Mapping Chinese Rice Suitability to Climate Change

Climate change has the potential to affect Chinese rice production; however, the rice crop could become more suitable to new climatic conditions because of benefits derived from new agricultural technologies. In this paper, a county-level dataset and crop model were used to analyze actual rice yield suitability by measuring the yield gap and yield stability from 1980 to 2011 in 1561 counties of China. The results showed that the national yield gap between the actual and potential yields was approximately 23.0%, which is close to the threshold for profitable planting. However, a number of counties in the northeastern and southwestern regions showed a 30 to 50% yield gap, which indicates a relatively lower suitability of the rice. The rice yield stability results indicated that the actual stability has exceeded the potential stability in most of the counties of China, thus indicating a high level of suitability. Temporally, a decreasing trend was observed for both the yield gap and stability, suggesting that the suitability of rice in China has improved, which might be associated with the development of agricultural technology. The only noteworthy locations presenting a high yield gap and yield instability were several counties in the northeastern region. Since the northeastern region accounts for a significant proportion of China's rice production, further investigations should be conducted to identify the underlying causes of the yield gaps and determine methods of increasing the yield stability. The implementation of more suitable agricultural technology in the area is also suggested to improve the rice suitability in the region.


Introduction
In China, rice is one of the most important crops and accounts for approximately 35% of the total cereal crop production (FAOSTAT, 2013).Changes in the rice yield have the potential to significantly affect food security in China.
Climate change caused by rise in CO 2 concentration has significantly alters the temperature and moisture regimes in China (Ding et al., 2006).A number of modeling studies have suggested that those changes have the potential to adversely influence Chinese rice growth (Yao, Xu, Lin, Yokozawa, & Zhang, 2007;Chavas, Izaurralde, Thomson, & Gao, 2009;Masutomi, Takahashi, Harasawa, & Matsuoka, 2009).For example, climate change has been estimated to reduce the Chinese rice yield by 0.3 to 7% and increase the yield variability by 3 to 9% (Yao et al., 2007).In addition, future rice yield projections indicate regional variability, with rice production in northeastern China predicted to benefit from CO 2 fertilization effects and rice production in the east and south predicted to experience a declining trend in which harmful effect of warming is dominated (Masutomi et al., 2009).
Despite the negative potential consequences of climate change, several recent studies focusing on historical rice data found that growers appear to have successfully adapted to these negative climate effects by planting suitable rice cultivars (Liu, Wang, Zhu, & Tang, 2012;Tao et al., 2013), rearranging rice production areas (Lin, 2005;Yang et al., 2007) and providing better water management (T.Zhang, Zhu, Yang, & X. Zhang, 2008;Deng et al., 2010).Thus, possible negative outcomes have not been observed to a large extent.Accessibility to new agricultural technology and updated agricultural infrastructure have enabled Chinese rice to become more resilient to climate change and extreme weather shocks and more suitable to growing under new climate conditions (Fraser et al., 2008;Simelton, Graser, Termansen, Forster, & Dougill, 2009).The progressively greater suitability of rice in the context of climate change is undoubtedly good news to Chinese rice production.However, to our knowledge, a national assessment of rice suitability under climate change has not yet been performed.
Evaluating the historical suitability of rice to climate change at a full national scale would provide new insights into how rice cultivation has adapted to past climate change, which could be used to determine whether the suitability of rice has improved in relation to climate change at a national scale and identify locations of vulnerable hotspots in need of further adaptations.These historical insights could inform our baseline capacity to adapt to climate change, which could be used as a guide for Chinese agricultural policymakers.
Conventional crop suitability assessments are generally based on yield gap analyses (Fischer, Velthuizen, Shah, & Nachtergaele, 2002), which assess the actual yield relative to the potential yield to determine the suitability of the crop yield according to climatic factors.However, this analysis only considers the yield level and does not consider yield stability, another aspect of food security that is becoming increasingly recognized (Schmidhuber & Tubiello, 2007).Low yield stability often results in unpredictable food shortages, which threaten food supplies and farmers' livelihoods (Schmidhuber & Tubiello, 2007).However, the consequences of variations in crop yields have received little attention.
The objective of this study is to 1) provide a national assessment on historical rice yield suitability over the period from 1980 to 2011 by quantifying the actual yield and yield stability relative to their potential levels and 2) identify regional variability in yield and stability, which will be used to support rice production security under changing climatic conditions.This study uses a high resolution county-level rice yield dataset for China to simulate the potential yield for each county; therefore, this study provides a comprehensive evaluation of China's rice yield suitability on a national scale.

