Civil Engineering June 2022 | Vol 30 No 5

Civil Engineering June 2022 39 including neighbouring Uganda. These observations include: Q Q No increase in runoff coefficient with return period, which is a common step in the standard application of the rational method, particularly in Southern Africa Q Q Ground slope, soil permeability and land use (vegetation) are considered twice in the calculation of the runoff coefficient (C), which can be inter- preted as duplication. Since limited hydrological studies are available for Rwanda itself, a desktop study was conducted to find other methods in the region where there are similar hydrological characteristics. The only available methods that could be found were the TRRL East African Flood Model and the Generalized Tropical Flood Method, which are limited to specific return periods, 24-hour storm durations and specific catchment sizes. Consequently, the results yielded from these methods were well outside a reasonable range of the modified rational method described above. STATISTICAL ANALYSIS OF AVAILABLE PEAK FLOW DATA The biggest challenge in conducting a hydrological analysis typically is obtaining reliable and complete measured flow data. Two river gauges, one upstream and one downstream of the bridge site, were found which had publicly available streamflow data. This data was used to conduct a statistical analysis which yielded 1:100- year flood peaks at the downstream gauge that were one third of those at the upstream gauge. There was initially no obvious explana- tion for this difference. However, upon further evaluation, the upstream gauge was found to be more accurate in terms of peak flows due to inconsistencies in the stream- flow data from the downstream gauge. LACK OF CONFIDENCE IN EXPLORED METHODS For the reasons given previously, the team lacked confidence in the flood peaks predicted by the available deterministic and statistical methods. To resolve this, additional sources of information were investigated, including flood studies, flood registers documented in the Rwanda Atlas, satellite imagery available in Google Earth and even social media (videos uploaded on Twitter). AVAILABLE INFORMATION ON THE OLD GAHIRA BRIDGE Rainfall data and peak flood levels at the bridge site for the May 2020 event that resulted in failure of the old Gahira Bridge were unfortunately not available from the relevant Rwandan authorities. Other de- sign avenues therefore had to be explored. One of these was to investigate/determine the age of the old Gahira Bridge and the floods that occurred during its lifetime. No details were available for the design or installation of the old Gahira Bridge and history had to be explored to obtain to an estimated age. From previous site photos it was determined that the Gahira bridge was a Bailey type bridge. Bailey bridges, which were developed in 1940/41 by the British engineer Donald C. Bailey, were used throughout the world from 1945 to 1960 due to their relatively simple construction method, as illustrated in Figure 4. We also found historical information confirming that, during the agricultural change period in Ruanda-Urundi (1945 to 1960), the transportation network was rapidly expanded in this region. By 1960, a network of 11 000 km of roads had been developed, including some permanent bridges (Leurquin, 1963). Since the sources of information for both the construction of Bailey type bridges and the expansion of the trans- portation network point to the period from 1945 to 1960, the old Gahira bridge was estimated to be 60 to 75 years old at the time of its failure. BRIDGE FAILURE EVENT OF MAY 2020 Video footage of the bridge failure was posted on Twitter, and this proved to be very useful in terms of quantifying the flood levels at the time of failure. The river water level at this time was recorded as being ankle deep for the people standing on the bridge approaches, as shown in Figure 5. This water level was then related to the project survey that was conducted in September 2021, which in turn al- lowed the flow at the time of failure to be approximately determined using the HEC-RAS hydraulic model that was set up as part of the hydraulic design study. The capacity of the old Gahira Bridge was also estimated not to be greater than 750 m 3 /s. RETURN PERIOD OF THE BRIDGE FAILURE EVENT Multiple flood events of the Nyabarongo River are listed in the Catalogue of Maximum Observed Floods, and in the Rwanda’s National Risk Atlas. The dif- ferent return periods of the floods were also documented in Rwanda’s National Risk Atlas and this information could be used to confirm that the May 2020 flood event had a return period of at least 50 years. A diagram showing the rationale Figure 4 Construction of a Bailey type bridge during the 1940s Figure 5 A snippet taken from the video footage of the Gahira Bridge during the May 2020 flood event