MCS colloquium presents concerns on filtering data

Global climate change is a large area of research interest, and has been for several years. On September 2, Dr. Juan M. Restrepo of the University of Arizona spoke on the patterns climate change in his lecture titled “Climate, Assimilation of Data, and Models.”

Restrepo started by making a distinction between climate and weather with his lecture focusing on the former, defining it as patterns in large areas over long periods of time.

Problems of prediction fall into a few categories. The first is actually finding the trend; the question of “Is it warming,” according to Restrepo. Being able to make predictions, as in “How hot will it get,” is the next question. Determining how sensitive the system is, how “human fault” might impact it, plays another role. Lastly, there is the question of whether the data is coincidental or whether it there is an underlying cause-and-effect relation.

The purpose of Restrepo’s work is to determine which method is best for interpreting the massive amount of data. These methods might include processes like Least-Square Regression. Because of the processes involved, each method can produce radically different results.

The challenge today is to calculate the uncertainty when a filtered pattern is created. This challenge is due to trying to define a pattern as simply as possible without losing too much information.

Restrepo’s work added to an already favored method and gave input on ways to show stronger correlation amongst the data. He did this by making uncertainty as minimal as possible and then maximizing predictions.

Restrepo stated that one must “keep an eye on the big picture” and that the predicting problem will depend upon methodology and that the numerous amounts of variables that must be included is a curse and unavoidable.