Remotely-sensed Enviroclimatic patterns and Ebola outbreaks: linkages and early warning Dan Slayback Jorge Pinzon Compton Tucker 8 September 2004 Biospheric Science Branch, Code 923 NASA Goddard Space Flight Center Greenbelt MD USA
Outline Ebola outbreaks: facts and hypotheses Environmental links to remotely sensed data Spatial & temporal specificity of environmental trigger events Conclusions From: Trigger events: enviroclimatic coupling of Ebola hemorrhagic fever outbreaks JE Pinzon, JM Wilson, CJ Tucker, R Arthur, PB Jahrling, and P Formenty, American Journal of Tropical Medicine and Hygiene. (in press)
Many Ebola outbreaks have occurred in African closed tropical forestMany Ebola outbreaks have occurred in African closed tropical forest Some Ebola outbreaks have occurred in African gallery tropical forest within a savanna matrixSome Ebola outbreaks have occurred in African gallery tropical forest within a savanna matrix
Outbreak Severity
Outbreak Locations
Transmission Scenarios
Possible Ebola Transmission(s) no usual suspects! It is unknown where the virus dwells…
NOAA 11 AVHRR NOAA 7 AVHRR NOAA 9 AVHRR NOAA 14 AVHRR SeaWiFS SPOT MODISes NOAA-16 AVHRR NPOESS NOAA9 Moderate-resolution (1-10’s kms) environmental satellites Environmental Links with Remotely Sensed Data: available datasets
NOAA AVHRR 8-km NDVI Data Set Radiation
VIS/NIR/SWIR Band Comparison AVHRR SeaWiFS SPOT-VGT MODIS
New, improved 8-km AVHRR NDVI data set 1981-present 2003
Major Dataset Differences Global NDVI anomalies
Are there unique environmental characteristics at outbreak sites, during the outbreak year?
Hypothesis: very extreme change from rainy to dry season
1994 Ebola Outbreak Locations NDVI Time Series
Mean time series and anomalies
Spatial signatures and risk
CCA is a method that maximizes the variance between two datasets. Here, we use CCA to identify areas and times that exhibit enviroclimatic signals typical of those at known outbreak sites and dates. CCA(A) = [U,S,V] A: paired-mode correlation matrix between yearly NDVI signals and corresponding NDVI signals from the outbreak sites at outbreak years. U: orthonormal vector of satellite NDVI signals S: percentage of covariance explained by canonical factors V: orthonormal vector of Ebola sites The analysis conducted only over areas with high enviroclimatic correlation to previous outbreak sites (r 2 > 0.95) Canonical Correlation Analysis (CCA)
Trigger event summary
the hot zone