OBJECTIVES

OBJECTIVESEvaporation, precipitation and atmospheric heating communicate SSTA to the atmosphere, driving changes in temperature, precipitation and storminess. This is why the SST is one of the dominant factors that controls climate.

Previous research has revealed some skill to forecast rainfall in areas of the North Atlantic. (Philips and McGregor, 2002; Philips and Thorpe, 2006; Rodriguez-Fonseca et al., 2006).

The aim of this study is to explore the potential of SST as a predictor of monthly rainfall anomalies in Galicia (NW Spain).*

DATA AND METHODSMethods:We calculated the Pearson product-moment correlation coefficient r. The significance of the coefficient was assessed at the 99.0% by means of Students t test.We applied a test for field-significance considering the properties of finiteness and interdependence of the spatial grid. (see Philips and McGregor, 2002) The SST data: Monthly data with a 2 resolution between the years 1950-2006.Provided by NOAA/OAR/ESRL PSD (http://www.cdc.noaa.gov/).Rainfall: The Rainfall was obtained from the database CLIMA of AEMET (Agencia espaola de Meteorologa) and also from data of METEOGALICIA. We used the rainfall anomaly index SWER *

*LAG 0: only 5 months (February, April, May, October and December) present statistically significant correlation. LAG 1-3: The main aim of this work is to explore the ability of SST to forecast rainfall. We explore correlations for February, April, May, October. MONTHLY RESULTS

MonthPercentage significant(LAG 0)Percentage significant(LAG 1)Percentage significant(LAG 2)Percentage significant(LAG 3)January13.5032.0*6.016.2February32.2*11.215.714.3March16.623.5*31.2*7.8April22.2*20.5*11.80.6May27.8*7.06.915.2June11.112.29.26.8July13.811.87.935.0*August5.91.357.8*14.1September3.067.8*12.21.8October68.7*6.43.05.0November7.55.56.725.7*December20.6*5.839.7*14.8

SSTA:SWER CorrelationsAreas with significant SSTA:SWER positive (red) or negative (blue) correlationFebruaryAprilMayOctober*

Linear RegressionsIn order to assess the predictability we have defined clusters. We defined two clusters for each month. Cluster were used as the input variables of a linear-regression models that specify rainfall anomaly from SSTA. *

Equations and coefficients Considering only one month of lag:Considering two months of lag:*

CONTINGENCY TABLES The potential predictability ranges from 64% to 86%*

FebruaryObserved (-)Observed (+)Forecasted (-)100Forecasted (+)38AprilObserved (-)Observed (+)Forecasted (-)83Forecasted (+)28MayObserved (-)Observed (+)Forecasted (-)84Forecasted (+)46OctoberObserved (-)Observed (+)Forecasted (-)92Forecasted (+)39

MONTHLY CONCLUSIONSThis work has investigated links between SST variations in North Atlantic and rainfall in NW Iberian peninsula. We have obtained that 5 months (February, April, May, October and December) satisfied the finiteness and interdependence criteria for field significance for concurrent SSTA:SWER.The interdependence criteria is only satisfied for February, April, May and October for one and two-monthly lagged analysis.The Atlantic area were clustered to be used as input variables for rainfall anomalies forecast. Regression equations provides correlations up to 0.59 between observed and predicted anomalies.The potential predictability ranges from 64% to 86% when considering rainfall as a discrete predictand.*

For the seasonal study we grouped months considering winter as JFM, spring as AMJ, summer as JAS and autumn as OND. We studied the 4 seasons with lags from 0 to 4. SEASONAL RESULTS*

*

Linear Regressions: Equations and coefficients *

Area SSTPeriodEquationCorrelationNio31951-2006Prec=a*SST3L1+b*SST3L2+c*SST3L3+d*SST3L4a=0.1103; b=-0.0053; c=0.3864; d=0.07040.4230Nio1+21951-2006Prec=a*SST1+2L1+b*SST1+2L2+c*SST1+2L3+d*SST1+2L4a=0.0124; b=0.5214; c=0.1739; d=-0.00630.4498Nio3&Nio 1+21951-2006Prec=a*SST3L1+b*SST3L2+c*SST3L3+d*SST3L4+ e*SST1+2L1+f*SST1+2L2+g*SST1+2L3+h*SST1+2L4a=-0.0969; b=-0.4360; c=0.4925; d=0.0267; e=0.0673; f=0.8479; g=-0.1161; h=-0.04750.4606

CONTINGENCY TABLES * La Nia years almost always announces dry spring in NW of the Iberian Peninsula (between 83% and 100% of hit rate). El Nio years do not preclude the appearance of wet spring (around 55% of hit rate).

Nio3Forecast (-)Forecast (+)Observed (-)90Observed (+)58Nio1+2Forecast (-)Forecast (+)Observed (-)102Observed (+)57Nio3 & Nio1+2Forecast (-)Forecast (+)Observed (-)101Observed (+)45

*

SEASONAL CONCLUSIONSIn this work we have proven the ability of SSTA in the Equatorial Pacific to forecast rainfall anomalies in spring season in the area of NW Iberian Peninsula. This ability seems to be mediated by the appearance of a blocking high centred at North Sea, extending from Ireland and Great Britain to Central Europe. Results show significant correlation higher than 45% if we combine different index and lags. If we consider only the possibility to forecast tendencies in the precipitation, results show that La Nia years almost always announces dry spring in NW Iberian Peninsula (between 83 and 100% of hit rate). Nevertheless, El Nio years (around 55 % of hit rate) do not preclude the appearance of wet spring. *

FUTURE RESEARCHNew research are being conducted to explain the mechanism involved in the correlations explained.EMICs and GCMs will be used to understand the Dynamical links observed.*

Thank you for your attention*

LINEAR REGRESSIONS: RESULTS Time series of rainfall anomaly observed (circles) and forecasted (squared) from 1951 to 2006 for the considered months, using the stepwise regression model of one and two months of lag.*

*********