Catarina Dinis Cavaleiro

First of all I would like to thank Glomar/MARUM for the financial support, which allowed for my participation in the Advanced Course on Applied Paleoclimate Time Series Analysis. This course happened on the 9th to the 13th of February 2015, in Heckenbeck (Germany).

When doing paleoclimate or paleoceanographic research, we often have hypotheses that we want to test, associated with time-dependent series or processes. To characterize these processes we tentatively investigate the parameters that define them. Ultimately, and to better estimate these parameters, we can (and should) use statistical methods taking uncertainty into account, achieving smaller error bars and narrower confidence intervals. Being said like this, it is sounds easy to understand. However, the importance of this uncertainty accountability can be lost in translation, between common language and mathematical or statistical language.
I wanted to take this course because I consider spectral analysis a powerful and informative tool when applied to climate records, such as marine sediment records, as the ones I’ve been using in my PhD project.
Besides this analysis, I also learnt very useful tools during this course. I even heard and understood for the first time fundamental concepts, such as persistence or bootstrap. Time series characteristics, such as the probability density function, its persistence, and spacing must be taken into account when further statistical analysis is needed, such as regression, correlation or spectral analysis.
It was sometimes hard to grasp some of the mathematical or statistical language or the reasoning for choosing the most suitable method, according to either what you want to know or the characteristic of your own data. In that sense, I believe that Dr. Manfred Mudelsee’s book “Climate Time Series Analysis – Classical Statistical and Bootstrap Methods” can be a good resource to go back to in the future, to help me make these choices.
Besides discussing and improving the spectral analysis for my own work I had the opportunity to analyze and discuss the data from other students. This made me think not only in my own analysis and analytical methods, the spectral analysis, but also to understand other researchers work and applied analytical methods, such as regression and correlation.
It is important to bear in mind when taking all of these concepts into account, that using these statistical and mathematical tools can be very time consuming. It made me wonder if most scientists are properly balancing the compromise between how important it is to have the most accurate possible measurement or time series analysis under the need to finish their goals or publish their results in a certain amount of time.
I therefore think that attending such a course is of high interest to paleoclimate researchers, especially those who are looking for climate mechanisms, feedbacks and lead and lags of different climatic components. However, I would advise to have beforehand all the statistical and mathematical concepts readily “charged” and fresh in your brain. Therefore one doesn’t loose too much time trying to remember what’s variance, how is the Pearson’s coefficient calculated, what t-student stands for or even what characterizes a Gaussian curve.
Finally, it was a privilege to work and receive advice from Dr. Manfred Mudelsee, to focus on my and participate on other colleagues problem solving and always in a very comfortable, familiar and cozy environment. Thank you Dr. Manfred and family!