Changes in climate zones across Turkey

Climate is a natural resource vital to our well-being, health and prosperity. Climate change can cause climatic hazards such as droughts and floods in the world. The information gathered, managed and analysed by National Meteorological and Hydrologic Services (NMHS), in collaboration with other regional and international stakeholder organizations and programs, helps decision-makers and end users in their activities and projects for planning and adaptation to expected conditions. As stated in the World Meteorological Organization report (WMO 2016), in this way, decisions may be taken into consideration for planning which reduce risks and optimize socio-economic benefits.

According to the results of the Intergovernmental Panel on Climate Change (IPCC), late 20th century levels of temperature will increase about 2 °C or more in some parts of the world if climate change adaptation cannot be achieved as it is expected. Some parts of the world may gain benefit from this temperature increase. However, if this increase reaches 4 °C or more, some other parts would be faced with risk of agricultural productivity, which in turn would negatively impact global food security. Moreover, renewable water resources would reduce and competition for water among sectors would increase starting from most dry regions (IPCC 2014).

The focus area of this study is Turkey, located in the eastern part of the Mediterranean region, which is particularly vulnerable to climate change (Figure 1). The Mediterranean region has been identified as one of the main climate change hotspots responsive to climate change due to water scarcity, concentration of economic activities in coastal areas, and reliance on climate-sensitive agriculture (EEA 2017).

Turkey is characterized by highly variable climatic conditions, and experiences semi-dry Mediterranean, semi-humid, and humid conditions. For the country as a whole, there is no dry zone. Most of the central Anatolia and some parts of the east and southeast of Turkey have experienced semi-dry periods, whereas the northeast Black Sea coastal areas have a relatively wetter climate (Deniz et al. 2011).

The coastal areas of Turkey, bordered by the Aegean Sea and the Mediterranean Sea, are characterized by Mediterranean climate with hot, dry summers and mild to cool, wet winters. The Black Sea region has a temperate oceanic climate with warm, wet summers and cool to cold, wet winters. Climate conditions can be harsh in the more arid interior regions. Mountains close to the coastal zone prevent propagation of maritime impacts into inland areas, resulting in a continental climate with sharply contrasting seasons in the central Anatolian plateau of the interior of Turkey. Winters on this plateau are especially severe. In this context, the coastal areas are characterised by milder climates, the inland Anatolian plateau experiences extremes of hot summers and cold winters with limited rainfall (Sensoy et al. 2016).

Both the observed decreasing trends in the Mediterranean Basin and Turkey's precipitation in winter and the rising trend in frequency for the occurrence of low intensity precipitation events in Turkey are the most substantial points in terms of the precipitation changes and variability (Turkes et al. 2009). These have also been indicating increased drying and desertification in the western and southern regions of Turkey, characterized generally by the Mediterranean climate.

Future climate conditions of Turkey are projected to be worsened in the fourth Assessment Report of IPCC. The report mentions already worsened climate conditions in Europe. According to the findings, Europe's average temperature has already risen 1 °C, which is above the global average. Moreover, the southern part of Europe has dried 20% in the last century. Water balance has changed as a result of the binary effects of climate change and changes in land use patterns. Thereof, the stream flow would decrease in south-east Europe, including Turkey (IPCC 2007). The Mediterranean climate zone is one of the most vulnerable zones to the climate change. Based on the IPCC findings, Turkey's climate change impacts are mainly characterized by an increase in temperature and a decrease in precipitation. The land use patterns of Turkey have been changing due to overpopulation and economic circumstances. As a result, the decrease of water resources may cause considerable water shortage, and Turkey is expected to be a ‘water scarce’ country.

The results of a recent study (Selek & Tuncok 2014) show that average temperatures within the Seyhan basin, which is located in the southern part of Turkey, can increase as much as 5 °C within the 2010–2100 time frame. In this context, it is apparent that many parts of the basin will suffer from significant water shortages. This will be further exacerbated by rising demand of limited water resources in many sectors, particularly irrigation water demand of the agricultural sector. Precipitation over the basin is expected to drop by 35% in the lower-segments, 23.5% in the middle-segments and 17% in the upper-segments within the 2010–2100 time frame, which will worsen the conditions for local communities and wildlife.

