Estimating the impacts of warming trends on wheat and maize in China from 1980 to 2008 based on county level data


  • Tianyi Zhang,

    Corresponding author
    1. State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China
    • State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China.
    Search for more papers by this author
  • Yao Huang

    1. State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, Beijing 100093, China
    Search for more papers by this author


This study estimated the past warming impacts on wheat and maize from 1980 to 2008 based on data from 2339 counties in China, providing a finer portrait of climatic effects than earlier assessments that often focused on a provincial scale. We detected an overall negative effect of the average temperature (Tavg) and diurnal temperature range (DTR) on wheat and maize yields on a national level, but our results also indicate notable beneficial effects of the Tavg in the northcentral region for wheat and the eastern parts of northeast for maize that, in most previous studies, were overlooked due to the coarseness of provincial data. Because wheat cultivation is highly intense in the north, the positive Tavg effect in the region has offset losses from warming in other areas, improving national production by up to 1.6% relative to the average level over the study period. For maize, unfortunately, approximately 5.8% production was lost due to increases in the Tavg as the dominant adverse effect. Given the reduced DTR in the major producing regions, the past trends of the DTR have boosted wheat production by up to 2.1% and maize by up to 1.4%. As a result, the combined effects of the Tavg and DTR have increased wheat production by up to 3.7%, but decreased maize production by 4.4%, equivalent to approximately 10% of the actual change in Chinese wheat and maize production from 1980 to 2008. Copyright © 2012 Royal Meteorological Society

1. Introduction

China has been experiencing pronounced climate change over the last century. The average temperature (Tavg) has increased by 1.1° C from 1951 to 2001 (Ding et al., 2007), and the warming trend has been particularly significant since the 1980s (Edition broad of national assessment report for climate change in China, 2007). Wheat and maize are the major upland crops cultivated in the entire region of China and offer staple foods for the Chinese people. Thus, the yield response to future warming trends has been an important focus in China.

Besides consistent increases in the Tavg, another important feature of climate change is the reduction of the diurnal temperature range (DTR, defined by the difference between the maximum and minimum temperature) that has been widely observed in China (Edition broad of national assessment report for climate change in China, 2007; Liu et al., 2009). The trend indicates that the increase in temperature is greater at night than in the daytime. The uneven increase is of particular concern because the effects of warming during the day and night differ in terms of magnitudes on crop yields (Wheeler et al., 2000; Lobell et al., 2005; Lobell, 2007). For example, Lobell (2007) examined the empirical relationship between the DTR and yields in the major cereal crop-producing countries and found that warming during the daytime is more harmful than at night.

A large body of studies has investigated the effects of climate on crop yields and production by quantifying past climate/yield relationships using statistical models (Kucharik and Serbin, 2008; Lobell et al., 2008; Tao et al., 2008; You et al., 2009; Zhang and Huang, 2012). In the case of China, such assessments are typically executed using national or provincial level data that are easy to access. For example, Tao et al. (2008), based on provincial crop and climate observations from 1979 to 2002, estimated that warmer climate has reduced the production of wheat and maize by 1.2 × 105 and 21.2 × 105 ton decade−1, respectively. The study (Tao et al., 2008) and other literatures (Lobell, 2007; You et al., 2009; Lobell et al., 2011) appear to have reached the unanimous conclusion that Chinese wheat and maize are two cases that are significantly vulnerable to the past warming trend.

However, a recent study (Li et al., 2010) investigated the relationship between climate and wheat yields at different spatial scales of China, finding that some unclear climate/yield relationships at a large scale would emerge if they were observed at a smaller scale. This raised the spectre that provincial data are too coarse to capture significant climatic effects occurring at the sub-provincial scale because the sizes of provinces are typically large in China and thus bias assessments. Additionally, it is also noteworthy that recent studies based on site observations in China reported that actual crop yields are not always negatively associated with an increased Tavg (Liu et al., 2010), but sometimes respond positively to warming trends (Xiao et al., 2008; Wang et al., 2009). However, such observations are inherently inconclusive and dependent on the selection of sites (Tao et al., 2006), thus making them difficult to represent a regional situation.

