WP3a: WEATHER REGIMES

Homogenization of daily data series at the hungarian meteorological service

More information: Tamás Szentimrey

The detailed investigation of regional climate requires at least daily resolution of good quality, homogeneous time series, however the instrumental data series are usually affected by inhomogeneities (artificial shifts), due to changes in the measurement conditions (relocations, instrumentation). As the artificial shifts often have the same magnitude as the climate signal, a direct analysis of the original, raw data series can lead to wrong conclusions about climate change. At our examinations the homogenization of data series was performed by the method MASH developed at the Hungarian Meteorological Service.

Long time daily maximum and minimum temperature data series were analyzed at 15 Hungarian stations for the period 1901-2005. The examination procedure consisted of the data compilation and homogenization by method MASH with the following main steps:

  • Derivation of monthly and annual averages from the daily data.
  • Homogeneity test for the monthly and annual series.
  • Homogenization procedure for monthly series, estimation of monthly inhomogeneities.
  • On the basis of estimated monthly inhomogeneities, estimation of daily inhomogeneities, and correction of daily data.
  • Missing daily value completion.

It is emphasized that the examined maximum and minimum temperature data series are affected by very strong inhomogeneities as it can be clearly seen in Figures 1-2.

Long time daily precipitation sum data series were analyzed at 37 Hungarian stations for the period 1901-2005. The examination procedure consisted of the data compilation and homogeneity test and data completion by method MASH with the following main steps:

  • Derivation of monthly and annual sums from the daily data.
  • Homogeneity test for the monthly and annual series.
  • Missing daily value completion.

It is admitted that the daily precipitation data series were not homogenized since the present version of MASH is suitable only for temperature elements in daily resolution. However according to the homogeneity test results the precipitation series can be accepted much more homogeneous than the temperature ones.

According to our examinations most of the available temperature series proved to be extraordinarily inhomogeneous, therefore it is absolutely necessary to homogenize them prior to further processing, because otherwise false conclusions can be obtained by their direct analysis. The tested precipitation series are much less problematic from that point of view, therefore they might be already used in their raw format for climate assessment studies.

Figure 1: The original (ori) and the homogenized (hom) annual average minimum temperature series (1901-2005) at station Szeged(10)Figure 1: The original (ori) and the homogenized (hom) annual average minimum temperature series (1901-2005) at station Szeged(10)

Figure 2. The original (ori) and the homogenized (hom) annual average maximum temperature series (1901-2005) at station Szeged(10)Figure 2. The original (ori) and the homogenized (hom) annual average maximum temperature series (1901-2005) at station Szeged(10)

Interpolation software MISH

More information: Tamás Szentimrey and Zita Bihari

The MISH (Meteorological Interpolation based on Surface Homogenized Data Basis) method for the spatial interpolation of surface meteorological elements was developed at the Hungarian Meteorological Service. This is a meteorological system not only in respect of the aim but in respect of the tools as well. It means that using all the valuable meteorological information – climate and supplementary model or background information – is intended.

In practice many kinds of interpolation methods exist therefore the question is the difference between them. According to the interpolation problem the unknown predictand value is estimated by use of the known predictor values. The type of the adequate interpolation formula depends on the probability distribution of the meteorological elements! Additive formula is appropriate for normal distribution (e.g. temperature) while some multiplicative formula can be applied for quasi lognormal distribution (e.g. precipitation). The expected interpolation error depends on certain interpolation parameters as for example the weighting factors. The optimum interpolation parameters minimize the expected interpolation error and these parameters are certain known functions of different climate statistical parameters e.g. expectations, deviations and correlations. Consequently the modelling of the climate statistical parameters is a key issue to the interpolation of meteorological elements.

The various geostatistical kriging methods applied in GIS are also based on the above mathematical theory. However these methods use only a single realization in time for modelling of the necessary statistical parameters that is neglecting the long data series which series form a sample in time and space alike. The long data series is such a speciality of the meteorology that makes possible to model efficiently the climate statistical parameters in question.

The MISH method has been developed according to the above basic principles. The main steps of the interpolation procedure are as follows.

  • To model the climate statistical parameters by using long homogenized data series.
  • To calculate the modelled optimum interpolation parameters which are certain known functions of the modelled climate statistical parameters.
  • To substitute the modelled optimum interpolation parameters and the predictor values into the interpolation formula.

The software MISH consists of two units that are the modelling and the interpolation systems. The interpolation system can be operated on the results of the modelling system.

Modelling System for climate statistical (deterministic and stochastic) parameters:

  • Based on long homogenized monthly series and supplementary model variables. The deterministic model variables may be as height, topography, distance from the sea etc..
  • Benchmark study, cross-validation test for representativity.
  • Modelling procedure must be executed only once before the interpolation applications!

Interpolation System:

  • Additive (e.g. temperature) or multiplicative (e.g. precipitation) model and interpolation formula can be used depending on the climate elements.
  • Daily, monthly values and many years’ means can be interpolated.
  • Few predictors are also sufficient for the interpolation and no problem if the greater part of daily precipitation predictors is equal to 0.
  • The representativity values are modelled too.
  • Capability for application of supplementary background information (stochastic variables) e.g. satellite, radar, forecast data.
  • Data series complementing that is missing value interpolation, completion for monthly or daily station data series.
  • Interpolation, gridding of monthly or daily station data series for given predictand locations. In case of gridding the predictand locations are the nodes of a relatively dense grid.

