Monitoring climate change, drought conditions and wheat production in Eurasia: the case study of Kazakhstan.

Wheat is an important global food security commodity. Kazakhstan is currently a producer and exporter of high-quality wheat to global markets. The most important wheat-growing regions, which lie in the northern part of Kazakhstan, are based on spring-sown rain-fed cultivation and are susceptible to climate change and drought. Using the monthly surface air temperature and precipitation data from 1950 to 2020 from 110 meteorological stations over Kazakhstan and in addition wheat cultivation data, the research aims to analyze climate change, drought occurrence, and wheat cultivation trends in Kazakhstan in recent 70 years and investigate relationships between wheat productivity and drought. The linear method and two drought indices (Standardized Precipitation Index and Standardized Precipitation Evapotranspiration Index) and in addition, Pearson's correlation coefficient have been used to characterise the climate change trends and vulnerability of agriculture in Kazakhstan to drought. The geographic information system (GIS) was applied to display climate change, drought, and wheat referenced information. The research has shown that the 70-year (1950-2020) linear rates of annual mean surface temperature in Kazakhstan have significantly increased (on average 0.31 °C per decade) with the precipitation trends are not obvious and fluctuated trends of drought. The wheat yield demonstrates strong internal variability and wheat yields were significantly correlated with 3-month June and July drought indices over the period of 1950-2020. The results underline the potential susceptibility of wheat yields in Kazakhstan to any future reductions in precipitation and increase in drought occurrence and intensity.

Introduction

Research rationale

Climate change is characterized by increasing temperatures and more extreme climatological events (Trenberth et al., 2014). Drought is a slowly developing extreme climatological event, which often has the longest duration, but the lowest possibility to be predicted at the moment (Mishra & Singh, 2010, 2011). Drought can be meteorological, agricultural, hydrological, socioeconomic based on the number of days below a particular rainfall threshold, soil moisture deficits, or on the levels of surface and groundwater storage, respectively (Pielke et al., 2005; Naumann et al., 2018; Pan et al., 2018). It has been reported in several studies that drought conditions affect agriculture in several ways, one of which is its direct impact on wheat productivity (Arora, 2019; Javadinejad et al., 2021) and as a consequence hindering the prospects of achieving some of the Sustainable Development Goals (SDG) including eradication of poverty and hunger and improvement of health and sustainability (Yusa et al., 2015). Studies have shown that droughts were common during the 20th century and evidence from global and regional climate models has pointed to the increase in droughts in terms of their duration and spatial extent (Spinoni et al., 2015).

The most common practice for studying droughts involves the use of indices that allow the comparative severity of a particular drought event to be established and allow the spatial and temporal variability of drought to be evaluated (Heim, 2002; Zargar et al., 2011; Wei et al., 2021). To date, many indices have been introduced and utilized (Guo et al., 2018; Kim et al., 2021). The Palmer Drought Severity Index (PDSI) has been widely used but requires temperature and precipitation data, as well as empirical relationships, to define a climate weighting factor (Yu et al., 2019; Yang et al., 2020). The Standardised Precipitation Index (SPI) is simpler to calculate, is based on the analysis of monthly rainfall series, and provides results at the multiple timescales relevant to soil moisture, streamflows, and groundwater conditions (Jenkins and Warren, 2015; He et al., 2015; Tigkas et al., 2019). The SPI is now widely used to characterise vulnerability to drought as well as a potential predictor of wheat production (Livada and Assimakopolous, 2007; Noori et al., 2012; Anshuka et al., 2019). In addition, the Standardized Precipitation-Evapotranspiration Index (SPEI) has been developed which factors in monthly temperature data and day length while maintaining the multi-scalar approach of the SPI (Vicente-Serrano et al., 2010; Peña-Gallardo et al., 2016). This SPEI is essentially based on water balance and should therefore provide a more appropriate and sensitive tool for characterizing comparative water availability during the short growing season of spring-sown wheat in marginal areas such as Kazakhstan (Mahmoudi et al., 2019).

This study attempts to analyse the relationship between climate change, drought conditions, and wheat production in Kazakhstan. In Kazakhstan, there have been several studies conducted on this issue. Recently, Schierhorn et al. (2020) used the econometric regression method to estimate the impact of the average temperature, heat, and precipitation on wheat and barley yields in the northern part of Kazakhstan. According to the results computed by Schierhorn et al. (2020), climate change has reduced wheat and barley yields between 1980 and 2015. The strongest reduction was observed in Kostanay region. Previously, Pavlova et al. (2014) simulated process-based crop growth by a mechanistic model and reported the impact of climate variability on wheat productivity in the steppe zone of Kazakhstan, highlighting that average wheat production from 2000 to 2010 was 1 ton ha−1, with high inter-annual fluctuations due to a shorter growing season, lower water supply and higher heat stress. Regarding the level of climate change impact on wheat production, Bobojonov and Aw-Hassan (2014) used modeling approaches and the authors came to the result that wheat production in Kazakhstan is moderately affected by climate change. Furthermore, it is now clear that El Niño through teleconnections can trigger noticeable anomalies in precipitation and temperature patterns around the globe, including Central Asia (Hu et al., 2017; Chen et al., 2021). It is found that strong El Niño events persistently enhance precipitation in Kazakhstan, while during weak El Niño events, significant wet anomalies have been observed (Hu et al., 2017). Estimating the impact of El Niño in the Kostanay region of Kazakhstan, Shmelev et al. (2021) used an econometric method and confirmed the high correlation between wheat yield fluctuations and El Niño.