Therefore, the goal of the current study is to provide a further quantification on the past warming impacts on wheat and maize from 1980 to 2008 based on yield and climate data from 2339 counties covering most of the Chinese territory. Such a practice would reconcile the aforementioned dilemma between coarseness and regional representativeness. In this study, the data on past variations in growing season climate and county yields are first used to calculate the strength of climatic effects. The quantified effects are then used to evaluate the consequences of wheat and maize production caused by the change in the individual Tavg, DTR and their combined effect from 1980 to 2008.

2. Materials and methods

2.1. Study region

Wheat and maize are produced in a wide range of locations and under a variety of climatic conditions in China. The average sowing areas for these crops during 1980 to 2008 were approximately 28 and 22 million hectares, respectively. Wheat and maize occupied a high portion of the area in the north, northeast and northwest provinces, and the cultivation was less in other regions. The study regions are highlighted in Figure 1. Winter wheat is common in most of the area, and spring wheat is grown in the northeast and northwest regions. The majority of maize was cultivated in the summer. From 1980 to 2008, both wheat and maize productions have experienced a substantial improvement by 38% and 95%, respectively.

Figure 1.

Sowing area of wheat and maize in China; NE, northeast; N, north; NW, northwest; E, east; C, central; SW, southwest; S, south; The study areas are highlighted. Each province is labelled with numbers corresponding to those in Table I

Table I. Growing season of wheat and maize in China
  • a

    ID corresponds to the numbers shown in Figure 1.

Northeast (NE)
North (N)
Inner Mongolia4AprJulMaySep
Northwest (NW)
East (E)
Central (C)
Southwest (SW)
South (S)

2.2. Data collection

This study was based on yield statistics at the county level, collected from the Chinese Academy of Agricultural Sciences. The dataset consisted of wheat and maize yield data from 2339 counties from 1980 to 2008. Additionally, the crop-growing seasons for wheat and maize were derived from the Chinese Agricultural Phenology Atlas (Zhang, 1987) (Table I). Climate data from 1980 to 2008 were obtained from the China Meteorological Administration, which included a total of 566 weather stations distributed over the study region. The climate data consisted of the Tavg, DTR and precipitation (Prcp). These climate data were firstly interpolated to the whole study region based on the algorithm previously described by Thornton et al. (1997), and they were then averaged over the growing season to derive the growing season average Tavg and DTR in addition to the growing season total Prcp in each county/crop combination for each year.

2.3. Statistical analysis

2.3.1. Identifying climate effects on yield

The grain yields (Yield) and climate variables in a time series (Tavg, DTR and Prcp) were converted to first-difference values by subtracting the value from the prior year for each year. Calculating the first-difference value is a common and necessary de-trending technique to establish climate/yield relationships (Lobell et al., 2008).

When constructing the statistical model, Equation (1) was used to describe the climate/yield relationship with the first-difference values of yield (ΔYield) as the response variables and the first-difference values of climate variables (ΔTavg, ΔDTR and ΔPrcp) as explanatory variables.

equation image(1)

where ΔYield, ΔTavg, ΔDTR and ΔPrcp are the first-difference values of crop yield, growing season average Tavg, growing season average DTR and growing season total Prcp, respectively; β0 represents the model intercept; βmath image, βDTR and βPrcp are the regression coefficients for climate variables; and ε is the model error.

Radiation is not taken into account in our model to facilitate the comparison with previous estimates that also excluded radiation (Lobell, 2007; Tao et al., 2008). Another reason for the omission is that significant radiation effect is often observed for rice (Dobermann et al., 2000; Zhang et al., 2010) since plenty of irrigation applied that meets crop water demands and reduces drought risk (Zhang et al., 2008) while Chinese wheat and maize are conventionally cultivated under rainfed and upland conditions, where Tavg and Prcp effects outstrip radiation (You et al., 2009). Moreover, to determine the uncertainties due to a finite sample size, we use a bootstrap re-sampling approach in which we randomly choose years with replacement. A new regression model was then computed, and this process was repeated 1000 times in each time series for calculating a 95% confidence interval (95% CI).

