Quantile regression modeling to predict extreme precipitation

Quantile regression is an extension of the median regression that analyzes various quantile values. This method is used to predict the relationship between the response variable (Y) and the predictor variable (X) on the conditional quantile function. Quantile regression can be used to detect extreme conditions, either extreme dry (quantile 5) or extreme wet (quantile 95). Extreme precipitation often occurs in Indonesian territory because the area is surrounded by oceans. High frequency of extreme precipitation may trigger disasters, one of which is flooding. In 2017, there were floods in Sidoarjo area with a loss of up to 2 billion. In an effort to anticipate the adverse effects of extreme precipitation, forecast information abaut extreme precipitation is needed, one of which is using quantile regression. The objectives of this research was to determine the best quantile regression model for predict extreme precitipation. The data of this study were secondary data obtained from BMKG with a data length of 30 years. Modeling of extreme precipitation in Sidoarjo area involved variables of humidity, temperature and air pressure for model 1. Whereas model 2 involved 3 variables in model 1 plus the month and month square variables. The addition of the month variable and the month quadratic variable in model 2 was based on the precipitation data plot which formed a quadratic trend. Based on the Pseudo R 2 value, it can be concluded that the best model to predict extreme precipitation in Sidoarjo is Model 2.