Malaria Alerts From Space

Documenting some of these factors, such as soil type and local bucket-leaving habits, requires initial groundwork by researchers in the field, Welch notes. This information is plugged into a computerized mapping system called a Geographical Information Systems database (GIS). Fieldwork is also required to characterize how the local species of mosquito behaves. Does it bite people indoors or outdoors or both? Other factors, like the locations of cattle pastures and human dwellings, are input into the GIS map based on ultra-high resolution satellite images from commercial satellites like Ikonos and QuickBird, which can spot objects on the ground as small as 80 cm across. Then region-wide variables like temperature, rainfall, vegetation types, and soil moisture are derived from medium-resolution satellite data, such as from Landsat 7 or the MODIS sensor on NASA's Terra satellite. (MODIS stands for MODerate-resolution Imaging Spectrometer.)

In the Mewat region of India livestock are an important part of the predominantly rural, subsistence economy. The presence of livestock facilitates the spread of malaria and other mosquito-borne diseases, because the insects prefer to feed on the animals' blood.

Scientists feed all of this information into a computer simulation that runs on top of a digital map of the landscape. Sophisticated mathematical algorithms chew on all these factors and spit out an estimate of outbreak risk.

The basic soundness of this approach for estimating disease risk has been borne out by previous studies. A group from the University of Nevada and the Desert Research Institute were able to "predict" historical rates of deer-mouse infection by the Sin Nombre virus with up to 80% accuracy, based only on vegetation type and density, elevation and slope of the land, and hydrologic features, all derived from satellite data and GIS maps. A joint NASA Ames / University of California at Davis study achieved a 90% success rate in identifying which rice fields in central California would breed large numbers of mosquitoes and which would breed fewer, based on Landsat data. Another Ames project predicted 79% of the high-mosquito villages in the Chiapas region of Mexico based on landscape features seen in satellite images.

Perfect predictions will likely never be possible. Like weather, the phenomenon of human disease is too complicated. But these encouraging results suggest that reasonably accurate risk estimates can be achieved by combining old-fashioned fieldwork with the newest in satellite technologies.

"All of the necessary pieces of the puzzle are there," Welch says, offering the hope that soon disease outbreaks that seem to come "from out of nowhere" will catch people off guard much less often.