We did well but we definitely have to do better: four critical points about fluxnet

Abstract:

Fluxnet is a real success story of data integration. The scientific outcome is overwhelming. Nevertheless: in a time of methodological consolidation and transfer of the networks to technically more integrated infrastructures, a critical view on its weak points may strengthen the future success and our position within biogeochemical science. Four points should be discussed:

1. We have to select our sites more thoroughly.
2. We need better data curation.
3. We should think about ‘forgetting’ some of the older datasets.
4. We have responsibility for the results of integration studies.

ad 1: We had to learn during the past years that the EC is not applicable in all terrains. Slope and footprint problems are widespread and sites have to be critically scrutinized before being sure that we submit valuable ecological information. This is time consuming and may be frustrating since we have to accept that we had sometimes invested lots of work and money for building a flux tower at a site that is not suitable for the method. Nevertheless, a clear site quality policy should be developed among infrastructures and integrating activities.

ad 2: In some cases it has turned out that the information about different steps leading from the raw data to a number in integrated scientific papers has been lost. This is a big challenge to research infrastructures that should develop common rules for data curation to increase trust in integration activities.

ad 3: In the first approach Fluxnet left the responsibility for site and data quality to the site PI and accepted more or less all data submitted. Further approaches and in particular long-term infrastructures have to develop strategies to reject (or at least flag) data from sites that are prone by terrain problems. This includes that in future integration studies we should stop using some of the datasets from the ‘wild old times’ when we did not know better.

ad4: We need a strategy to communicate with data users that are far away from practical problems at the field sites. The amount of users without experience about possible problems related to terrain or technical limitations of eddy covariance will be dramatically increasing with open data policy, high performance computation and automated processing routines. The question arises how we can avoid misuse or misinterpretation of the data.