2.3.2. Statistical analysis approaches

We executed multiple linear regression (MLR) and principal component regression (PCR) analyses to each county time series using the R version 2.12.2 software (R Development Core Team, 2011). Comparison of the results derived from the two methods serves to detect any bias due to possible collinearity between the explanatory variables, which has been thought to potentially affect MLR model establishment (Lobell, 2010).

MLR is based on the method of ordinary least squares, which estimates the regression coefficient by minimizing the sum of squared error between the prediction and observation. MLR was conducted using the lm() function in the R software by inputting Equation (1), which calculated the regression coefficients of climate variables for the regression model.

PCR is an alternative regression method for untangling collinearity, which reduces the dimensionality of a dataset that consists of correlated variables. First, PCR identifies patterns between explanatory variables by calculating the eigenvector loadings and eigenvalues using the covariance matrix of explanatory variables. The step was performed using the princomp() function in the R software. Second, PCR selects components, forming orthogonal feature vectors (referred to as principal components in PCR) that successively account for the greatest variation in the explanatory variables. We retained the components to explain up to at least 90% of the variation. Finally, a multiple regression analysis was run on the retained orthogonal components to calculate the regression coefficients of each component, which were then translated to original explanatory variables by Equation (2).

equation image(2)

where β is the regression coefficient of the original explanatory variable (ΔTavg, ΔDTR or ΔPrcp);β* is the regression coefficient of the principal component; a is the eigenvector loadings of principal components; SD is the standard deviation of the original explanatory variable; n is the number of principal components retained and i denotes the principal components from 1 to n.

The percent regression coefficients were calculated to show the percentage yield change for each additional climate variable derived by the MLR and PCR methods. Equation (3) was shown to calculate the percent regression coefficient.

equation image(3)

where βpercent and β are the percent and absolute regression coefficients of certain climate variables (ΔTavg, ΔDTR or ΔPrcp), respectively; MeanYield is the mean yield during the period of 1980 to 2008.

3. Results

3.1. Climate change during the wheat and maize growing seasons

Figure 2 illustrates the time trends of the Tavg, DTR and Prcp during the growing seasons of wheat and maize from 1980 to 2008. Only the trends with statistical significance are shown here. There were general warming trends for both wheat (Figure 2(a)) and maize (Figure 2(d)) growing seasons in most of the study region. In contrast to the widespread upward trends for Tavg, the DTR exhibited a clear geographical variability (Figure 2(b) and (e)). During the wheat growing season, there is a general narrowing trend for the changes of minimum and maximum temperatures, resulting in a decrease in the DTR with a rate of − 0.6 to 0.0° C per decade, in the northern and western regions. However, an increasing trend was observed for the DTR in the central and southern area with a rate of 0.0–0.6° C per decade. The maize growing season showed a broadly similar feature with the exception of some portions of the eastern Inner Mongolia Province (No. 4 in Figure 1), where the DTR has increased. As to the Prcp, there was a significant reduction in Prcp in the northeast and southwest regions in the maize growing season (Figure 2(f)), and the Prcp time trends for the wheat and maize growing seasons were insignificant in the major areas (Figure 2(c) and (f)).

Figure 2.

Linear time trend of the Tavg, DTR and Prcp during wheat (a, b, c) and maize (d, e, f) growing seasons. Only trends with statistical significance are shown (p < 0.05)

3.2. Effects of climate variables on wheat and maize yields

Using the bootstrap method, the uncertainties for each regression coefficient due to the finite sample size were estimated, and they are quite small in most counties (Figures S1–S4, Supporting Information). Thus, for brevity, only the average values of the regression coefficients are presented in this part.

3.2.1. Wheat

The estimated effects of the climate variables on wheat yields as derived by MLR and PCR are shown in Figure 3. A negative value indicates an inverse response of yield to additional climate variables, and a positive value reflects a coincident pattern between the yield and climate variables.

Figure 3.

