Why can China maintain a high crop diversity? A spatial-temporal dynamic analysis

II Data and methods

2 Crop diversity index

The crop diversity index is designed to capture the richness of crops. Richness refers to the number of different crops in a given area: the greater the number is, the higher the diversity is (Aguilar et al., 2015; Duflot et al., 2015; Zhou et al., 2014). In ecology, the Shannon’s index is used for showing the variance in the distribution of species abundance. This study adopted it to show variance in the proportion of area covered by each crop. The Shannon diversity index (SHDI) is expressed as

SHDI = − ∑ i = 1 m P i ln ( P i )

(1)

ESHDI = e SHDI = e − ∑ i = 1 m P i ln ( P i )

(2)

3 Spatial and statistical analysis

The least square regression was used to quantify the trends in crop diversity. A negative slope coefficient (S) indicated a decrease in crop diversity and a positive coefficient indicated an increase (at p < 0.05). The slope was calculated as follows

S = n ∑ t = 1 n t E S H D I t j − ( ∑ t = 1 n t ) ( ∑ t = 1 n E S H D I t j ) n ∑ t = 1 n t 2 − ( ∑ t = 1 n t ) 2

(3)

To analyse whether crop diversity agglomerates in space or not, the hot-spot statistical analysis is used to identify areas with statistically significant clustering. In this study, the Getis–Ord local statistic was employed to detect the hot/cold spots of crop diversity and their distribution, the formula is as follows

G i ∗ = ∑ j = 1 n w i , j ESHDI j − ESHDI ¯ ∑ j = 1 n w i , j S n ∑ j = 1 n w i , j 2 − ( ∑ j = 1 n w i , j ) 2 n − 1

(4)

ESHDI ¯ = ∑ j = 1 n ESHDI j n

(5)

S = ∑ j = 1 n ESHDI j 2 n − ( ESHDI ¯ ) 2

(6)

The G i ∗ statistic is a z-score represented by GiZScore. We can divide the value of GiZScore into seven types by standard deviation (Std. Dev.) to represent a hotspot (red in map) or cold-spot (blue in map).

4 Semiparametric geographically weight regression analysis

The Semiparametric geographically weight regression (SGWR) is an improvement of geographically weight regression and can generate local parameters to reflect spatial heterogeneity of derived statistics, including local R ² , local model residuals and local coefficients (Gao and Li, 2011). Because the adjusted R², coefficient and residuals for each county from SGWR express the fitting effect at different locations, the complex spatial heterogeneity of crop diversity can be identified, mapped and simulated as follows

y i = ∑ l γ l z i , l + ∑ k β k ( μ i , v i ) x i , k + ε i

(7)

1 Trends in crop diversity

China’s farming maintains a high crop diversity during the past 35 years. The crop diversity index averages 7.39 with SD of 0.97, indicating that at least 7 major crops have undergone massive planting process. The highest crop diversity (8.64) occurred in 2002 and the lowest (4.79) in 1986. When averaged for the whole country, crop diversity showed an overall increasing trend by a rate of 0.07/yr during 1980–2014 (defined as the period P0). However, the overall trend consists of downward and upward trends at the shorter-temporal scale. In specific, the crop diversity decreased by a rate of 0.36/yr before 1986 (defined as the period P1) and then it increased by a rate of 0.087/yr till 2003 (defined as the period P2), after which it decreased by a rate of 0.065/yr (defined as the period P3) (Figure 1(a)). Moreover, only the North-east and North Plateau region have a significant negative trend in crop diversity, and the Huang-Huai-Hai region has a nonsignificant negative trend. Meanwhile, the rest of other cropping regions experienced significant positive trends during past 35 years (Figure 1(b)–(k)).OPEN IN VIEWER

