Mapping potential risk of Rift Valley fever outbreaks in African savannas using vegetation index time series data

Rift Valley fever (RVF) outbreaks in East Africa are closely coupled with above normal rainfall that is associated with the occurrence of the warm phase of the El Niño/Southern Oscillation (ENSO) phenomenon. Outbreaks elsewhere in central and southern Africa are also linked to elevated rainfall patterns. Major RVF activity has been reported to occur throughout much of sub-Saharan Africa, except in areas with extensive tropical forest. In this study we used normalized difference vegetation index (NDVI) time-series data derived from the Advanced Very High Resolution Radiometer (AVHRR) instrument on polar orbiting National Oceanographic and Atmospheric Administration (NOAA) satellites to map areas with a potential for an RVF outbreak. A 19-year NDVI climatology was created and used to discriminate between areas with tropical forest, savanna, and desert. Because most RVF outbreaks have occurred in regions dominated by savanna vegetation, we created a mask to identify those areas where RVF would likely occur within the savanna ecosystems. NDVI anomalies were then calculated for the entire time series from July 1981 to the July 2000. Subsequently, we developed a methodology that detects areas with persistent positive NDVI anomalies (greater than + 0.1 NDVI units) using a three-month moving window to flag regions at greatest risk. Algorithms were designed to account for periods of extended above normal NDVI (by inference rainfall) and to consider the complex life cycle of mosquitoes that maintain and transmit RVF virus to domestic animals and people. We present results for different ENSO warm- and cold-event periods. The results indicate that regions of potential outbreaks have occurred predominantly during warm ENSO events in East Africa and during cold ENSO events in southern Africa. Results provide a likely historical reconstruction of areas where RVF may have occurred during the last 19 years. There is a close agreement between confirmed outbreaks between 1981 and 2000, particularly in East Africa, and the risk maps produced in this study. This technique is adaptable to near real-time monitoring on a monthly basis and may be a useful tool in RVF disease surveillance.