IDŐJÁRÁS Quarterly Journal of the Hungarian Meteorological Service Vol. 118, No. 4, October December, 2014, pp - PDF

IDŐJÁRÁS Quarterly Journal of the Hungarian Meteorological Service Vol. 118, No. 4, October December, 14, pp Estimation of future precipitation conditions for Hungary with special focus on dry

Please download to get full document.

View again

of 17
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Books - Non-fiction

Publish on:

Views: 18 | Pages: 17

Extension: PDF | Download: 0

IDŐJÁRÁS Quarterly Journal of the Hungarian Meteorological Service Vol. 118, No. 4, October December, 14, pp Estimation of future precipitation conditions for Hungary with special focus on dry periods Rita Pongrácz *, Judit Bartholy, and Anna Kis Department of Meteorology, Eötvös Loránd University, Pázmány Péter sétány 1/A, H-1117 Budapest, Hungary *Corresponding author (Manuscript received in final form August 25, 14) Abstract In this paper, estimated trends of precipitation- and drought-related climate indices and the return period of the daily precipitation amount are analyzed. For this purpose 11 regional climate model (RCM) simulations from the ENSEMBLES project with 25 km horizontal resolution for the emission scenario A1B are used after applying a bias-correction procedure. According to the results, the summer - and -year return periods will increase by a factor of by the late 21st century relative to the reference period. The projected changes are considerably smaller for the other three seasons compared to future summer changes. Furthermore, drought-related climate indices in summer are projected to increase significantly in Hungary as well as in Central/Eastern Europe by the end of the 21st century. Additionally, precipitation-related indices are projected to decrease in summer by compared to Key-words: precipitation index, dry period, return period, bias correction, regional climate model simulation 305 1. Introduction Climate change is most often referred as higher temperature values and more frequent heat waves (e.g., Pongrácz et al., 13). However, it usually involves more intense and more frequent extreme events related to excess or lack of precipitation (e.g., severe dry spells, heavy precipitation, intense thunderstorms), too (IPCC, 12). This highlights the importance of climate research in quantifying the detected past and the projected future changes from global to local scales. Frequent hot weather in summer and overall increasingly warm climatic conditions are quite straightforward consequences of global warming. Global and regional warming induced effects on precipitation are not as clear as on temperature, because the higher spatial and temporal variabilities might hide any robust changing signal. Nevertheless, precipitation is one of the most important meteorological variables, since it considerably affects natural ecosystems and cultivated vegetation as well as most of human activities. Extreme precipitation events both excessive, intense rainfalls and severe droughts may result in several environmental, agricultural, economical, and natural disasters. The lack of precipitation for extended period and coincidental intense heat wave often lead severe drought events. For instance, in 03 a long-lasting, devastating heat wave occurred throughout Europe (Stott et al., 04), causing death of hundreds of people (Bouchama, 04). In Hungary, the year 03 was generally dry with 17% less annual precipitation than the average (Schirokné Kriston, 04). The Europe-wide heat wave in the summer superposed to these overall dry conditions, resulting in severe drought. The estimated monetary damage in the Hungarian agriculture reached billion HUF by the end of the year (Faragó et al., ). Another hot and dry summer from the past decades occurred in 07, this drought resulted in reduced harvest of maize in Hungary and caused at least 80 billion HUF loss (Faragó et al., ). On the contrary, in May, the total rainfall in Hungary largely exceeded the average monthly precipitation of the baseperiod for May, namely, almost three times more precipitation occurred than usual (Móring, 11). The excessive precipitation led to inland inundation and floods on Sajó, Hernád, Bodrog, and Bódva rivers resulting in more than billion HUF of defence and recovery costs (KSH, 11). Overall, the year became the wettest year in Hungary since 1901 with 959 mm annual precipitation amount exceeding the annual mean of the period by 65% (Móring, 11). Besides Hungary, a large majority of the Central/Eastern European region was hit at the same time by severe floods (Bissolli et al., 11; WMO, 11). After the year of excessive precipitation, Hungary experienced the driest year in 11 since 1901 with only 407 mm annual total precipitation amount, being only 72% of the annual average in the period (Móring, 12), which affected the agricultural production quite negatively. The very 306 next year, 12 was also dry in Hungary, the annual total precipitation was only 470 mm (Horváth et al., 12; Rajhonáné Nagy, 13) resulting in more losses in agriculture than in 11 (e.g., by 