: Statistical prediction of quantity of days with daily mean temperature below 10 degrees in Mar del Plata (Argentina)

DOMINGUEZ DIANA; SKANSY, MARIA DE LOS MILAGROS; GONZÁLEZ, MARCELA HEBE

NATAL

Congreso; VI Simposio Internacional de Climatologia; 2015

SBMET

ABSTRACT: One of the indicators associated with energy demand in winter is the number of days with less than or equal to 10 ° C (ndT10) daily mean temperature. This parameter is monitored operatively by National Meteorological Service in some of the most densely populated cities in Argentina but there is still no model to make a projection of this parameter. This analysis attempt to reach to a prediction model for the variable ndT10 June-July-August (JJA) in the city of Mar del Plata (MDP). MDP is analyzed as the city that has most ndT10 from other cities densely populated of Argentina (Buenos Aires, Rosario, Cordoba, Mendoza, Resistencia, Santa Fe and Salta) in the 1961-2012 period. It is also highly correlated with cities as Buenos Aires, Rosario, Santa Fe and Córdoba.Analysis of ndT10 linear trends in 1961-2012 was performed and showed a straight flat statistically significant which gives evidence of having no significant increase ndtT10 in the period analyzed. July is the month with most ndT10 in that JJA period.In order to generate a forecast of JJA ndT10, the existence of predictors in May prior to the occurrence of cold period was investigated. The forcing of interannual variability ndT10 June to August in Mar del Plata were studied by calculating correlations between anomalies ndT10 and temporary anomalies of the following variables: geopotential height at 1000, 500 and 200 hPa zonal and meridional wind at 850 hPa, temperature at 2 meters and sea surface temperature (SST) obtained from the NCEP / NCAR reanalysis. The composites of the years considered below normal and above normal were also calculated. The existence of global circulation patterns associated with interannual variability both simultaneously as in the previous month was analyzed. This analysis allowed defining several predictors. With them a model of multiple linear regression was generated and 23% of variance of ndT10 were explained. The final model was associated with SST in the equatorial zone and with air temperature over the Pacific in mid-latitudes. The scheme was validated using the method of crossvalidation and 41.6% of the total cases were predicted as actually occurred. The subnormal category was the better predicted