MOSAiC Data for Process-Based Model Evaluation – When and How?
RV Polarstern heading towards the North Pole during the MOSAiC expedition (photo: Alfred Wegener Institute/Steffen Graupner (CC-BY 4.0)).
by Gunilla Svensson, Stockholm University, on behalf of the YOPP Process Task Team
With the extra attention on environmental prediction in the polar regions over the last decade, the polar prediction community is very excited about the observational data that was collected during the MOSAiC expedition. This extraordinary Arctic field campaign, led by atmospheric scientist Markus Rex, and co-led by Klaus Dethloff and Matthew Shupe, was completed successfully with the return of the research icebreaker Polarstern to Bremerhaven in October 2020. With 389 operating days and numerous observations gathered by scientists studying the atmosphere, ocean, sea ice, ecosystems and biogeochemistry, MOSAiC will provide a wealth of data for many years to come.
It is important to realize that the terabytes of data were gathered by nearly five hundred scientists directly involved in the expedition and more than three hundred colleagues participating additionally in the background. Although the data is collected for the benefit of science – and all data will be available for the community eventually – it is important to recognize the importance for the MOSAiC scientists to first work with their hard-earned data. They need time for processing, quality assurance and publication of descriptive science studies. The MOSAiC Data Policy states that the all data will be freely and publicly available on 1 January 2023. However, a significant amount of data will be publicly available well before that date, and some significant data is already available.
The Process, Data and Verification Task Teams of the WMO WWRP Polar Prediction Project are actively trying to facilitate the community’s access to this unusual and extremely valuable data by several means to pave the way for making the observational data accessible also for modelers. In this effort, our focus is to work together with MOSAiC scientist, on collaboration projects and in continued dialogue about how to improve the integration of process observations into numerical weather prediction and coupled models, while offering co-authorship to scientists responsible for observations and with proper acknowledgements following the MOSAiC Data Policy.
Already up and running is the verification of numerical weather prediction (NWP) prime variables lead by Amy Solomon, mainly using the observations distributed over the Global Telecommunication System (GTS) thus primarily weather-station and radiosonde data. The exchange of model data is accomplished using the standards developed by the three tasks teams together within the framework of the the YOPP Supersite Modelling Intercomparison Project, in short YOPPsiteMIP. The goal of YOPPsiteMIP is to facilitate process evaluation of models by organizing observational data at so-called YOPP supersites – with MOSAiC being of of the important supersites for YOPP.
Merged Observatory Data Files
Figure 1. The creation of Merged Observatory Data File is facilitated by Python scripts that converts the observed variable name to the standardized one with all netCDF attributes needed. The usability is thereby enhanced as files for all observatories have the same structure and variable names. The files are developed in Phases that successively include data from more instruments and parameters for more advanced process evaluation (source: Gunilla Svensson).
Producing observational files of process-oriented data for modelers is a complicated task that takes time, more than anticipated when the Task Teams embarked on this route. But we find this effort extremely important as it is trying to break a barrier so more observations can be used to better understand the biases in terms of processes in models. Even if Merged Observatory Data Files (or MODFs) were created for all supersites now available in the world, it is still an extremely limited part of the global surface that is covered by measurements. However, if we manage to do this for all YOPP supersites around the Arctic Ocean including MOSAiC, this joint activity can already have a substantial impact for model development work for years to come. The MOSAiC site is exceptional, as it will eventually include information on the atmosphere, sea ice, and ocean jointly. The most important characteristics of the MODFs is that all supersites follow the same data protocol so that it is easy to add more locations to model evaluation (Figure 1).
At the moment, we are trying to accelerate the creation of MODFs, with a specific focus on the YOPP supersites Utqiagvik (former Barrow) in Alaska, Sodankylä in Finland, and the two Canadian sites Whitehorse and Iqaluit where data has been gathered during the YOPP Arctic winter Special Observing Period (NH-SOP1; from 01 February to 31 March, 2018). This data will be analyzed and transferred into MODFs along with data from the MOSAiC site for the winter season (October 2019 to March 2020). We anticipate that first versions of data, i.e. Phase 1/2 MODFs (see Figure 1) will be available on the YOPP Data Portal (yopp.met.no) by the end of February 2021. Updated versions with increased quality and additional parameters will follow eventually, with the next deadline tentatively set to end of March 2021.
As mentioned above, the additional data generated from MOSAiC can be used for specific projects in collaboration with observationalists. As part of the YOPPsiteMIP activitiy, which is endorsed by MOSAiC, interesting periods will be selected for multi-model process evaluation and detailed analysis.
Anyone interested in contributing with model results or coordinate an intercomparison case, please contact Gunilla Svensson gunilla@misu.su.se.