24% less harvested cereal and % less production of sunflower and grape) (KSH, 13). Due to the large temporal variability of precipitation, after two consecutive very dry years, in late May and early June in 13 large precipitation occurred again in Central Europe and resulted in extreme water levels, with record high peak levels on several Central European rivers, i.e., the Danube, the Elbe, and the Vltava (BBC News, 13; van der Schrier et al., 13; WMO, 14). Besides the great amount of precipitation, the large spatial extension and the strong intensity (exceeding 0 mm/24 hours) also contributed to this extreme event (Horváth et al., 13). Overall, this flood affected several countries in Central and Eastern Europe (e.g., Germany, Austria, Czech Republic, Hungary, Serbia) with 16 billion EUR losses and 22 deaths altogether (Munich Re, 13). In order to avoid or at least reduce the effects of these precipitation related hazards, national and local communities need to develop regional adaptation strategies (IPCC, 12; Motha, 09; Sivakumar and Stefanski, 09; Anwar et al., 13), and then, act according to them. For this purpose, results of global climate model (GCM) simulations must be downscaled to regional and local scales, hence better serving end-users needs. Downscaling of coarse resolution GCM simulation outputs is especially important in case of precipitation because of the large temporal and spatial variabilities, and consequently, since appropriate precipitation impact assessment studies require fine resolution information (e.g., Marengo and Ambrizzi, 06; Fowler et al., 07; Maurer et al., 07; Serinaldi and Kilsby, 14). From the agricultural point of view, especially potential dry conditions induce longterm planning, for which estimation of precipitation is evidently the key element. Sheffield and Wood (08) analyzed global and regional trends of drought using a moisture-based drought index for According to their results, soil moisture has increased globally with regional differences. In Africa, a significant drying can be identified, whereas increasing trend is detected in North America. The annual precipitation sum in Hungary decreased in the period; in Budapest the mean change is.5%, which is statistically significant (Lakatos and Bihari, 11). Precipitation measurements in the Carpathian Basin suggest that both the overall intensity and frequency of extreme precipitation events related to both excess and lack of precipitation increased in the th century, whereas the mean climate became slightly drier (Bartholy and Pongrácz, 05; Lakatos et al., 11). For the future, 50 km horizontal resolution regional climate model (RCM) experiments of the PRUDENCE project (Christensen et al., 07a) suggest that the annual distribution of precipitation will be totally restructured 307 in Hungary both in case of A2 and B2 emission scenarios (Nakicenovic and Swart, 00), namely, the wettest season (currently summer) will become the driest, and the driest season (currently winter) is likely to be the wettest by the end of the 21st century (Bartholy et al., 08). The projected changes for Central/Eastern Europe involve large uncertainity, therefore, further analysis is necessary. In order to successfully adapt to the changing climatic and environmental conditions, appropriate assessment of possible changes is essential. In the current study, fine (25 km) resolution RCM experiments of the ENSEMBLES project are analyzed taking into account the A1B intermediate emission scenario for the entire 21st century. First, the data and the bias correction method applied to the raw RCM outputs are presented. Then, the precipitation-related characteristics, return period of daily precipitation, and various climate indices are defined. Section 3 discusses projected changes in the seasonal return period of daily precipitation, and estimated seasonal changes of climate indices with special focus on dry conditions. Finally, Section 4 summarizes the main conclusions Data used in the analyses 2. Data and methods In this paper, simulations of 25 km horizontal resolution RCMs nested in coarse resolution GCMs are used to estimate the future precipitation- and drought-related climatic conditions in Central/Eastern Europe covering the region N, E. The assessment focuses on analysis of daily precipitation outputs of 11 RCM simulations (listed in Table 1) from the ENSEMBLES project (van der Linden and Mitchell, 09). This European Union funded project aimed and successfully completed to run several climate models between 04 and 09 in order to improve the reliability of climate projections, measure uncertainty, and help decisionmakers with reliable information. All of the RCM simulations selected for this study cover the entire period and apply the intermediate A1B emission scenario, according to which the estimated CO 2 concentration level will be 532 ppm and 717 ppm by 50 and 20, respectively (Nakicenovic and Swart, 00). The necessary initial and boundary conditions are provided by three different GCMs: ECHAM (Roeckner et al., 06) developed at the Max Planck Institute, HadCM (Gordon et al., 00) developed at the UK MetOffice, and ARPEGE (Déqué et al., 1998) developed at Météo-France. 