Secular changes in Costa Rican rainfall: correlation with elevation

Abstract

In 1982 and 1983, Costa Rica, like many other places in the world, suffered from the effects of an extremely intense El Nino Southern Oscillation (i.e. major dislocations of tropical rainfall regimes) (Rasmusson 1985, Schreiber & Schreiber 1984). Guanacaste Province, located in Costa Rica's dry northwestern corner, received little rain from October 1982 until June 1983, and rainfall for the rest of 1983 was far below long-term averages. With the cessation of El Nino in mid-1983, Guanacaste and its farmers, cattle ranchers, and biota could expect rainfall levels to return to 'normal' levels (1500-1600 mm rainfall annually in the lowlands). However, this may not be the case because, as I herein report, annual rainfall has been declining in lowland Guanacaste over the last 40-50 years. To determine the generality of this trend, I analysed long term rainfall statistics from several Costa Rican localities and, using regression analysis, discovered a non-random relationship between rainfall trends and elevation. Data for this report come from 10 Costa Rican meterological stations located throughout the country at elevations ranging from 5 to 1380 m above sea level (Figure 1). The number of years of continuous rainfall records varied from 16 to 95, and mean annual rainfall varied from 1617 to 4554 mm (Table 1). Temporal trends in rainfall at the 10 sites are illustrated in Figure 2. As is typical of rainfall data (Gaskell & Morris 1979, Nieuwolt 1977), considerable year-to-year variation exists. To reduce this variation, I transformed the raw data with a binomial smoothing operator with weights of 1, 4, 6, 4 and 1 for years t-2, t-l, t, etc., which improves the data's response to spectral analyses (J. M. Mitchell, in litt.). As Portig (1965) has noted, part of this variation in Costa Rica and elsewhere in Central America appears to be cyclical. Years of low rainfall tend to be associated with the occurrence of El Nino (Schreiber & Schreiber 1984), but this is not invariably the case (Figure 2).