| << | Profiling of Clouds | >> |
Abstract
Geostatistical methods (kriging) aim at estimating the average value. In case of sparse measurements, such fields are too smooth. This can lead to biases in radiative transfer calculations on such a kriged field. Stochastic modelling, e.g. surrogate data, aims at reproducing the structure of data. Surrogate clouds from (profiling) measurements enable us to perform studies on empirical clouds that otherwise may be performed with clouds from numerical models.
Surrogate clouds are well-suited for 3D radiative transfer studies. However, up to now we could only achieve good results for the radiative properties averaged over the field, but not for a radiation measurement located at a certain position. Therefore we have developed and tested a new (so-called conditioned) algorithm that combines the high-quality structure of stochastic (surrogate) modelling with the positioning capabilities of kriging.
Preliminary results on pseudo profiling measurements simulated on LES clouds show that these new surrogate clouds reproduce the structure of the original clouds very well and the minima and maxima are located where the pseudo-measurements sees them. The root mean square error is reduced by a factor three compared to unconditioned surrogate clouds; that means that the number of case studies can be reduced by about a factor nine. The main limitation seems to be the amount of data, which is especially very limited in case of just one zenith-pointing measurement; scanning profiling cloud measurements are clearly very valuable.