=Paper=
{{Paper
|id=Vol-1498/HAICTA_2015_paper98
|storemode=property
|title=Fine Scale Climate Change Analysis: from Global Models to Local Impact Studies in Serbia
|pdfUrl=https://ceur-ws.org/Vol-1498/HAICTA_2015_paper98.pdf
|volume=Vol-1498
|dblpUrl=https://dblp.org/rec/conf/haicta/VukovicVDCRPRK15
}}
==Fine Scale Climate Change Analysis: from Global Models to Local Impact Studies in Serbia==
https://ceur-ws.org/Vol-1498/HAICTA_2015_paper98.pdf
Fine Scale Climate Change Analysis: from Global
Models to Local Impact Studies in Serbia
Ana Vukovic1,2, Mirjam Vujadinovic3,2, Vladimir Djurdjevic4,2, Bojan Cvetkovic2,
Zorica Rankovic-Vasic3, Zoran Przic3, Mirjana Ruml3, Aleksandra Krzic2
1
Faculty of Agriculture, University of Belgrade, Serbia, e-mail: anavuk@agrif.bg.ac.rs
2
South East European Climate Change Center, RHMSS, Belgrade, Serbia
3
Faculty of Agriculture, University of Belgrade, Serbia
4
Faculty of Physics, University of Belgrade, Serbia
Abstract. Climate change studies involve changes in climate system of global
long-term scales with effects on regional and local climates, down to short-
range 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.
Keywords: climate change, climate model, bias correction
1 Introduction
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
892
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 (Nakicenovic
eta al., 2000, IPCC, 2007) with statistical bias correction using EOBS gridded data
are applied so far in impact studies in Serbia, and documented in Initial and Second
National Communication of Republic of Serbia (Djurdjevic et al., 2011, Rajkovic et
al., 2014, and references within).
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 (Djurdjevic and Krzic,
2013a). Future climate simulation is done according to extreme RCP8.5 scenario
(IPCC, 2013). Period of simulation 1971-2000 is used for model bias correction of
daily temperature data. Interpolation of daily temperature data is done for the
territory of Serbia using all available observations from national network, and near
boundary available data from border countries. Observations are interpolated on
model grid. The same dataset of observations and interpolation method is used for
climate analysis in project related to the renewal of viticulture zoning (Ivanisevic et
al., 2015). Using model and observed data, for each month and for each grid point are
created correction functions, which are applied on the past and future climate
simulation results. After applying bias correction we can assume that model bias is
reduced to minimum. Use of corrected model data is shown with short analysis of
annual and seasonal temperature change and change of several climate indices related
to temperature extremes.
2 Model Simulation
Regional climate model simulation is done for the period 1971-2100, using
RCP8.5 scenario (IPCC, 2013). Global climate model CMCC-CM (Scoccimarro et
al., 2011) simulations are used as boundary conditions. GCM resolution is 0.75°.
Dynamical downscaling with NMMB model (Djurdjevic et al., 2013b) is done to the
0.06° resolution, for the SEE region presented in Fig. 1. High-resolution model
simulation can resolve fine scale features and increase quality of the summer
convective precipitation, which is related to the non-hydrostatic mode of simulation.
893
Fig. 1. NMMB domain of simulation for the period 1971-2100 using RCP8.5 scenario (purple)
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 (Djurdjevic et
al., 2011, Rajkovic et al., 2014). Fig. 2 evidently shows deficit of GCM resolution
and large improvement in increasing RCM resolution, having in mind complexity of
Serbian terrain. High resolution approach enables climate change analysis on local
scales. More on NMMB performance in climate simulations can be found in
Djurdjevic and Krzic (2013a).
894
Fig. 2. Topography on GCM (upper left), coarse RCM (upper right), fine RCM (lower left)
and DEM (lower right) resolutions.
3 Bias Correction
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 (Cressman, 1959). This method is in use
usually for numerical weather forecast purposes and represents optimal combination
of simplicity, computational efficiency and quality of the obtained interpolated data.
It is used for the climate analysis in project national viticulture zoning (Ivanisevic et
al., 2015).
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.
895
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 30-
years periods, base period 1971-200 and two future climate periods 2011-2040 and
2071-2100.
Fig. 3. Probability density functions (a), cumulative density functions (b) for the observed and
model mean daily temperature data, and the correction function (c).
4 Results and Discussions
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 (Ruml et al., 2012).
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.
