Winds from the North-West quadrant and lack of precipitation are\udknown to lead to an increase of PM10 concentrations over a residential neighborhood\udin the city of Taranto (Italy). In 2012 the local government prescribed\uda reduction of industrial emissions by 10% every time such meteorological\udconditions are forecasted 72 hours in advance. Wind forecasting is addressed\udusing the Weather Research and Forecasting (WRF) atmospheric simulation\udsystem by the Regional Environmental Protection Agency. In the context of\uddistributions-oriented forecast verification, we propose a comprehensive modelbased\udinferential approach to investigate the ability of the WRF system to\udforecast the local wind speed and direction allowing different performances for\udunknown weather regimes. Ground-observed and WRF-forecasted wind speed\udand direction at a relevant location are jointly modeled as a 4-dimensional\udtime series with an unknown finite number of states characterized by homogeneous\uddistributional behavior. The proposed model relies on a mixture of joint\udprojected and skew normal distributions with time-dependent states, where\udthe temporal evolution of the state membership follows a first order Markov\udprocess. Parameter estimates, including the number of states, are obtained\udby a Bayesian MCMC-based method. Results provide useful insights on the\udperformance of WRF forecasts in relation to different combinations of wind\udspeed and direction.
展开▼
