2016/03/04 by Michael Weniger, Weniger, Michael, Petra Friederichs +1
Computer Science · Engineering · Environmental Science · #Atmospheric aerosols and clouds #Atmospheric and Environmental Gas Dynamics #Atmospheric and Oceanic Physics (physics.ao-ph) #FOS: Physical sciences #Geochemistry and Geologic Mapping #Remote Sensing in Agriculture #Remote-Sensing Image Classification
paper · pdf · doi:10.48550/arxiv.1603.01656
openalex publication_date 2016/03/04 · openalex created_date 2022/10/05 · openalex updated_date 2026/07/28
The feature based spatial verification method SAL is applied to cloud data,\ni.e. two-dimensional spatial fields of total cloud cover and spectral radiance.\nModel output is obtained from the COSMO-DE forward operator SynSat and compared\nto SEVIRI satellite data. The aim of this study is twofold. First, to assess\nthe applicability of SAL to this kind of data, and second, to analyze the role\nof external object identification algorithms (OIA) and the effects of\nobservational uncertainties on the resulting scores.\n As a feature based method, SAL requires external OIA. A comparison of three\ndifferent algorithms shows that the threshold level, which is a fundamental\npart of all studied algorithms, induces high sensitivity and unstable behavior\nof object dependent SAL scores (i.e. even very small changes in parameter\nvalues can lead to large changes in the resulting scores). An in-depth\nstatistical analysis reveals significant effects on distributional quantities\ncommonly used in the interpretation of SAL, e.g. median and interquartile\ndistance. Two sensitivity indicators based on the univariate cumulative\ndistribution functions are derived. They allow to asses the sensitivity of the\nSAL scores to threshold level changes without computationally expensive\niterative calculations of SAL for various thresholds. The mathematical\nstructure of these indicators connects the sensitivity of the SAL scores to\nparameter changes with the effect of observational uncertainties.\n Finally, the discriminating power of SAL is studied. It is shown, that - for\nlarge-scale cloud data - changes in the parameters may have larger effects on\nthe object dependent SAL scores (i.e. the S and L2 scores) than a complete loss\nof temporal collocation.\n