Heidke Skill Score (HSS) values for each verification season from Oct-Dec 1999 onwards for the Interior Southwest domain Each core region was verified as a "Hit" if the correct tercile was predicted, and as a "Miss" if the wrong tercile was predicted, or if the predicted tilt in the odds was under 3% (be it for above-normal, near-normal, or below-normal moisture in the tercile sense), or if the prediction was ambiguous (3% or higher tilt in BOTH tail terciles). There are two skill scores: if all missed forecasts (in the above sense) are counted as a total loss, the Heidke Skill Score is labeled "non-EC HSS", and is followed in the table below by its areal coverage (=number of cores that had an unambiguous tilt of 3% or more, as percent of total number of cores in that season). If all ambiguous or small-tilt (under 3%) predictions are considered "climatological forecasts", they are scored with a 1/3 hit rate, and added into the "all HSS" score (=number of "Hits", plus 1/3 times the number of small-tilt (and ambiguous) predictions, divided by the total number of cores in that calendar season, times 100%). Skill is computed at 1/2 month lead-time (issued in the month just prior to the target season). Since not all of the station data is in hand at this point, some skill scores may still change down the road. However, that should not result in a systematic bias. The chosen lead-time reflects the best skill performance from July through December, while longer-lead forecasts appear to work better for the first half of the year. Areal coverage is computed by assuming that each forecast region is about the same size, hence, e.g., six out of nine regions translate into coverage of 67%, etc. YEAR TARGET FORECAST SKILL SEASON non-EC areal all HSS coverage HSS 1999 OND +25 67% +17 2000 JFM +06 89% +06 2000 AMJ 0 67% 0 2000 JAS +36 70% +25 2000 OND -50# 83% -42# 2001 JFM +06 89% +06 2001 AMJ 0 67% 0 2001 JAS +33 90% +30 2001 OND +100* 67% +67* 2002 JFM 0 100% 0 2002 AMJ +14 78% +11 2002 JAS -13# 80% -10# 2002 OND -50# 67% -33 2003 JFM -13# 89% -11# 2003 AMJ +36 78% +28 2003 JAS -13# 70% -10# 2003 OND 0 50% 0 2004 JFM -13# 89% -11# 2004 AMJ -13 89% -11 2004 JAS +44 80% +35* 2004 OND +10 83% +08 2005 JFM 0 100% 0 2005 AMJ +44* 89% +39* 2005 JAS +33 90% +30 2005 OND +63 67% +42 2006 JFM +50* 100% +50* 2006 AMJ +40 56% +22 2006 JAS +50* 60% +30 2006 OND 0 50% 0 2007 JFM +17 100% +17 2007 AMJ -29# 78% -22# 2007 JAS +10 50% +05 2007 OND +10 83% +08 2008 JFM -13# 89% -11# 2008 AMJ +25 67% +17 2008 JAS +10 100% +10 2008 OND -20 83% -17 2009 JFM +06 89% +06 2009 AMJ -25 67% -17 2009 JAS +25 80% +20 10yr AVG OND +09 70% +05 10yr AVG JFM +05 93% +05 10yr AVG AMJ +09 73% +07 10yr AVG JAS +22 78% +16 For each calendar season, the best (worst) individual forecast is flagged by an asterisk (*) and pound symbol (#), respectively. Reminder: HSS = -50 = every single prediction was in the wrong tercile = WORST possible case, HSS = 0 = one third of all predictions were in the correct tercile = "coin-toss", HSS = +50 = two out of three predictions were correct = "very good", HSS =+100 = every single prediction was a "hit" (OND2001!) = BEST possible case. Alternate computation of skill, averaging only those forecasts that were tilted by at least 3%, and computed for each region individually, then averaged across the domain - note that earlier forecasts often have less frequent tilts at that level, leading to an inflated score; this is for verified forecasts through JAS'09, or ten cases: OND: +09 JFM: +08 (at 3.5: +22 - the best score at any lead or for any season) AMJ: +09 JAS: +20