What do you need to know about climate in order to be in the best position to adapt to future change? This question was discussed in a European workshop on Copernicus climate services during a heatwave in Barcelona, Spain (June 12-14).
The answer is not clear-cut, even after having some information about user requirements from a survey to identify a direction for data evaluation for climate models (DECM). The survey is still being carried out.
Some of the key issues concerning user requirements include essential climate variables (ECVs), climate data storage (CDS), evaluation and quality control (EQC), and fitness for purpose (F4P). I include their acronyms here since they often appear in reports and discussions and their meaning is not always obvious.
The ghost that keeps coming back is “uncertainty”. The data give an incomplete description of the world, and include some inaccuracies. How significant are these, and how closely do they represent the aspects which they are meant to describe?
The Copernicus climate services will be able to provide a lot of data and information, which includes observations of past climate, seasonal forecasts, and projections for the future. It will provide both global and regional/local data in addition to metadata and information about their quality.
There will also be a set of tools to search, sort, visualise, process and access the data. Exactly what the tool will look like is still not determined, although it is likely to be partly based on some tools which already exist.
For future climate change, the climate data store will provide a large number of simulations derived with different climate models.
Users, however, often do not want a large number of model simulations, but results from a “best” model. This requirement is problematic, and the question how to solve this took up some of the time (as it often does).
In my opinion, we need to emphasize the information that can be distilled from the data. Information should be distinguish from data, and there are statistical methods that can find the most robust information in large data sets. Furthermore, the information is often more robust if it is drawn from multiple and independent sources.
It is important to understand that one simulations cannot give a reliable indication of the future state on a regional scale because the future outcome is subject to pronounced natural variations. This would be true even if the climate model gave a perfect representation of the climate system. The reason is that it is impossible to predict the exact course of such natural variations due to their chaotic nature.
We can nevertheless estimate statistics describing the range of likely future outcomes. Large ensembles of climate simulations may serve as a basis for estimating their statistical characteristics. The ensembles are not perfect, but do nevertheless give a reasonable indication.
Another question is whether it is feasible with a general framework for describing the quality of the wide range of data types such as observations, seasonal forecasts, and projections. In many cases, the question is context dependent. The accuracy of satellite data and the realism of model results are two different examples.
Transparency, openness, and provenance (the history of data processing) will ensure that the data can be trusted, and a unique digital object identifier (DOI) will make the replication of results easier. It is essential that people know exactly what the data really represent, their limitations, and how to interpret them. There are many instances where data have been misinterpreted.
If you have some experience with using climate data and requirements concerning climate data, I’m sure the people who carry out the survey would like to know.