INTRODUCTION
From 1906 to 2005, the earth's average surface temperature rose by 0·74 °C, with the increase in temperature from 1956 to 2005 contributing c. 0·65 °C, i.e., the rate of warming over nearly 50 years was almost twice that seen over the last 100 years. In the next 100 years the global average surface temperature may rise by 1·1–6·4 °C (IPCC Reference Solomon, Qin, Manning, Chen, Marquis, Averyt, Tignor and Miller2007).
In the past 100 years, China has experienced significant climate change. The average air temperature has increased by 0·5–0·8 °C over the last 100 years in China, which is slightly higher than the global average (Ding et al. Reference Ding, Ren, Shi, Gong, Zheng, Zhai, Zhang, Zhao, Wang, Wang, Luo, Chen, Gao and Dai2006). During 1951–2001, the average air temperature increased by 1·1 °C in China (Ding et al. Reference Ding, Ren, Zhao, Xu, Luo, Li and Zhang2007), and the increase in the average air temperature has been particularly significant since the 1980s (NARCC Committee 2007). The diurnal temperature range (DTR, the difference between the maximum and minimum temperature) shows a downward trend in most areas of China (Liu et al. Reference Liu, Xu, Henderson, Qi and Li2004; NARCC Committee 2007), which is because the increase in temperature is greater at night than in the daytime. Precipitation has increased by 10–15% per decade in western China, but decreased in most of northern China. A climatic model has projected that by 2050 the annual mean temperature might rise by 2·3–3·3 °C and precipitation might increase by 5–7% in China (Qin et al. Reference Qin, Ding, Su, Ren, Wang, Wu, Rong, Yang, Wang, Liu, Dong, Lu, Huang, Du and Luo2006).
A number of studies show that climate change has affected crop yields (Chmielewski & Potts Reference Chmielewski and Potts1995; Nicholls Reference Nicholls1997; Peng et al. Reference Peng, Huang, Sheehy, Laza, Visperas, Zhong, Centeno, Khush and Cassman2004; Tao et al. Reference Tao, Yokozawa, Xu, Hayashi and Zhang2006, Reference Tao, Yokozawa, Liu and Zhang2008a; Schlenker & Roberts Reference Schlenker and Roberts2009; Liu et al. Reference Liu, Wang, Yang and Wang2010; Zhang & Huang Reference Zhang and Huang2012) and will influence global crop production now and in the near future (Parry et al. Reference Parry, Rosenzweig, Iglesias, Livermore and Fischer2004; Schmidhuber & Tubiello Reference Schmidhuber and Tubiello2007; Lobell Reference Lobell2007; Lobell et al. Reference Lobell, Burke, Tebaldi, Mastrandrea, Falcon and Naylor2008). Currently, many studies focus on the effects of increasing temperature on yield. However, the impacts of climate change on crop yields from those studies are very different. Peng et al. (Reference Peng, Huang, Sheehy, Laza, Visperas, Zhong, Centeno, Khush and Cassman2004) reported that increasing the minimum temperature would result in a decrease of rice yield, but Sheehy et al. (Reference Sheehy, Mitchell and Ferrer2006) re-analysed the experimental data of Peng et al. (Reference Peng, Huang, Sheehy, Laza, Visperas, Zhong, Centeno, Khush and Cassman2004) and found that the decrease in rice yield might be a combined effect of increasing minimum temperature and decreasing solar radiation. Therefore, changes in crop yield are the result of interacting climatic factors (Lobell & Oritiz-Monasterio Reference Lobell and Ortiz-Monasterio2007; Zhang et al. Reference Zhang, Zhu and Wassmann2010). Moreover, climate-yield relationships are scale-dependent: large-scale statistical data and regional climate datasets are important for investigating general response patterns of crop production to climate change and variability (Lobell Reference Lobell2007; Lobell et al. Reference Lobell, Burke, Tebaldi, Mastrandrea, Falcon and Naylor2008).
