Research
Our Project
Rift Valley Fever (RVF) is a viral mosquito-borne zoonosis. Though it primarily affects animals, humans can also contract the disease, usually through contact with infected livestock. The counties of Wajir, Garissa, Isiolo and Marsabit in Northern Kenya have experienced RVF outbreaks, which can cause serious disruptions to economic and agricultural systems, as well as loss of animal and human lives.
This project is using Earth Observation (or "EO") data from aerial drone and satellite photos, so we can study how animal distributions and landscape features influence RVF transmission. Drones can be used to investigate the environmental factors that lead to RVF outbreaks, and monitor small-scale changes which can’t be picked up when looking at satellite data due to frequent cloud cover and how often images are taken. Drones can also be used to map wildlife and livestock that carry disease. This project will explore how new sources of EO data and advances in machine-learning methods of data analysis can help us understand RVF virus transmission and develop more accurate risk maps of where Rift Valley Fever outbreaks are likely to take place.
Check out our "Education" page to learn more about Rift Valley Fever and the technologies we are using!
Objectives and activities
1) Evaluate the ability of aerial images to identify distributions of livestock and wildlife.
We’re taking aerial photos from each of our study sites and stitching them together to create “orthomosaics” – drone maps that show the whole area that the drone captured. These huge images are split into small tiles, each a 15m x 15m square, before being uploaded here to Zooniverse. Using our field guide and talk boards for help, we’re asking the citizen scientists at Zooniverse to count and locate animals appearing in the drone images. We’ll then look at these results of your classifications and use them to estimate the distribution of livestock and wildlife during our study period.
2) Develop fine-scale maps of water bodies and land cover characteristics influencing mosquito densities and livestock and wildlife distribution.
We’re going find the land features and waterbodies that play a role in the transmission of Rift Valley Fever with the help of people from the study communities, as well as mosquito surveys conducted by scientists in Kenya. We’ll plug these results into a model built using machine learning methods to identify the most important features. We’ll also look at including additional data from satellite images, to help us refine our land cover maps and to look at changes in water bodies over time.
3) Predict areas with high overlap between mosquito vectors, livestock, and wildlife.
We’ll then develop models to predict the distribution of livestock, wildlife, and the mosquito vector-species within selected study sites. We’ll look at how landscape affects animal and vector abundance, including variables like climatic conditions, disease control activities, and human population distributions. Next, we’ll estimate how Rift Valley Fever virus moves between wildlife, livestock, and mosquito populations, identifying the areas and times where exposure to the disease is highest. To do this we’ll develop a model using methods that can learn from the data to infer unaccounted for factors, like how much Rift Valley Fever there is in our study areas, and how infectious animals are.