Spatial-temporal Variation Characteristics of Grain Yield per Unit Area and its Balanced Increasing Potential in China

The aim of this study is to provide assistance decision for grain yield in China, which can provide basis for reasonable layout of grain production project. Based on the statistical data of 2301 counties in China, using spatial autocorrelation analysis method, spatial changes of grain yield per unit area at county level in China during 1990-2010 were discussed and then the increase potential of grain yield per hectare and total yield at regional scale were calculated. The results showed that: (1) Grain yield per unit area at county level showed the evident pattern “High in the northern while low in the southern” and “High in the eastern and western while low in the middle”; The average grain yield per hectare increased by 1040.74 kg/hm during 1990-2010 and the increment of grain yield per unit area descended from North to South at county level. (2) Grain yield per unit area at county level in China had a strong spatial autocorrelation. The counties with "High-High" and "Low-Low" correlation were the majority. Counties with significant "High-High" correlation in 2010 were mostly located in plain area, while counties with significant "Low-Low" correlation were mainly distributed in Hengduan Mountainous Area, Inner Mongolia steppe Area, etc. (3) Two thousand three hundred and one counties were divided into 41 first-grade regions and 115 secgrade regions according to the coupled conditions of cultivation system regionalization and LISA cluster map. The total potential output of China was 1.77×10 tons.


INTRODUCTION
Grain yield continues to increase steadily is the basic guarantee for food security, which is important to world stability and social development (Rondaninia et al., 2012).On the basis of solid sown area, increasing per unit area yield of grain is the main way to increase grain output in the future (Liu et al., 2013).In order to provide scientific basis for grain yield planning and efficient utilization of agricultural resources, systematic analysis of spatial-temporal differentiation of grain yield per unit area, scientific calculation of improvement in potential yield of regional grain should be taken (Bai et al., 2013).
In recent years, domestic and foreign scholars have made depth analysis of spatial-temporal pattern, influencing factors and yield-increasing potential of per unit area from different perspectives (He et al., 2008;Cheng, 2009).With the diversification of data acquisition methods and more attention paying to "spatial interaction", more researches are focusing on per unit area yield of grain and spatial heterogeneity of input factors by spatial economic model (Cheng et al., 2007).According to literature search, the researches on per unit area yield of grain have converted their research scales from national scale or provincial scale to municipal scale or county scale.Research methods mainly adopt variation coefficient, multiple regression analysis, ridge regression analysis, etc., but more and more researches adopt spatial econometric model.Research contents concerning about the status of per unit area yield of grain, while paying attention to reveal spatial temporal evolution rule and spatial agglomeration characteristics of per unit area yield.
Researches on yield-increasing potential adopt mechanism method (Zheng et al., 2013) and focus on the region itself, but think less about the relationship between regions.The unbalanced development of per unit area yield of one same grain in different regions caused by significant differences in cultivated land resource endowment and social economic development level (Bi et al., 2008;Zhang and Qin, 2011;Zhang et al., 2006), thus affecting the overall increase in regional grain yield.So it is an effective method to measure regional yield-increasing potential through type partition based on spatial agglomeration model of per unit area yield of grain and cultivation system regions with measurement of regional yield-increasing potential by regional linkages.Therefore, in order to provide assistance support for grain yield in China, this study adopts per unit area yield data in 1990 and 2010 at county scale with application of spatial autocorrelation analysis to explore the spatial characteristics of per unit area yield of grain and with cultivation system regions to estimate regional yieldincreasing potential.

