CONICET | Buscador de Institutos y Recursos Humanos

This work investigates short range predictability associated with a cut-off low (COL) event that took place in 2007 over southern South America. This event produced heavy rainfall over central and Northern Argentina, with maximums exceeding 100 mm/day at some particular locations. Numerical weather forecasts produced by different models are intercompared to asses to what extent, this particular event, could have been adequately forecasted. Some of the ensembles that are part of the TIGGE grand ensemble as well as NCEP Reforecasts, are used to asses errors associated to the initial conditions and sensitivity of the system evolution to the use of different models. Additional numerical experiments are performed using the WRF model as a global model and as a limited area model. Forecasts are verified in terms of how they represented the COL event using an object oriented verification method. QPF produced by the GFS and WRF models for this event are also verified using traditional scores and the Fraction Skill Score. Results show that most of the operational ensembles have a much better performance than WRF model in both the representation of the COL evolution. Moreover, the spread of the TIGGE ensembles and reforecasts suggest that the sensitivity to the initial conditions is not too large while WRF model exhibits significant changes in skill from one initialization to the other. With respect to the forecast of the precipitation, GFS produces better forecasts than WRF for all the initialization times. To explore the reasons behind the differences in skill among WRF and other global models, a series of experiments is carried out where the sensitivity of WRF forecast to changes in horizontal and vertical resolution, and to different physics is investigated. Better results are achieved when WRF is run as a limited area model using the GFS control forecast as boundary conditions.