Urban spatial ground heat island may be detected from airborne or satellite infrared images. Such an image was utilized to study the structure of the city center of Johannesburg. This was done by applying the Distance Correlation Matrix (DISTCORMAT) method. However, the thermal input contains values which represent the ground (roads, pavements, vegetation etc.) and roof levels alike. Partitioning between these two levels is not straight-forward, and some suggestions to differentiate between them have been applied.
Lately, an image processing technique was presented which removed the roof top of buildings. It is estimated that the partitioning accuracy is between 81% and 92%. The results are published elsewhere.
The present presentation deals with the mid-day heat island of Munich, Germany. Since the IR spatial resolution was greater for Munich (6.3m at Nadir) than for Johannesburg (1.6m), different methods of ground/roof partitioning were applied. The results of the DISTCORMAT analysis of both levels and for each separately revealed that there is quite a substantial distance between the centers of these three DISTCORMAT outputs.