Cb-Tcu classification based on radar and satellite observations

Cumulonimbus and towering cumulus clouds form a threat for save aviation conditions. Their 
detection in the vicinity of airports is an ICAO requirement. The detection is still predominantly 
done by human observers. At KNMI an automated Cb detection algorithm is developed based on 
radar and satellite observations. It became operational in 2011 at three airports in the Netherlands.
This report describes the derivation and performance of the Cb-classification software as used at 
the KNMI. It details the 2014 version of the software. The 2014 version substitutes the 2010 
version, De Valk and van Westhrenen, 2010. This manual should be read in conjunction with the 
2010 report and the operational manual, De Valk and van Westrhenen 2011. 
The report and especially its appendices should enable the reader to develop an automatic Cb-classification from scratch based upon radar and satellite observations at any location within the viewing geometry of the SEVIRI instrument on board of METEOSAT. It requires the coverage of the area of interest by both radar and satellite observations and a supervised Cb classification over the same area. They serve as input for predictor and their coefficients determination which is based upon an algorithm that determines the predictor values over the areas of interest. The 
algorithm should produce contours, averages, minimum and maximum values of the observed 
values.
To derive the predictors and their coefficients an open source statistical software package “R” is 
used. The probabilistic equation based on the predictors and their coefficients enables a Cb 
classification. The choice by the developer of probability thresholds will determine the performance characteristics.
Improvements of the 2014 version in comparison to the De Valk and van Westhrenen, 2010 
version are:
- extension of the number of Aerodromes to 26 for which a classification is produced
- slight modifications of the predictors and their coefficients in comparison to the 2010 version. 
Even though the modifications were slight the underlying research was thorough.
- clutter, i.e. false radar echos, impacted the results over sea in the summer. A rudimentary clutter removal has been applied to reduce the impact on the performance