Data sources and Preparation
A rice yield dataset at the county level was collected from the Agricultural Information Center at the Chinese Agricultural Academy of Sciences.In this study, these data represent the actual rice yields of Chinese farmers, which cover 1561 counties (Figure 1a) in 28 provinces (Figure 1b).Climatic data were downloaded from the China Meteorological Data Sharing Service System (http://cdc.cma.gov.cn/).Climatic observations include daily minimum and maximum temperatures, sunshine hours, vapor pressure, wind speed and rainfall for the period from 1980 to 2011.Because the dataset only included 756 stations and did not satisfy the climate input requirement of the model for each county, we estimated the daily climate data using the algorithm presented by Thornton, Running, & White (1997); this algorithm interpolates the abovementioned data of the 756 climate stations into 10 km grid cells and then extracts climatic information from the grid data that corresponds to the locations of the 1561 counties.(1996( -1999( ) Lishui (1994( , 1995( , 1999)), Lonogquan (1994Lonogquan ( , 1996Lonogquan ( , 1997Lonogquan ( , 2002) ) Zhejiang 17 Late 2You92 Lishui (1996,2000,2001) Longquan ( 2001 (1994)(1995)(1996)(1997) Nanning (2000)(2001)(2002)(2003)(2004) Note. 1 The ID number corresponds to Figure 1b.
In this study, the ORYZA2000 model was used as a tool to evaluate potential yields under optimum supplies of water and nitrogen, and these yield values were used to represent yield changes exclusively caused by climatic conditions.Other management information (i.e., emergence dates, seed-bed duration and planting density) was assumed to be province-specific and set to the average value of the experimental rice data.Crop coefficients were derived from the representative rice cultivars and assumed to be province-specific.For certain provinces in the northwest, we lacked experimental rice data; in these cases, the cultivar coefficients from nearby provinces that share a similar cropping system were used.Moreover, because the county-level yield dataset did not provide separate yields for the early and late rice seasons, the simulated early and late rice yield potentials were averaged and weighted by the sowing areas for early and late rice according to provincial level data from a statistics yearbook and from the National Bureau of Statistics of China website (http://www.stats.gov.cn/tjsj/ndsj/).We assumed that the province-specific percentages of early and late season rice were consistent between years for all counties in the province, and the yield values for each county-year combination were estimated by multiplying the area-weighted percentage for the simulated early and late rice yields.Although clearly not ideal, such practices allowed us to determine a first-order approximation of the yield potential for all of China that could be compared with the available county yield data.

Index of the Yield Gap and Variation Differences
To achieve the objectives of this study, two indices were used to quantify the actual rice yield suitability in relation to climate change: a yield gap index and yield variation difference index.All of the statistical analyses were executed in R version 3.02 (R Core Team 2013).
Yield gap.The yield gap was defined as the percentage difference between the potential and actual yield (Equation 1), and the index was used to quantify the similarity between the actual yields and their potential levels in each county.A lower value of the index indicates that the actual yield is closer to its potential level and more suitable for the climatic conditions.
Where, YG denotes the yield gap (%) and Y p and Y a denote the potential and actual yield, respectively.Yield variation difference.As mentioned in the introduction section, yield stability is another important index for crop suitability.In this study, we adopted the yield variation index used by Reilly et al. (2003) to quantify the yield stability (Equation 2).
Where, V is the yield variation difference (%), Y t is the rice yield in year t (ton ha -1 ) and Y trend is the fitted yield in each series, which is determined using a smoothing spline method and indicates the yield trend over time (ton ha -1 ).
The yield gap was similarly defined by quantifying the suitability of the rice cultivar relative to the yield stability by determining the difference between the actual and potential yields (Equation 3).
Where, VD is the yield variation difference (%), V a is the yield variation calculated from the observed county yield data, and V p is the yield variation calculated from the simulated potential yield data.