Percent regression coefficients of the Tavg, DTR and Prcp on the yield of wheat derived by MLR (a, b, c) and PCR (d, e, f). Only counties passing the significance test are shown (p < 0.05). The scale for Prcp has been reversed so that positive Prcp effect is represented by red, a conversional colour for drought

According to the results calculated by the two methods, there were several counties in the far north and northwest regions where the Tavg posed a strong negative effect on yields. Each additional degree of Tavg reduced yields by 0–30% (Figure 3(a) and (d)). However, the counties in the northcentral region, for example, some counties in the Hebei (No. 6 in Figure 1), Shandong (No. 8 in Figure 1) and Henan (No. 9 in Figure 1) Provinces, saw a notable positive effect ranging between 0 and 20%° C−1 (Figure 3(a) and (d)). The vulnerable regions for DTR increases include a number of counties in the northeast, north, northwest and southwest regions with 0–30% yield losses for each additional degree of DTR, while several counties in the east showed a positive effect with 0–20% yield increases per degree increase of DTR (Figure 3(b) and (e)). For Prcp, the yields were positively associated with a higher Prcp in the north, northwest and southwest, while the effects turned negative in the northeast and east (Figure 3(c) and (f)). Therefore, the calculated regression coefficients in spatial distribution are broadly in agreement between the two methods. The only place that the two methods disagree is in the northeast where MLR calculated a positive effect of the Tavg, but PCR suggested a negative one (Figure 3(a) and (d)).

3.2.2. Maize

Figure 4 illustrates the percent regression coefficients of Tavg, DTR and Prcp for the maize yields as calculated by the MLR and PCR methods. The two methods show that the north, northwest, southern parts of northeast and parts of the southwest were vulnerable to a higher Tavg with yield reductions of 0–30% per one-degree temperature increase (Figure 4(a) and (d)). Yield increases were associated with a higher Tavg in the eastern parts of northeast, with approximately 10–30% increases for each additional degree (Figure 4(a) and (d)). The DTR increases led to lower yields in large number of counties in the north, northwest and southwest, varying between 0 and 20% yield losses per degree, but an inverse effect was observed in the northeast, where yields were improved by approximately 10–30% per each additional degree of DTR (Figure 4(b) and (e)). For the Prcp, the PCR analysis produced a homogeneous positive regression coefficient in the major area where yields were improved by 0–30% for Prcp increases of 100 mm (Figure 4(f)), while some counties showed a negative effect calculated by MLR.

Figure 4.

Percent regression coefficients of the Tavg, DTR and Prcp on the yield of maize derived by MLR (a, b, c) and PCR (d, e, f). Only counties passing the significance test are shown (P < 0.05). The scale for Prcp has been reversed so that positive Prcp effect is represented by red, a conversional colour for drought

3.2.3. Wheat versus maize

Table II shows the calculations of the mean values and 95% CI of the regression coefficients for wheat and maize as a reflection for entire China and compares their quantitative differences based on t-test.

Table II. Mean percent regression coefficients of the climate variables, 95% confidence interval and the t-test results between wheat and maize
CropsClimate variableWheatMaizet-test
  Mean95% CIMean95% CItp-value
MLRΔTavg (%/° C)− 2.1− 3.3 to − 0.9− 8.8− 10.5 to − 7.26.440.00
 ΔDTR(%/° C)− 6.6− 8.4 to − 4.8− 4.6− 6.9 to − 2.3− 1.390.08
 ΔPrcp(%/100 mm)0.4− 3.1 to to 4.5− 1.520.06
PCRΔTavg (%/° C)− 2.9− 4.0 to − 1.8− 8.8− 10.1 to − 7.36.480.00
 ΔDTR(%/° C)− 4.1− 5.5 to − 2.8− 4.6− 5.6 to − 3.70.620.27
 ΔPrcp(%/100 mm)2.40.0 to to 4.1− 0.770.22

