Flood risk and mitigation strategies in Derna (Libya) (2023-10-09)

Activation date: 2023-10-09

Event type: Flood

Activation reason:

On 10 September 2023, storm Daniel reached its peak in north-eastern Libya, causing much damage and major destruction. One of the worst affected cities was Derna, at the time of the event home of 90,000 people. In the aftermath of the heavy rainfall the two ageing dams located in the Wadi Derna valley, one just close to the city itself and one about 12 km southern along the river, collapsed not being able to resist the huge volume of water. This led to a catastrophic event in which large parts of the city were destroyed and thousands of people lost their lives.One of the consequences of this event is that the city of Derna is finding itself once more exposed to floods, without mitigation systems installed to protect the inhabitants and the town. To better understand the flood disaster risk of the area, some analyses based on three different scenarios are mandatory to retrieve insights into the possible impact of future events.Five products in total were delivered for this activation:P1 – Flood hazard modelling;P2 – Reference mapping (of the most current situation);P3 – Exposure assessment of citizens and infrastructures;P4 – Mitigation measures;P5 – Flood hazard modelling considering the mitigation measures.The  impacts of the following scenarios were elaborated for the products (except P2):Floods with a low probability (return period ≥ 500 years or extreme event scenarios);Floods with a medium probability (return period ≥ 100 years);Floods with a high probability (return period of five years).P01: Flood hazard modellingThe Euro Maps 3D DSM with a resolution of 5m was pre-processed to improve the quality for easier visual interpretation and data extraction as well as for run-off characteristics for the subsequent hydro modelling. The processing steps were:Conversion into a Digital Terrain Model (DTM) by removing all surface structures except for buildings or other infrastructure relevant for the hydro modelling. To visualize the flow direction in the urban areas the actual state of the buildings were integrated by their footprints. The two dams which were destroyed by the flood were removed from the DTM.The derivation of the Hydro-DEM for the modelling included further processing steps, e.g., removing depressions, stream burning (slope-based filtering), editing the river flow paths to catch the correct outlet point, editing outliers and unwanted elements.A 2D rain-on-grid model has been performed for flood hazard modelling. A high mesh resolution was considered for the upstream hydrological catchment from the main dam up to the outlet of 10m, and the rest of the AOI was modelled with a mesh resolution of 40m. The model has been set up to account for the current situation, where the two dams and bridges at the downstream were removed from the DSM and was run taking into account the three different scenarios: 5y, 100y and 500y return period (RP). The result for the modelling shows a maximum flood depth of 8.60m, 8.72m and 9.28 respectively, these values have been noticed at the two tributaries just upstream of Derna waterfall, where the relief characteristic is very steep with very narrow water courses. Furthermore, the results for the rainfall event modelling also show a flood depth of around 2m (5y RP), 3m (100y RP) and 4m (500y RP) in the main Derna river in the urban area, with relatively high values at different sections in the river where the water seems to be accumulated due to the presence of check dams in this section of the river, leading to a small increase of the water level just at the upstream of the check dams where the reservoirs are. The same phenomena occurred at the reservoirs of the two dams. Figure 1: Flood extent and maximum water depth for the three scenarios in the urban area of Derna.P02: Reference mappingAfter gathering all free available ancillary information with the potential to contribute to the reference dataset production, the data was validated concerning their level of completeness, thematic and geometric accuracy and then included to reference product vectors. Afterwards, a complete and exhaustive visual interpretation of the optical satellite image was performed to improve the completeness and correctness by adding missing geometries and attributes as well as deleting features that were no longer present. A statistic summarises the aera or length of all features per category.  Figure 2: Reference dataset example (P2) P03: Exposure assessment of citizens and infrastructuresThe exposure was calculated by intersecting the P1 flood hazard layers with the population and the P2 reference mapping information on buildings and infrastructure. The derived statistics reflect the hazard considering the current situation with the dams destroyed by the flooding on September 10th, 2023.The estimated population potentially affected (based on P1) is around the 9.5% (high probability scenario), 14.5% (medium probably scenario) and 15.5% (low probability scenario) out of a total population of ~72,900. P04: Mitigation measuresThe analysis of the flood results and infrastructure distribution within the AOI implies to consider rural and urban areas on two different scales. The urban area reaches from the Mediterranean Sea to Al Bilad Dam, based on the quality of the used data, the vulnerability of this area and the risk knowledge and awareness of the population. Nevertheless, the rural area is also considered, and mitigation measures are proposed. In cases where these seem not cost-effective it is highly advised to consider actions in the development plan that should be reviewed for future planning. For example, where a road is situated the same level with a riverbed, protecting measures do not seem suitable. Such