To compare the variation of crop diversity trend, the spatial distribution of crop diversity changing slope that differs among the periods in each county is shown in Figure 2. Generally speaking, crop diversity increased in more than two-thirds of the counties during 1980–2014. Spatially, the crop diversity increased significantly in the south but decreased significantly in the north of China (Figure 2). The counties showed increase in crop diversity accounted for 69.24%, of which 51.13% have a significant (p < 0.05) trend and are distributed mainly in southern China; The counties with significant negative trend accounted for 30.76%, 16.7% of which has a significant decreasing trend (p < 0.05), and are mainly distributed in north-eastern China and in the North China Plain where maize and wheat are dominated. Specifically, the variation in crop diversity trend varied over space and time. In the period from 1980 to 1986, 23.96% of positive counties had significant (p < 0.05) trend and they are mainly distributed in southern China. In contrast, only 9.69% of negative counties have significant trends and they are mainly scattered in Henan and Shaanxi province. During 1987–2003, 34.47% of counties showed a significant (p < 0.05) increasing trend and these counties are mainly distributed in Yunnan and Guangxi provinces located in Southern China as well as the Gansu and Xinjiang province in North-western China. In contrast, 12.77% of counties showed a significant (p < 0.05) decreasing trend and these counties are mainly located in the northern China. During 2004–2014, the counties with significant positive trends accounted for 31.27% and those with significant negative trends accounted for 24.78% of the total numbers of counties in China (Figure 2).OPEN IN VIEWER

2 Spatial pattern of crop diversity

To illustrate the spatial dynamics of crop diversity during the past 35 years, a visually image of annual crop diversity was created (See the Figure S1). And each period of average crop diversity was also mapped (Figure 3). Generally, crop diversity in China exhibited distinct spatial variation at the county scale. It is higher in hilly and mountainous areas and lower in the plains and the Tibet-Qinghai Plateau. The mean crop diversity at the county level were 4.10 ± 1.08 in the period of P0, 3.47 ± 1.16 in the period of P1, 4.22 ± 1.14 in the period of P2 and 4.26 ± 1.32 in the period of P3, respectively. The average crop diversity at the county level is lower than the corresponding value at the national level. It also reveals that most of counties in China has a complex cropping system to maintain a high crop diversity. Spatially, the crop diversity is found to be relatively high in a zone extending from north-east to south-west in China, and this zone covers 14 provinces during the period of 1980–2014 (Figure 3(a)). The spatial distribution of crop diversity partly is attributable to the crop specialization, mostly of rice, maize, wheat, or soybean in northern China and partly because the southern China has a reflection of flexible cropping patterns, where farmers have a wider choice of crops through crop rotations and multiple cropping on smaller holdings.OPEN IN VIEWER

In more details, over half of counties in China has a high crop diversity value greater than 4 during past 35 years, indicating that Chinese farming system maintains a relative complex diversity to harvest more food. The crop diversity index higher than 4 occurred in 837 counites and these counties are mainly distributed in a zone extending from North-east to South-west, covering Heilongjiang, Liaoning, Shandong, Shanxi, Shaanxi, Henan, Hubei, Sichuan, Chongqing, Guizhou and Yunnan province of China in the period of P1. The maximum crop diversity occurred in the Loess plateau region and in mountain region of South-west China (Figure 3). In the period of P2, the mean crop diversity index is significantly higher than that in the period P1. The crop diversity index beyond 4 has 1434 counites and these counties are mainly distributed from the Mountain areas of Hebei and Shandong to Yunnan (Figure 3). In the period of P3, the mean crop diversity index is slightly lower than that in the period of P2. However, the crop diversity index higher than 4 has 1328 counites. The decreased diversity counties are mainly distributed in the three provinces of North-east China, the Loess plateau region of Shaanxi and Shanxi and some urban agglomerate region such as the Yangtze River Delta and the Pearl River Delta (Figure 3).

