Statistical Forecast Experiences of Seasonal Rainfall in Bermejo River Basin (Argentina

GONZÁLEZ, MARCELA HEBE

florianopolis

Congreso; VSIC Simposio Internacional de Climatologia,; 2013

sbmet

In this paper an statistical method was applied to predict summer rainfall (January to March, JFM) in upper (BS) and middle and low Bermejo river basin (BI) in Chaco Plains of Argentina. Bermejo river basin is highly sensitive to floodings specially with rainfall above normal in the upper basin. Mean areal rainfall series in both basins for JFM were calculated using 19 stations in 1982-2007 period in BS and 14 stations in 1968-2007 period in BI. Correlation between these series and previous month (December) sea surface temperature (TSM), 1000 Hpa (G10), 500 Hpa (G5) and 200 Hpa (G2) geopotential height, 850 Hpa zonal (U) and meridional (V) wind and specific humidity (H), were calculated. Some predictors were defined using these fields and multiple linear regression models were constructed using “forward stepwise” methodology. Rainfall in BS was affected by Indian and south Pacific oceans TSM and the model explained 50% of the summer rainfall variance; correlation between observed and predicted (by cross-correlation methodology) precipitation was 0,49. Predictors for BI rainfall were the southern annular mode and the intensity of the subtropical Atlantic High, resulting a model explaining 49% of the summer rainfall; the correlation between observed and predicted (by cross-correlation method) rainfall was 0,6. Some statistics were calculated to prove efficiency (probability of detection, false alarm ratio, etc). Results indicated that the method detects some cases of extreme rainfall, improving the climatology. Results in BI are better than in BS. The probability to detect above normal rainfall events is in general, better than the probability to detect below normal rainfall ones. The probability to give a false alarm in a below normal rainfall event is greater than in the above normal cases. Empirical probability functions of observed and predicted rainfall were similar at 95% confidence level. In both BI and BS, the derived prediction schemes tend to underestimate the extremes. However in BI basin, cases of precipitation greater than 300 mm and lower than 400 mm, were well detected. The observed rainfall gamma adjustment reveled that summer accumulated precipitation of 550 mm in BI and 633 mm in BS had a return period of 20 years.