Modelling of climate statistical parameters is a key issue to the interpolation of meteorological elements and that modelling can be based on the long data series. In the following Figure we present a block diagram to illustrate the possible connection between various important meteorological topics.

Assessment report on the statistics of weather patterns and their future evolution

More information: Andras Horanyi (horanyi.a@met.hu)

The objective of the recent work is to study the behaviour of weather patterns (circulation characteristics) for the recent past and for the forthcoming future. For that end, first, ERA-40 re-analyses data were investigated in order to understand the climate’s circulation behaviour for the past and then ECHAM model data were scrutinized in order to see whether the general circulation models are able to reflect the observed circulation patterns of the past and what evolution they do predict for the future. Different meteorological variables (and trend coefficients) were studied for the periods of 1961-2000 and 2011-2050 respectively.

The basic mean sea level characteristics are rather similar for the past and for the future as well: the Icelandic low and the Azorean high pressure systems can be clearly identified. Nevertheless the strength of these important pressure features are slightly different for ERA-40 and ECHAM and there are even more differences between the trend coefficients. For the past ERA-40 basically shows increasing zonality on annual average and particularly it is quite strong in the winter season, which is gradually weakening during spring. Unfortunately ECHAM is unable to capture this important phenomenon. Comparing the past and future tendencies of ECHAM it can be seen that there are very small changes in the annual mean, with small seasonal changes. So it seems that ECHAM doesn’t foresee important changes in the main circulation patterns (certainly this statement should be treated with care due to the fact that ECHAM was not really successful to reproduce the past trends, therefore the future evolutions might be also doubtful).

The main patterns of the 500 hPa geopotential fields are similar between ERA-40 and ECHAM, however in ECHAM the geopotential absolute values are bigger all over the domain. Increasing zonality is indicated for the past by ERA-40, which is partially confirmed by ECHAM as well (where the southern increase does appear, but not the northern decrease on the annual mean). The strongest changes in trends are identified in winter (again) for ERA-40, however ECHAM also shows significant changing patterns especially for the future. For the future this change is positive almost all over the domain, which further increases the
absolute geopotential values, but basically keeping its gradient unchanged. Unfortunately ECHAM is not fully convincing again for the past, especially if the winter and spring trends are considered (while the annual trends seem reasonable for the past).

The absolute patterns of the 10m zonal wind components are rather similar between ERA-40 and ECHAM with the difference that ECHAM considers significantly stronger zonality over the mid-latitudes as it is the case for the ERA-40 data (this is true not only annually, but also seasonally as well). On the other hand for the past trends ERA-40 provides more robust evolutions, which are just in partial agreement with the ECHAM fields (the differences are especially pronounced during the autumn). As far as the future is concerned the annual trends are rather small, but the seasonal ones are significant (for instance at the mid-latitude cyclone track area there is decreased zonality during summer and spring and increased one in autumn and winter). The decreased summer zonality for summer might indicate that the humid and cool air masses will reach the central part of the continent with lesser extent resulting in warmer summers, while on the other hand the increased zonal wind in autumn just means the opposite that the mild air can penetrate into the continent causing less cold winters than before (all this is in good agreement with the subjective “feelings” about the changes in the main weather systems over Central-Europe).

The absolute values of the East-West zonality index are rather similar between ERA-40 and ECHAM for the past and also between the two ECHAM periods (past and future). The differences are again in the trend fields, however the trend values are rather small all over the year with the exception of spring and winter. The trends are also characterised by lots of local features indicating some local characteristics in the zonal wind patterns. For the future the summer decrease and the winter increase is much less pronounced as it was the case, when one directly scrutinized the zonal wind component fields. The mid-latitude increasing zonality
for spring and winter is not very well captured by the ECHAM data.

The general patterns of the North-South zonality index are similar between ERA-40 and ECHAM with a bit of more pronounced mid-latitude features in the ECHAM model data. The differences are again in the trend fields: while the annual trends patterns are more or less similar between ERA-40 and ECHAM, seasonally important differences can be found (the only exception is spring, where there is a rather acceptable agreement). The significant increased winter zonality is not captured by ECHAM. For the future ECHAM anticipates small annual changes with more important seasonal variations (which are of opposite signs for the same regions for the different seasons). It is interesting to see for instance that in summer there is a decreased zonality over the Atlantic ocean, while an increased one over the continent (the winter changes are not significant).

The 700 hPa zonal wind maps are quite similar to those of the 10m wind with more smooth patterns (no influence coming from the orography). Certainly the strength of the wind is bigger at 700 hPa and more zonal (positive values all over the domain). The general characteristics are again rather similar with a bit of stronger zonal wind fields for the ECHAM model. The weak trends are quite in agreement annually, however the differences are quite important especially in summer and autumn. The annual trends for the future are rather the opposite than in the past, which would indicate no change with respect to the initial period (the 1950s).