This research makes additional contributions to current knowledge. First, it demonstrates the suitability of utilizing drought indices to improve agriculture planning and management. No study was found on the assessment of a relationship between climate change, drought conditions, and wheat production using droughts indices for Kazakhstan. The previous studies primarily rely on econometric regression methods and used modelling approaches that include main meteorological conditions but frequently suffer from the lack of specific data on soil moisture, streamflows, and groundwater conditions. The SPEI, a variant of SPI index uses these specific data, in addition to considering the influence of global warming to some extent. Second, this research focuses on examining the level of agreement between SPI and SPEI conducting the Pearson correlation method. Previous performance comparison research between SPI and SPEI indicated that SPEI performs better than SPI on most occasions (Gurrapu et al., 2014). Regardless of this fact, however, SPI continues to be widely used in different parts of the world (Peña-Gallardo et al., 2016). Finally, it applies drought indices over the whole wheat-producing regions of Kazakhstan for the period from 1950 to 2020, in contrast with previous studies, which focused on few locations and shorter periods.

The study area: Kazakhstan

Kazakhstan is located in heart of the Eurasian continent, it is nine largest country in the world and the country demonstrates stable economic growth. Kazakhstan is currently in the top ten of global wheat producers with the majority of the area sown (ca. 17.5 million hectares in 2020) relying on rain-fed spring-sown wheat. The Food and Agriculture Organization of the United Nations (FAO) estimates that the combined wheat production from Kazakhstan, Ukraine, and Russia will account for 25–30% of the world's wheat exports by 2030 (FAO, 2021a). Currently, all together Kazakhstan, Ukraine, and Russia contribute 21% of world exports of wheat, in particular Kazakhstan (4%), Ukraine (6%), and Russia (11%). Among these countries, Kazakhstan accounts for approximately 15% of wheat production (FAO, 2021a). However, any increased reliance on Kazakhstan wheat production on the part of global markets raises issues of food security since the land productivity decline (Swinnen et al., 2017; Hu et al., 2020), temperature change (Sommer et al., 2013a, Sommer et al., 2013b; Reyer et al., 2017; Schierhorn et al., 2020) and spring-sown wheat yields exhibit strong inter-annual variation and appear vulnerable to drought (Lioubimtseva et al., 2013; Zampieri et al., 2020). Although wheat was produced in Kazakhstan prior to the Second World War, it accounted for less than 3% of the total production of the the Union of Soviet Socialist Republics (USSR) (Durgin, 1962; McCauley, 2016). A change occurred in the post-war period when large areas of steppe grasslands in Northern Kazakhstan were ploughed for wheat cultivation. This change was the result of a strategic political decision, taken in the early 1950s, as part of Khrushchev's Virgin Land Programme to achieve a step-change in soviet grain production (McCauley, 2016). This was globally the largest ecosystem conversion of grassland after World War II and it had an important demographic dimension as well (Tokbergenova et al., 2018; Nyussupova et al., 2019). In the period since 1953, the area cultivated for wheat in Kazakhstan has varied from 19.6 to 8.7 million hectares while yields have varied from 0.09 to 2.52 tons per hectare (Almaganbetov, 2005; FAO, 2021a). Poor wheat harvests in Kazakhstan in the 1950s, 1960s, and early 1990s were blamed on climate change and drought (Meng et al., 2000; McCauley, 2016; Dubovyk et al., 2019). The extent to which soviet period wheat yields were influenced by meteorological factors, rather than agricultural policy, has remained contentious (Desai, 2013; Zampieri et al., 2017).

Objective and outline

The main objective of this study is to analyse historical and recent climate change and drought trends and its impact on wheat yield in Kazakhstan. The first step of this study is to analyze wheat cultivation data and monthly surface air temperature and precipitation data from 1950 to 2020 from meteorological stations of Kazakhstan. The second step is to calculate Standardized Precipitation Index and Standardized Precipitation Evapotranspiration Index to characterize drought occurrence and investigate relationships between wheat productivity and the Standardized Precipitation Index and Standardized Precipitation Evapotranspiration Index values. The final step is to discuss the potential consequences of future increases in temperature and decreases in precipitation.