With regard to wheat, the MLR approach inferred a negative Tavg effect of − 2.1%° C−1 (95% CI of − 3.3 to − 0.9%° C−1), and the PCR method estimated a quantitatively similar yield reduction of 2.9% for each degree increase of the Tavg with 95% CI of − 4.0 to − 1.8%° C−1. Both methods estimated a negative yield response for a higher DTR, but the yield losses were approximately 1.5 times greater in the MLR analysis (−6.6%° C−1) than the reductions in PCR (−4.1%° C−1). On average, the Prcp effect was 0.4% with each additional 100 mm (−3.1 to 3.8% per 100 mm with 95% CI) based on the MLR analysis, and the PCR analysis estimated 2.4% per 100 mm increase of Prcp (95% CI of 0.0–4.7% per 100 mm). For maize, the regression coefficients calculated by MLR and PCR are almost identical. Yields were decreased by 8.8 and 4.6% for each additional degree of the Tavg and DTR, respectively. Approximately 3.2–3.3% yield increases per additional 100 mm of Prcp were estimated.

Comparing the two crops, the differences in the estimated percent regression coefficients of the Tavg are statistically significant (P < 0.05), with more harmful effects for maize than wheat according to the t-test results. However, there is no statistically significant difference (P > 0.05) detected for DTR or Prcp.

3.3. Production change caused by recent warming

We used the regression coefficients calculated above with the actual Tavg and DTR trends at the county level as input to quantify the relative production change caused by the Tavg, DTR and their combined effect over 1980–2008 (Table III).

Table III. Relative production change and 95% CI caused by the Tavg, DTR and their combined effect from 1980 to 2008
CropsMethodsTavgDTRTavg + DTR
  Relative production (%)95% CIRelative production (%)95% CIRelative Production (%)95% CI
  1. All counties were involved in the calculation.

WheatMLR1.60.9 to to to 4.8
 PCR0.6− 0.1 to to to 1.7
MaizeMLR− 5.8− 6.7 to − to 1.6− 4.4− 5.5 to − 3.2
 PCR− 5.1− 5.9 to − to 0.8− 4.4− 5.2 to − 3.5

For wheat, despite some differences, the two methods estimated that recent warming (Tavg+ DTR) resulted in an increased wheat production for the period between 1980 and 2008; the value using MLR is 3.7% (95% CI of 2.7–4.8%), and the value using PCR is 0.8% (95% CI of 0.0–1.7%). As to the individual Tavg and DTR effects, both have caused an increased production with values of approximately 1.6% (95% CI of 0.9–2.4%) in MLR or 0.6% (95% CI of − 0.1–1.3%) in PCR due to changes in the Tavg, and 2.1% (95% CI of 1.8–2.4%) in MLR or 0.2% (95% CI of 0.0–0.4%) in PCR due to changes in the DTR. The most important contributor to the overall increases is north where increase of wheat production is around 75% of projected total change in production in MLR (151% in PCR) caused by combined effects of Tavg and DTR (Table SI, Supporting Information).

For maize, the MLR and PCR estimated a quite consistent result that approximately 4.4% reduction in maize production (95% CI of − 5.5 to − 3.2% in MLR and − 5.2 to − 3.5 in PCR) could be related to the combined effects of the Tavg and DTR from 1980 to 2008. Maize production was improved by approximately 1.4% in MLR (95% CI of 1.2–1.6%) or 0.7% in PCR (95% CI of 0.6–0.8%) as the result of changes in the DTR, which is consistent with wheat in the change of direction. However, maize production was reduced by 5.8% in MLR (95% CI of − 6.7 to − 4.8%) or 5.1% in PCR (95% CI of − 5.9 to − 4.3%) with increases in the Tavg, opposite to the result in wheat. The regions with the most decreases are northeast and northwest, making around 30–40% of the projected total decrease of maize production due to Tavg and DTR combined effects in the past 29 years (Table SI, Supporting Information).