ORYZA2000 Calibration and Evaluation
The ORYZA2000 model was calibrated based on the observations on day after emergence (DAE) of flowering and maturity dates and rice yields.Based on Figure 2 and Table 2, there is a good agreement between the simulated and observed yields and phenology for both the calibration and validation datasets.The normalized root-mean-square error (NRMSE) for the yield simulation was approximately 13%, and the NRMSE for phenology varied from 3.7-5.1% (Table 2). (1) (2) www.ccsen Note.The After obta potential g the represe

Yield G
The mean distributio ha -1 ) were whereas th southern a (5.9 ton ha were also northeaste between 7 The yield Lower yie region (Hu counties in and Heilo higher yiel      4a) and actual yields (Figure 4b) averaged over the study period.One of the reasons for the observed variability may be the longer growing season in the north, which causes a greater potential for variability under adverse climatic conditions.However, the actual yield variation was generally lower than the potential yield variation, which caused a uniform negative yield variation difference except in certain counties in the northeast (Figure 4c).The negative values indicate that the actual rice yield has already become more stable than the values derived under potential climate change conditions.Higher yield variability under potential conditions is not uncommon and has been observed in American (Hansan & Jones, 2000) and European maize (Reidsma, Ewert, Boogaard, & Diepen, 2009).In our region, the local adaptive responses to adverse climatic impacts not accounted for in the crop model may further reduce the influence of climatic factors and promote even lower observed yield variability.Such inconsistency between the potential and actual yield variations was also observed temporally (Figure 6).Compared with the spatial distribution for the potential yield variance trends (Figure 6a), the actual yield variance trends exhibited a uniform decreasing trend over time in most counties (Figure 6b).The more stable trend of the actual yield was consistent with several earlier empirical studies (Simelton, Graser, Termansen, Forster, & Dougill, 2009;Zhang et al., 2008), which attributed this trend to progressive improvements in several socio-economic factors in China, including technical inputs, breeding investments and mechanization.As a result, the yield variation differences also experienced a downward trend, indicating improved rice suitability because of the current agricultural technological developments in China.Therefore, our results suggest a positive outcome for yield stability because Chinese rice cultivars were able to maintain a stable yield according to average values and temporal trends.
By combining the yield gap index and variation difference index, this study mapped the actual yield suitability from 1980 to 2011 (Figure 7).Over most of the study area, the actual rice yields exhibited a high level of suitability in relation to the local climate.On average, the actual yields over the study period were close to the potential values and exhibited a high level of stability (Figure 7a), especially for the eastern and southern regions, which are the traditional rice production regions in China.Temporally, the yield gap between the actual and potential rice yields was reduced and became more stable over the majority of the study regions (Figure 7b).At a national level, several large yield gaps in the northeast and southwest regions should be given a higher priority and investigated in further studies because the actual rice yield has performed well in terms of stability, and the yield gaps have the potential to decrease.Finally, the only locations that showed a high yield gap and high variance differences were located in the northeast and northwest regions.Because of recent increases in the rice planting areas (Lin, 2005;Yang et al., 2007) and the greater production demands on the northeastern region to meet China's food self-sufficiency (Simelton, 2011), further investigations are recommended to maintain yield improvements and ensure yield stability in the northeastern region of China.

Conclusions
In order to understand how yield suitability changes in China, we investigated rice yield gap and stability over 1980 to 2011 in 1561 counties of China.The level and variance of yields relative to their potential ones indicates whether the historical climate change and technology improvement has made the yield more suitable for their growing environments or not.Potential yield changes due to climate were calculated by crop model, which compared with actual yield statistics. Results suggest that the national yield gap between the actual and potential yields was approximately 23.0%.However, this presents a regional heterogeneity; a number of counties in the northeastern and southwestern regions showed a 30 to 50% yield gap, which indicates a relatively lower suitability of the rice.On the other hands, actual rice yield stability has exceeded the potential stability in most of the counties of China, thus indicating a high level of suitability.There is a decreasing trend for both yield gap and stability, indicating that the suitability of rice in China has improved, which might be associated with the development of agricultural technology.The only locations showing a high yield gap and yield instability were several counties in the northeastern region.Therefore, we conclude that the region of northeastern in China has a potential to improve the level and stability of yields.Given the northeastern region accounts for a significant proportion of China's rice production, further investigations should be conducted to identify the underlying causes of the yield gaps and determine methods of increasing the yield stability.

Figure 1
Figure 1 Figure ◆ denotes the Figure 4.A

Fig
Fig

Table 1 .
Representative cultivars for experimental stations in the

Table 2 .
E and maturi , with the rice yield in the northern region more variable than that in the southern region for both the potential (Figure