The presentation of this study is organized in the following way. The next section describes data sets and methods used in the present research. The third section of research deals with current wheat cultivations trends in Kazakhstan. This section also analyses recent climate change observations in Kazakhstan and particular attention is given to annual and seasonal mean surface air temperature and precipitation during 1950–2020. Furthermore, it integrates climate change and drought trends and analyses its impact on wheat production. The last section gives a discussion, research limitation, and conclusion.

Data sets and methods

Wheat yield data

The state-level and oblast-level crop statistics (production, harvested areas, and yield) from 1950 to 2020 were collected from the Kazakhstan Ministry of Agriculture. Additionally, annual wheat production data for north-central regions of Kazakhstan (Akmola, Kostanay, North Kazakhstan, and Pavlodar) were derived from Kazakhstan's Agricultural Atlas (Koshim et al., 2018) and the Kazakh National Agrarian University, and this data was compared with Kazakhstan Ministry of Agriculture's data. Kazakh National Agrarian University has recently conducted a project on the improvement of drought and water monitoring for supporting agriculture management and mitigation of risks related to extreme weather conditions. The project provides an analysis of the wheat sector in Kazakhstan including present market position and expected future demand, market size, statistics, trends, Strength, Weakness, Opportunity, and Threat (SWOT) analysis and forecasts.

Meteorological data

Data sets used in the present research are daily mean temperature, daily maximum temperature, daily minimum temperature, and precipitation. The meteorological data was collected from 110 meteorological stations with the support of the national hydrometeorological service «Kazhydromet» (Kazakhstan Ministry of Ecology, Geology, and Natural Resources). The length of the record varies between stations, with some precipitation records going back to the late 19th century (e.g. Petropavlosk, 1893–2020; Kostanay, 1903–2020; Akmola, 1893–2020; Pavlodar, 1920–2020). The station data were checked for homogeneity and complete sets of records of both surface air temperature and precipitation for the period 1950–2020 were compiled for meteorological stations. Based on available data, climate change trends in Kazakhstan in recent 70 years have been evaluated by linear trends and a series of maps have been created by using a geographic information system.

Computation of the SPI drought index

Based on climate and wheat data, SPI drought index was calculated using the multi-scalar algorithm, which was applied in previous studies (Abiodun et al., 2013; Beguería et al., 2014; Zhang et al., 2016). The detailed methodology of calculating SPI and SPEI values was described by McKee et al. (1993), Vicente-Serrano et al. (2010), Teuling et al. (2013), Beguería et al. (2014), Bacanli (2017). The SPI was formed by McKee et al., 1993, McKee et al., 1995 to indicate meteorological drought, and it defines the number of standard deviations that the observed cumulative rainfall at a given time scale would deviate from the long-term mean (Surendran et al., 2017). The only input for calculation of SPI is long term precipitation records. Computation of the SPI involves fitting a gamma probability density function to a given time series of precipitation, whose probability density function, , is defined by Eg. (1) as:where, is a shape parameter, is a scale parameter, and is the amount of precipitation. is the gamma function, which is defined by Eg. (2) as:

Fitting the distribution to the data requires and to be estimated by Eg. (3) as follows:where, is the number of observations. This allows the rainfall distribution at the station to be effectively represented by a mathematical cumulative probability function as given by Eg. (4):

It is possible to have several zero values in a sample set. In order to account for zero value probability, since the gamma distribution is undefined for , the cumulative probability function for gamma distribution is modified by Eg. (5) as:where, is the probability of zero precipitation. Finally, the cumulative probability distribution is transformed by Eg. (6) into the standard normal distribution to yield the SPI by fitting the log-normal distribution with the sample mean and variance of the logarithmic transformed data and , the SPI becomes:

Because the gamma distribution tends towards the normal as the shape parameter tends to infinity, it is possible to use the normal probability distribution instead of gamma, which is computationally easier to calculate and maybe more accurate, due to a better fitting to the data. In this case, the SPI index becomes simply to calculate by using Eg. (7):

Computation of the SPEI drought index

The SPEI was developed by Vicente-Serrano et al. (2010, 2015), which reflects the monthly climatic water balance. The SPEI was applied in previous studies (Xu et al., 2018; Tong et al., 2018; Chen et al., 2020). Since the SPEI represents the normalized precipitation minus evapotranspiration it can be used to examine the rates of onset and recession of drought at the different frequencies of relevance to agriculture (soil moisture availability) and water supply (surface flow, groundwater). Computation of the SPEI stars from defining potential evapotranspiration by using Eg. (8):where is the monthly mean temperature (°C); is a heat index, which is calculated as the sum of 12 monthly index values; is a coefficient depending on and is a correction coefficient computed as a function of the latitude and month. Deficit or surplus accumulation of a climate water balance at different time scales with a value for , the difference between the precipitation and for the month is calculated by using Eg. (9):

The calculated values are aggregated at different time scales, following the same procedure as that for the SPI. The difference in a given month and year depends on the chosen time scale . For example, the accumulated difference for 1 month in a particular year with a 12-month time scale is calculated by using Eg. (10):where is the − PET difference in the first month of year , in millimetres. The log-logistic distribution was selected for standardizing the series to obtain the SPEI. The probability density function of log-logistic distributed variable is expressed by Eg. (11) as:where , , and are scale, shape, and origin parameters respectively, for values in the range (). Thus, the probability distribution function of the series is given by Eq. (12) as:With the SPEI can easily be obtained as the standardized values of by using Eq. (13):where for and is the probability of exceeding a determination value, . If , then is replaced by and the sign of the resultant SPEI is reversed. The constants are , , , , , and .

