Industrial Maths study group for AI and Health: can AI and EO help predict malaria outbreaks?
One of the problems for KTN’s Industrial Maths in AI and Health event has been announced and researchers are encouraged to apply to take part.
This is one of several problems that will be discussed at a three-day study groups in Cardiff from 22-24 May.
If you are a researcher working in a UK university who would like to work on this problem, please register here.
The organisers, KTN, alongside the University of Cardiff, are looking for researchers to work on the following conundrum:
AI and EO for the prediction of malaria outbreak risk presented by isardSAT
Malaria is still regarded as a major public health challenge undermining development throughout the world. The primary aim of the Malarsat Project was to provide maps of suitable areas for mosquitoes reproduction in views of global malaria early warning, based on Earth Observation (EO) data. This would help prevent and support epidemic episodes, allowing for improved efficiency of insecticide programs, vaccine campaigns and epidemic treatment support.
Whilst meteorological maps play a key part in the development of these maps, they alone are not sufficient to provide accurate and reliable data. Malaria-bearing mosquitoes primarily breed in new water puddles under a meter wide, making it especially difficult to detect with Earth Observation satellites. However, when considering a group of these puddles, their detection becomes much higher in possibility. The use of radar altimetry data (ie. Sentinel-3), coupled with optical data (ie. Sentinel-2, cloud dependent) and the added input of meteorological data, opens up the possibility of being able to produce near real-time malaria maps. Soil moisture is a recently available EO variable that has been shown to be a reliable proxy for precipitation and has been successfully used for desert locust management.
In places where meteorological data is not available or not precise enough, additional EO data (i.e. land surface temperature, precipitation) could also play a role to find suitable breeding places for mosquitoes. The main challenge here is making the optimal use of all data and developing a viable, robust and reliable method to create the malaria risk infection maps. Previous attempts of this have proven difficult, primarily due to the size of the water puddles and lack of signal retrieved from them.
Resources available for challenge
The addition of AI could prove valuable to being able to produce the malaria risk infection maps. The solution would be able to distinguish radar echoes coming from areas with water puddles, using the shape of the waveform from Sentinel-3 to do so. Ideally, it would have the capability to combine different datasets and make a classification system, detecting pools of water, taking into account local conditions, and subsequently creating higher accuracy malaria risk infection maps. These maps could become invaluable to forecasting where malaria outbreaks will happen, monitor their progress and potentially prevent or minimise their impacts, helping streamline and save on- the-ground resources, and most importantly, save lives.
If you’re a university researcher and think you could play a part in helping with this challenge, sign up for here.