To show the spatial cluster dynamics of crop diversity during the past 35 years, a visually image of annual hot and cold-spot regions of crop diversity was mapped (See the Figure S2). The identification of the hot and cold spots of crop diversity in China during different periods were also mapped. The results show that the hotspots of crop diversity gradually moved to the south-west mountainous area (Figure 4). Spatially, the whole China can be divided into three parts, consisting of one hot-spot and two cold-spot regions. The hot-spot region extends from the north-east to the south-west and comprises 16 provinces. The two cold-spot regions with low crop diversity consist of four north-western provinces and six south-eastern provinces (Figure 4). The hot spots are mainly distributed in a zone extending from the north-east to the south-west of China during the P1 period (Figure 4). Then, the zone with high hot-spot value retreated gradually toward the southwestern China during the P3 period, in which the dominant crops are rice (Figure 4). Overall, the number of hot-spot counties showed a significant downward trend, and decreased from 52.54% during the P1 period to 31.70%during the P3 period. In contrast, the number of cold-spot counties showed an upward trend, and increased from 34.22% to 40.75% (Figure 4).OPEN IN VIEWER

3 Driving forces for crop diversity

The mean crop diversity during 1980–2014 is used as the dependent variable. Seven physical and socio-economic environment characteristics factors, namely, cropland elevation (ELE), slope of cropland (SLOPE), road density (RD), population density (PD), irrigation index (IRR), mean annual precipitation (MAT) and mean annual temperature (MAP), were used as the independent variables to explain the distribution of crop diversity. Table 1 shows the estimated global and local coefficients based on the GWR between factors and crop diversity. The seven explanatory variables listed above explained about 85% of the spatial heterogeneity of crop diversity. The AICc of SGWR model has been improved with the value of 161.05 compared to that based on local GWR model, suggesting that the constructed model is a good fit to changes in crop diversity in different counties. Overall, SLOPE is the major natural factor controlling the spatial distribution of crop diversity while PD is the major social factor that has significant negative influence on crop diversity.

OPEN IN VIEWER

Table 1.

Estimated global and local coefficients by the semiparametric geographically weight regression.

Fixed (global) coefficients
Variable	Estimate	Standard error	t (Estimate/SE)
PD	−0.90	0.32	−2.79
RD	0.01	0.24	0.02
SLOPE	1.71	0.26	6.63

Geographically varying (local) coefficients
Variable	Mean	STD	Min	Max	Range	Lwr quartile	Median	Upr quartile	Interquartile R	Robust STD
Intercept	6.27	8.32	−18.66	58.20	76.86	1.50	4.65	8.45	6.95	5.15
IRR	−0.08	1.38	−5.07	6.25	11.33	−0.91	−0.08	0.65	1.56	1.16
ELE	−1.33	18.65	−245.74	144.28	390.02	−6.09	−2.06	2.77	8.86	6.57
MAT	−2.43	12.83	−80.14	36.74	116.87	−5.12	0.19	3.96	9.08	6.73
MAP	0.17	6.58	−24.25	24.47	48.72	−3.31	−0.55	3.01	6.32	4.68

Specifically, PD, RD and SLOPE have global effects on crop diversity in China. Particularly, slope is a significant (p < 0.05) positive factor and plays the greatest impact on the crop diversity in China. For example, the counties with higher slope have high crop diversity. Population density has a significant (p < 0.05) negative impact on the crop diversity in China, which means that the high population density counties have lower diversity. Road density has a slight positive impact on the diversity.