Maize and wheat are the major cereal crops cultivated in China and offer staple foods for the Chinese people. Research on how climate change will influence the yields of maize and wheat in China will be helpful for understanding the potential effects of future climate change and for developing scientific countermeasures. The aim of the present study was to systematically investigate climate-yield relationships and the impacts of climate trends over the last 50 years on maize and winter wheat yields at provincial levels throughout China in a spatially explicit way.
MATERIALS AND METHODS
Study region and data sources
Maize and winter wheat are grown throughout China: the regions studied in the current paper are shown in Fig. 1. Maize was common in most of the area and the majority was cultivated in the summer. Winter wheat occupied a high portion of the area in the north and northwest provinces.
Fig. 1. Distribution of maize and winter wheat cultivation area in China. The provincial ID is numbered as: 0, non-study region; 1, Heilongjiang; 2, Jilin; 3, Liaoning; 4, Beijing; 5, Hebei; 6, Tianjin; 7, Shandong; 8, Henan; 9, Shanxi; 10, Shaanxi; 11, Ningxia; 12, Gansu; 13, Neimenggu; 14, Xinjiang; 15,Anhui; 16, Jiangsu; 17, Zhejiang; 18, Shanghai; 19, Hubei; 20, Hunan; 21, Jiangxi; 22,Chongqing; 23, Sichuan; 24, Guizhou; 25, Yunnan; 26, Guangxi; 27, Guangdong; 28, Fujian.
Crop data were obtained from the Data Sharing Infrastructure of Earth System Science (Chinese Academy of Sciences 2011) and China's agricultural statistics (MOAPRC 2009), including provincial yields of maize and winter wheat. Additionally, the growing seasons of maize and winter wheat were derived from the Chinese Agricultural Phenology Atlas (Zhang Reference Zhang1987) (Table 1). The climate data were obtained from the National Meteorological Information Centre, China Meteorological Administration, which included data from 536 weather stations distributed over the study region. Climatic variables included daily average temperature (T avg), DTR and precipitation (Prcp). The average of each climate variable during the crop growing season was computed for each province.
Table 1. Growing seasons of maize and winter wheat for different regions of China
* ID corresponds to the numbers shown in Fig. 1.
To avoid the confounding influence of long-term variations such as changes in variety and crop management, and to focus on the relationships between climatic factors and yield, the climate variables and crop yield were converted to first-difference values by subtracting the prior year's value from each year (Nicholls Reference Nicholls1997; Lobell et al. Reference Lobell, Burke, Tebaldi, Mastrandrea, Falcon and Naylor2008).
Winter wheat is grown throughout China, but spring wheat is only grown in Northeast China. Climate-yield relationships were not studied for spring wheat.
Linear time trend
The linear time trend represents the change for each variable (Wei Reference Wei2007). A simple linear regression equation between X i and t i can be expressed as follows:
$$X_{\rm i} = a + bt_{\rm i} (i = 1,\,2, \ldots, n)$$where X i is the variable (e.g. climatic variables), t i is time corresponding to X i, n is the number of samples and a and b are regression coefficients. The trend rate of variables is represented by 10 × b.
Effects of climate variables on yield
To avoid the confounding influence of long-term variations such as changes in crop management, and to focus on the relationships between climatic factors and yield, the climate variables (T avg, DTR and Prcp) and crop yield (Yield) were converted to first-difference values by subtracting the prior year's value from each year (Nicholls Reference Nicholls1997; Lobell et al. Reference Lobell, Burke, Tebaldi, Mastrandrea, Falcon and Naylor2008). A multiple linear regression model could be used to compute the climate-yield relationship with the first-difference values of the climate variables (ΔT avg, ΔDTR and ΔPrcp) as explanatory variables and yield (ΔYield) as the response variable:
$$\eqalign{\Delta Yield &= \beta _{\rm o} \!+ \beta _{T_{{\rm avg}}} \Delta T_{{\rm avg}} \!+ \beta _{{\rm DTR}} \Delta DTR \cr & \quad+ \beta _{{\rm Prep}} \Delta Prep + \varepsilon} $$where ΔYield, ΔT avg, ΔDTR and ΔPrcp represent the first difference values of crop yield and growing season climate variables; β o represents the model intercept; β Tavg, β DTR and β Prcp represent the regression coefficients for climate variables; ε represents the model error.
