JIA-2018-09

2098 WANG Di et al. Journal of Integrative Agriculture 2018, 17(9): 2096–2106 (Frutos et al. 2007; Le Rest et al. 2013), fishery resources (Jardim and Ribeiro 2007), forest community structures (Gilbert and Lowell 1997) and land use (Overmars et al. 2003). There are no reports on the spatial autocorrelation of crop acreage including the sampling units, moreover, factors affecting the spatial autocorrelation of crop acreage have not previously analyzed. This complicates the application of spatial sampling in crop acreage monitoring in conjunction with remote sensing. The main objectives of this study were to: (i) investigate the spatial autocorrelation among crop acreage included in sampling units of different sizes; (ii) analyze the effects of different stratification criteria on the spatial autocorrelation level of crop acreage within the sampling units; (iii) suggest reasonable stratification criteria and sampling unit sizes for spatial sampling scheme to estimate crop acreage. 2. Materials and methods 2.1. Analysis process The analysis process consisted of seven steps. The first step was the preparation of the basic data used to formulate the spatial sampling scheme for crop acreage estimation. Step two was the selection of the spatial autocorrelation index. The global Moran’s I was used to evaluate the degree of spatial autocorrelation of crop acreage within the sampling units. The third step was the design of the sampling unit sizes and the construction of sampling frames with different unit sizes. Step four was the analysis of spatial autocorrelation of crop acreage within the different sized sampling units. The fifth step was the formulation of the stratified sampling strategy, including the design of the stratification criteria, strata number and intervals. Step six was the spatial autocorrelation analysis of crop acreage in the sampling units belonging to different strata. The last step was the optimization of the stratification criteria and sampling unit size in the spatial sampling scheme for crop acreage estimation. Fig. 1 shows the overall flowchart for this study. 2.2. Study area Dehui County in Jilin Provine, China, was chosen as the study area because of the availability of high-quality crop distribution, cultivated land and digital elevation model (DEM) data covering the whole region, at similar times. Moreover, the topography and spatial distribution of crops that are found in Dehui County are representative of a wider area and, therefore, indicative for analysis of spatial correlation of variables used in defining sampling units for crop acreage estimation. Dehui County is located in the north central part of Jilin Province, Northeast China (44°02´–44°53´N latitude, 125°14´–126°24´E longitude), with a total land area of 3460 km 2 (Fig. 2). It includes a total cultivated land area of 2540 km 2 . It has a sub-humid continental monsoon climate with four distinct seasons. The annual average temperature is 4.4°C, the average annual hours of sunshine are 2695.2 h, the average annual frost-free period is 139 days, and the average annual precipitation is 520 mm. Within the agricultural sector, priority is given to food crops, while cash crops are complementary. Food crops include maize, rice, soybean and sorghum. Maize and rice are two particularly important food crops in Dehui County. 2.3. Data preparation First, the administrative boundary (from 1:10000 mapping) of Dehui County was acquired to define the extent of the sampling survey for maize and rice acreage estimation and to help construct the spatial sampling frame. Second, cultivated land plots covering the study area were extracted from the Second National Land Survey, finished by China in 2009, and used to calculate plot fragmentation. Third, maize and rice growing areas were extracted from three Systeme Probatoire d’Observation de la Terre (SPOT-5) images (track numbers: 296-260, 296-261 and 297-260, date: 08-01-2009, spatial resolution: 10 m), to determine the planting intensity within each sampling unit. Finally, for the study area, we downloaded the freely available ASTER Global Digital Elevation Model (GDEM) (version 2) with a resolution of 30 m in GeoTIFF format, and from this computed the ground slope within each sampling unit. Figs. 3 and 4 show the spatial distributions of cultivated land plots, and maize and rice distributions. Fig. 5 shows the land elevations (from the GDEM). 2.4. Selection of spatial autocorrelation index The global spatial autocorrelation index (Moran’s I ) quantitatively describes the spatial dependence existing between variable values, and has been widely used to evaluate the spatial autocorrelation of variables of interest. Compared with Geary’s C , the distribution characteristics of Moran’s I are easier to depict with an estimator (Cliff and Ord 1981). Hence, we selected this method to evaluate the spatial autocorrelation of the sampling units. Moran’s I ranges from −1 to +1. The closer a value of Moran’s I is to +1, the stronger is the positive spatial autocorrelation of the variable being analyzed among the sample plots. When the value of Moran’s I is equal to or approximately zero, there is no spatial autocorrelation and the values of the analyzed variable in the sampling units are random (not spatially