Point LRF Forecasting Method By: Brad Satkowski
Main Topics Explanation of method Forecasts Verifications Forecast evaluation Recommendations
Explanation of Method 3 sites in United States and 3 in Europe used –Pittsburgh, Dallas, Chicago, Lisbon, Rome, Oslo Recent 30 day temperature departure from normal NAO and AO values documented during these times Find 5 other time spans with similar AO and NAO trends Past anomalies used to make forecast for March 2003 based on recent AO and NAO values
Forecasts Forecasts made on February 25, 2003 AO value on this date was approximately –0.500 NAO value on this date was approximately –1.000 Both had been negative for most of winter Assumed trend would continue Wrong assumption!
Forecasts CityForecast Pittsburgh3-4ºF below normal Dallas1-2ºF below normal Chicago1-2ºF below normal Lisbon1-2ºF below normal Rome1-2ºF above normal Oslo2-3ºF below normal
Verifications CityForecastVerification Pittsburgh3-4ºF below normal+1.2ºF Dallas1-2ºF below normal-1.2ºF Chicago1-2ºF below normal-0.6ºF Lisbon1-2ºF below normal+1.9ºF Rome1-2ºF above normal-1.4ºF Oslo2-3ºF below normal+3.2ºF
Forecast Evaluation Negative AO and NAO assumption DateAONAO Feb February March
Forecast Evaluation Dallas verifies, Chicago close All other cities not even close to verifying Persistent trough in Midwest allowed Dallas and Chicago to verify Trend seems to exist between AO and NAO values, and temperature anomalies Difficult to forecast AO and NAO
Recommendations Avoid forecasting extreme anomalies –Pittsburgh was big mistake –Large anomalies occur less frequently –Uncertainty also a factor Use more than just recent trends –Assumption of negative AO and NAO trend was wrong –Different forecasts would have been made if entire month of February was analyzed
Conclusions Relationship between AO and NAO values and temperature anomalies seems to exist Must be careful when looking at AO and NAO trends More accurate predictions of AO and NAO can lead to better forecasts, especially for Europe and Eastern United States