Thank you dear @ Daniel Wright. A very interesting comment. In fact I have a research article concerning monthly stream flow for the period 1931-1999, which was analysed by Box-Jenkins ARIMA models, the reviewrs rejected it asking me to compare the results with Machine Learning NN.
I am also having a similar problem trying to generate future offshore wind scenarios based on ARIMA modeling and spatio-temporal correlations but I dont't think this is the most appropriate way (for long-term predictions and future scenarios generation). Therefore, I guess this is the reason that the reviewers ask to compare the results with Machine Learning NN or I guess with MCMC (Markov Chain Monte Carlo implementations).