It is important to implement a structured approach to evaluate variations in climatic conditions. In similar applications, a climate index driven approach (Deniz et al. 2011) was used to characterize variations in climate. Typically, climate indices which are diagnostic tools to define the state of a climate system and understanding of the various climate mechanisms are described through climate variables. In parallel with this methodology, Turkes et al. (2016) examined consecutive periods of 1950–1980 and 1981–2010 via statistical comparative analysis and showed an increase in air temperature and a decrease in precipitation through central Anatolia.

There are many climate classification methods dating back to the 1900s which use spatial variations of hydro-meteorological parameters to differentiate climatic zones. For example, Köppen (1923), De Martonne (1926), Thornthwaite (1948), Erinc (1965) and Trewartha (1968) are only a few of the methodologies to be applicable. However, some difficulties have been experienced to apply any specific method to the study area due to lack of representative data sets. The respective formulations require the use of data sets, which are at times difficult to access and represent at the river basin scale. In this context, the Thorthwaite precipitation efficiency index, which can be accurately represented with available data sets, was used in this study as an indicator for the differentiation of climatic zones in time across Turkey.

The main objectives of this study were classification of climatic zones in Turkey with respect to their climatic and meteorological characteristics and delineation of temporal and spatial changes in the index values for water resources planners, irrigation water managers, water authorities, decision makers, etc. For this purpose, the areal change of climatic zones was evaluated in time by using observed historical hydro-meteorological data. In this study, the Thornthwaite index was used to identify aridity and humidity characteristics of basins and to map temporal and spatial changes. Two distinct periods (1950–1980 and 1981–2010) were used to assess the climatic conditions and evaluate historical changes.

In order to determine climate change with its past, present and future, there have been a number of methodologies used including statistical analysis of datasets, marine biology, tree ring chronologies, etc. (Brown et al. 2011; Gholami et al. 2015; Turkes et al. 2016).

The Thornthwaite climate classification index (Thornthwaite 1948) was used to represent aridity and humidity characteristics, through use of precipitation, evapotranspiration and temperature. Thornthwaite index values were used to define climate zones and associated climate classes. Moreover, water budget components so called surplus and deficit were calculated by a water balance model. In order to interpret results with hydrological and regional perspectives, river basin and country scales were used respectively.

To determine the change of climate zones in time across Turkey, two distinct periods (1950–1980 and 1981–2010) were selected. The 30-year climate normal could be interpreted as the true background state, offset by decadal and longer-term tendencies, and further biases are introduced by inter-annual variability as well as random and systematic errors and for stationary time series. Since 1956, the WMO has recommended that each member country re-compute their 30-year climate normal during 10-year periods (Arguez & Vose 2011). It is important to note that a time interval of 30 years was selected in-line with international standards. The main driver behind the use of a 30-year time frame is the ability to set climate ‘normal’ as reference points, and to compare current climatological trends to that of the past or what is considered as ‘normal’. WMO (2016) stated clearly that the 30-year period is long enough to filter out any inter annual variation or anomalies, but also short enough to be able to show longer climatic trends.

As stated in the objectives, this research was focused to identify the areal changes in climate zones, both at the basin- and country-scales. Basin scale was used because it is the most representative geographical definition to evaluate hydrological processes such as droughts, floods, hydro-climatology, etc. Country scale was then used to evaluate the results of the study at a macro level.

Climate classification assessments require not only long-term historical records but also representative spatial coverage across geographic regions. In this context, systematic measurements administered by the Turkish State Meteorological Service were evaluated to select the optimum number of meteorological stations. The systematic meteorological measurements started in the 1930s, but the number of stations were not adequate in determining meteorological characterization of the whole country. That is why a total of 107 stations, which had continuous coverage over the 60-year time frame between 1950 and 2010 was selected. The selected stations also have good geographical representation at the basin and country scales. The meteorological parameters recorded at the stations included precipitation and temperature which are the key inputs to the Thornthwaite precipitation efficiency index. Therefore, temperature and precipitation series of meteorological observation stations were used in the study. To this end, quality controlled data sets were obtained from Turkish State Meteorological Service data gaps were completed using historical statistics at respective stations. The list of stations used in the analyses is given in Table 1, including their geographical coordinates across Turkey.

