Good, bad and the ugly: Seasonal filtering and Autoregressive Moving Average (ARMA) models for detecting and replacing spikes in velocimetric profile data