Predicting malaria seasons in Kenya using multitemporal meteorological satellite sensor data.
Hay SI., Snow RW., Rogers DJ.
This article describes research that predicts the seasonality of malaria in Kenya using remotely sensed images from satellite sensors. The predictions were made using relationships established between long-term data on paediatric severe malaria admissions and simultaneously collected data from the Advanced Very High Resolution Radiometer (AVHRR) on the National Oceanic and Atmospheric Administrations (NOAA) polar-orbiting meteorological satellites and the High Resolution Radiometer (HRR) on the European Organization for the Exploitation of Meteorological Satellites' (EUMETSAT) geostationary Meteosat satellites. The remotely sensed data were processed to provide surrogate information on land surface temperature, reflectance in the middle infra-red, rainfall, and the normalized difference vegetation index (NDVI). These variables were then subjected to temporal Fourier processing and the fitted Fourier data were compared with the mean percentage of total annual malaria admissions recorded in each month. The NDVI in the preceding month correlated most significantly and consistently with malaria presentations across the 3 sites (mean adjusted r2 = 0.71, range 0.61-0.79). Regression analyses showed that an NDVI threshold of 0.35-0.40 was required for more than 5% of the annual malaria cases to be presented in a given month. These thresholds were then extrapolated spatially with the temporal Fourier-processed NDVI data to define the number of months, in which malaria admissions could be expected across Kenya in an average year, at an 8 x 8 km resolution. The resulting maps were compared with the only existing map (Butler's) of malaria transmission periods for Kenya, compiled from expert opinion. Conclusions are drawn on the appropriateness of remote sensing techniques for compiling national strategies for malaria intervention.