In contrast, IRR, ELE, MAT and MAP have local effects on crop diversity distribution during the past 35 years. When the spatial non-stationary was considered in the SGWR model, the spatial distribution of local R ² has a higher value in mountainous counties and a lower value in the plains (Figure 5(e)). When it comes to each driving factors of local effects, the MAT, ELE and IRR have the mean value of −0.08, −1.33 and −2.43, respectively. In contrast, the MAP has a mean value of 0.17(Table 1). From the spatial perspective, MAT has a negative effect in Xinjiang, Inner Mongolia, Jiangxi, Hunan, and Guangdong, largely because MAT determines the area under farming and limits the range of crops that can be profitably grown through heat (Figure 5(a)). IRR has the slight negative impact on crop diversity because irrigation contributes to large area cropland that can be specialization, which, in turn, reduces crop diversity (Figure 5(c)). ELE has a negative effect on diversity in the plain and plateau, indicating that the large cropland can only plant one or two crops (Figure 5(d)). MAP is also important for crop growth. In areas with low rainfall, farmers tend to grow a variety of crops to obtain greater yields, which diversifies agricultural production (Figure 5(b)).OPEN IN VIEWER

1 The evolution of crop diversity in China

In this study, the results indicate that the crop diversity process has undergone an increase in most of the counties in China. The majority have a crop diversity value over 4, which means there are at least four types of crops in most counties. The crop diversity of the United States and India, which have similar cropland size to China, has experienced a reduction in crop diversity (Aguilar et al., 2015; Smith et al., 2019). Compared to the U.S., China maintains a higher crop diversity at county level than most of U.S. counties. American crop diversity fluctuated around 4 from 1978 to 2012 at national level, which is half of Chinese crop diversity (Aguilar et al., 2015). In contrast, crop diversity at district level also was stable at approximate 6 during 1947–2014. It is similar to Chinese counties crop diversity, but higher at national levels (Smith et al., 2019). In the future, it is necessary to conduct comparative studies of the driving mechanism among regions or countries of similar size such as Europe, America and India .

The crop diversity change was due to aggregation, specialization and the accompanying homogenization of cropping systems. Our results indicate that the crop diversity increased in southern China and decreased in northern China, thereby leaving the impression that the crop diversity in China did not change significantly at the national level. Crop diversity can be aggregated based on data of varying spatial-temporal resolution (Lin, 2011), which can be characterized by crop mixtures comprising different varieties (Hossain et al., 2001), or may be represented by areas practising multiple production systems (Deng et al., 2017; Lin, 2011). Moreover, there are two turning years in crop diversity change at national level. They corresponded, respectively, well to the timing of two important agricultural policies, that is, the household responsibility system (HRS) and the agricultural tax abolition (ATA) launched by Chinese government in 1980 and 2003, respectively. However, the impacts of these policies on the crop diversity over time still need further study (Boody et al., 2009; Renard and Tilman, 2019).

Many factors influence crop diversity in China, and a comprehensive examination of all of them is difficult. Some studies indicate that cultivated land area, labour force, irrigation, increased consumption of non-food products and decline in the consumption of staple foods are key factors influencing crop diversity (Kearney, 2010; Rahman, 2016); other studies show that crop diversity is influenced by the choice made by farmers and natural conditions (Duflot et al., 2015). The spatial distribution of crop diversity in China may be controlled by natural conditions, whereas crop diversity at local scale may be controlled by the farmer’s choice and socio-economic conditions (Lin, 2011). By entrusting the choice of crops and cropping systems to individual households, the increasing prices of selected agricultural commodities, and relaxing restrictions on the trade of agricultural products, China’s food supply received a substantial boost in recent decades (Huang and Rezelle, 2015). When conditions favour crop specialization, crop diversity decreases (Matsushita et al., 2016; McDaniel et al., 2014). The relationship between crop diversity changes and crop specialization still needs further study in the future.

2 Crop diversity sustainable strategy

Changes in crop diversity are very sensitive to spatial distribution of crops (Liu et al., 2018). Crop aggregation demonstrates a decreasing trend in crop diversity and an increasing trend in the homogenization of cropping systems, which may have far-reaching implications for food production, and ecosystem services associated with agricultural systems in China (Yu et al., 2018; Zuo et al., 2018). From a policy-making perspective, the resilience conferred by crop diversity deserves greater attention in the development of agriculture throughout the country and in promoting both crop specialization and crop diversity. Given the importance of crop diversification, it is important to promote homogenization of agriculture at the county level but do not overlook crop diversity at the national level (Hijmans et al., 2016). How to use the agricultural zoning policy, which link spatial differentiation with cropping sustainability management, to maintain the high regional crop diversity still need the attention of policymakers (Peng et al., 2020).

