Most climate variability studies have been conducted at a regional level in which it was assumed that the climate was uniform for each individual climatic zone. The objective of this study was to determine climate variability at a local level for the humid zone in Argentina and to develop statistical linear yield models using meteorological variables. Monthly weather data (minimum and maximum temperatures, rainfall and humidity) and soybean yields were analyzed for the period 1973-2000. Minimum temperature in November was the only variable than showed a statistically significant positive trend in the whole region. A similar tendency was found for soybean yields in the 75% of the stations. Agrotechnological trend in soybean yield variations and trends in weather data were removed. Interannual variations of rainfall were studied giving emphasis to the wet and dry periods and their relationship with the soybean yields. The correlation between the soybean yields and monthly meteorological variables were calculated. For monthly rainfall, the correlation coefficients were positive, during November to March, showing higher rainfall in general was favorable for higher yields, meanwhile in April and May this relationship was negative. Linear relationships between detrended soybean yield anomalies and anomalies of meteorological variables were developed, using stepwise multilinear regression. Taking into account the growing season for soybean, the meteorological variables included in the statistical models depended on the region.