4. Discussion

This paper estimated the effects of the Tavg and DTR on wheat and maize yield and production based on data from 2339 Chinese counties during 1980–2008 using the MLR and PCR approaches. Despite slight differences, the results obtained by the two methods are broadly in agreement and support each other. Averaged on the national level, the increases of Tavg and DTR were determined to adversely affect wheat and maize yields (Table II), which is in line with an earlier national assessment (Lobell, 2007) that presented that wheat and maize yields were often negatively associated with the Tavg and DTR. A comparison between the two crops revealed that the increase in Tavg exhibits an approximately four times greater negative impact on maize than on wheat (Table II), and their difference reaches the statistically significant level (P < 0.05), indicating a greater vulnerability of maize than wheat to the warmer climate. The negative effect of the Tavg on crops has been widely reported because the higher temperature often accelerates respiration rates (Abrol and Ingram, 1996; Lobell et al., 2005; Prasad et al., 2008), limits grain development (Cheikh and Jones, 1994; Cantarero et al., 1999; Asseng et al., 2010), and shortens crop duration (Wheeler et al., 2000; Pathak et al., 2003; Fang et al., 2010). The negative DTR effect indicates more harmful warming impacts during the daytime than at night-time. The result is consistent with Zhang and Huang (2012) who, focusing on a provincial scale, found that wheat and maize yield variability showed a closer affinity with maximum temperature than with minimum temperature in China. The negative response of yield to the DTR was also reported in Australia by Nicholls (1997), who believed the reduction in frost occurrence associated with warming is the major reason.

Despite the overall corroboration of the effects of the Tavg and DTR with previous studies, our results also emphasize the importance of geographical variability on these climatic effects. The negative Tavg effect is majorly exhibited in the far north, northwest and east for wheat and the north, northwest, southern counties in northeast and some portions of the southwest for maize. However, there is a significant positive effect of the Tavg on wheat and maize yields in the northcentral and eastern part of northeast regions (Figures 3(a) and 4(a)), respectively. It is noteworthy that these regions with a beneficial warming effect often missed capture by most of the previous studies exemplified by Tao et al. (2008) who, based on provincial data, found three and seven provinces susceptible to warming for wheat and maize, respectively, but very few cases showing a significant positive effect (one province for wheat and none for maize). The difference confirms our initial hypotheses that using provincial data might merge the areas with significant and insignificant climatic effects and therefore veil statistical significance. Supporting our result, several recent analyses have noted that positive yield responses to the Tavg do exist in China based on long-term field trials or farmer surveys in the north and northeast and attributed the positive effect to associated lower risks of chilling injury (Wang et al., 2009) and a longer growing season (Yang et al., 2007; Liu et al., 2010).

The northern region is the most dominant wheat-producing region in China, accounting for 43% of the national wheat cultivation area (Figure 1). Therefore, it appears that trends caused by increases in the Tavg that have an overall negative impact on yields have been majorly counterbalanced by the positive Tavg effect in the region with intense wheat cultivation. According to our results using MLR, the increases in the Tavg have led to wheat production improvements of approximately 1.6% with 0.9–2.4% of 95% CI (0.6% with − 0.1 to 1.3% of 95% CI in PCR) from 1980 to 2008, if other factors were held constant (Table III). However, in the case of maize, the potential gains from the Tavg in the eastern parts of northeast are not able to offset the negative effects of the Tavg in other regions, like the northwest and southern regions of northeast provinces. We estimated that maize production has decreased as a result of increases in the Tavg by approximately 5.8% with − 6.7 to − 4.8% of 95% CI based on the MLR approach (−5.1% with − 5.9 to − 4.3% of 95% CI in PCR).

In contrast to the unanimously decreased annual DTR trends in China (Liu et al., 2004), the changes in the DTR showed a clear geographical variability in the wheat and maize growing season with decreased trends in the northeast, north and west, but an increased trend in the southern regions (Figure 1(b) and (e)). Therefore, because the major production regions for the two crops are based in the north and northeast, the negative DTR effect was estimated to boost wheat and maize production by 2.1% with 1.8–2.4% of 95% CI in MLR (0.2% with 0.0–0.4% of 95% CI in PCR) for wheat, and 1.4% with 1.2–1.6% of 95% CI in MLR (0.7% with 0.6–0.8% of 95% CI in PCR) for maize from 1980 to 2008, when controlling other factors (Table III).

