Predictive Models Using JMP® for the Disappearance of Arctic and Antarctic Sea Ice Due to Climate Change
Manny Uy, PhD, Principal Staff, Johns Hopkins University Applied Physics Josefino Comiso, PhD, Chief Scientist, NASA Goddard Space Flight Center
Satellite data of the ice extent, ice area and ice concentration for the Arctic and Antarctic was compiled from 1978 through 2012. While least squares linear regression models can be – and often are – used for short-term extrapolation, it is not the best statistical technique for making forecasts over a long period of time – especially in the presence of a periodic phenomenon like seasonal variations. The foundational statistical technique for seasonal time series modeling is the Box-Jenkins Auto-Regression with Moving Average (ARIMA) method. Using the Box-Jenkins ARIMA time series model in JMP 10, the results of the Arctic data analysis show that seasonal variation affects the amount of ice extent, ice area and ice concentrations at the Arctic much more strongly than the yearly variation. The seasonal and decreasing yearly trends are statistically significant and will be discussed and explained in terms of climate change. The same Box-Jenkins ARIMA method is used to forecast the behavior of ice in the Antarctic, which turns out to be more complex than the case of the Arctic. The time for an ice-free Arctic, which will probably occur during the month of September, is estimated to be between 75 and 105 years from the present time.