Computation of Pearson's correlation

In order to examine the impact of droughts on wheat yields and advantages of using the SPEI over the more widely used SPI, 3-month and 6-month values were correlated by Pearson's coefficient with wheat yields for the four most important wheat producing regions in Kazakhstan (Akmola, Kostanay, North Kazakhstan and Pavlodar regions). Pearson's correlation coefficient is a measure of the linear correlation between two variables (Pearson, 1895). It is calculated by using Eq. (14), in which x represents the independent variable and y represents the dependent variable.

The complete dependency between two variables is expressed by either -1 or +1, and 0 represents the complete independency of the variables. Pearson's correlation analysis has been employed in many studies to examine the correlation among drought indices as well as their relationships with wheat yield (Leilah and Al-Khateeb, 2005; Qaiser et al., 2021).

Analysis of key-finding results

Wheat cultivation trends

The growing season for spring wheat is short. Sowing cannot take place until the winter snow has melted, in May, while harvesting tends to start in late August and can continue for up to two months (Lioubimtseva and Henebry, 2012). Soil moisture levels are affected by both winter snowfall and summer rains (Koshim et al., 2018). Nonmeteorological factors affecting wheat yields include length, and sequence of fallow rotation, application of fertilizers, and herbicides, use of certified seed, and efficiency of the harvesting process as well as farm capital stock, labour, and government economic subsidies (Almaganbetov, 2005). The sown area for wheat in Kazakhstan has changed over time with a peak value of 19.6 million hectares in the late 1960s declining to 8.7 million hectares by 1999 and recovering to 17.5 million hectares by 2020 (FAO, 2021b). As shown in Figure 1, the north-central regions of Kazakhstan (Akmola, Kostanay, North Kazakhstan, and Pavlodar regions) together account for about 80% of the sown area and 85–90% of the tonnage harvested (USDA, 2020).

Figure 1

Wheat producing areas of Kazakhstan.

Spring wheat is dominant in total wheat production in Kazakhstan, accounting for 15.5 million tons, while durum wheat comprises roughly 8.5% of total wheat production in Kazakhstan - nearly 1.4 million tons. Some winter wheat is grown in the south of the country but represents a minor share of Kazakhstan's total wheat production, nearly 0.3 million tons of production. Figure 2 shows the wheat productivity data for the period 1950–2020 for all regions of Kazakhstan and these time series show strong interannual variability. Given the market dominance of Akmola, Kostanay, North Kazakhstan, and Pavlodar regions there appears to be little scope for compensating poor yield in one oblast with good yield in another as can happen at the supra-national scale with the combined outputs from Ukraine, Kazakhstan, and Russia (Fehér and Fieldsend, 2019). Nine productivity minima are evident in the case of the Akmola, Kostanay, North Kazakhstan and Pavlodar regions with yields lower than 1.1 ton per hectare.

Figure 2

Spring wheat production and the overall wheat yields in Akmola, Kostanay, North Kazakhstan and Pavlodar regions for period of 1950–2020.

Climate change trends

The climate of Kazakhstan is continental, characterized by intensely cold winters with January air temperatures ranging from -18.5 °C, in the north of the country, to -1.8 °C in the south, and hot summers with July air temperatures ranging from 19.4 °C in the north to 28.4 °C in the south (Salnikov et al., 2015). It has been observed for the period 1950–2020 that the linear rates of mean surface air temperature have increased at the rate of 0.31 °C per decade. Annual and seasonal changes in mean surface air temperature over territory of Kazakhstan are shown in Table 1 and Figure 3. It could be seen from Table 1 and Figure 3 that all trends in the series of annual and seasonal values of surface air temperature are positive and statistically significant, indicating a steady increase in air temperature in Kazakhstan from 1950 to 2020. More tangible increase of the mean surface air temperature is observed in the southwestern regions of Kazakhstan, from 0.32 °C to 0.50 °C per decade, minor changes in the mean surface air temperature are observed in the north-, north-eastern and central regions, from 0.19 °C to 0.23 °C per decade (Kazhydromet, 2020).

Table 1

Linear trends of mean surface air temperature during 1950–2020, °C per decade.

