Understanding of machine learning models is crucial for Artificial Intelligence acceptance, particularly for domains focused on highly interconnected and biological processes. In this manuscript, our experience in improving readability of Random Forest regression results via surrogate models, i.e. models trying to replicate non transparent fitting results in an interpretable framework, is reported. A Random Forest regression model describing the abundance of Anopheles mosquitoes in the northern Italy has been implemented. Then, surrogate models have been built and analysed to improve readability of the original model. Adequacy of surrogate models is stated on the basis of both classical coefficient of determination R2 and Variable Importance ranking comparison.