Betting reg

from Wikipedia, the free encyclopedia

WETTREG ( own writing in capital letters , portmanteau word : weather conditions-based regionalization method) is a system developed in Germany statistical method for calculating climate variables.

The model is being developed by Climate & Environment Consulting Potsdam GmbH on behalf of 15 state authorities. In contrast to dynamic regional climate models such as CCLM , which attempt to infer local climate variables by solving systems of physico-chemical equations, WETTREG creates statistical relationships between global and local climate variables.

Basic assumptions

There are basically five basic assumptions behind the way the model works:

  1. Global climate models are able to describe the climate over a large area with sufficient quality, since regional data are derived from them
  2. There are semi-stable patterns in the atmospheric field variables (e.g. circulation, humidity, vorticity , etc.), which repeatedly cause a certain class of local consequences (high / low temperature, heavy / low precipitation, etc.)
  3. under the drive of emission scenarios, the frequency distribution of the atmospheric patterns changes
  4. The current relationship between atmospheric patterns and local consequences will remain valid in the future as well
  5. the representation of the changing climate in the local simulation time series generated by WETTREG is statistically correct, so that statements about mean values, variance and extremes of weather elements at the locations of the climate measurement series are possible

Working principle

Simply put, known data from local climate stations are statistically associated with known large-scale weather patterns. If one now knows the large-scale weather conditions of the future, for example from global climate models , this relationship can be applied back to the individual stations.

Establishing relationships between known variables

In order to keep the complexity and thus the computing effort low, only two key variables are used for the calculation on the station side: temperature and precipitation. If further values ​​were recorded at a station ( air pressure , humidity, etc.), these remain attached to the key variables and are therefore not lost for later projections.

The station values ​​of the two leading variables are now divided into classes of predefined size. Originally there were ten temperature classes (differentiated from cold to warm) and eight precipitation classes (dry to moist). The classes were also differentiated according to the season, so that there are 40 temperature and 32 precipitation classes, which was considered the optimal number for the planned model complexity. For reasons described in section WETTREG 2010 , two more had to be added to the temperature classes later.

Each class now contains a variety of station values. The individual classes are now associated with certain atmospheric patterns, which are classified according to a method for objective circulation pattern recognition (according to a so-called K-Means cluster method ). The weather conditions are defined by 43 potential predictors (e.g. air pressure, temperature gradients, thermal winds ...) which are located on a distortion-free horizontal grid. The data of these weather situations typically come from meteorological reanalysis data, whereby both NCEP / NCAR reanalyses and ERA-40 data from the ECMWF were used. These reanalysis grids can be available in resolutions similar to those used by today's global climate models (approx. 100 × 100 km). Compounds are now generated from the mean values ​​of the daily station values ​​of a class, which are associated with these grid-based weather conditions.

Global model and resampling

In principle, a single class can now be assigned to any weather situation using the same statistical relationship. Each class contains a large pool of days, with each day containing all the variables that are measured at this station. This is where the above assumption comes into play of trusting the results of global models. These are used to identify future weather conditions. Initially, ECHAM 4 was mostly used for this, later ECHAM  5. A class can then be assigned to each weather situation at each station. From the pool on days of a class, one is now chosen at random.

Since the days can have very different climatic sizes, the mean value from a larger number of days is usually chosen. This resampling typically involves 10 or 20 repetitions. During this process, care is also taken to ensure that the transition probabilities known today between two classes are adhered to (extremely cold days, for example, are very rarely followed by extremely hot days). Since it can be assumed not only that the frequency of the patterns changes, but also their amplitude, the data are better adapted to these potential extremes using a regression method.

With this method, the potential regional resolution of the climate variables is only limited by the number of existing measuring stations. For example, 1977 stations (282 climatic and 1695 precipitation stations) were used for the first Germany-wide runs.

COMPETITION REGULATION 2010

After the first successful runs with the model, now known as WETTREG 2006, some observations were made:

  • Although the frequency of the circulation pattern and the modeled physical conditions of a future climate from ECHAM 5 are evaluated as the driving variable by WETTREG, the amplitude of the temperature signal simulated by Wettreg is lower than in the driving ECHAM model from around the middle of the 21st century - including that of ECHAM Driven dynamic regionalizations with Remo or CCLM have higher signal amplitudes.
  • The frequency distributions of the circulation patterns that are recognized in ECHAM-5 scenarios have a tendency to degenerate ; H. towards the end of the 21st century, the patterns associated with low local temperature decrease until they disappear, and a growing proportion of the circulation patterns accumulate in the patterns associated with particularly high local temperature.
  • The quality of recognition diminishes over time - so there are patterns that cannot be easily reconciled with the specifications.

The new WETTREG 2010 therefore contains some adjustments. The most important of these concerns the introduction of so-called transweather situations (TWL). It is assumed that new atmospheric patterns or extreme forms of known patterns will occur more frequently in the future. Therefore the temperature classes have been extended by 2. It was shown that these weather conditions increased sharply, especially towards the end of the century, and that a few days that were previously assigned to class 10 were now in classes 11 and 12.

Other adjustments concern the use of anomalies instead of absolute values, as this has a statistically stabilizing effect over all periods of the year. The distribution of cold and warm days is not even over time. In spring, for example, cold days tend to appear at the beginning, warm days more towards the end. As the climate warms warmer days come now to the past, so in the row formation about warmer days (approximately from May) in the early April hike . However, May days have different conditions, such as the position of the sun and consequently the irradiation, which led to inconsistencies in the data. Using anomalies greatly reduces this data inconsistency.

Another change concerned the transition between the seasons, for which separate classes are formed as described above. This resulted in jumps in the climate signal at these transitions, which is why a buffer zone has now been introduced between two seasons, within which episodes of the ending season can be continued in the next or episodes of the new season are allowed to start early.

literature

  • J. Degener: Effects of regional climate change on the development of biomass yields of selected agricultural crops in Lower Saxony . 2013, p. 42–52 ( ediss.uni-goettingen.de ).

Web links

Individual evidence

  1. ↑ a b c d e F. Kreienkamp, ​​A. Spekat, W. Enke: Further development of WETTREG with regard to new types of weather . 2010 ( TWL_Laender (PDF; 2.9 MB)).
  2. ↑ W. Enke, A. Spekat: downscaling climate model outputs into local and regional weather elements by classification and regression . In: Climate Research . tape 8 , 1997, pp. 195-207 .
  3. ↑ a b W. Enke, T. Deutschländer, F. Schneider, W. Küchler: Results of five regional climate studies applying a weather pattern based downscaling method to ECHAM4 climate simulation . In: Meteorological Journal . tape 14 , no. 2 , 2005, p. 247-257 .
  4. ↑ a b c A. Spekat, W. Enke, F. Kreienkamp: New development of regional high-resolution weather conditions for Germany and provision of regional climate scenarios based on global climate simulations with the regionalization model WETTREG based on global climate simulations with ECHAM5 / MPI-OM T63L31 2010 to 2100 for SRES scenarios B1, A1B and A2 . 2007 ( Umweltbundesamt.de (PDF; 7.4 MB) - research project on behalf of the Federal Environment Agency R&D project, funding number 204 41 138).
  5. ↑ U. Cubasch, D. Wuebbles: Chapter 1: Introduction. In IPCC, ed., Working Group I contribution to the IPCC fifth assessment report (AR5): A report accepted by Working Group I of the IPCC but not approved in detail . 2013 ( AR5 - Introduction (PDF; 2.7 MB)).