Real-time Verification of CPC Long-Lead Forecasts Using Heidke Skill Scores from Station Data All Stations Non-EC Only % Coverage Season T 4.3 11.4 37.6 MAM 05 P -8.5 -20.0 42.7 Month (LL) T 0.4 1.1 37.6 MAY 05 P 0.9 10.0 8.5 Updated Mo. T 1.7 3.8 45.3 MAY 05 P 10.7 43.1 24.8 Change T +1.3 +2.7 +7.7 (U-LL Mo.) P +9.8 +33.1 +16.3 Brief Discussion of Forecasts and Verification ---------------------------------------------- The temperature forecast for MAM 2005 had modest positive Heidke skill scores of 4.3 for all stations and 11.4 for non-EC only, with a fairly larger coverage of 37.6 %. These were both however slightly below the long-term averages of 9.0 for all stations and 19.1 for non-EC only. Coverage for this forecast was slightly less than the average of 47.0% for the entire period of record. The MAM 05 seasonal forecast was based on a variety of models, including the new CFS model from NCEP, and multi-model ensembles from IRI and CDC, as well as CPC's CCA and OCN statistical tools. Since the influence of the weak central Pacific Basin El Nino was expected to be gradually fading during the season, trend-adjusted ENSO composites, which would have favored the below normal temperatures that were observed in the Southeast, were not as heavily weighted into the forecast for that area. It is likely that the cool conditions in the Southeast were aided by both persistently below normal SSTs in the area and the relatively wet conditions that developed during the early spring and recurred throughout the season, particularly in the Florida Panhandle. Below normal temperatures over the rest of the East were due largely to frequent strongly negative phases of the NAO and AO, which of course could not be predicted with certainty for a whole season. With generally good agreement from both statistical and dynamical tools that the West and Alaska would have above normal temperatures, these areas were largely correctly forecast to be warmer than normal, but in the Southwest, the probabilities were reduced to less than what the strong trend signal alone would have indicated, due to the expectation of residual El Nino influences and wetter than normal ground and heavy seasonal snowpack in the higher elevations. The forecast of above normal temperatures in Alaska was very successful, scoring 76.3 for all stations and 85.7 for non-EC only. Strong troughs coming in from the Pacific from time to time did lower the average seasonal temperatures to the near normal range at many locations in the western CONUS. Nearly all of Alaska had above normal temperatures, with some record early season heat in late April and early May contributing to a rapid loss of snow cover. Since there was fairly good agreement from a majority of the tools that below normal temperatures were likely to occur over the southern and central Great Plains, this area was predicted to have below normal temperatures, but only parts of southern and western Texas and couple of stations in the lower Mississippi Valley verified as correct. Part of the central Great Plains, which was becoming increasingly dry, actually had slightly above normal temperatures. (As has often been the case, the seasonal forecast skill scores would have been slightly better, 5.6 for all stations and 14.8 for non-EC only, if the area forecast to have below normal temperatures had been forecast to be near normal instead.) The precipitation forecast had negative skill, with both the predicted wet and dry areas doing poorly. The all-station score was -8.5 and the non-EC only score was -20.0, with a fairly generous coverage of 42.7%. These forecasts were based primarily on a consensus of the models and statistical tools, which favored a continuation of the pattern of wet weather in the Southwest and dry in the Northwest that had prevailed for much of the winter, This forecast was also consistent with an expected weak El Nino signal continuing relative wetness over southern California and the southern Plains. However, the circulation changed in late March, and storms began to come into the Pacific Northwest on a regular basis for the first time this cold season. The developing dryness over the central Mississippi and Ohio Valleys and wetness along the east Gulf Coast and Florida that are favored by weak El Nino composites actually were observed, even as the weak El Nino observed for the past several months was fading away. However, the Southeast was left as EC in the forecast because there was a conflict between most of the models, which predicted drier than normal conditions in the Southeast, and the statistical tools and El Nino composites, which indicated a preference for above median rainfall. The 0.5-month lead temperature forecast for May 2005 showed no significant skill overall, scoring 0.4 for all stations and 1.1 for non-EC only, with a fairly large coverage of 37.6%. The areas of below normal temperatures forecast for the southern Great Plains and Florida were both based on indications from the dynamical models, with support from persistently below normal SSTs in the latter area, and showed positive skill overall. Conflicting signals for warmth in the Rio Grande Valley and parts of the Southeast from OCN and CCA were correctly disregarded in this case. The larger area of above normal temperatures forecast to arch from the interior Southwest across the central Great Plains and Mississippi Valley into the interior of the central Gulf States had negative skill overall, and was correct only the interior Southwest and a couple of stations in Kansas. The Southwest part was indicated by highly significant signals from OCN, while warmth was favored over a large area in the middle of the country by many of the models and the CAS model due to developing dryness. The entire Midwestern, central and eastern part of the CONUS was observed to have below normal temperatures, in large part due to a strongly negative NAO and AO which repeatedly deflected unseasonably cold Canadian air into those areas. Mutually agreeing indications from both the models and statistical tools for warmth in most of Alaska led to a highly successful forecast of above normal temperatures for that area, giving Heidke skill scores of 82.6 for all stations and 100.0 for non-EC only in the state. The all-station skill of the 0.5-month lead precipitation forecast was also not significantly different from zero, scoring 0.9 for all stations, although the non-EC only score was amplified to 10.0 due to a rather small coverage of only 8.5%. Only one small area centered over the upper Mississippi Valley was predicted to have above median precipitation, on the basis of fairly consistent signals from the CFS and some of the other models. Any possible residual forcing from the fading El Nino was considered too weak to be used, especially on the monthly time scale with evidence of continuing MJO activity. Alaska was left as all EC due to lack of any statistically significant signals from any of the tools. The revised updated temperature forecast for May issued at the end of April and incorporating input from several dynamical models used for medium range forecasting resulted in a very slight improvement to the temperature forecast, raising the scores to 1.7 for all stations and 3.8 for non-EC only. Indications for ridging in the middle of the country that led to a slight expansion of the area of predicted warmth were wrong, but forecast above normal heights over the Pacific Northwest were largely correct, while model indications for troughs approaching the central Pacific coast, although correct, did not lead to the expected area of cooling in northern and central California, perhaps due to the unexpected onset of above normal SSTs in that area. A slight expansion of the area of forecast below normal temperatures in the Southeast did however add skill Model guidance for Alaska, however, resulted in the removal of a couple of stations on the northwestern coast from the area forecast to have above normal temperatures, leading to a slight degrading of the all-station score to a still very good 73.9. Since above normal temperatures were observed over the whole state, the non-EC only score remained at 100.0 . The revisions to the precipitation forecast resulted in a considerable increase in the skill, to 10.7 for all stations and 43.1 for non-EC only, with an increased coverage of 24.8%. Short- and medium-range model forecasts favored retention of the wet area over the upper Mississippi Valley and the addition of a large area of above median for the central Pacific coast extending into the western Plateau, as well as the addition of an area of above median to the east Gulf States including Florida. All these added areas showed significant positive skill. No areas were forecast to have below median precipitation, and Alaska was left with a forecast of all EC. --- EOF