The traditional test for the presence of first-order autocorrelation is the Durbin–Watson statistic or, if the explanatory variables include a lagged dependent variable (or is normally distributed), Durbin's h statistic.

I strongly recomend you the Variance Inflation Factor (VIF) analysis. You can calculate the VIF using the R package "usdm".

References:

James, G., Witten, D., Hastie, T., Tibshirani, R., 2013. An Introduction to Statistical Learning, with Applications in R. Springer. doi:10.1007/978-1-4614-7138-7

Naimi, B., Araújo, M.B., 2016. sdm: a reproducible and extensible R platform for species distribution modelling 39, 368–375. doi:10.1111/ecog.01881

Dear Cristian

You can check the paper: Jou et al. (2014). A VIF-based optimization model to alleviate collinearity problems in multiple linear regression. Comput. Stat. 29:1515–1541.

Basically the authors propose that the VIF-based model can mitigate the harm from data collinearity through the reduction in both the condition number and VIFs, meanwhile improving the statistical significance.

Kind regards

Fredy

Dear Andrew, Oliver and Fredy thanks a lot for your answers.

I will read your suggestions as soon as possible.

Best regards

Cristian 

Further to Fredy and Oliver, see an empirical example attached.

Article Transferability of species distribution models: The case of ...

Thanks a lot dear Hein. Your paper is very useful.

Cristian Martínez

Hello all there!

I was looking at several options to eliminate the problem of autocorrelation, I observed that the statistical package "usdm" for "R software" is an excellent tool. This package, show you which are the autocorrelated variables, also eliminates them by two methods, reducing to the suitable variables for the linear models only.

Best regards and thank you very much for your help.

Cristian Martínez