River Flood

Floodplains of Northern Ghana. Photo: Neil Palmer/CIAT

ARC has been developing a pan-African flood model which will be used to underpin parametric flood insurance offered to its Member States. The parametric insurance product for floods will be the first of its kind globally. ARC has partnered with leading US-based atmospheric and environmental research firm AER to develop the underlying flood model, while ARC will create the overlying insurance product as part of Africa RiskView. At the end of 2016, AER completed the pan-African Flood Extent Depiction (AFED) product for ARC using historical daily data starting in 1992. The product includes automated near real-time AFED processing streams to produce Africa-wide daily updates of AFED.

The insurance product will work in a similar manner to ARC’s current drought insurance product. The model will convert a daily Africa Flood Extent Depiction into an economic impact value in US dollars creating an Africa RiskView flood index. The insurance policies will be issued on the basis of this index. Due to the complexity of flood modelling, the model only focuses on a subset of flood risks:

  • Extensive river floods that last at least several days in duration
  • Insurance will target the largest of these events, that would have the most direct impact on people

How does the ARC Flood Model work?

The ARC Flood Extent Depiction Model (AFED) is a daily depiction of temporarily flooded areas everywhere in Africa. AFED uses satellite remote sensing from microwave sensors to map floods. The sensors measure the intensity of microwave radiation emitted naturally from the earth, with a decrease in radiation signalling that flooding is occurring. The microwave sensors provide twice-daily measurements and make measurements day and night with the ability to see through non-precipitating clouds and to sense flooding beneath vegetation. The current data comes from three satellite sensors: SSM/I; AMSR2 and GMI, with older SSM/I data and another sensor AMSR-E adding to the historic data record.

The microwave data does however have some limitations in terms of the limitations in resolution and rain interference with the signal. To overcome this, the model uses elevation data to downscale the microwave data and depict floods at a finer scale (higher resolution). The model also uses rain detection to produce flood depictions from the best-quality data. AFED also includes a land-water mask to exclude persistent open water. The final AFED is a daily set of georeferenced images tiling together to cover all of Africa, with each image pixel (3 arcseconds or about 90m x 90m) or
grid cell representing whether an area was flooded or not that day. The model includes sets of images which date back to January 1988. The model has some limitations as some floods are too small or short-lived to detect definitively. A
small flood may not be picked up if the model cannot distinguish the flood’s microwave signal from the natural variation in dry-land conditions.

Pilot Countries

In 2017, the Africa Flood Extent Depiction model was piloted in Ghana, the Gambia, and Cote d’Ivoire. Dedicated technical experts formed in-country technical working groups that modelled and validated AFED in their countries, fine-tuning the model and paving the way for a robust flood parametric insurance product across Africa.