Weather analysis Vs. forecast: 20 dBZ threshold comparison
Fig.16 Comparison Real (analysis, cyan) and forecasted (1h before) reflectivity A widely-used index for measuring accuracy in a region recognition in images is the JaccardTanimoto Index. Given the exact shape on grid of the region (in this case, the current weather analysis) and its approximation (in this case the forecast), the Tanimoto Index TI is defined as TI = TP/(FP+FN+TP) or, in other words, the number of “pixels” of intersection on the number of pixels of the union of the two images. A TI of 85-90% or above is usually considered, in image segmentation, a very accurate result. We compared each one of the weather analysis (current weather) in the 4 days considered (96 hours total) with the forecasts for that time from 1 to 6 hours before; then we computed Tanimoto index for each and the total clouds coverage. These calculations were made for the data with threshold at 20 dBZ (level 1 or more clouds) and for the data with threshold at 36 dBZ (level 2 or more clouds). In table below are reported the average results: Jaccard-Tanimoto Index – 20 dBZ threshold 1.Hour 2.Hour 3.Hour Clouds Forecast Forecast Forecast 0,094122 0,910662 0,596208 0,469113
Clouds
Jaccard-Tanimoto Index – 36 dBZ threshold 1.Hour 2.Hour 3.Hour
4.Hour Forecast 0,391596
.5Hour Forecast 0,33863
6.Hour Forecast 0,301704
4.Hour
5.Hour
6.Hour
71