Billabong Posted December 4, 2011 Share Posted December 4, 2011 I can infer the higher the numbers, the better the performance.... but can someone give a good explanation on how to properly read these verification charts and what the numbers mean? Also, what model are the blue plots from? Lastly, does anyone have the link to the main page for these charts? Thanks in advance! Link to comment Share on other sites More sharing options...
isohume Posted December 4, 2011 Share Posted December 4, 2011 The charts are showing the correlation of the respective modeled anomalies to a 30-year climo anomalies (1959-1988) for the specific variable selected. In this case, H5. Basically, how well the model did outside of the climo mean. The FNO is the NOGAPS. These two pages have the charts and more info... http://www.emc.ncep....TATS/STATS.html http://www.emc.ncep....gmb/STATS_vsdb/ Link to comment Share on other sites More sharing options...
Billabong Posted December 4, 2011 Author Share Posted December 4, 2011 Great, thanks! Link to comment Share on other sites More sharing options...
kylemacr Posted December 4, 2011 Share Posted December 4, 2011 To add a little to that... an anomaly correlation coefficient (ACC) greater than 0.6 is usually considered skillful. This idea came around when the ECMWF started looking at ACC and they determined that the number which separated subjectively good forecasts vs. bad forecasts was 0.6. Link to comment Share on other sites More sharing options...
Billabong Posted December 4, 2011 Author Share Posted December 4, 2011 So it's basically a percentage. The 0.6 usefulness threshold is just 60% of a perfect forecast? Link to comment Share on other sites More sharing options...
dtk Posted December 4, 2011 Share Posted December 4, 2011 The charts are showing the correlation of the respective modeled anomalies to a 30-year climo anomalies (1959-1988) for the specific variable selected. In this case, H5. Basically, how well the model did outside of the climo mean. The FNO is the NOGAPS. These two pages have the charts and more info... http://www.emc.ncep....TATS/STATS.html http://www.emc.ncep....gmb/STATS_vsdb/ Also, note that the figures/numbers from the old (first) website and second ("vsdb") are not the same. The newer site (which we have going back to 2006) uses the actual 30 year climo (the old statistics do not) from reanalysis as well as unfiltered/untruncated forecast output. There is a powerpoint linked from the first URL above that describes how the methodology has been update from the old to new method/webpage. Link to comment Share on other sites More sharing options...
dtk Posted December 4, 2011 Share Posted December 4, 2011 Also, description of AC from ECMWF webpage here: http://www.ecmwf.int/products/forecasts/guide/Measure_of_skill_the_anomaly_correlation_coefficient.html Link to comment Share on other sites More sharing options...
kylemacr Posted December 4, 2011 Share Posted December 4, 2011 It's also important to note that this is not a perfect measurement of verification. Nothing is. The ACC rewards spatial similarities more than value similarities, which leads to higher scores when the flow is amplified (waves). Link to comment Share on other sites More sharing options...
isohume Posted December 4, 2011 Share Posted December 4, 2011 It's also important to note that this is not a perfect measurement of verification. Nothing is. The ACC rewards spatial similarities more than value similarities, which leads to higher scores when the flow is amplified (waves). Haha...yeah that'd be impossible. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.