Demonstration of a Lagrangian Re-Analysis

Meteorological re-analyses of the atmospheric conditions over the last few decades, for which sufficient measurements of meteorological variables (e.g., pressure, temperature, humidity, etc.) are available, have become a standard tool in the atmospheric sciences. They are arguably the most-used data sets in meteorological research. All existing re-analyses store these data on a regular grid. This is suitable for many research purposes (e.g., analysing temperature time series at a certain location; providing weather maps at any time during the last decades, etc.). However, transport processes cannot be analysed so well with these data sets. Questions like where air pollution observed in one location originated from, where the moisture came from during a certain rain event, or how energy is transported in the atmosphere between different regions, cannot easily be answered with these grid-pint data sets. For this, meteorological data stored along so-called trajectories (i.e., the paths of infinitesimally small air parcels) would be very useful. In LARA, we will develop such a Lagrangian re-analysis the first of its kind on the basis of an existing grid-point re-analysis and utilizing a Lagrangian transport model. At several examples, we will demonstrate the usability of the Lagrangian re-analysis and we will make the re-analysis data set available to other researchers.

Global re-analyses of the atmosphere are important for studies of the weather in the past, climate change diagnostics, as input data sets for chemistry transport models, and have many other applications. All existing such data sets are Eulerian, meaning that they provide meteorological data on a regular three-dimensional grid spanning the whole atmosphere. LARA is the first Lagrangian re-analysis providing meteorological data at "particle" locations. These virtual particles are advected with the winds and thus follow the movement of air. This makes such a data set extremely useful for studying transport processes in the atmosphere. We traced 6 million "particles" continuously over an 80-year period (1940-2023) and produced hourly output of particle positions and meteorological data at these positions, resulting in a data set of some 250 terabytes, just small enough to share it publicly. The data set was created by running the Lagrangian model FLEXPART with input data from the latest existing Eulerian re-analysis from ECMWF (European Centre for Medium-Range Weather Forecasts), ERA5, converting this gridded data set into an airmass-following one. To demonstrate the usefulness of the LARA data set, we considered a few use cases: we identified certain air streams in the atmosphere (so-called warm conveyor belts), for which previous climatologies exist for comparison. We calculated the continentality of air based on transport times of air from the ocean. We investigated the residence time of air in the Arctic and showed that it decreased significantly since the 1980s, making the Arctic more mid-latitude-like, especially in spring and fall. We traced air from the tropical Pacific and investigated how it impacts other regions, and how this depends on the climate phenomenon El Niño. In an ongoing study, we compare the moisture sources for the extreme precipitation event in September 2024 in Austria to the climatology. There is great potential for many other applications, and several students are currently using LARA data for their master and PhD theses. We also hope that in the future groups worldwide will start exploiting the LARA data set. While the LARA data set is the main project deliverable, we also improved the underlying model FLEXPART, of which we released an updated version (v11), as well as a corresponding documentation and publication. Since FLEXPART is a community model used by many researchers and with operational applications (e.g., for emergency response in Austria and other European countries), this is an important side product of the project. We also explored the potential of artificial intelligence techniques to improve interpolation of meteorological data.