Atmospheric reanalyses: a new tool for climate research

17. 06. 2022

Three scientists from the Institute of Atmospheric Physics of the CAS helped to significantly refine the use of so-called atmospheric reanalyses in examining the causes and consequences of extreme precipitation. Knowing the inaccuracies of these "backward" analyses compared to the real situation will help in estimating future weather changes. However, the use of reanalyses is much broader as the provided data are the basis for hydrological applications and climatological studies.

In our area, precipitation is the most spatially and temporally variable of all meteorological phenomena. Even the very dense network of stations that monitor them cannot reliably record showers and thunderstorms during the summer months. Atmospheric reanalyses are one way of recovering information from the locations where measurements are not available. Reanalyses are able to simulate past atmospheric conditions very well. Therefore, they are often used to study processes in the free atmosphere and at ground level, to analyze the causes of extreme events such as heavy rainfall or droughts, and last but not least, they provide input to hydrological and growth models. In the near future, they will also be available as reference databases. Expected climate changes will be reported as changes compared to reanalyses.

Reliability of reanalyses tested in the Czech Republic

If reanalyses are to complement observations at locations and time periods where they were not available, it is necessary to know the possible inaccuracies of the reanalyses and also the causes of these inaccuracies. Otherwise, their use could lead to incorrect conclusions.

The ability of the latest reanalyses to correctly simulate the temporal and spatial distribution of rainfall over the area of the Czech Republic has been tested in the PERUN and Remote Sensing of the Earth projects; the results of the study have now been published in the Journal of Hydrology: Regional Studies.

"Reanalyses are based on numerical prediction models whose outputs are refined by observations. However, the preparation of observational data is very time-consuming and technically demanding," says the study´s co-author Lucie Pokorná from the Institute of Atmospheric Physics of the CAS. "Therefore, there are currently only a few global reanalyses with a spatial resolution of 30 to 50 km. These cover the entire Earth, or rather the lower atmosphere that surrounds it. In recent years, regional reanalyses with a spatial resolution of less than 10 km have also been created. In our study, we focused on two global and one regional reanalysis to find out if they are able to correctly simulate the occurrence and amount of precipitation from April to October in the Czech Republic," adds Lucie Pokorná.

The results of the study show that while reanalyses can simulate the monthly and seasonal rainfall totals with good accuracy, the distribution of rainfall into individual days differs to some extent from reality. "In the lowlands, most days during the warm months are without precipitation. The series of days when it does not rain is then usually interrupted by a significant convective storm, as we experienced this year in the second half of May. That is what reanalyses have a problem with. Compared to observations, they simulate up to twice the number of days with weak precipitation, while the totals dropped during strong episodes are often underestimated. Another problem with reanalyses is incorrect timing or location of significant precipitation," says another co-author of the study, Zuzana Rulfová from the Institute of Atmospheric Physics of the CAS.

Using specific examples, Lucie Pokorná explains the importance of understanding the inaccuracies in reanalyses. "There are places, such as south-eastern Moravia or some mountains, where reanalysis overestimates precipitation. If a future calculation compared to this reanalysis shows that precipitation will not change, the conclusion without knowing the errors will be that nothing will happen to precipitation. However, that will be a wrong conclusion. If there is less precipitation in reality than in reanalysis, the conclusion should be that in the future, precipitation will be decreasing at that particular location. But the article also shows the opposite errors, e.g. in the Giant Mountains where precipitation is underestimated. When using reanalyses in hydrological models, we can see another risk there: if it rains more often in reanalyses while less in heavy rainfall, then the flood damage calculated by the model will not correspond to reality and will usually be less serious than in reality. This is something to watch out for."

Reanalyses vs. radars

"Over the last 20 years, important information about the amount and distribution of rainfall in the summer months have been provided by weather radars. By combining the observed radar reflectivity at a certain altitude level and precipitation totals measured by a network of rain gauge stations on the Earth's surface, precipitation totals can be derived at a high spatial and temporal resolution," explains the study's lead author Vojtěch Bližňák from the Institute of Atmospheric Physics of the CAS. "The calculation process is based on radar measurements at five-minute intervals. During the adjustment, the sum of radar-detected precipitation over 24 hours is then refined by station data. Thus, it is possible to obtain precipitation totals in a regular grid and then to use them to assess the correctness of reanalyses."

While the precipitation field computed from radar measurements is obtained independently of other characteristics, atmospheric reanalyses provide interconnected fields of meteorological features in grids with a regular spatial step at several tens of levels from the Earth's surface to the stratosphere. They are thus a tool to investigate the behaviour of the atmosphere, which manifests itself as variable weather, including extremes.

Therefore, it follows from the Czech scientists´ study that reanalyses are suitable for climatological studies and detection of long-term drought or relationships between circulation and precipitation. In their use for analyzing extreme rainfall episodes and hydrological modelling, however, we must take into account their inaccuracies. Therefore, they are also not a tool intended for uninformed users.

Atmospheric reanalyses: a new tool for climate research

Read also

Contacts for Media