Photo: Colourbox/John Stick

Climate scientists and statistics—for improvement

mandag 02 okt 17

Kontakt

Martin Drews
Seniorforsker
DTU Management Engineering
46 77 51 55

Facts about the research

  • The article New vigour involving statisticians to overcome ensemble fatigue in Nature Climate Change is the result of a Nordic research project called eSACP.
  • Apart from DTU, the participating institutions include the meteorological institutes in Norway, Finland, and Denmark as well as Norsk Regnesentral (Norwegian Computing Center, NR) and the Bjerknes Centre in Norway.
  • The project is financed by NordForsk, an organization under the Nordic Council of Ministers.

Read more about the project on NordForsk’s website.


The volumes of climate data have exploded, but this does not necessarily provide us with an insight into the climate of the future. If climate scientists used statistics to a greater extent, we could gain greater insights and better predictions, according to a recently published article in Nature Climate Change.

Generally, climate scientists use multiple climate models for calculating, for example, how high the seas will rise, how much precipitation we can expect, or how often we will experience extreme weather conditions such as storms, storm surges, and torrential rain. The more models, the more data. But if the data are to become useful, enabling, for example, the right decision-making in relation to climate-proof cities and coast lines, climate scientists need to delve further into the statistical toolbox. That is the opinion of a group of Nordic scientists who just published the article New vigour involving statisticians to overcome ensemble fatigue in Nature Climate Change

Senior Scientist Martin Drews, DTU Management Engineering, is one of the authors behind the article:

“The climate models are our best estimate of how the world will look when the climate changes. But it’s often necessary to boil down the increasing volumes of data so they become user relevant. Here, advanced statistics can help, and this is currently an untapped potential within climate research.”

The hunt for the right information
Martin Drews explains that a part of the climate research these years is moving in a more action-oriented direction. The research is less about whether climate change is man-made, and more about what we should do about it. Climate scientists are, among other things, seeing an increasing demand for ultra-local climate projections:

“Today, authorities or urban planners, for example, want to know the future climate of a single city or harbour. This means that the complexity and the requirements for data have increased strongly within climate research. In this context, the solution is not solely to perform more climate model runs, but also to become better at analysing the runs with the right statistical tools. The statistics can help us organize and find the right information in a giant sea of data and make it useful, so that society can assess the risks and plan for climate change adaptation,” says Martin Drews.

"Climate models are our best estimate of how the world will look when the climate changes. But it’s often necessary to boil down the increasing volumes of data so they become user relevant. "
Martin Drews, senior scientist, DTU Management Engineering

Statistics attacking uncertainties
A well-known problem with the climate models is the uncertainty in the calculations, i.e. how likely it is that the model calculations are actually spot on in relation to the real world situation. This is one of the reasons why researchers generally use many climate models rather than relying on a single one. Here, climate research may also benefit from using more advanced statistical tools that are better at taking account of different climate physics-related contexts than just the calculation of average values and standard deviations.

“The uncertainty surrounding the results that users are typically confronted with is, in reality, the sum of a multitude of uncertainties. One uncertainty may be the calculation of how global temperature increase affects the global sea level. Another uncertainty is associated with the calculations when we go from a global sea level to a local sea level. These are uncertainties which we, to a high degree, can extract from the results from the climate model runs and examine separately, using advanced statistical methods which are already available. This will increase the credibility of our calculations and means that we can be more precise in our projections,” says Martin Drews.

Statistics reinforce Denmark’s climate change adaptation
The Danish Meteorological Institute (DMI) is co-author of the article in Nature Climate Change. DMI can use the statistical analyses of the climate model results in connection with the climate atlas which DMI is to prepare in order to assist Danish society in adapting to the climate of the future, explains Peter Langen, daily manager of the climate research at DMI, who also has the scientific responsibility for the new climate atlas.

“In future, our climate atlas will provide local authorities with information on developments in temperature, precipitation, and sea levels with a special focus on extreme events. This might, for example, be the height of future storm surges or the frequency and strength of heavy rain and cloudbursts,” says Peter Langen.

“It’s crucial for the quality of the climate atlas that we have statisticians on the team. Only with their assistance will we be able to deliver data which make it possible to weigh risks and potential costs relative to the likelihood of a specific event occurring. That’s what makes the new work in Nature Climate Change particularly interesting,” says Peter Langen.