To evaluate the percentage yield change for each additional climate variable, the percent regression coefficients were calculated using Eqn (3):
$$\beta _{{\rm percent}} = \displaystyle{\beta \over {Mean_{{\rm Yield}}}} $$where β and β percent represent absolute and percent regression coefficients of certain climate variables (ΔT avg, ΔDTR and ΔPrcp), respectively. Mean Yield represents the mean crop yield in 1961–2010.
RESULTS
Climate change and variability during the maize and winter wheat growing seasons
Maize
Table 2 illustrates the average and linear time trend in climate during the maize growing season for the period 1961–2010. The average (T avg) for the maize growing season was >17 °C in China over this period with the highest values in the areas of central and east China. There were general warming trends in most regions, and the warming was more significant in northern parts of China than in southern parts during the maize growing season from 1961 to 2010. Spatial distribution characteristics of the average DTR showed as a band, gradually increasing from the southeast to the northwest. There were general decreasing trends for DTR in most regions with a rate of −0·39 to 0 °C per decade during this period. For Prcp, spatial distribution characteristics of the average Prcp was also revealed as a band, gradually increasing from the northwest to the southeast. The Prcp showed an increasing trend in the south, central, east and Xinjiang regions and a decreasing trend in other provinces over the period.
Table 2. Average and linear time trend in climate during the maize growing season over the period 1961–2010
* T avg represents average temperature.
† DTR represents diurnal temperature range.
‡ Prcp represents precipitation.
Winter wheat
Table 3 illustrates the average and linear time trends in climate during the winter wheat growing season over the period 1961–2010. The average T avg for the winter wheat growing season was >3 °C in China during this period with the highest values occurring in the areas of south China. There were general warming trends in the entire study region with a rate of 0 to 0·53 °C per decade in the winter wheat growing season from 1961 to 2010. The spatial distribution characteristics for average DTR showed as a band, gradually increasing from the southeast to the northwest. There were general decreasing trends for DTR in most regions with a rate of −0·47 to 0 °C per decade during this period. The spatial distribution characteristics of the average Prcp also showed as a band, this time gradually increasing from north to south. The Prcp showed an increasing trend in the north, east, south and Xinjiang regions of China and a decreasing trend in other provinces during the study period.
Table 3. Average and linear time trend in climate during winter wheat growing season over the period 1961–2010
* T avg represents average temperature.
† DTR represents diurnal temperature range.
‡ Prcp represents precipitation.
Effects of climate change on maize and winter wheat yields
Maize
Figure 2 shows the effects of climate variables on maize yield in China from 1961 to 2010. A positive value of the percent regression coefficient indicates a coincident pattern between the climate variables and yield, and a negative value reflects the opposite response of yield to additional climate variables.
Fig. 2. Percent regression coefficients of (a) T avg, (b) DTR and (c) and Prcp on maize yield over the period 1961–2010.
The T avg showed a negative effect on yields in most regions of China: each additional 1 °C in T avg decreased maize yields by 0–17%. However, the provinces in the northeast and northwest regions showed a positive effect ranging between 0 and 7·5%/°C (Fig. 2(a)). The regions with increasing DTR included a number of provinces in the northeast, north, east and southwest regions with 0–22·4% yield loss for each additional degree celsius of DTR, while several provinces in the northwest and south showed a positive effect with 0–7% yield increases per 1 °C increase of DTR (Fig. 2(b)). As for Prcp, yield increases were associated with a higher Prcp in parts of northwest, north, Sichuan, Yunnan and Zhejiang, with 0–6·5% increases seen for Prcp increases of 100 mm, while the effects were negative in the south, central and northeast , varying between 0 and 7·3% yield loss per 100 mm (Fig. 2(c)).
