Evaluation of Markov Chain Based Drought Forecasts in an Andean Regulated River Basin Using the Skill Scores RPS and GMSS

Abstract

On behalf of the decision-makers of Andean regulated river basins a drought index was developed to predict the occurrence and extent of drought events. Two stochastic models, the Markov Chain First Order (MCFO) and the Markov Chain Second Order (MCSO) model, predicting the frequency of monthly droughts were applied and the performance checked using two skill scores, respectively the ranked probability score (RPS) and the Gandin-Murphy skill score (GMSS). Data of the Chulco River basin (3200–4300 m.a.s.l.), situated in the Ecuadorian southern Andes, were employed to test the performance of both models. Results indicate that events with greater drought severity were more accurately predicted. The study also revealed the importance of verifying the quality of the forecasts and to have an assessment of the likely performance of the forecasting models before adopting any model and accepting the resulting information for decision-making.