Location	Regions	Annual	Winter	Spring	Summer	Autumn
North	Akmola	0.25	0.03	0.61	0.01	0.37
	Kostanay	0.34	0.11	0.54	0.15	0.55
	Pavlodar	0.19	-0.10	0.64	0.03	0.25
	North Kazakhstan	0.21	-0.02	0.44	0.00	0.44
Center	Karaganda	0.23	0.03	0.70	0.01	0.19
West	Aktyubinsk	0.40	0.21	0.55	0.32	0.48
	Atyrau	0.42	0.32	0.46	0.46	0.42
	Mangistau	0.32	0.25	0.37	0.46	0.23
	West Kazakhstan	0.50	0.40	0.53	0.57	0.49
South	Almaty	0.25	0.07	0.58	0.20	0.21
	Zhambyl	0.27	0.12	0.59	0.18	0.23
	Kyzylorda	0.40	0.25	0.77	0.26	0.32
	Turkestan	0.32	0.23	0.54	0.21	0.28
East	East Kazakhstan	0.20	-0.07	0.62	0.16	0.17
Country level	Kazakhstan	0.31	0.11	0.59	0.21	0.32

Figure 3

Linear trends of annual and seasonal mean surface air temperature during 1950–2020, °C per decade: (a) annual; (b) winter; (c) spring; (d) summer; (e) autumn.

As for seasonal changes, the highest rate of increase in the mean surface air temperature is observed in the spring period (0.59 °C per decade), the lowest - in the winter period (0.11 °C per decade). In winter, the highest rate of increase in the mean surface air temperature was recorded in the southern and western regions from 0.21 °C to 0.51 °C per decade. In February, a tendency in air temperature increase was observed from 0.37 °C to 0.98 °C per decade overall territory of Kazakhstan. In December, a decrease in air temperature from -0.02 °C to -0.77 °C per decade was observed in the north-eastern, central, and southern regions. In spring, the most intense and statistically significant warming is observed from 0.30 °C to 0.93 °C per decade. The highest rate of increase in air temperature was noted in March from 0.58 °C to 1.79 °C per decade. In summer, the stable trend of temperature increase was observed in the east, as well as in the southern and western regions, from 0.15 °C to 0.93 °C per decade. In autumn, over the past seven decades, a steady increase in surface air temperature was observed in the north-west, west, and south, from 0.37 °C to 0.69 °C per decade.

In contrast to air temperature, the change in the regime of atmospheric precipitation on the territory of Kazakhstan for the studied period is not obvious. It could be seen from Figure 4 that the annual precipitation in Kazakhstan spatially varies between 100-200 mm in Western Kazakhstan and 350–600 mm in Northern Kazakhstan; average rainfall was between 130 mm in central and southwestern regions to 350 mm in Northern Kazakhstan (Karatayev et al., 2017; Valeyev et al., 2019). There have been positive and negative atmospheric precipitation anomalies since 1950 over Kazakhstan (Table 2). Annual and seasonal precipitation changes in Kazakhstan are visualized in Figure 5 and statistically reported in Table 2. It could be seen from Figure 5 that, there is a tendency of a nonsignificant increase in the annual amount of precipitations by 5.5 mm per decade for the period 1950–2020, and an increase in the amount of precipitation by 1.3–3.8 mm per decade is observed for all seasons with the length of three months, except for the autumn season, in which the decrease in the amount of precipitation was observed at rate 1.0 mm per decade (Kazhydromet, 2020).

Figure 4

Station-based distribution of annual and seasonal mean precipitation during 1950–2020, mm per decade: (a) annual; (b) winter; (c) spring; (d) summer; (e) autumn.

Table 2

Trends of annual and seasonal precipitation for period of 1950–2020, mm per decade.

Location	Regions	Annual	Winter	Spring	Summer	Autumn
North	Akmola	13.4	3.5	3.6	5.5	-0.8
	Kostanay	2.8	-0.8	8.3	0.2	-4.7
	Pavlodar	8.8	0.1	5.7	3.8	-0.1
	North Kazakhstan	15.4	1.1	10.2	4.7	-0.1
Centre	Karaganda	4.7	-0.2	1.8	6.0	-2.4
West	Aktyubinsk	-1.2	-0.3	6.2	-3.1	-3.9
	Atyrau	4.4	3.1	6.2	-3.4	-1.3
	Mangistau	-1.4	3.6	-4.3	1.1	-1.5
	West Kazakhstan	-2.3	-3.3	6.7	-5.1	-1.1
South	Almaty	12.4	4.8	3.9	2.9	1.2
	Zhambyl	0.2	1.6	-2.9	2.9	-0.9
	Kyzylorda	-5.7	-1.1	-1.1	-0.4	-3.1
	Turkestan	9.4	3.5	3.3	2.2	0.6
East	East Kazakhstan	6.4	0.7	3.7	2.8	-0.1
Country level	Kazakhstan	5.5	1.3	3.8	1.8	-1.0

Figure 5

Linear trends of precipitation for period of 1950–2020, % per decade: (a) annual; (b) winter; (c) spring; (d) summer; (e) autumn.

