Impact of misspecifying spatial exposures in a generalized additive modeling framework: with application to the study of the dynamics of Comandra blister rust in British Columbia