Thornthwaite climate classification index values were used to analyse spatial and temporal changes of climate zones. A representative set of climate classes were used to identify distinct climate driven pressures in 25 river basins. The geographical setting of river basins across Turkey is depicted in Figure 2.

The Thornthwaite index and associated climate types are documented in Table 2. Letter designation reflects that Thornthwaite's original description of climate classes varies between very humid (A) and dry (E).

The point meteorological data sets obtained from meteorological stations, as listed in Table 1, were used to calculate associated values for the Thornthwaite index in a geographical information system (GIS) setting by using ArcGIS software. An inverse distance interpolation is one of the simplest and most popular interpolation techniques. It combines the proximity concept with the gradual change of the trend surface. An inverse distance weighted (IDW) interpolation is defined as a spatially weighted average of the sample values within a search neighborhood (Shepard 1968; Franke 1982; Diodato & Ceccarelli 2005).

The optimal power is determined by minimizing the root mean square prediction error (RMSPE). In order to represent the conditions realistically at the river basin scale, the (p) value was selected as 2, in-line with the recommendations by Cetin & Diker (2003). The (p) value of 2 reflects the outcome of an iterative process to properly represent spatial distribution of the Thornthwaite index across the river basin systems in Turkey.

It was noted that the index-based approach had similar findings in spatial and temporal representation of climate parameters and associated aridity conditions in comparison to climate change studies. The index based approach was used to realistically estimate climate conditions at various geographical scales in an accurate and representative manner. The outcomes were also assessed in the context of readily available climate change models and their spatial and temporal changes in the study area. In this study, climatic zones of Turkey were defined with respect to the Thornthwaite index based on their climatic and meteorological characteristics. In this context, climate driven conditions were first evaluated using climate parameters recorded at meteorological stations. The average representative observed values of meteorological stations for temperature and total precipitation during the 30-year periods of 1950–1980 and 1981–2010 are mapped as shown in Figure 3.

The changes between the two distinct 30-year periods (1950–1980) and (1981–2010) were also mapped, as shown in Figure 4. Average precipitation showed an increasing pattern along the northeastern part of Turkey, with a maximum change of 23.2%. In contrast, the southeastern part experienced a drop with a maximum change of 23.2%. Average temperatures showed a decreasing pattern, up to 1.0 °C along the northern part of Turkey and an increasing pattern, up to 1.1 °C, along the southeastern part of Turkey. The combined impact of temperature and precipitation will potentially have multiplier impacts on various human needs (mainly potable water) and various strategic sectors including but not limited to agriculture, industry and energy. The ecosystems and associated environmental services will also be severely impacted.

The outputs of a study based on A2 scenario, which is a high emission scenario that was used in the fourth Assessment Report of IPCC, presented by Sen (2013), indicate that temperatures in Turkey are projected to increase between 1.0 and 2.5 °C by the mid-21st century and between 2.5 and 5.0 °C by the end of the century with the 1961–1990 period as reference (Figure 5). The changes are not uniformly distributed. The eastern and southeastern parts of Turkey illustrate comparatively larger increases in temperatures. On the other hand, annual precipitation is expected to decrease in the southern parts of Turkey while it will tend to increase in the northern parts, especially in the northeastern parts. The reductions along the Mediterranean coastal line could be as large as 20% by the mid-century and 30% by the end of the century. Similar magnitudes could be stated for the increases along the northeastern coastal areas of Turkey.

The outputs of the present study conducted by observed hydro-meteorological parameters (precipitation and temperature) show a significant consistency with the results of the previous study carried out by Sen (2013).