The regression coefficients calculated with actual trends of T avg, DTR and Prcp at the provincial level were used to evaluate the relative production change caused by climate variables over the period 1961–2010.
The warming caused a 0–13% reduction in maize production in most regions from 1961 to 2010. However, increased production of c. 0–11·6% was seen in the northeast, northwest and Guangxi regions (Fig. 3(a)). For DTR, the decreasing trends in most regions resulted in decreased production (c. −8·2 to 0%) in the west and central regions, while several provinces in the northeast, north, east and southwest showed increased production with values of c. 0–21% (Fig. 3(b)). For Prcp, the changes increased production by 0–4·5% in the northeast, Xinjiang, Zhejiang and Guizhou regions, while the changes in other provinces decreased production by −5·3 to 0% (Fig. 3(c)).
Fig. 3. Relative production change caused by (a) T avg, (b) DTR and (c) Prcp for maize over the period 1961–2010.
Table 4 illustrates the average percent regression coefficients of climate variables and production change caused by climate variables for maize in the whole study region from 1961 to 2010. The results showed that a negative T avg effect of −3·8%/°C (95% CI of −7·4 to −0·2%/°C), and a yield decrease of 0·8% for per degree increase of DTR with 95% CI of −7·3 to 5·6%/°C. On average, the Prcp effect reached 5·5% per additional 100 mm (1·0–10·0% per 100 mm with 95% CI).
Table 4. Average percent regression coefficients of climate variables and 95% confidence interval(CI), average production change and 95% CI caused by climate variables for maize over 1961–2010
The regression coefficients calculated with actual trends of T avg, DTR and Prcp at the national scale were used to evaluate the relative production change and actual production change caused by climate variables in the whole study region over the period 1961–2010. For maize, T avg caused decreased production with values of c. −3·4% (95% CI of −6·5 to −0·2%), and the actual production change was c. −114·0 kg/hm2 (95% CI of −220·7 to −7·2 kg/hm2). The production was increased by c. 0·6% (95% CI of −4·0 to 5·2%) as the result of changes in DTR, and the actual production change reached c. 20·3 kg/hm2 (95% CI of −135·8 to 176·4 kg/hm2). However, the effect of Prcp on maize production was not significant at the national scale (Table 4).
Winter wheat
Figure 4 shows the percent regression coefficients of T avg, DTR and Prcp on the yield of winter wheat in China from 1961 to 2010. The results showed a negative effect of increasing temperature on yield in the north and southwest of China: each additional 1 °C of T avg decreased winter wheat yield by 0–7·6%. However, the provinces in the west and central regions showed a positive effect ranging between 0 and 7·7%/°C (Fig. 4(a)). The changes in DTR led to higher yield in most regions except Xinjiang and Jiangxi, varying between −3·6 and 10·0% yield increase per degree (Fig. 4(b)). For Prcp, yield increases were associated with a higher Prcp in most areas, with 0–13·6% increases for a 100 mm increase in Prcp (Fig. 4(c)).
Fig. 4. Percent regression coefficients of (a) T avg, (b) DTR and (c) Prcp on the winter wheat yield over the period 1961–2010.
The increase in T avg decreased winter wheat production by −10·5 to 0% in the Huang-Huai-Hai and southwest regions from 1961 to 2010. However, the increase in T avg in the northwest and central regions caused an increased production with values of c. 0–13·6% (Fig. 5(a)). For DTR, the general decreasing trend resulted in increased production with values of c. 0–2·7% in the northwest and south regions, while other provinces of the study region showed decreasing production with values of c. −14·2 to 0% (Fig. 5(b)). For Prcp, the changes resulted in increased production by 0–1·6% in the north, Sichuan, Guizhou, Shanghai and Jiangsu, while changes in other provinces decreased production by −5·7 to 0% (Fig. 5(c)).
