Mapping of spatiotemporal changes across the East African Rift System to identify geothermal anomalies using MODIS land surface temperatures

<div>A range of satellite datasets, including MODIS land surface temperature (LST), are</div><div>used to identify geothermal anomalies associated with rift basins across the East African</div><div>Rift System. Monthly and yearly absolute LST means are generated from January 2003</div><div>to December 2013 and show regions of warmer LSTs in relevant basins. However,</div><div>without auxiliary data it is not possible to show that these are related to geothermal</div><div>anomalies. Two approaches are applied to delineate the LST more clearly - principal</div><div>component analysis (PCA) and normalisation of the LST with respect to elevation. The</div><div>first technique uses PCA to delineate the known physical parameters influencing LST</div><div>and reveals elevation to be dominant. Consequently, steps have been taken to minimise</div><div>the effects on LST. This has been achieved via normalisation, whereby absolute LST is</div><div>recalculated, using linear regression analysis, to equivalent normalised LST at an</div><div>elevation of 0 m. Several previously masked areas, including the Ethiopian Dome, have</div><div>since been revealed as warmer and with an increased likelihood of relationship to</div><div>geothermal heat flux since they correspond to emissivity and tectonic patterns. Note the</div><div>impressive manner in which volcanoes including Mount Elgon, cold in absolute LST</div><div>because of elevation, are also identified as warmer post normalisation. Caution must</div><div>still be exercised with respect to the warm anomalies in normalised LST, as these can</div><div>still not be conclusively confirmed as geothermal anomalies. A restricted PCA of the</div><div>normalised LST shows that these are still sensitive to emissivity as expected but</div><div>particularly in a well-defined region around Lake Turkana. In conclusion, the likelihood</div><div>of identifying a geothermal anomaly is best associated with the normalised LST and</div><div>where high frequency spatial structure is observed. Identified regions should be checked</div><div>against the restricted PCA. Future work should incorporate the use of other indicators of</div><div>geothermal activity or heat flux to better identify the LST variance that corresponds to</div><div>geothermal anomalies.</div>