The 700 hPa East-West zonality index (type 1) maps show quite similar general features (also for the trends) like it is the case for the 10m fields, with the difference that it is a bit smoother being less influenced by the orography. Therefore the absolute values are similar between ERA-40 and ECHAM, moreover the trends are also in more or less in agreement. Nevertheless the trends are not really significant except during the spring and winter for ERA-40. Regarding the ECHAM data the trends are mostly significant in summer and in a lesser extent in autumn, but a robust conclusion cannot be drawn for the future.

The absolute temperature fields are rather similar between the ERA-40 and ECHAM data indicating that the “thermal” behaviour of the model is rather correct. The annual trends are rather similar too, however seasonally the warming tendency of the continent is judged differently by ECHAM with respect to the re-analyses data (for instance the intensive Western-European winter warming is not indicated by the ECHAM model; probably it is in agreement with the changes in the zonality patterns). It is interesting to see that the warming tendency for the future is increasing according to the ECHAM model and also it is going to be extended over the Atlantic region as well.

The 500/1000 hPa relative topography patterns show the average temperature of this layer, so it can be a good indicator of the thermal behaviour of the lower and middle troposphere. The absolute fields are again qualitatively in good agreement with the difference that the model (ECHAM) provides bigger values (but similar gradients). The agreement on the trends can be seen mostly in summer, but not in the other seasons. Interestingly the annual twofold (warming and cooling at different locations of the domain) feature of the re-analyses data cannot be seen at the ECHAM patterns (where basically only the warming tendency is identified). In spite of this fact the Western-European warming is less pronounced in ECHAM. Regarding the future trends the significant warming all over the domain at every season is apparent (which is a bit less intense over the British Isles).

As a summary of the achievements for the weather pattern investigations it can be said that less robust conclusion (than originally expected) can be drawn based on the above described synoptic-climatological study. It is due to the fact that the circulation patterns of the ECHAM model are not in a good agreement with the ERA-40 re-analyses data, therefore the interpretation of the results for the future should be done with extreme care as well. On top of that it was seen that in spite of the fact that the thermal characteristics are well simulated by ECHAM the circulation-related fields are in lesser extent and with more persistency, which means that the model mostly doesn’t indicate significant circulation changes for the future. This deficiency might come from the fact that the chosen future period is too near from the present and no robust changes are anticipated until that time and/or that the resolution of the ECHAM model doesn’t permit to have a meaningful thorough investigation of the weather patterns over the Atlantic-European region. All this means that further studies are really needed for instance in extending the trend investigations for the entire 100 years and see the significantly higher resolution REMO regional climate model patterns over the Central-European region.

ERA-40 1961-2000 ECHAM 1961-2000 ECHAM 2011-2050
Figure: Annual mean maps (first row) and linear trend coefficients (second row) for mean sea level pressure (hPa) for the periods ERA-40 1961-2000 (left column), ECHAM 1961-2000 (middle column) and ECHAM 2011-2050 (right column) /full year/

 

ERA-40 1961-2000 ECHAM 1961-2000 ECHAM 2011-2050
Figure: Annual mean maps (first row) and linear trend coefficients (second row) for 500 hPa geopotential (gpm) for the periods ERA-40 1961-2000 (left column), ECHAM 1961-2000 (middle column) and ECHAM 2011-2050 (right column) /full year/

 

ERA-40 1961-2000 ECHAM 1961-2000 ECHAM 2011-2050
Figure: Annual mean maps (first row) and linear trend coefficients (second row) for 10m zonal wind component (m/s) for the periods ERA-40 1961-2000 (left column), ECHAM 1961-2000 (middle column) and ECHAM 2011-2050 (right column) /full year/

 

ERA-40 1961-2000 ECHAM 1961-2000 ECHAM 2011-2050
Figure: Annual mean maps (first row) and linear trend coefficients (second row) for zonal wind component at 700 hPa (m/s) for the periods ERA-40 1961-2000 (left column), ECHAM 1961-2000 (middle column) and ECHAM 2011-2050 (right column) /full year/

 

ERA-40 1961-2000 ECHAM 1961-2000 ECHAM 2011-2050
Figure: Annual mean maps (first row) and linear trend coefficients (second row) for 2m temperature (°C) for the periods ERA-40 1961-2000 (left column), ECHAM 1961-2000 (middle column) and ECHAM 2011-2050 (right column) /full year/

 

ERA-40 1961-2000 ECHAM 1961-2000 ECHAM 2011-2050
Figure: Annual mean of mean sea level pressure and its linear trend for Hungary (upper curve) and Iceland (lower curve)

 

Figure: Annual mean of mean sea level pressure and its linear trend for Hungary (upper curve) and Iceland (lower curve) for 1951–2050 40°N and 70°N for latitude averages between given longitudes

 

ERA-40 1961-2000 ECHAM 1961-2000 ECHAM 2011-2050
Figure: Zonality index 1 calculated at 10 m and linear trends for Budapest

 

ERA-40 1961-2000 ECHAM 1961-2000 ECHAM 2011-2050
Figure: Zonality index 2 calculated at 10 m and linear trends for Budapest