Fig. 5. Relative production change caused by (a) T avg, (b) DTR and (c) Prcp for winter wheat over the period 1961–2010.
Table 5 illustrates the average percent regression coefficients of climate variables and production change caused by climate variables for winter wheat in the study region from 1961 to 2010. The results showed a negative T avg effect of −2·5%/°C (95% CI of −5·8 to 0·7%/°C), but a yield increase of 3·9% for per degree increase of DTR with 95% CI of −1·4 to 9·2%/°C. On average, the Prcp effect reached c. 7·5% with per additional 100 mm (3·4–11·6% per 100 mm with 95% CI).
Table 5. Average percent regression coefficients of climate variables and 95% confidence interval (CI), average production change and 95% CI caused by climate variables for winter wheat over 1961–2010
For winter wheat, T avg caused a decrease in production with values of c. −5·8% (95% CI of −13·1 to 1·6%), and the actual production change reached c. −138·5 kg/hm2 (95% CI of −316·5 to 39·5 kg/hm2). Production was reduced by c. −2·9% (95% CI of −7·0 to 1·1%) because of changes in DTR, and the actual production change reached c. −70·8 kg/hm2 (95% CI of −167·2 to 25·6 kg/hm2). Production improved by c. 1·6% (95% CI of 0·7–2·4%) due to the changes of Prcp, and the actual production change was c. 37·3 kg/hm2 (95% CI of 16·8–57·8 kg/hm2) (Table 5).
DISCUSSION
The present study shows the general response patterns of agricultural productivity to climate change; however, the key response mechanisms still need to be researched further. The physiological mechanisms by which climate variables may affect crop yields have been partly revealed (Matsui & Horie Reference Matsui and Horie1992; Tao et al. Reference Tao, Yokozawa, Xu, Hayashi and Zhang2006, Reference Tao, Hayashi, Zhang, Sakamoto and Yokozawa2008b). Global warming may have harmful or beneficial impacts on crop production in low or high latitude areas, depending on whether the current temperature is generally more or less than the optimal temperature for crop production (IPCC 2001). This mechanism underlies crop response patterns to temperature change in the current paper. Taking maize as an example, the present study showed that growing season average temperature was positively related to maize yield in some provinces of the northeast and northwest regions. This suggests that the present growing season average temperature might be in the optimal temperature range for maize production and maize yield could be increased due to a warming trend, which is consistent with some recent research (Yang et al. Reference Yang, Peng, Laza, Visperas and Dionisio-Sese2008; Liu et al. Reference Liu, Wang, Yang and Wang2010). In contrast, maize yield in most of the study regions showed a negative relationship with growing season average temperature, suggesting that the present average temperature is at or above the optimum for maize production and that yield could be reduced due to a warming trend, which is consistent with other studies (Peng et al. Reference Peng, Huang, Sheehy, Laza, Visperas, Zhong, Centeno, Khush and Cassman2004; Tao et al. Reference Tao, Yokozawa, Xu, Hayashi and Zhang2006; Sheehy et al. Reference Sheehy, Mitchell and Ferrer2006).
Evaluation of the effects of climate change on maize and winter wheat yields should consider the effects of all climatic factors (Lobell & Oritiz-Monasterio Reference Lobell and Ortiz-Monasterio2007; Zhang et al. Reference Zhang, Zhu and Wassmann2010). Firstly, the impact of DTR was bilateral. An increase in DTR during the growing season, which has warming days and cooling nights, is beneficial for crop growth because warming days will increase photosynthetic rate and cooling nights will reduce the respiration rate (Leopold & Kriedemann Reference Leopold and Kriedemann1975). However, an increase in DTR may also reduce yield because the associated increase in maximum temperature results in reductions in net photosynthetic rates or increasing water stress (Dhakhwa & Campbell Reference Dhakhwa and Campbell1998; Tao et al. Reference Tao, Yokozawa, Xu, Hayashi and Zhang2006). In the present study, the decreases in DTR during the growing season increased maize production on a national level by causing a decrease in cold damage, but decreases in DTR during the growing season reduced winter wheat production at a national level by causing an increase in night-time maintenance respiration rates (Ryan Reference Ryan1991) and consequently biomass consumption. Secondly, precipitation during the maize growing season had a negative relationship with yield in most areas of southern China, but a positive relationship with yield in most regions of northern China. These results were consistent with the ground truth in the past few decades. In southern China, the climate is humid and too much precipitation often occurs during the maize growing season leading to yield losses from insects, disease and insufficient sunshine (Tao & Yokozawa Reference Tao and Yokozawa2005). In northern China, the climate is arid and semi-arid; water stress is the most important limiting factor for rainfed crop yields (Tao et al. Reference Tao, Yokozawa, Hayashi and Lin2003).
