Optical Earth observation data have been used frequently in agricultural applications. The problem of cloud cover found with optical data can be overcome by using radar Earth observation data, but in turn these data have a speckle problem. This paper describes six types of speckle reduction filter and assesses the capability of each to improve classification accuracy in agricultural applications. The best filters are Lee-Sigma, Gamma MAP (Maximum a Posteriori) and simulated annealing. The filters are assessed in the context of identifying fields growing potatoes in the UK using Earth observation data, and results for other agricultural crops are also included.