We are using julian days to represent phenological information (i.e. date of the start of spring) for weveral years in a predictive model. Our problem is that when a phenological event occurs in December 31st or January 1st the absolute diffence between both is 364 days. However, it should be just 1 day. How could we manage this???