Uncertainties of the results as mentioned above may come from several sections. Firstly, the data in the present study are from the provincial scale over 50 years. However, research using data collected on different spatial or temporal scales may give different conclusions (Tao et al. Reference Tao, Yokozawa, Xu, Hayashi and Zhang2006; Liu et al. Reference Liu, Wang, Zhu and Tang2012). Therefore, research on the response of crop yield to changes in climatic factors should be extended to these different spatial and temporal scales. Secondly, it is very difficult to fully consider the effects of technology and management, as well as weather variability, disease, insects, etc., occurring in the five decades of observations. To some extent, the effects could disturb the results of trend analysis. For instance, crop yields may have increased because of the increasing use of modern technology and newer higher yielding cultivars during the study period. If so, the changes in crop productivity resulting from climate change may actually be larger or smaller than those estimated in the present study. In addition, the effects of extreme climatic conditions (high and low temperatures, and floods) on crop yield are not considered in the present study. With the increasing frequency of extreme climate events and frequent agricultural meteorological disasters, research on the effects of extreme climate events on crop production should be emphasized in the future.
CONCLUSIONS
The present study revealed the evidence of climatic effects on maize and winter wheat over 1961–2010 and highlighted the changes in maize and winter wheat production because of the past warming trend using climatic and crop data in China at the provincial level.
Results indicated that warming trends were observed over 1961–2010 during maize and winter wheat growing seasons. The warming resulted in a decrease of maize yield in most of the study regions but an increasing yield in the northeast, northwest and Guangxi regions: the increases in T avg reduced maize production on average by c. 3·4% for the whole country. Moreover, the warming adversely affected winter wheat yield in the Huang-Huai-Hai and southwest regions and produced an overall loss in production at a national scale by c. 5·8%.
There were general decreasing trends for DTR in most regions over 1961–2010 in maize and winter wheat growing seasons. It was found that decreases in DTR during the growing season adversely affected maize yield in the west and central regions but a beneficial DTR effect was observed in the northwest, north, east and southwest regions: the changes in DTR resulted in an increased production on average by up to approximately 0·6% for maize in the whole study region. However, the general decreasing trends for DTR resulted in an increasing winter wheat yield in the northwest and south regions but a decreasing yield in other provinces. The production of winter wheat at a national scale was reduced by c. 2·9% because of changes in DTR.
For precipitation, there was an increasing trend in some provinces but a decreasing trend in other provinces during the maize and winter wheat growing seasons. Results indicated that the change in precipitation increased maize yield in the northeast, Xinjiang, Zhejiang and Guizhou regions but caused decreasing yield in other regions. There was no significant effect of precipitation on maize production at a national scale over the period 1961–2010. However, the change in precipitation increased winter wheat yield in the north, Sichuan, Guizhou, Shanghai and Jiangsu but caused decreased yield in other provinces, and the increasing average production reached c. 1·6% due to changes of precipitation in the whole study region.
This study was partially funded by the National Basic Research Program of China (2010CB951303), the Special Fund for Climate Change of CMA (CCSF201422) and the Southwest Regional Major Scientific and Operational Projects of CMA (2014-8).