At the regional level, an increase in precipitation was observed in most regions of Kazakhstan, with the exception of Kyzylorda, Aktobe, West Kazakhstan, and Mangistau regions, where precipitation decreases by 1.2–5.7 mm per decade. Steady tendencies of increase in atmospheric precipitation are observed in the winter season, mainly in the northeast (3–14% per decade), southeast (9–15% per decade), and west (16–22% per decade). In the summer season, the highest rate of decrease in the amount of precipitation is observed in June and August (4–20% per decade). The negative trends for precipitation amounted to 0.3–13% per decade in the western region of the country. In autumn, a negative trend in precipitation is recorded in most of the territory of Kazakhstan (2–20% per decade). As can be seen from Figure 6, the Northern, Eastern, and Sothern Kazakhstan are wetter in the spring and summer seasons, while the Western and Central parts of Kazakhstan are extremely drier in the summer and autumn seasons.

Figure 6

Linear trends of dry and wet periods during 1950–2020, % per decade: (a) annual; (b) winter; (c) spring; (d) summer; (e) autumn.

Drought trends

Annual SPEI time-series for representative meteorological stations from the main wheat-producing regions are shown in Figure 7. In the past 70 years, each oblast (Akmola, Kostanay, North Kazakhstan, and Pavlodar) shows fluctuated trends of drought. In all cases, major drought conditions are apparent in the 1950s and at sub-decadal levels through to all periods of 1950 and 2020. Two SPEI time-series, 12 and 24 months show that 1955–1960, 1991–1995, and 2010–2015 as extreme drought periods, and 48-month SPEI shows similar dry periods few differences between 1985 and 2000. The lowest 3-month SPEI (extreme drought) was found in 1955, 1991, 1998, and 2014. The short-term severe drought conditions continued during 1955–1960, 1965–1985, and 1995–2020. In the years 1998, 2010, and 2012, more than 50% of the area of the country was affected by drought conditions of different severity with the largest droughts in terms of the areal spread occurring in 1998 and 2012.

Figure 7

Temporal variations of drought trends in Kazakhstan during 1950–2020.

Correlation analysis

Yield minima in the wheat data are all related to minimum values in the annual occurrence 3-month SPEI series (Table 3) and fit within periods of longer-term drought identified in the 24-month SPEI series. Pearson's correlations between grain yield and 3-month SPEI values for the months of March to August are given in Figure 8 for the four most wheat-producing regions. In order to examine the comparative sensitivities of different drought indices, yields were correlated with 3- and 6-month SPEI and SPI values for the main spring wheat producing regions of Akmola, Kostanay, North Kazakhstan and Pavlodar (Figure 8).

Table 3

Wheat yield minima and corresponding 3-month SPEI values.

Year	Minimum yield (tons per hectare)	3-month SPEI value	Drought category based on SPEI
1955	0.20	-2.10	Extreme drought
1957	0.50	-1.54	Severe drought
1963	0.42	-0.88	Moderate drought
1965	0.49	-1.93	Severe drought
1975	0.57	-1.57	Severe drought
1984	0.63	-1.58	Severe drought
1991	0.40	-2.54	Extreme drought
1995	0.52	-1.89	Severe drought
1998	0.60	-2.12	Extreme drought
2010	0.70	-1.65	Severe drought
2012	0.64	-1.37	Moderate drought
2014	0.18	-2.05	Extreme drought

Figure 8

Correlation between wheat yields, 3- and 6-month SPEI and SPI indices: (a) Akmola; (b) Kostanay; (c) North Kazakhstan; (d) Pavlodar regions.

All four indices (3- and 6-month SPEI and SPI) show significant correlation values with wheat yields. However, different patterns are apparent: in Akmola, the June–July 6-month SPEI had a higher correlation than the 3-month SPEI; whereas for Kostanay, the June SPIs were higher than the SPEIs, in North Kazakhstan the highest correlations were with the July 3-month SPEI value and in Pavlodar, the July 6-month SPEI had a higher correlation than the 3-month SPEI. Whichever high occurrence index is dominant, the highest correlations are for June and July.

The subdivision of the records of yield and 3-month SPEI for the four regions into two distinct blocks of 35 years (i.e. 1950-1985 and 1985–2020) again show different degrees of correlation (Figure 9). The hypothesis that the variance explained by meteorological factors might decrease with time can be supported in North Kazakhstan, the data for Akmola, Kostanay, and Pavlodar show an increased correlation with July 3-month SPEI while all three series show highly significant correlations for August SPEI values in the period 1985–2020 compared with the 35 years 1950–1985. Given that the correlations between yield and SPEI and SPI indices for June or July were significant at the 99% level, this may indicate the potential of using of SPEI and SPI indices for June or July values as predictors of annual wheat yields.

Figure 9

Correlations between wheat yields and 3–month SPEI index: (a) Akmola; (b) Kostanay; (c) North Kazakhstan; (d) Pavlodar regions.

