A recursive change point estimate of the wind speed and direction

Mean wind speeds and mean wind directions are important input parameters for environmental models and applications. In many practical cases, anemometer sensor provides measurements of direction and wind speed, supposed to be Gaussian and piecewise stationary. In this work we use a von Mises distribution and a Log-normal distribution in order to model sensor data with distributions having a physical reality. We derive from these distributions an ¿on-line¿ recursive estimate of the change point in the wind speed and direction processes. We present numerical results obtained using synthetic and real data to demonstrate the ability of these estimates.