308 Table 1. List of the selected RCMs, their main references, their driving GCMs, and the responsible institutes used in this analysis. RCM (Reference) Driving GCM Institute HadRM3Q0 (Jones et al., 1995; 04) HadCM3Q HC (Hadley Centre), United Kingdom RCA3 (Samuelsson et al., 11) CLM (Böhm et al., 06) RCA3 (Samuelsson et al., 11) RACMO (van Meijgaard et al., 08) REMO (Jacob and Podzun, 1997) RegCM (Giorgi and Bi, 00) HIRHAM (Christensen et al., 07b) ALADIN (Radu et al., 08) HadCM3Q (high sensitivity version) HadCM3Q HadCM3Q (low sensitivity version) ECHAM5 ECHAM5 ECHAM5 ECHAM5 ECHAM5 ARPEGE ARPEGE C4I (Community Climate Change Consortium for Ireland), Ireland ETHZ (Eidgenössische Technische Hochschule Zürich), Switzerland SMHI (Swedish Meteorological and Hydrological Institute), Sweden KNMI (Koninklijk Nederlands Meteorologisch Instituut), Netherlands MPI (Max Planck Institut), Germany ICTP (International Centre for Theoretical Physics), Italy DMI (Danmarks Meteorologiske Institut), Denmark CNRM (Centre National de Recherches Météorologiques), France 2.2. Bias correction of RCM outputs The evaluation of raw precipitation outputs of RCMs for suggests that simulated values usually significantly overestimate the observations in Central/Eastern Europe, except in summer when mostly underestimations were found (Pongrácz et al., 11). In case of precipitation indices associated with specific thresholds, it is particularly important to use the most accurate simulations, as close to measurements as possible. For this purpose, before the analyses, a bias correction method should be applied to the raw simulated data. The biases of the raw RCM outputs are corrected using quantile matching technique. This is based on the assumption that two datasets are considered similar if their distributions are close to each other (and the closer is the more similar), therefore, the monthly empirical distribution functions of daily precipitation at each grid cell should be fitted (Formayer and Haas, ) to the observed distribution represented by the gridded E-OBS (Haylock et al., 08) data for a baseperiod, i.e., in this study. These fitting procedures provide the multiplicative bias-correcting factors for each month, for each grid cell. Then, these calculated factors are applied to the raw daily outputs of RCM experiments both for the past ( ) and the target (00 20) period. 309 Fig. 1 illustrates the successful fitting of the bias-correction for January for a selected grid cell, where the percentile values of the raw and bias-corrected simulations are compared to the percentiles of E-OBS data. The Q-Q plot clearly shows that after the correction, the distribution of the simulated precipitation fits perfectly to the distribution of the reference data (i.e., all the percentile value pairs are located along the y = x line). Simulated daily precipitation (mm) 30 0 Adatsor3 Bias-corrected RCM Raw simulation 0 30 Daily precipitation (mm) from E-OBS Fig. 1. Q-Q plot for raw and bias corrected simulation data, Results for January daily data from the grid cell located at N, E using the ARPEGEdriven HIRHAM experiment are shown The return period and the selected climate indices After the bias-correction, both the - and -year return periods of the daily precipitation amount are analyzed. The return period (τ) is defined as the inverse of the expected average number of occurrences (P) in a year (τ=1/p). Fig. 2 shows an example for how to determine the change of the -year return period. First, the 90th percentile of the daily precipitation (P 0.9 ( )) is calculated for the reference period ( ) in each grid cell. Then, this daily precipitation amount should be compared to the future (71 20) percentile values, and that one (P X (71 20)) is selected, which equals to this P 0.9 ( ) daily precipitation. In the example of Fig. 2, X = 0.94 since the 94th percentile value of the future period equals to P 0.9 ( ). So τ years, = 0/(0 94) = years, which implies a substantial increase of the return period, and hence, drier climatic conditions. 3 Relative frequency 0% 90% 80% 70% 60% 50% 0 30 Precipitation (mm) Fig. 2. Calculation of the projected value of the -year return period (τ years, ). Empirical distributions of summer daily data from the grid cell located at N, E using the ARPEGE-driven HIRHAM experiment are shown. In order to assess future climate tendencies in the Central/Eastern European region, several precipitation-related indices are also analyzed on seasonal scales. Table 2 lists the names, definitions, and units of the selected climate indices. Three indices are directly related to drought (DD, MDS, CDD), the other three indices refer to wet conditions using small precipitation thresholds (RR1, RR5, MWS). The grid cell values of all the six indices are calculated from the bias-corrected simulated precipitation data sets for the entire selected domain covering the latitude N and longitude E for the whole simulation period ( ) using all the 11 RCM experiments. Overall projected