Real-time Verification of CPC Long-Lead Forecasts Using Heidke Skill Scores from Station Data All Stations Non-EC Only % Coverage Season T 6.8 34.8 19.7 SON 05 P -3.4 -20.0 17.1 Month (LL) T 12.8 100.0 12.8 NOV 05 P 0.0 ----- 0.0 Updated Mo. T 36.8 87.8 41.0 NOV 05 P 0.0 0.0 5.1 Change T +24.0 -12.2 +28.2 (U-LL Mo.) P 0.0 ----- +5.1 Brief Discussion of Forecasts and Verification ---------------------------------------------- The temperature forecast for SON 2005 was the fifth consecutive seasonal forecast to have positive skill. Heidke skill scores were 6.8 for all stations and 34.8 for non-EC only. Both scores were not as high as those for the two previous seasons, as smaller areas were forecast and with somewhat less skill as the seasons progress into a time of year with lower a priori skill. Although the area of above normal temperatures forecast for the Pacific Northwest coast was all wrong, the other two areas of Above normal forecast had positive skill sufficient to give a net modestly good score for the CONUS as a whole. In fact, every station in the interior Southwest that was predicted to be above normal verified as correct. The tool relied on most strongly was trend, as reflected in the OCN tool, which along with CCA gave strong indications for the warmth in the Southwest. The forecast of relative warmth in the Pacific Northwest was based primarily on the CAS and SMLR tools that incorporate soil moisture. The Pacific Northwest had been in the midst of a serious long-term drought for the past several years, but a more active regime with typical fall onset of rains developed this year, and cooler air from the Gulf of Alaska often came in behind the storms. Various climate models gave mixed forecasts, with some indicating warmth and some cooler than normal conditions, so as a whole they were not weighted very heavily in the forecast. A new tool that is a consensus of several of the better individual temperature tools was also used, but it was not weighted heavily due to lack of a track record. A prediction of the above normal category that was made for the southern coastal areas of Alaska, which was based primarily an extensive areas of above normal SST that developed under persistent anomalous ridging that had been in place most of the summer, had only modest success, with an all-station score of 2.3 and a non-EC score or 4.6 with a coverage of 50.0%. Both September and October were well above normal, but at the end of October the circulation regime changed and a very cold November knocked many stations down into the near normal category even with the first two months of the fall being well above normal. Only one very small area of above median precipitation was forecast over Florida and the South Atlantic Coast, on the basis of long-term trends and the expectation that the active hurricane season would remain that way into the fall. Even though the hurricane season did remain active beyond expectations with a record number of named storms by the end of November, including two major systems hitting the U.S. mainland in the fall, it was dry most of the rest of the time and only one station in Florida reported above median precipitation for the fall as a whole. The area predicted to have below median precipitation in the interior Pacific Northwest and northern and central Great Basin was based primarily on indications from the OCN tool that reflects long-term trends, with some support from the CAS tool. Unfortunately less than a third of the stations predicted to be dry in this area were correct, resulting in a negative overall score on the below median and the precipitation forecast as a whole. Model forecasts of precipitation were also in some disagreement and were not weighted heavily in the final forecast. Alaska was given a forecast of EC everywhere for precipitation. The forecast for November 2005 made at approximately 0.5-month lead on Oct.20 was calling for above normal temperatures over a medium-sized area extending east-northeastward from most of New Mexico and the northern half of Texas to Missouri and Arkansas, on the basis of primarily CCA, with weak support from OCN and SMLR. The CFS model, which indicated widespread warmth centered in the south-central part of the country, was also considered and used Every station forecast to be above normal was correct, resulting in skill scores of 12.8 for all stations and a perfect 100.0 for non-EC only in the CONUS. The temperature forecast for Alaska was however a disaster, with the entire state being predicted above normal, but only two stations in the southern part of the Panhandle that were still influenced primarily by maritime flow actually being above normal. The forecast was based on a consensus of weak but agreeing statistical tools, and unfortunately the CFS model, which indicated below normal temperatures for the entire State, was disregarded, perhaps because of previous tendencies to have a cold bias. Heidke still scores for the Alaska temperature forecast were -36.4 for both all stations and non-EC only, since every stations was forecast to have above normal temperatures, giving a coverage of 100%. The 0.5-month lead precipitation outlook for November was given EC everywhere in both the CONUS and Alaska, due to weak, conflicting, or non-existent indications from both statistical and dynamical tools. Scores were therefore 0.0 for all stations and indeterminate for non-EC only. The revised forecast made on October 31 with input from several dynamical models used for medium range forecasting produced what at first might seem to be contradictory results, as far as skill is concerned. The all-station temperature skill increased dramatically to 36.8 due to the much larger area skillfully forecast, but the non-EC only score decreased slightly from a perfect 100 to a still fantastically good 87.8 since it would be almost impossible to still have every station correct in a large area including 41.0% of the stations in the CONUS. The area of predicted anomalous warmth was correctly extended northward and eastward to the Great Lakes Region and the Appalachians, based on very good indications from the GFS models used in medium range forecasting that the first two weeks of November would be warm over much of the eastern and central U.S. Since these models also indicated below normal temperatures in Alaska, in agreement with the CFS model's indications from ten days earlier, the forecaster removed the above normal from Alaska and gave the entire State EC., which was a big improvement from the earlier long-lead forecast, even though no below normal temperatures were forecast due to previous experience that models often overpredicted cold regimes in Alaska. The revisions to the precipitation forecast, on the other hand, were based primarily on the basis of good medium-range forecast guidance for the first few two weeks of the month that correctly predicted and active trough over the Gulf of Alaska that generated storms hitting both the Pacific Northwest and the Alaska Panhandle. In spite of this being basically correct, the precipitation skill scores for the CONUS remained at 0.0 for all stations and were also 0.0 for non-EC only, since only two of the six stations forecast to be wet in the Pacific Northwest verified as correct. All three of the stations predicted to be wet in the Alaska Panhandle were correct however, resulting in scores of 13.1 for all stations and 100.0 for non-EC only in Alaska. --- EOF