RESEARCH METHODS AND DATA SOURCES
Spatial autocorrelation analysis model: ESDA method discovers spatial agglomeration and spatial anomaly through visualization of spatial distribution pattern of objects, which provides an effective method to investigate spatial distribution pattern of different regional attribute values and relative changes in spatial differences.Spatial autocorrelation analysis can be divided into global spatial autocorrelation analysis and local spatial autocorrelation analysis.Global spatial autocorrelation analysis model can be used for analysis of overall regional spatial association and difference degrees; Global Moran's I index is more commonly used (Lu and Mei, 2007;Wang et al., 2009): wherein I t represents global Moran index of per unit area yield of grain in the year of t, while its value range is [-1, 1], I<0 represents negative spatial autocorrelation, I = 0 represents spatial uncorrelation, I>0 represents positive spatial autocorrelation, while the bigger the absolute value of I, the stronger the spatial autocorrelation.x it and x jt represent per unit area yield of grain in County i and County j in the year of t. w ij represents spatial weighted matrix of research object.Local spatial autocorrelation analysis can measure local spatial association and difference degrees between each area and its surrounding areas, while visualization of difference patterns can be achieved with Moran scatter diagram, etc. Local Moran's I is always used for measurement of spatial difference degrees and significance between Area i and its surrounding areas: The meaning of symbols in the formula is the same as that in Formula (1), drawing LISA aggregation diagram according to Local Moran's I index.Measurement of regional equilibrium potential yield increase: Regional equilibrium potential yield increase is a kind of potential to achieve regional equilibrium development and scale benefit based on the foundation of interregional spatial relationships with the basically same premise of light, heat, water, soil, natural resources and topographic and geomorphic conditions, while development goal of grain yield is relatively clear (Bai et al., 2013).When agricultural production level has developed to a certain degree, grain yield will certainly transforms to regional development, scale development and centralized development (Deng et al., 2013), per unit area yield of grain in one region will be significantly affected by per unit area yield of grain in surrounding areas (Liu et al., 2011;Deng et al., 2013).Therefore, this study makes comprehensive consideration of LISA diagram for per unit area yield and divides yield-increasing type areas according to cultivation system regions, sets the highest per unit area yield of the county in each type area as its target per unit area yield of the type area and then estimates yieldincreasing potential according to yield gap between actual per unit area yield and target per unit area yield of each county: where, Z p = Equilibrium yield-increasing potential N 1 = The quantity of equilibrium yield increase type areas N 2 = The quantity of counties in type area x kMax = Target per unit area yield of type area k x ik = Actual per unit area yield of County i in type area k M ik = Grain acreage of County i in type area k Data source and data processing: This research reveals spatial-temporal characteristics of per unit area yield with its yield-increasing potential at county scale and secondary regions of cultivation system.After carding, taking the year of 2010 as standard to correct county-level administrative division unit, uniformly merging the area under jurisdiction of prefecture-level cities into urban areas of prefecture-level cities, while 2301 units at county scale in total, not including Hongkong, Macao and Taiwan.Grain yield and grain acreage data at county scale in 1990 comes from Thematic Database for Human-earth System (http://www.data.ac.cn/newzrzy), while relevant data in 2010 comes from statistical yearbooks of provinces, cities and countries.Standard cultivation system refers to farming methods that have been used or will be widely used under current socio-economic level, production condition and technical level; these methods are able to satisfy social needs and good for production or to maximize the productive potential of lands without causing ecological damage.Cultivation system district based on heat, moisture, topography, social and economic conditions as well as other indexes, the whole country is divided into 12 primary regions and 40 secondary regions (Liu, 2002; GB/T 28407-2012, 2012), while Hengduan Mountainous region is regarded as a secondary region because Hengduan Mountainous region in primary region doesn't divide into secondary regions, therefore 41 secondary regions in total, with basically the same natural resources and environment in each secondary region of cultivation system.Grain total output, grain acreage and per unit area yield for each secondary region of cultivation system can be obtained respectively by combining all the data.

RESULT ANALYSIS
Spatial-temporal variation of per unit area yield at county scale: In 2010, total grain output in China was 5.46×10 8 t, increasing by 1.22×10 8 t in 20 years, average annual growth rate was 1.02%; per unit area yield increased from 3932.84 kg/hm 2 in 1990 to 4973.58 kg/hm 2 in 2010, average annual growth rate was 1.18%.Partial correlation analysis was made for relevant data relating to grain from 1990 to 2010 by SPSS 15.5.Correlation coefficients between grain yield, per unit area yield and grain acreage were 0.8648 and -0.3299, respectively, which mean increasing per unit area yield was an effective method to increase grain yield during the period, while decline in grain acreage limited steady increase in grain yield.
As apparent from Table 1 and Fig. 1, 1243 counties' per unit area yield in 1990 were lower than national average value (3932.84kg/km 2 ) in 1990.One hundred and nineteen counties' per unit area yields were equal to or greater than 6000 kg/hm 2 and 605 counties' per unit area yields were between 4500 and 6000 kg/hm 2 .In 2010, average per unit area yield was 4973.58 kg/hm 2 , the highest average yield was 12071 kg/hm 2 , the lowest average yield was 230 kg/hm 2 , while counties with relatively high per unit area yield were dominant.Counties with relatively high per unit area yield were mainly distributed in two kinds of regions: one kind of regions were located in plain area in mideast region with flat terrain, relatively thick soil layer, plentiful water resources and suitable light and heat conditions, these regions have built relatively complete agricultural infrastructure with relatively high per unit area yield based on relatively developed socioeconomic level; the other kind of regions were located in plateau area in Northern China and Western China, such as Northern Xinjiang Irrigation Area, Hetao-Yinchuan Plain, while the counties in these areas were in high elevation with longer illumination time, large temperature difference between day and night, therefore grain crops always had relatively high per unit area yield and good quality with the premise of sufficient water resources for irrigation.Counties with relatively low per unit area yield were centralized distributed in first ladder (Qinghai-Tibet Plateau) and second ladder (excluding Northwest China) with large undulating terrain, thin soil layer, frequent natural disasters, incomplete agricultural infrastructure, as well as low and unstable per unit area yield.
The change characteristics of per unit area yield at county scale between 1990 and 2010 were as follows (Fig. 1):   which mean spatial autocorrelation was significant under confidence value of 99.9%.Local differentiation of per unit area yield manifested as four spatial agglomeration types: HH Type area (LL Type area) represented small spatial differences in per unit area yield at county scale, while both per unit area yields of the county and its surrounding counties were relatively high (low); HL Type area (LH Type area) represented large spatial differences in per unit area yield, per unit area yield of the county was relatively high (relatively low), but per unit area yields of its surrounding counties were relatively low (relatively high), which formed local heterogeneous "hot spot" ("cold spot").
As apparent from Fig. 2a, 1734 counties were distributed in the first quadrant and the third quadrant, which showed significant spatial correlation of per unit area yield in local scope and spatial aggregation structure of high-value (low-value) area of per unit area yield; a part of counties were distributed in HL quadrant and LH quadrant, which showed spatial heterogeneity of per unit area yield of some counties.Combing Moran scatter diagram with LISA significance level to obtain Local Moran's I diagram with significance level of 5%, which in turn contributed to further analysis of spatial autocorrelation degree of per unit area yield.As seen in Fig. 2b, significant HH areas were distributed in Songnen Plain Area, Huang-Huai Plain Area, Hebei-Shandong-Henan Low-lying Plain Area, etc. Significant LL areas were mainly distributed in Hengduan Mountainous Area, Shanxi-Shaanxi Hilly Valley Area, Inner Mongolia Steppe Area, etc.; there were small amount of significant LH (HL) areas with scattered distribution; other counties were not representative.

