When parameters of wireless communication channels vary at a fast rate, simple estimation algorithms, such as weighted least squares (WLS) or least mean squares (LMS) algorithms, cannot estimate them with accuracy needed to secure reliable operation of the underlying communication systems. In cases like this, the local basis function (LBF) estimation technique can be used instead, significantly increasing the achievable tracking accuracy. We show that when some prior knowledge of statistical properties of parameter changes is available, such as the bandwidth of their variation, both the type and the number of basis functions, used in the LBF approach to approximate time-varying channel parameters, can be optimized using the Karhunen-Loève (KL) decomposition based technique.