seasonal changes by and periods relative to the reference period are also calculated. Furthermore, spatial average changes for Hungary represented by the grid cells located within the country border are estimated for all the seasons both for mid to late 21st century. Table 2. Drought- and precipitation-related climate indices used in the current analysis Index Definition Unit DD Number of dry days (R day 1 mm) day MDS Mean length of dry spell (R day 1 mm) day CDD Maximum length of dry spell, i.e., maximum number of consecutive dry days (R day 1 mm) day RR1 Number of precipitation days exceeding 1 mm (R day 1 mm) day RR5 Number of precipitation days exceeding 5 mm (R day 5 mm) day MWS Mean length of wet spell (R day 1 mm) day 311 3. Results and discussion First, we focus on the -year and -year return periods of the daily precipitation amount. The projected seasonal changes generally show similar patterns for the whole selected domain. According to our results, a slight decrease of the return period is likely to occur in winter, namely, the -year return period may change to 8 9 years by the end of the 21st century (Fig. 3). This implies wetter climatic conditions for winter. In spring and autumn, individual RCM experiments suggest slightly more diverse changes than in winter, which results in larger uncertainty but very small changes overall. In case of summer, the results for the period clearly suggest that the return period of daily precipitation occurred once in a decade on average in the recent past is very likely to increase by a factor of 1.2 2, so drier climatic conditions are projected. Larger increase of the -year return period is estimated in the southern parts (exceeding 8 years) of the selected domain than in the northern subregions (less than 4 years). MAM SON DJF JJA years Fig. 3. Composite maps of 11 RCM simulations indicating the estimated seasonal mean changes of the -year return period by relative to the reference period Besides the average return periods, the seasonal uncertainties for nine subregions are also determined (Fig. 4). Whisker-Box plot diagrams are used for 312 indicating the highest (maximum) and the lowest (minimum) values, and the lower and upper quartiles, i.e., the 25th and 75th percentiles of the -year return period of daily precipitation amount for each subregion based on the 11 individual RCM simulations. According to these results, the return period increases in summer, thus implying an overall future drying trend by almost all of the RCM simulations in every subregion (only a few RCM simulations project slight decrease in the northwestern subregions). Although the projected tendency is clear, the RCM-based projections cover a wide range of return periods, thus, the uncertainty of the estimation is quite large. The estimated changes are clearly larger as proceeding from the northwestern to the southeastern part of the domain. In Hungary and Slovenia, the doubling of the return period is estimated by only a couple of RCM simulations (using CLM for instance), whereas in the southern subregions (Romania, Croatia, and northern Serbia) 25% of the RCM simulations suggest larger increase than by a factor of 2. In the other three seasons, the overall uncertainties of the projections are smaller than in summer, however, even the signs of the estimated changes are not identical, especially in spring and autumn. In winter, most of RCM simulations suggest considerable decrease of the return period, thus implying wetter conditions in all subregions (only two RCM simulations project increase of winter return periods, namely, ALADIN and HIRHAM driven by ARPEGE) DJF Years 15 Years 15 JJA 5 CZ AT SK UA SI HU RO CR SR 5 CZ AT SK UA SI HU RO CR SR MAM Years 15 Years 15 SON 5 CZ AT SK UA SI HU RO CR SR 5 CZ AT SK UA SI HU RO CR SR Fig. 4. The maximum, minimum, upper, and lower quartile values of the -year seasonal return period of the daily precipitation amount for nine subregions (CZ: southeastern Czech Republic, AT: eastern Austria, SK: Slovakia, UA: southwestern Ukraine, SI: Slovenia, HU: Hungary, RO: Romania, CR: Croatia and SR: northern Serbia). 313 The projected seasonal changes of - and -year return periods are compared for Hungary in Fig. 5. In general, the signs of the projected changes by one particular RCM simulation are identical for both return periods. It can be clearly seen that all the RCM simulations suggest clear increasing return period in summer. Most of the RCM simulations project similar rate of changes, except three RCM simulations (HIRHAM driven by ARPEGE, CLM driven by HadCM, and HadRM3Q driven by HadCM), when, when extremely large changes (larger than twofold increasing) is projected for Hungary in case of the -year return period of daily precipitation sum. The projected changes are considerably smaller for the other three seasons than for summer. Nevertheless, the estimated changes of the -year return period are slightly larger than the changes of the -year return period in winter and autumn. DJF Projected value of τyears, (year) Pro
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks