Strengths: A coherent project, trying out many techniques from the class and some from independent study. Good use of group work to try multiple approaches; perhaps more collective time could have been spend critically editing the combined work.
Points for consideration:
A series that has occasional periods different from others is best described as showing non-stationary behavior.
ARIMA(5,1,4) is a big model. There is evidence of numerical instability in your results - how do you take this into account? Among other things, the Fisher CIs produced by R may work badly.
ACF plot interpretation: only high at lag 0.
ARFIMA model is defined only by reference.
For looking at trend, maybe look for mean after differencing. Modeling the non-stationary (un-differenced) data using stationary models could cause problems.
Does non-normality affects conclusions for ARIMA? Maybe a CLT applies?
Could plot ACF of absolute log difference.
More text and less code would be better.
First graph: x-axis label should have units of years.