Climate change studies involve changes in climate system of global long-term scales with effects on regional and local climates, down to shortrange time scales, like weather hazards and extremes. Climate studies significantly contribute to the future strategic planning in economic development and thereby must produce results with high level of confidence. Global climate models simulations of the past and future climate are initial step in climate change research, but their coarse resolution data are unable to provide useful information for analysis on national levels. In this paper is presented workflow algorithm of the procedures necessary to convert coarse global climate system projected changes to fine scale data with included effect of small scale features and reduced model bias impact.
Global climate models (GCMs) are designed to resolve large scale physical processes of the atmosphere-ocean-land system on coarse-grid resolution, and obtained data are insufficient for regional or local impact assessments. Downscaling of the GCM data can be statistical or dynamical. Statistical downscaling considers creation of statistical relations between simulated and observed data, and their application on the future climate projections. Dynamical downscaling of GCM data is much more complex and involves climate simulations with the nested regional climate model (RCM) with finer resolution. The second approach is much more expensive in computing time but enables smaller scale features of the climate system to react on global scales climate changes, which makes this method more reliable for the use in climate change studies. Model bias, i.e. systematic model error, is reduced applying downscaling on higher resolutions, but still remains with considerate effect on results quality. Usual approach in presenting model data, to avoid impact of model bias, is “delta” approach (difference of climate values obtained from the model future climate and past climate simulations). For impact studies this approach is insufficient, and it usually requires use of daily model data as input for other models (crop models, hydrology models, etc.) or for calculation of different indices, which are not linearly dependent on model output data. In this case bias correction of daily model data must be applied. Model bias is a consequence of different parameterizations used in model and imperfect input data. It changes in time and space and is different for each model.
Coarse RCM simulations using SRES/IPCC scenarios A1B and A2
In this paper is presented downscaling of the global model CMCC simulation for
the period 1971-2100 to fine scale using nested model NMMB for the South East
Europe region, performed under the ORIENGATE project
Regional climate model simulation is done for the period 1971-2100, using
RCP8.5 scenario
Comparison of the GCM and RCMs resolutions is presented in Fig. 2, together
with original 0.004° DEM topography dataset. RCM 0.25° resolution is used for
regional coupled atmospheric-ocean EBU-POM model simulations using SRES A1B
and A2 IPCC scenarios, which is so far used for regional and national climate studies
and documents Initial and Second National Communication for Serbia
Bias correction of the model data in time and space it is necessary to produce daily
data interpolated on model grid for the reference period 1971-2000. For interpolation
are used all available observations for daily maximum, minimum and mean
temperature from 103 sites from national network and 32 sites near border from
surrounding countries. Interpolation of daily data is done for the territory of Serbia
using method of successive corrections
Statistical bias correction is done creating correction functions for each model grid point, for each month and for each temperature separately, under the assumption that cumulative density functions of the model and observed data have the same values. Model and observed temperatures is assumed to follow Gaussian distribution. In Fig. 3 is presented example for one month mean temperature probability and cumulative density functions of model and observed data, and obtained correction function, which has function to assign to model data appropriate observed value. More about statistical bias correction can be found in Ruml et al. (2012).
After creating correction functions they are applied on model data for three 30years periods, base period 1971-200 and two future climate periods 2011-2040 and 2071-2100.
Corrected model results have model bias reduced to the minimum. Such data are
can be used in different impact studies, under the assumption that model bias will not
have impact on the quality of the derived conclusions
In Fig. 4 and Fig. 5 are presented mean annual and seasonal temperature changes, for the periods 2011-2040 and 2071-2100 with respect to the period 1971-2000 according to the RCP8.5 scenario, respectively. Temperature change during the period 2011-2040 shows increase in all seasons. Mean annual temperature change is in the range 1.6-2.0°C, but with larger values over lower altitudes and during the winter period with increase over 2.6°C in the large part of the country. During the period 2071-2100, in other words during after one century period, expected mean annual temperature change is over 5.6°C over the large part of the country with maximum changes going over 6.4°C. Again, largest projected change is for the winter season. In both periods fall season has larger increase in temperature than spring and summer.
Several extreme temperature indices related to temperature extremes are selected for analysis in this paper. Number of ice days (ID, maximum temperature Tx<0°C) and frost days (FD, minimum temperature Tn<0°C) changes are shown in Fig. 6. During the period 2011-2040 decrease in ID and FD with respect to 1971-2000 is significant. Decrease of FD is larger in lower altitudes, which is related to temperature decrease pattern. ID decrease is higher in higher altitudes because it already reached point of very rare event in lower altitudes. During the period 20712100 decrease in ID and in FD is significantly larger in higher altitudes. This is related to the fact that by the end of the century can be accepted that in lower altitudes temperatures bellow 0°C almost never appear.
Number of summer days (SU, Tx>25°C) and tropical nights (TR, Tn>20°C) are shown in Fig. 7. During the period 2011-2040 increase in SU over 16 days in most part of the country shows significant expansion of the summer season. TR is appearing over lower altitudes and increase is up to 10 days. By the end of the century SU increase goes over 60 days, which suggests expansion of summer season by two months. TR increase is over 30 over the large part of the country, which means that climate over the territory of Serbia by the end of the century will reach subtropical/tropical climate characteristics.
Other analysis (not shown here) shows extension of the growing season length 2030 days during the period 2011-2040 and over two months by the end of the century. Increase in tropical days (Tx>35°C) also confirms climate change pattern toward much lower latitudes climate. Precipitation change in general shows some increase in annual cumulative values with highest increase during the spring season during the period 2011-2040. During the other seasons in some areas is expected small reduction. During the period 2071-2100 annual precipitation is decreasing over 10%, with significant reduction during summer season (over 40%), which is the season with highest precipitation values. This suggests high risk of summer draught hazards.
In this paper is given short review of the current work on high resolution climate change analysis with the main goal to obtain reliable model data. They can be used in development of national adaptation strategies and planning of the future economic development, which is mostly related to the agriculture and thereby closely connected to the climate factors. Serbia is a country with small-scale terrain features with economy mostly based on local landowners agricultural production, which is the reason for the high resolution approach. Following work will be focused on improvement of the interpolation method with further quality check of the collected observations. Database of corrected model daily precipitation data will be created. Having high resolution of daily temperature and precipitation data with reduced bias enables model data use by experts from other disciplines (biology, medicine, hydrology, agriculture, forestry, etc.) and production of the reliable information for the decision makers.
Acknowledgments. The authors acknowledge to ORIENTGATE project (co-funded by the South East Europe Transnational Cooperation Programme) for providing and sharing data on climate change scenarios. This paper was realized as a part of the project “Studying climate change and its influence on the environment: impacts, adaptation and mitigation” (43007) financed by the Ministry of Education and Science of the Republic of Serbia within the framework of inte- grated and interdisciplinary research for the period 2011–2015.