Exploring relationships between regional climate and Atlantic Hurricanes Mark R. Jury Physics Department University of Puerto Rico - Mayagüez Mayaguez, PR, 00681
Data employed: hurricane index: 1850-2004 (60-80W,15 80W,15-35N) monthly NCEP atmospheric and oceanic variables 1948-2004 Methods of Analysis: wavelet spectrum of 150 yr record correlation to evaluate regional indices composite patterns for high and low years (50,55,61,95,04 & 72,77,82,83,94) hovmoller analysis to study zonal propagation The aim is to find predictable interactions.
Earlier studies have revealed links between Pacific ENSO and Atlantic hurricanes N-S correlation: Nino3 vs zonal wind 60-80 W long. 1948-2004 3 pressure from W +.4 suppression 20 S 10 0 10 20 30 40 50 N 12
TS H1 Hurricane tracks are concentrated as intensity increases H4 H5 pressure N-S section correlation 1948-2004 NAO and geopt. ht. 60-80 W long. +.4 N. Atlantic High strengthens in +NAO phase: deflects the jet stream. 20 30 40 50 N -.4 Result : westward tracks for the most intense hurricanes
Wavelet analysis Wavelet analysis of smoothed hurricane index: 8 and 50 year cycles.
Hurricane cycles and seasonal timing 10 September is the annual peak. Are they phase-locked? -0.5 0 +0.5 SST anomaly over the western Atlantic W C W C W C W C W W Caribbean upper wind ~ 8 year cycle westerly easterly 1965 1970 1975 1980 1985 1990 1995 2000 2005
Correlation of Atlantic hurricanes and climate indices 1948-2004, r 0.27 : 95%. soi pna nao nino3 nino1 wp qbo nino4 censo o pdo npac 0.08 0.06 0.04-0.41-0.58-0.03-0.03 0.04-0.08-0.03 0.28 ao trend tni pacw eofpa atlt solar mjo m indm sahl braz 0.04 0.03-0.07 0.05-0.07 0.05-0.01-0.27 0.50-0.01-0.02 mei swus gmst epo tnh amo noi esp aao -0.07 0.30 0.01 0.01 0.01 0.00 0.00-0.10-0.04 Six indices are significant: Pacific SSTs: nino1, nino3 and north. Indian monsoon rainfall, SW US rainfall, Madden-Julian Oscillation
Links between Pacific Ocean and equatorial Atlantic atmosphere push pull Atlantic overturning Pacific upwelling Indian monsoon Pacific thermocline bold Atlantic zonal wind thin
Is the ocean-atmosphere atmosphere coupling transient? Are changes abrupt or wave-driven? Which oscillations are relevant to Atlantic hurricanes and which occur naturally? Is upper level forcing evident? >15 m s -1 >15 m s -1 One signal in the composite analysis : the stratospheric wind
Composite SST pattern favoring Atlantic hurricanes: pre-la Niña +1% c = - 0.27 m s -1-10% BEFORE +1% -10% DURING based on high minus low hurricane index
Usually, a Pacific El Niño causes warming in the tropical North Atlantic through the reduction of trade winds. Pacific Niño3-33 month correlation with SST, 1948-2004, r 0.27 sig. at 95% zonal wind anom. +0.4 >+0.6 Here the hurricane composite SST pattern indicates that Pacific La Niña precedes Atlantic warming The anti-phase relation is a valuable predictor
Composite tropopause temperature: westward movement c = - 0.26 m s -1 BEFORE DURING based on high minus low hurricane index
Upper level zonal winds shift across the tropical Atlantic ATLANTIC BEFORE F M A M c = - 0.28 m s -1 J DURING J A S O 50% 90 80 70 60 50 40 30 20 10 0 10 Composite hovmoller analysis on 5-10 N from hurricane index
Abrupt change of low level circulation northwest of Puerto Rico 3 2 Anticyclonic Divergent Cyclonic Convergent 10 6 m 2 s -1 1 0-1 F M A M J J A S O PSI CHI -2 BEFORE DURING MONTH
Discussion: transient evolution Atlantic hurricanes enhanced by: Anti-phase Atlantic - Pacific SST pattern. Pacific upwelling and Atlantic upper winds spreading westward prior to La Niña. Propagation is consistent with a Pacific ocean Rossby wave: c = βg H f 2 = - 0.27 m s -1 uncoupled, at 10 S
Conceptual model of Pacific Atlantic interaction Signals transmitted weak Kelvin Wave strong Rossby Wave COLD WARM Westward spreading ENSO and Atlantic circulation
Discussion: predictor signals Hurricane region SST and SLP change only 1% East Pacific and West Africa SST change 10% Upwelling Regions Tropical Atlantic tropospheric and stratospheric wind change 50% Hurricane region low level Ψ and Χ change 100% from Feb to Aug Using this knowledge a group of predictors were assembled, keeping in mind the need to limit co-linearity and artificial skill Ensuring the training period is 3 x greater than the number of candidate predictors The predictors: NAO, QBO, Au2, a-pslpa pslp,, a-pssta where a=atlantic Preceding March-April April-May p=pacific The targets: #Hur1-5, #Hur3-5, ACE Multi-variate step-wise regression is performed
Model outcome: Artificial skill +.3(QBO)+.3(a-pSST) = #Hur3-5 Atlantic Hurricane Prediction at 3 month lead (adj. r 2 = 30%) Obs Model 1951 1954 1957 1960 1963 1966 1969 1972 1975 Years 1978 1981 1984 1987 1990 1993 1996 1999 Preliminary findings are useful, further research is needed!