Monitoring of climate variables such as air temperature is gaining increasing importance under climate change. This study aimed at developing an hourly gridded 3.5 x 3.5 km air temperature (T-air) data set for entire South Africa. In a Random Forest approach, MSG SEVIRI data from 2010 to 2014 were used and related to T-air measured by 78 weather stations. An external validation on new climate stations and years that were not used for model training indicated the ability of the model to predict T-air with a RMSE of 2.61 degrees C and a R-2 of 0.89. The resulting model can be applied to the entire MSG SEVIRI time series since 2004. It hence allows for spatiotemporal pattern analysis as well as for the detection of trends which is relevant in the context of climate change.
Key words
Air temperature/Climate/Machine leaming/Meteosat/Random Forest/South Africa
引用本文复制引用
出版年
2019
International journal of applied earth observation and geoinformation
NSTL
Elsevier