896
Fig. 4. Mean annual temperature change for the periods 2011-2040 and 2071-2100 with
respect to the period 1971-2000, according to the RCP8.5 scenario.
Fig. 5. Mean seasonal temperatures change for the periods 2011-2040 and 2071-2100 with
respect to the period 1971-2000, according to the RCP8.5 scenario.
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
897
already reached point of very rare event in lower altitudes. During the period 2071-
2100 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.
Fig. 6. Number of frost (FD) and ice days (ID) change for the periods 2011-2040 and 2071-
2100 with respect to the period 1971-2000, according to the RCP8.5 scenario.
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.
898
Fig. 7. Number of summer (SU) and tropical nights (TR) change for the periods 2011-2040
and 2071-2100 with respect to the period 1971-2000, according to the RCP8.5 scenario.
Other analysis (not shown here) shows extension of the growing season length 20-
30 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.
5 Conclusions
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
899
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.
References
1. Cressman, G. P. (1959) An operational objective analysis system, Monthly
Weather Review, 87, p.367-374.
2. Djurdjevic, V., Rajkovic, B., Vujadinovic, M. and Vukovic, A. (2011) Сlimate
Change Scenarios, In: Initial National Communication of the Republic of Serbia
under the United Nations Framework Convention on Climate Change, edt. D.
Bozanic and M. Gasperic, The Ministry of Environment and Spatial Planning of
Republic of Srebia, p.150.
3. Djurdjevic, V. and Krzic, A. (2013a) Analysis of the downscaled ERA40
reanalysis performed with the NMMB model, ORIENTGATE project report,
p.20.
4. Djurdjevic, V., Janjic, Z., Pejanovic, G., Vasic, R., Rajkovic, B., Djordjevic, M.,
Vujadinovic, M., Vukovic, A. and Lompar, M. (2013b) NCEP’s multi-scale
NMMB model in the Hydrometeorological Service of Serbia: experiences and
recent model developments, EGU General Assembly, Geophysical Research
Abstracts, Vol. 15, EGU2013-8217, 7-12 April 2013, Vienna.
5. Ivanisevic, D., Jaksic, D. and Korac, N. (2015) Atlas of Viticulture (in Serbian).
Belgrade: Statistical Office of the Republic of Serbia.
6. IPCC (2007) In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M.,
Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change.
Cambridge University Press, Cambridge.
900
7. IPCC (2013) Climate Change 2013: The Physical Science Basis. Contribution of
Working Group I to the Fifth Assessment Report of the Intergovernmental Panel
on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen,
J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge
University Press, Cambridge, United Kingdom and New York, NY, USA, 1535
pp, doi:10.1017/CBO9781107415324.
8. Nakicenovic, N., Alcamo, J., Davis, G., de Vries, B., Fenhann, J., Gaffin, S.,
Gregory, K., Grüber, A., Jung, T.Y., Kram, T., la Rovere, E.L., Michaelis, L.,
Mori, S., Morita, T., Papper, W., Pitcher, H., Price, L., Riahi, K., Roehrl Rogner,
H.H., Sankovski, A., Schlesinger, M., Shukla, P., Smith, P., Swart, R., van
Rooyen, S., Victor, N. and Dali, Z. (2000) Special Report on Emissions Scenarios
(SRES). Contribution to the Intergovernmental Panel on Climate Change.
Cambridge University Press, Cambridge.
9. Rajkovic, B., Vujadinovic, M. and Vukovic, A. (2014) Report on revisited
climate change scenarios including review on applied statistical method for
removing of systematic model errors, with maps of temperature, precipitation and
required climate indices changes, In: Second national communication of the
Republic of Serbia under the United Nations framework convention on climate
change, MERZ, Belgrade, Serbia.
10. Ruml, M., Vukovic, A., Vujadinovic, M., Djurdjevic, V., Rankovic-Vasic, Z.,
Atancakovic, Z., Sivcev, B., Markovic, N., Matijasevic, S., and Petorvic, N.
(2012) On the use of regional climate models: Implications of climate change for
viticulture in Serbia, Agricultural and Forest Meteorology, 158, p.53-62, doi:
10.1016/j.agrformet.2012.02.004.
11. Scoccimarro, E., Gualdi, S., Bellucci, A., Sanna, A., Fogli, P.G., Manzini, E.,
Vichi, M., Oddo, P. and Navarra, A. (2011) Effects of Tropical Cyclones on
Ocean Heat Transport in a High Resolution Coupled General Circulation Model.
Journal of Climate, 24, p.4368-4384.
901