AEMET-γ-SREPS: convection-permitting EPS at the Spanish Meteorological Agency

Alfons Callado Pallares
Spanish Meteorological Agency
Noon Mar 1 in Room 2155

Since 2004 the Spanish Meteorological Agency (AEMET) has been running operationally a limited area ensemble prediction system (LAM-EPS) for the Iberian Peninsula. Until 2014 and nearly for a decade, this 25-km LAM-EPS called AEMET-SREPS (AEMET Short Range EPS) provided probabilistic products to forecasters with a performance which largely overcome ECMWF-EPS.

Since 2014 a 2,5-km convection-permitting LAM-EPS named AEMET-γ-SREPS has been developing with the main goal to forecast mesoscale high impact events as heavy precipitations estimating their uncertainty.

The AEMET-γ-SREPS, as its AEMET-SREPS predecessor, is a multi-model multi-boundary conditions EPS. This configuration has turned out to be the best EPS technique based on probabilistic verification when compared with other single-model techniques. AEMET-γ-SREPS initial and boundary conditions’ synoptic uncertainties are dealt through analyses and forecasts coming from 5 deterministic global NWP models with independent assimilation processes: IFS from ECMWF, GFS from NCEP, ARPEGE from MeteoFrance, GSM from JMA (Japanese) and GEM from CMC. Model errors and uncertainties are taken into account using HARMONIE (2 configurations), WRF-ARW and WRF-NMM convection-permitting models. As a result a 20 member EPS is achieved, combining 5 BCs with 4 NWP models.

AEMET-γ-SREPS is currently daily running without assimilation and the work plan is to provide the first probabilistic forecast products during the forthcoming spring.

Future improvement prospects for 2016 are to use LETKF (Local ensemble Kalman Filter) assimilation technique and complement model uncertainties with SPPT (Stochastic Perturbations of Parameterisations’ Tendencies) and surface parameters perturbations.