construction should rather be banned in the development plan to avoid flooding and disruption of the transportation network.Further it needs to be mentioned that the urban discharge sewage system is not considered in the modelling, results of P5 might show the same flooded zones in the urban area. This is more related to sewage evacuation capacity than the present activation.The flood modelling results for all three scenarios highlighted:The submersion of bridges in the city of Derna reaching 3.5m in some zones;Water overflowing roads and cart tracks close by;Overflow of the urban channelized river near the bridges and check dams due to sediment loads;High water depth in the areas near Derna waterfall because of the topography. Visual exposition shows no risk, but since it might be a touristic area, further investigation needs to be considered. Accumulation of water in some zones of the urban area related to urban drainage, but not to a river flooding.To protect some sensitive infrastructure like bridges and to decrease the water depth and extent, the following measure were considered and implemented if suitable:Reprofiling of the Derna river in the urban area to restore the natural flow path and adjust the riverbank slopes that were damaged by the floods or by erosion. This includes removing sediments, decreasing the slopes of the riverbanks as well as minoring and deepening the riverbed.Adjustment of culverts and the elevation of certain bridges to avoid the direct strike of water to the bridges.Reconstruction of Al Bilad Dam as an overflow dam to the same height, or 20m higher for the extreme scenario. This dam has shown its effectiveness in reducing the impact during the historical floods of November 1986 and September 2011. It should be designed only for protection and not for storage taking in consideration its position at the entrance of the city and as a potential area for the city to be extended. Construction of a small retention dam just upstream of Derna waterfalls: Here flows one of the important tributaries of Derna River. This will reduce the time needed for water to join the main river, hence decreasing the risk of additional flooding.Many check dams especially in the rural area needs to be rehabilitated to be more effective in reducing the flow velocity and discharge, preventing and decreasing the rate of sediment transport and enhancing the recharge of groundwater in this area.Reconstruction of the Abu Mansour dam is mandatory but was not implemented for this activation in terms of mitigation measure that feeds into the P5 modelling. This is reasoned by the reduced catchment AOI that does not include the upstream catchments of Abu Mansour. Nevertheless, the reconstruction of the dam is vital seeing its valuable impact in protecting the city from pervious flood events. It is recommended to plan the wastewater sewage system according to the volume of run-off water during urban flash floods. This measure is not implemented in this activation due to the origin of flood and lack of data. However due to the frequency of flash flood, it needs to be considered to a higher level in the next urban development plan.P05: Flood hazard modelling with mitigation measures of P04A remodelling of the rainfall event-scenarios for flood hazard was carried out after implementing the mitigation measures proposed in P04 into the DSM. The Al Bilad dam (upstream the city), and the dam upstream of Derna waterfall have been implemented in the 2D hydro model as weirs, allowing for the water to release from the crest of the dam wall in case of an overtopping. The culverts have been integrated as well and their dimensions have been estimated based on the estimated flow discharge of P01 at the cross sections upstream of the bridges. The results showed a decrease in the flood extent and the flood depth at the main river where the water depth declined by around 0.5m up to 2m. The dam upstream of Derna waterfalls helps slightly reducing the flow discharge at this section of the river of around 3%. The culverts managed to evacuate flow through the bridges holding water back from accumulating and flowing over the bridges. As culverts almost always cause a contraction of the flow at the entrance and an expansion of flow at the exit, an increase of the back water at the entrance of the culverts have been occurred due to the sudden change of the river geometry.  Figure 6: Flood hazard considering mitigation measures in the urban area of Derna (P5) It is of importance to mention here that because of instability issues of the 2D hydraulic model when integrating the hydraulic structures of P4, a reducing of the computational time step was necessary to keep the hydraulic computations stable and continue the 24h of simulation.For P01 an interval of (max=1, min=0.23s) was used to generate the flood hazard for the 100y while an interval time step of (max=30s, min=30s) was used for P05. The computational time step refers to the time the model spends to compute changes from one cell to another, therefore reducing it leads to an integration of more cells into hydraulic computations which may lead at the end to additional cells to receive a very slight flow volume. This difference has been occurred in the model leading to an increase of the flooded cells in the urban areas (around the buildings), and therefore the potentially exposed buildings has been slightly increased in the P05 scenario. It is worth mentioning that this relates only to the change of this parameter and not to the actual mitigation measures.

European Commission, Joint Research Centre (JRC) (2023): Flood risk and mitigation strategies in Derna (Libya) (2023-10-09). European Commission, Joint Research Centre (JRC) [Dataset] PID: http://data.europa.eu/89h/65909314-0273-575c-9a7e-9126289b8af3

No additional information