Distribution characteristics of grain yield between secondary regions of cultivation system:
• In 2010, Songnen Plain Area and Huang-Huai Plain Area were highest-yield areas to produce total grain output and per unit area yield, total grain outputs of Songnen Plain Area and Huang-Huai Plain Area were accounted for 11.91% and 11.66% of total output in China, respectively; while total

DISCUSSION AND CONCLUSION
Per unit area yield of grain increased from 3932.84 kg/hm 2 in 1990 to 4973.58 kg/hm 2 in 2010, per unit area yield of 2046 counties increased, increment of per unit area yield showed the law of "lower in the south and higher in the north" clearly.Spatial agglomeration characteristics of per unit area yield of grain at county scale were significant.Whole country was divided into 41 primary regions and 115 secondary regions based on LISA distribution diagram of cultivation system regions and per unit area yield of grain at county scale, total yield potential in China was 1.77×10 8 t.These results prove to conform with actual situations in China fairly, measurement results based on per unit area yield potentials at county scale and total yield potential will contribute to the distribution of yield-increasing projects and provide the basis for reasonably distribute high-standard basic farmland construction projects and low yield land improvement projects.This study reveals spatial-temporal differentiation characteristics and local homogeneity of change of grain yield per unit area from perspective of inter regional relationship by exploratory spatial data analysis method, which provides a fresh idea to divide yield-increasing type areas scientifically.While, per unit area yield of grain will be affected by a variety of factors, even increasing methods and specific measures to achieve grain yield increase in different kinds of type areas are quite different.Therefore, systematic research on driving mechanism for per unit area yield increase, clarify limiting factors for per unit area yield increase in different kinds of type areas and develop appropriate yield-increasing measures for different kinds of type areas are the issues that need deep exploration.

Fig. 1 :
Fig. 1: Spatial pattern of grain yield per unit area at county level in China Fig. 3: Regionalization of balanced potential to increase grain production in China grain outputs of Yanjiang Plain Area, Hebei-Shandong-Henan Low-lying Plain Area, Jianghuai Plain Area, East Sichuan Low Hilly Mountain Area, Shandong Hilly Area, West China Hilly and Mountainous Area, Yanshan-Taihang Mountain Piedmont Plain Area were accounted for 4% to 8% of total output; total grain outputs of Inner Mongolia Steppe Area, Hexi Corridor Area, Shanxi-Shaanxi Hilly Valley Area, Hengduan Mountainous Area, Houshan Bashang Plateau Area, South Tibet Plateau Valley, Southern Qinghai and Northern Tibet Plateau area, Alxa Plateau and other areas were accounted for 1% of total output in China; increase speeds of 18 areas were faster than national average level during 1990-2010, Songnen Plain Area, Huang-Huai Plain Area, Hebei-Shandong-Henan Low-lying Plain Area, Changbai Mountain Area in Sanjiang Plain, Northern Xinjiang Irrigation Area, Daxing'anling Mountain Area and Xiaoxing'anling Mountain Area and other areas had relatively large growth rates; while increase speeds of 23 areas were slower than national average level, which cause dramatic decline in proportions of total grain outpus of Yanjiang Plain Area, South China Low Plain Area, Eastern China Hilly and Mountainous Area, East Sichuan Low Hilly-mountain Area, West China Hilly and Mountainous Area, West Sichuan Plain and Nanling Hilly Area.•From perspective of per unit area yield, counties of relatively high per unit area yield were mainly distributed in Eastern and Northern China, per unit area yields of Songnen Plain Area, Changbai Mountain Area in Sanjiang Plain, Liaoning Plain