Discussion

Impact of drought

The longer-term patterns of water balance revealed SPEI analyses demonstrate that the 1950s was probably not the decade to choose for converting steppe grassland into arable land. Jackson, W. D. (1956) commented that dry farming in the chestnut-brown soil (of Kazakhstan) involves considerable risk. In view of past Soviet and Tsarist experience, the prospects for successful long-range cultivation do not appear good. Contemporary reports in the UK Manchester Guardian mention both low yield due to drought in 1955 (25th January 1956), and loss of grain due to lack of availability of harvesting equipment and labour in 1956 (3rd September 1956). Despite these setbacks, and the low yields compared to other gain producing areas in the former Soviet Union, Kazakhstan's profile as a producer and exporter of high-quality wheat has continued to rise.

In the absence of phenological and meteorological data at the level of individual farms or local areas, such as that used by Qian et al. (2009) for spring wheat in the Canadian prairies, relationships have been demonstrated between yield data and 3- month and 6-month SPEI and SPI data at a regional scale for the months of June, July and, in some cases, August. In the context of Kazakhstan, the comparative performances of the 3-month and 6-month SPI and SPEI values as predictors of wheat yields are interesting in that factoring in evapotranspiration via the SPEI algorithm does not necessarily improve performance compared to the SPI. This result appears to indicate the comparatively dominant role of precipitation in determining spring-sown wheat yields in Kazakhstan. In comparison, in their study of agricultural drought impacts on spring wheat yield and quality, Mkhabela et al. (2010) observe that water demand and water balance indices provide the most accurate assessment on both yield and quality. Vicente-Serrano et al. (2006) successfully used 1- and 3-month SPI to predict autumn-sown wheat and barley production in eastern Spain, but according to Mavromatis (2007), the SPI index performed poorly as a predictor of yield for yields of durum wheat in Greece compared with three variations of the Palmer drought severity index. Yield and quality are two different issues however and the quality of Kazakhstan wheat improves under conditions of moisture stress with, for example, 90% of wheat graded in classes 1–3 (milling quality) (FAO, 2021a).

Given the poor quality and management of much of the agricultural land-use in Kazakhstan, the area is potentially sensitive to future changes in precipitation and temperature and also the fact that farmers might shift land, other inputs, and yield-improving investments to crops with less climate change impact and volatile prices (Haile et al., 2016; Kim and Moschini, 2018). The correlations presented in this paper indicate that yield is sensitive to precipitation during, and immediately prior to, the growing season, with highest correlations in June and July. Ensemble projection for Kazakhstan, increase of temperature will be ranging from 1.8°С to 2.0°С by 2030 and will increase by 5.6°С by 2085 and there may be some indication of a reduction in summer precipitation (Kazhydromet, 2020). Under the projected climate conditions in 2030, spring wheat yield in Akmola, Kostanay, North Kazakhstan, Pavlodar regions is expected to be 63–82% of the current level, and in conditions projected for 2085–31-51% (Kazhydromet, 2020). Any increase in aridity is however likely to lead to a reduction in the areas in which rain-fed production of cereals can continue to be viable. Although the present study has focused simply on yield data, the nature of the wheat cultivars involved is likely to be very important. In a recent study of winter-sown wheat in Europe, Porter and Semenov (2005) stress that crop responses to climate forcing may be non-linear and that changes in both mean values and variability about mean values need to be taken into account, and Semenov and Shewry (2011) note that while new wheat cultivars may be more drought resistant, on account of ripening earlier in the summer, heat stress may affect crops at the point of flowering and therefore may be of greater significance in decreasing yields than lower precipitation during the growing season.

Impact of heat waves

Figure 7 demonstrates that the longer SPEI values have oscillatory trends. Positive 48-months SPEI in the mid-1990s, 2003 and 2010 correspond to periods with heat waves in Europe, Russia, and south China, which resulted in widespread droughts and wheat losses and these extreme weather conditions could affect wheat yields in Kazakhstan due to teleconnection factor over the region (Zampieri et al., 2009; Barcikowska et al., 2020). In 2003, Europe experienced an exceptionally warm and dry summer, with summer temperature anomalies, with respect to 1961–1990 mean, greater than 3 °C in a very large area (UNEP, 2004; Schär et al., 2004; García-Herrera et al., 2010). That extreme summer began with a heat wave in June, continued through July and was characterized by another heat wave in early August. In 2010, Russia experienced the hottest summer in the last 500 years. Temperatures soared above 40 °C in western and southern Russia from July through mid-August 2010 (Liu et al., 2020). In Moscow, where long-term daily average temperatures for July range from 16 °C to 18 °C, daily average July 2010 temperatures climbed to 24 °C (Revich et al., 2015). The maximum temperatures higher than 38–40 °C were documented for a large area of the south of China (Xu et al., 2003; Ding et al., 2010). Some studies confirm the impact of the 2003 heat wave in Europe, Russia, and China and highlight a high percentage of concurrence of heat waves and significant negative yield anomalies in wheat production areas (Ciais et al., 2005; Bastos et al., 2014; Hauser et al., 2016). Furthermore, longer data records show increasing trends in regional heatwaves (Perkins-Kirkpatrick and Lewis, 2020).