Crop diversification is critical to food production and helps to relieve hunger, malnutrition and even excess consumption (Massawe et al., 2016). Increasing crop heterogeneity is an effective way to enhance multi-trophic diversity (Sirami et al., 2019). However, using information on spatial and temporal heterogeneity of crop diversity to shape national policies for food stability has not attracted the attention of policymakers and farmers (Smith et al., 2019). One possible reason is that the crop diversity did not historically directly affect food supply. Another reason is the ignorance of policymakers on regional differences in crop diversity. Because crop diversity continues to be high when examined on the national scale, it is no longer seen as a major problem. We also need to develop better models of cropping systems, especially models of diversified cropping systems, to justify that crop diversity can not only enhance food security but also is profitable to farmers (Boody et al., 2009; Wu et al., 2018). Such models of agricultural land systems are particularly useful to decision-makers in planning long-term policies for land use.

3 Uncertainties and limitations

This study evaluated the spatial and temporal dynamics of crop diversity to understand why the Chinese agriculture maintain a high diversity of crop variety to sustainably feed more people during past 30 years. Most of previous researches have studied the crop diversity at global scale, but these studies considered a country or region as one sample, which is difficult to discuss the spatial heterogeneity impact of crop diversity (Jarvis et al., 2008; Khoury et al., 2014; Renard and Tilman, 2019). The cropping system in China is highly heterogeneous and spans a huge geographic space. Therefore, although this study is beneficial for revealing the spatial driving mechanism, it also incurs uncertainties in revealing the temporal driving dynamics. Moreover, because the Chinese crop diversity has a fluctuating upward trend, it is likely that the conclusions of this study are different from other studies in other countries (Aguilar et al., 2015; Smith et al., 2019). Specifically, the trend of crop diversity has an obvious difference in cropping regions. Thus, we can infer that crop diversity in different countries also shows difference in space and over time. However, this study successfully presents a relative long-term period case in China, and the findings can contribute to a comprehensive understanding of crop diversity changes and its drivers in cross physical geographical zone of worldwide.

In addition, this study focused on crop diversity at the county level rather than at the farm level. The underlying mechanism of crop diversity may differ. Because the crop diversity at farm level is a relative microscale research, it pays more attention to the behaviour of farmers and the regulation of the market-induces (Labeyrie et al., 2021; Song et al., 2021), while the macroscale study can observe the influence of natural conditions, cropping conditions, accessibility and policies. This study showed that the spatial pattern of crop diversity at the county scale is not static and governed by both anthropogenic and natural forces. China has a large population concentrated in rural areas, providing a foundation of planting options for crop diversity. Farmers will choose crops to plant according to their preferences and knowledge of crops. This is why population density determines crop diversity. Similarly, China’s arable land is widely distributed, both in the plains and in the mountains. For mountain farming, slope is an important factor affecting the crop diversity of cultivated land. For example, in southern and southwestern China, farmers plant rice in terraced fields, and cultivate corn and other economic crops in mountainous areas with steep slopes. In this study, the physical and socio-economic environment are the targets of macroscale analysis. As a consequence, it is necessary to conduct more studies on the behaviours of farmers or agricultural policies of counties from field census.

Finally, owing to limitations of data, this study focused on only the period of 1980–2014. And in terms of research driving forces, the selection of variables is limited by data acquisition, where many variables are not considered in the model. We only focus on the spatial drivers and use the average pixel data at county level, which has limitation. However, the period considered in this study was almost 35 years between 1980 and 2014. Many spatial variables such as elevation and slope changed relatively small, thus, the limitation does not have a significant impact.