The spatial distribution of the Thornthwaite index is presented in Figure 6. According to the results of the index approach, no area with ‘dry zone’ was detected in Turkey during the 60-year period evaluated in this study. On the other hand, inland regions of Turkey are characterized by semi-dry zones. In addition to the analysis within the respective 30-year periods, relative changes between these two 30-year time frames were also evaluated. This comparison was undertaken at both basin and country scales. The outcomes of country-scale analysis are described in Table 3. By comparing the two periods, it can be seen that there are significant changes, especially in semi-dry and very humid zones, in the order of 14.2 and 13.9%, respectively. On the other hand, semi-dry–less humid and semi-humid areas show a decreasing trend. These trends are consistent with the outcomes presented in Figures 4 and 5.

The trends at the country scale, specifically for two distinct classes of semi-dry and very humid zones, highlight the importance of evaluating spatial changes in climate conditions at the basin-scale. The basin-scale analysis will then allow identification of specific trends more accurately. The outcomes of the basin-scale analysis are presented in Figure 7.

In the context of semi-dry zones, Basin No. 11 (Akarcay basin), which is a closed basin, has experienced the highest level of increase (in the order of 35–45%) in semi-dry conditions. This is specifically important to evaluate the risks on the sustainability of the lake ecosystems (Eber and Aksehir lakes). Basin No. 21 (Euphrates-Tigris basin), which is a transboundary system accounting for approximately 30% of the water resources potential in Turkey, has experienced a high level of increase (in the order of 15–25%) in semi-dry conditions. This has important implications for both water resources management and hydroelectric power production. The two river systems in the basin account for approximately 40% of the hydroelectric power production potential in Turkey.

In the context of very-humid zones, Basin No. 22 (Eastern Black Sea basin), has experienced an increase in the order of 5% in very humid and a decrease in the order 5% in semi-dry–less humid conditions. Basin No. 23 (Coruh basin), which is a transboundary system with a share of 21% Turkey's hydroelectric potential, has also experienced an increase in the order of 5% in very humid and a decrease in the order 5% in semi-dry–less humid, semi-humid and B1-humid conditions. It is important to note that these two river basins account for 30% of the hydroelectric power production potential in Turkey.

In this study, a chain of structured processes were implemented by using an index based approach. In this context, the Thornthwaite index was introduced as a practical and robust approach to evaluate spatial and temporal changes in climate parameters, mainly driven by precipitation, temperature and evapotranspiration.

• A two-step approach was undertaken by using climate conditions at country and basin-scale conditions. This scale was used as an alternative to the global and continental scales at macro-level and municipality scale at micro-level. These scales allowed more representative reflection of the spatial and temporal variations without unrealistic distortions driven by the global and continental processes. The outcomes were then evaluated to local and regional vulnerabilities as reflected through use of the Thornthwaite index. The country-scale analysis was undertaken and changes in precipitation, temperature and evapotranspiration were evaluated for two distinct 30-year time frames (1950–1980 and 1981–2010). These historical periods were used to better reflect the meteorological conditions and associated trends. It was determined that the changes in climate as calculated through the Thornthwaite index were consistent with climatic and meteorological processes at the country scale.

• In the basin-scale analysis, it was determined that semi-dry and very-humid climate zones had experienced the largest amount of increase. This in turn has direct and indirect impacts on the sustainable use of water resources.

The results of the study indicate that during the 60-year time frame, no arid zones had been experienced in Turkey. On the other hand, an increase of semi-dry and dry humid zones and a decrease of semi-dry–less humid, semi-humid and humid zones had been experienced. The Thornthwaite index indicates that significant spatial and temporal variations of climate conditions have taken place across Turkey. Unless greenhouse gas emissions reduction measures are undertaken at the global scale, it will be challenging for Turkey to maintain a balanced portfolio of actions through use of climate change adaptation and mitigation measures. In this context, there is a high potential that semi-arid areas in Turkey will further increase in size in the coming decades.

We are grateful to Dr Hakan Aksu and Yeliz Tüzgen for helping us to use GIS applications and exchange thoughts through development of processes implemented in this study.