For Kazakhstan, the increase of intense heat waves (including those persisting for 3–5 days and longer) were observed in the whole territory of Kazakhstan during the period of 1950–2020. The highest frequencies of the absolute heat waves were about 9–10 days per decade in the Mangistau oblast, as well as in Kostanay and Pavlodar regions (Kazhydromet, 2020). In most regions of Kazakhstan, an increase of intense heat waves is observed during the summer period. The highest rate of intense heat waves (4–7 days per decade) in summer is observed in the West Kazakhstan, Atyrau, and Mangistau regions (Figure 10). In Kyzylorda and Turkestan regions, the trend of intense heat waves is 3–5 days per decade (Kazhydromet, 2020). Heat waves in Kazakhstan have the potential to significantly reduce wheat production.

Figure 10

(a) Annual station-based linear trends of heat waves, days per decade; (b) station-based linear trends of heat waves in warm seasons, days per decade.

Research limitation

It is fact that the most common practice to study the impact of droughts on agriculture involves the use of indices, which provides a clear, simple, and quantitative assessment of the major drought characteristics – intensity, duration, severity, and spatial extent. To date, many indices as drought assessment tools have been introduced and utilized (Li et al., 2015; Lee et al., 2017). They include Palmer Drought Severity Index, Standardized Precipitation Index, Standard Precipitation and Evapotranspiration Index, and Drought Severity Index, amongst others. A fundamental difference is that these indices are either rainfall-based or additionally factors in the soil-water balance. Other differences relate to the complexity of the index in terms of its computation and data requirements. Relatively simpler alternatives include the rainfall deficiency index, run analysis, and the percentile method. The Standard Precipitation Index and Standard Precipitation and Evapotranspiration Index are among the most commonly used drought assessment indices worldwide and although both these indices have more positive scientific reviews, there is no universally agreed method in which the impact of climatological events such as drought on agriculture can be evaluated – this presents both a limitation and possible avenues for further work. Further work to derive additional useful information could implement other approaches such as a yield response equation model as very often such model includes components as the climate, with its thermal regime, rainfall, evaporative demand, and carbon dioxide concentration; the crop, with its development, growth and yield processes; the soil, with its water (and salt) balance; and importantly it includes the management, with practices including irrigation and fertilization. The various Soviet and post-Soviet agricultural management approaches were not accounted for this research.

Conclusion

During the second half of the twentieth century Kazakhstan moved from being a minor supplier of wheat within the Soviet Union to a wheat producer of global importance. The most important wheat-growing regions lie in the northern part of Kazakhstan, are based on spring-sown rain-fed cultivation and are subject to drought. In addition to a clear relationship between yield minima and drought indices, wheat yields for the most producing regions (regions) were significantly correlated with 3-month June and July SPEI values over the period 1950–2020. Yield data for the most producing regions of Kazakhstan (Akmola, Kostanay, North Kazakhstan and Pavlodar) exhibit significant correlations with both 3-month and 6-month SPEI and SPI values, but with the most significant months remaining June and July. The level of agreement between the indices indicates that, contrary to expectations, the use of the SPEI does not significantly improve correlations with wheat yield and that, in some cases, the SPI performs marginally better. These results underline the potential susceptibility of Kazakhstan wheat yields to any future reductions in precipitation and increase in drought occurrence and intensity. Therefore, the adaptive strategy of agriculture to manage drought risks and climate change should be proactive, not merely reactive. The Government of Kazakhstan recognises the need for an interdisciplinary approach to curbing environmental degradation of the agricultural ecosystem driven by climatic and anthropogenic processes (such as excessive development of irrigation networks, poor land management, and inadequate drainage). The country adopted low carbon and sustainable transition concept, which aims by 2050 to reduce their dependence on fossil fuels from over 80% to around 50% in order for them to achieve significant negative carbon emissions. Agricultural practices need to be incorporated into mitigation policies and programmes. This will promote the measurement of carbon balances in agriculture and the search for synergies between mitigation and adaptation in the sector. This manuscript to interrogating climate change and its impact on agriculture in Kazakhstan will help inform the development of stakeholder policies and practice focused on future environmental sustainability, which engages both techno-scientific and community agendas.

Declarations

Author contribution statement

Marat Karatayev: Conceived and designed the experiments; Contributed reagents, materials, analysis tools or data; Analyzed and interpreted the data; Wrote the paper.

Michèle Clarke; Vitaliy Salnikov: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Wrote the paper.

Roza Bekseitova: Performed the experiments; Wrote the paper.

Marhaba Nizamova: Contributed reagents, materials, analysis tools or data; Wrote the paper.

Funding statement

This work was supported by British Council INSPIRE Programme «Climate change impacts on land degradation and society in Kazakhstan».

Data availability statement

Data included in article/supplementary material/referenced in article.

Declaration of interests statement

The authors declare no conflict of interest.

Additional information

No additional information is available for this paper.