Seasonal rainfall patterns classification, relationship to ENSO and rainfall trends in Ecuador

Abstract

Water is one of Ecuador's most important natural resources, whose management should rely on good information and adequate models for the water balance. Although there are many studies focusing on rainfall and temperatures, most of them are localized for specific projects or study basins. The convergence of factors like its location in the Intertropical Convergence Zone, the Amazon River basin, the complex topography of the Andes, and being next to the Pacific Ocean impacted by the El Niño Southern Oscillation, modelling climate in Ecuador is a challenging task. The aim of this research was to classify seasonal rainfall patterns, using robust procedures to deal with missing values and outliers. The selected database contained 319 stations with monthly rainfall from 1982 to 2011. A hierarchical clustering technique applied to the proportion of monthly rainfall percentiles allowed identifying four clusters that, when compared to a digital elevation model, resulted geographically related to Sierra, Coast, Amazon and Coast Orographic Sierra (COS). Coast shows strong seasonality with the dry season from June to November, while COS presents a similar pattern with moderate seasonality, the Amazon exhibits mild to no seasonality and the Sierra with a moderate seasonality. Interestingly, the Amazon seasonality enters rather deep into the Sierra through large river valleys. Cumulative differential plots of the standardized deleted deviations of monthly rainfall aided detecting cluster dependent El Niño effects in rainfall for Coast and COS. Correlations of rainfall versus SST at El Niño regions were also cluster dependent, with low to no significant correlations for Amazon, higher correlations for Coast and COS and moderate correlations for Sierra. Monthly rainfall trends assessed by the non-parametric method of Sen's slopes, showed overall decreasing trends for September and October rainfall and increasing from February to April, thus suggesting a strengthening of seasonality. © 2017 Royal Meteorological Society