Predictive ability of logistic regression, auto-logistic regression and neural network models in empirical land-use change modeling – a case study