Extremes of Monthly Precipitation in La Plata Basin

MOIRA E. DOYLE; RAMIRO SAURRAL; VICENTE BARROS

Paris

Workshop; Workshop on metrics and methodologies of estimation of extreme climate events; 2010

The La Plata Basin, extending over five countries in South America, namely Brazil, Bolivia, Paraguay, Uruguay and Argentina, is one of the most important water reservoirs in the World and is of great economic value to these countries. Under a context of global warming it has been established that rainfall regimes will be affected, particularly at mid latitudes. In this sense more intense and heavy precipitations are expected to be more frequent in the area as a result of increased water vapor availability. The analysis of precipitation trends presents evidence which indicates that during the second half of the twentieth century rainfall has been increasing in the area. It is of interest to understand how extreme monthly precipitation responds to this evidence.After a detailed analysis of the available data from the different countries in the Basin, 46 stations with less than 0.5% of missing daily data in the 1960 – 2005 period were selected. This study uses data from 24 Argentine raingauge stations, 1 station from Uruguay and 21 Brazilian stations, all of them inside the La Plata Basin area. Total monthly precipitation values were obtained after aggregating the daily values. The Generalized Extreme Values (GEV) and Generalized Pareto Distribution (GPD) functions were fitted to the time series, with time-fixed distribution parameters and with a positive trend in the location parameter (in the case of GEV) and in the scale parameter (in the case of GPD). A likelihood ratio test was then applied to assess which model (with or without temporal trend) fitted best each time series, so as to determine the presence of a trend in the intensity of the extreme monthly precipitation events. To analize the presence of trends in the frequency, a Poisson distribution was fitted to the series and the likelihood ratio test was again used to compare fittings with and without trends (only in those series to which Poisson fitted well). The Mann-Whitney-Pettitt was used to look for changing points in the series. Return values of extreme monthly precipitation were calculated using GEV for each station, for the whole (1960-2005) period and for the first 16 (1960-1975) and last 16 (1990-2005) years. Results suggest that extreme monthly precipitation behavior changed throughout the period, with a trend towards an increase in the frequency of these events. Return periods for a same precipitation amount are smaller during the last period than in the early part of the series, suggesting extreme events are now more frequent. The changing-point test also shows changes over the basin, mainly in the ‘70s and early ‘80s. However, the likelihood ratio tests indicate GEV and GPD fit best when no trend in the parameters is applied (i.e., when there is no temporal trend). This could suggest the presence of a cycle in the series (rather than a trend).