Generalized linear models and its extensions are widely used for analyzing non-normal data. But Poisson mixed model may exhibit inadequate fitting and inference when encounter excessive zero counts. Mixed hurdle model is a preferable method to solve the problem. Nevertheless, it is still a challenge to use the mixed hurdle model to deal with correlated data. There are a few computational procedure for hurdle model can be used to calculate, particularly for the model with random effects being correlated between non-zero and zero response parts. In our paper we display a method to fit the hurdle model with conditionally autoregressive random effects for the spatial data. Based on the extended algorithm, some modifications are made to the existing procedure in R to help us to fit the data. We conduct Monte-Carlo simulation to study the finite sample properties of our model. The result shows that the new procedure fit the model well. The estimation becomes better with the increase of measurement in each subject. At last, we apply the new procedure to a real problem. The dataset is about reindeer spatial distribution related to the wind power establishments at Storliden Mountain in North Sweden. The new procedure gives a better fit of the real problem than a usual Poisson mixed model