130 likes | 272 Views
THE NWP “BREAKTHROUGH” FOR CLIMATE ANALYSIS CENTER MONTHLY PREDICTIONS IN 1981. Bob Livezey Climate Services Division/OCWWS/NWS Symposium on the 50 th Anniversary of Operational NWP College Park, MD, June 17, 2004. OUTLINE. Introduction 1981 Implementation Somerville’s (1980) Experiment
E N D
THE NWP “BREAKTHROUGH” FOR CLIMATE ANALYSIS CENTER MONTHLY PREDICTIONS IN 1981 Bob Livezey Climate Services Division/OCWWS/NWS Symposium on the 50th Anniversary of Operational NWP College Park, MD, June 17, 2004
OUTLINE • Introduction • 1981 Implementation • Somerville’s (1980) Experiment • Verification Evidence • Description of Study for Monthly Mean Temperature Progs • Results • Monthly Mean 700-hPa Forecast Results • Discussion • Monthly Forecast Process • Impact on the Process • Possible Corroborative Study
INTRODUCTION • Aug. 12, 1980 NMC operational NWP suite change • Replacement of 7LPE (Shuman and Hovermale, 1968, JAM) with the Spectral Model (Sela, 1980, MWR). For 00 GMT • 0-48, global, 30 waves, 12 layers • 48-144, hemispheric, 24 waves, 12 layers • Mar. 18, 1981 NMC operational NWP suite change • Extension of the 00 GMT global Spectral Model • 48-84, 24 waves, 12 layers • 84-144, 24 waves, 6 layers • NMC Forecast Div. (now HPC) found that the new system increased “the number of cases with usefull skill” (Cooley’s Feb. 23, 1981 memo)
INTRODUCTION • Somerville’s results (1980, JAS) anticipated the improvements. • 500-hPa errors (30m contours) for 11 January 1973 case.
VERIFICATION EVIDENCEForecast and Control Forecast Definitions and Timing
VERIFICATION EVIDENCE • Verification of monthly mean U.S. surface temperature forecasts • Issued twice-monthly for three categories at “zero” lead (3-5 days) from 1973 through 1994 • Verified with a modified Heidke skill score for 102 mega-divisions • Verification repeated for “operational persistence” • Differences computed to normalize for “predictability” • Smoothed with a 25-forecast running mean
VERIFICATION EVIDENCE • Verification of monthly mean extratropical N.H. 700-hPa heights • Issued twice-monthly at “zero” lead (3-5 days) from 1973 through 1994 • Verified with the spatial anomaly correlation over the Pacific/North American region (Livezey et al., 1995) • Not normalized for “predictability” with “operational persistence” • Smoothed with a 25-forecast running mean
VERIFICATION EVIDENCE NH post-GNWP NH pre-GNWP
MONTHLY FORECAST STRATEGY(Kalnay and Livezey, 1985) • Step 1 of 3-step process, “prognosis of the 700 mb height anomalies” • For step 1, subjective blend of “three available indicators” • “kinematic/synoptic extrapolation of recent anomaly fields”; the five-day mean centered at D-3 • “statistical forecasts derived from lagged correlation fields”; effectively damped operational persistence • “a five-day averaged NMC forecast centered at the 1st day of the month”; D+3
MONTHLY FORECAST STRATEGY D-3 D+3
MONTHLY FORECAST STRATEGY(Kalnay and Livezey, 1985) • “Of these, experience indicates that the dynamical forecast [D+3] is the most skillful. Although the forecast extends only two or three days into the verification period, the five-day averaging process is an effective filter that seems to accurately reveal the low-frequency trend.” • A corroborating test: Comparison of the scores for 6-10 day forecasts (D+8)with persistence of D-3.
MONTHLY FORECAST STRATEGY D-3 D+3 D+8