Laserfiche WebLink
<br />categories (The "other" category included weak frontal <br />system, advection of tropical moisture, etc.) Now 78.3% of <br />the cases reside on the diagonal with no off-diagonal cell <br />having greater than 3.5% of the counts. <br /> <br />There is some over-forecasting of the "none" or no <br />front category (i.e., 58.2 versus 52%) and under-forecasting <br />of cold fronts and cutoff lows. Nevertheless, over all <br />forecasts there was indication of considerable forecaster <br />skill in that both the CPOD and the TSS values were positive <br />and sizable (i.e., .60 and .73 respectively). The breakout <br />by season also shows evidence of substantial and consistent <br />forecaster skill. <br /> <br />E. Forecast Inputs <br /> <br />One of the interesting questions in all forecasting <br />problems is how important were the various inputs to the <br />final forecasts? Table 6 presents the results for this <br />~~estion as assessed by the forecasters. For all days <br />combined, the satellite (45%), NMC numerical guidance (22%), <br />and radiometer inputs (12%) are the most important. If one <br />separates the local from the regional or national scale <br />i~puts then 75% of the forecast is based on regional or <br />national data. The more specific local inputs (e.g., <br />Sheridan radar, (4%) local rawinsondes (4%), icing probes <br />(3%) etc.) are not very important. <br /> <br />However, when the forecast days are partitioned by the <br />p~esence or absence of storms on the barrier, the results <br />are rather different. On Storm Days the synoptic or larger <br />scale data systems (e.g., satellite and numerical guidance) <br />get decreased attention and the more local systems (e.g., <br />She~idan radar, icing probes, etc.) get inc~eased attention. <br /> <br />12 <br />