VERIFICACIÓN ESPACIAL DE PRONÓSTICOS DE PRECIPITACIÓN EN ALTA RESOLUCIÓN PARA LA REGIÓN SUR DE SUDAMÉRICA

Abstract

Given that there is an increasing demand and use of numerical models to forecast precipitation events, it is essential to advance in the use of different verification methods to measure the quality of the forecasts with the evaluation of errors and biases. The method for object-based diagnostic evaluation (MODE) is a spatial verification method that identifies regions of interest, like precipitation, in the same way that a human would do. This method defines objects in the forecast and observation fields based on user-defined parameters. MODE was used to evaluate the performance of 4-km hourly precipitation forecasts from the Weather and Research Forecasting Model (WRF) over southern South America against the Global Precipitation Measurement (GPM) derived product IMERG Final Run version (IMERG-F). For a one month period, tests were performed to select the values for threshold and the radius of convolution parameters adequate for 3 and 24 hour accumulated precipitation. The whole verification period considered was 2017-2018 and furthermore, traditional verification statistics (eg, Probability of Detection, False Alarm Ratio) were used. Additionally, 24-hour accumulated precipitation forecasts from WRF were compared with those from the Global Forecast System (GFS). This study proved that traditional verification methods allow objectively to know the quality of precipitation forecasts. Conversely, object verification rather than making a pointwise evaluation of hits and misses, identifies precipitation patterns and compares attributes describing position, size and intensity of matched forecasted and observed objects. Regarding the analyzed models, although WRF and GFS present many surprises and false alarms, hit events present low errors associated with location and intensity of precipitation.