is found to be statistically not significant. From the statistical methods applied it is therefore not possible to say with confidence whether anthropogenic climate change made the 2020 Lake Victoria floods more likely or more intense.
By disentangling the role of anthropogenic climate change and natural variability in the 2020 high-impact flood event, this thesis contributes to a better understanding of impacts caused by climate and weather extremes in the Greater Horn of Africa. It also highlights the possibilities and limitations of applying a probabilistic extreme event attribution framework to study slow-onset climate extremes.
| Date of Award | 2023 |
|---|---|
| Original language | English |
| Supervisor | Wim Thiery (Promotor), Nicole Van Lipzig (Promotor) & Inne Vanderkelen (Advisor) |