Detecting geothermal anomalies and evaluating LST geothermal component by combining thermal remote sensing time series and land surface model data
This paper explores for the first time the possibilities to use two land surface temperature (LST) time series of different origins (geostationary Meteosat Second Generation satellite data and Noah land surface modelling, LSM), to detect geothermal anomalies and extract the geothermal component of LST, the LSTgt. We hypothesize that in geothermal areas the LSM time series will underestimate the LST as compared to the remote sensing data, since the former does not account for the geothermal component in its model.
In order to extract LSTgt, two approaches of different nature (physical based and data mining) were developed and tested in an area of about 560 × 560 km2 centered at the Kenyan Rift. Pre-dawn data in the study area during the first 45 days of 2012 were analyzed.
The results show consistent spatial and temporal LSTgt patterns between the two approaches, and systematic differences of about 2 K. A geothermal area map from surface studies was used to assess LSTgt inside and outside the geothermal boundaries. Spatial means were found to be higher inside the geothermal limits, as well as the relative frequency of occurrence of high LSTgt. Results further show that areas with strong topography can result in anomalously high LSTgt values (false positives), which suggests the need for a slope and aspect correction in the inputs to achieve realistic results in those areas. The uncertainty analysis indicates that large uncertainties of the input parameters may limit detection of LSTgt anomalies. To validate the approaches, higher spatial resolution images from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data over the Olkaria geothermal field were used. An established method to estimate radiant geothermal flux was applied providing values between 9 and 24 W/m2 in the geothermal area, which coincides with the LSTgt flux rates obtained with the proposed approaches.
The proposed approaches are a first step in estimating LSTgt at large spatial coverage from remote sensing and LSM data series, and provide an innovative framework for future improvements.