Consequently, the combined effect of the Tavg and DTR was estimated to have increased wheat production by 3.7% with 2.7–4.8% of 95% CI in MLR (0.8% with 0.0–1.7% of 95% CI in PCR), but decreased maize production by 4.4% (−5.5 to − 3.2% of 95% CI in MLR and − 5.2 to − 3.5% of 95% CI in PCR) between 1980 and 2008. This estimated increased value for wheat is approximately 10% in MLR (0.2% in PCR) of the actual total increase in national wheat production, and the estimated decrease for maize accounts for 10.1% of the actual change in maize production in China during the same period. It is noteworthy that there is a discrepancy in the estimated directional change in wheat production compared with previous studies. For example, around 1.2 × 105 ton decade−1 reduction in wheat production was estimated by Tao et al. (2008). The disagreement mainly reflects bias from data resolution. As mentioned above, using provincial data failed to detect a significant positive effect of the Tavg that benefits the wheat yields in the northcentral region, thus causing the omission of gain from warming in the largest wheat growing region.

The northcentral and eastern parts of northeast regions showing a positive warming effect are intriguing for further investigation. Yields would benefit from a continued increased Tavg and decreased DTR, a future climatic feature in China anticipated by climate models (Ding et al., 2007; Lobell, 2007; Gao et al., 2011), if the estimated regression coefficients hold constant. This offers a possible opportunity to rearrange crop distribution as an adaptive response to climate change. However, it must be considered whether the future climatological temperature would exceed certain thresholds and turn the positive effect of the present climate to a negative effect in the future because a number of experiments have shown that increases in temperature would be ultimately detrimental to yields (Cheikh and Jones, 1994; Cantarero et al., 1999; Prasad et al., 2008). Yield changes as related to Prcp are out of the scope of the study because for the growing seasons of wheat and maize, we barely detected a significant trend for the Prcp, but its importance should not be underrated when estimating future climate change impacts. The anticipated water stresses due to climate change would pose substantial threats to Chinese crops without adaptation (Thomas, 2008; Challinor et al., 2010).

5. Conclusion

This study, using crop and climatic data at the county level, (1) provides empirical evidence of the climatic effects on wheat and maize from 1980 to 2008 in the finest resolution and (2) estimates the change in crop production due to the past warming trend.

Results indicate that increases in the DTR are harmful to wheat and maize yields in China except in some portions in the far northeast where the DTR has a positive effect on maize. Given the dominant decreased DTR trends in the major wheat and maize producing areas since 1980, the changes in the DTR have helped to improve crop production by up to 2.1% for wheat and 1.4% for maize when compared with the national average over the study period. Additionally, we also found that increases in the Tavg during the growing season adversely affect wheat and maize yields on average, but a notable beneficial Tavg effect was observed in the northcentral region for wheat and eastern parts of northeast region for maize. In the case of wheat, the positive Tavg effect in the northcentral region, as the most productive wheat growing area, has offset losses from warming in other less important areas, producing an overall gain in production by up to 1.6%. However, in the case of maize, up to 5.8% production was estimated to be lost due to increases in the Tavg over the study period. Consequently, the combined effect of the Tavg and DTR has resulted in an improvement for wheat production by up to 3.7%, but a decline in maize production of approximately 4.4%, equivalent to approximately 10% of the actual total change in Chinese wheat and maize production from 1980 to 2008.

As further temperature stresses from climate change are posed on agriculture, we emphasize that future studies should analyse the potential suitability of several regions that seemed favourable to warming at the present climate, i.e. the northcentral regions for wheat and eastern parts of northeast regions for maize. There is a need for continued regional-scale research on these two regions about the probability that future climate change would break threshold and make these regions vulnerable to future climate change. If the calculated regression coefficients were to hold in the future climate, the two regions could be considered potential spots for wheat and maize production to mitigate the climate change impacts in China.


We appreciate the anonymous reviewer for insightful suggestions and the Chinese Academy of Agricultural Sciences for providing county level data. This research was supported by the External Cooperation Program of the Chinese Academy of Sciences (Grant No.GJHZ1204) and the National Natural Science Foundation of China (Grant No. 41021004).