Spatial autocorrelation and its influencing factors of the sampling units in a spatial sampling scheme for crop acreage estimation

摘要

One important feature of geography phenomenon or spatial object is that there exists the spatial autocorrelation between them. As for a spatial sampling scheme for crop acreage estimation, the spatial autocorrelation of the sampling units has an important effect on the design of the sampling scheme and the improvement of the sampling efficiency. While the related researches on spatial autocorrelation have been found in a variety of subject areas like ecology, hydrology, soil, forestry and so on, however, so far, little attentions have been paid to the sampling survey for estimating crop area at a regional scale. The object of this study is to quantitatively evaluate the spatial correlation characteristics between the sampling units for crop acreage investigation, and analyze the factors (e.g. sampling unit size, shape, crop cultivated abundance, cropland slope and fragmentation index) that have the impact on it. In this study, Dehui County in Jilin Province, China was selected as the study area. Systeme Probatoire d'Oberservation de la Terre (SPOT5) imagery from 2009 is used to extract maize and rice area and distribution, the cultivated land map is derived from the Second National Land Survey in China and land slope from the Digital Elevation Model (DEM) data with a spatial resolution of 30m. The global spatial autocorrelation index (Moran's I) was employed to evaluate the spatial correlation of the sampling units, and the impact of the sampling unit size, crop planting abundance, cultivated land fragmentation index on the spatial correlation was analyzed. The results of this study show that the spatial correlation of maize and rice decreases with the sampling unit size increasing, accordingly, Moran's I vary from 0.50 to 0.90. There is a high significant spatial autocorrelation found among those sampling units for maize and rice acreage estimation, due to that the Z-Score varies from 10 to 1000. The correlation coefficients between dryland, paddy land and maize, rice acreage increase with the sampling unit size. It can be also found that the spatial correlation degree of the sampling units in each stratum can be significantly reduced, even nearly to zero, when crop planting abundance or cultivated land fragmentation index is selected as the stratification criterion in the stratified sampling scheme for maize and rice acreage estimation. However, when the land slope is selected as the stratification criterion, the spatial correlation degree in the strata is found that there is no an obvious reduction. In addition, compared with the irregular sampling unit shape (that is township administrative boundary), the Moran's I from the square sampling unit is average 10% more than the former. In this way, this study can provide some approaches for formulating a reasonable sampling scheme and improving the survey efficiency for crop acreage estimation in those areas with a high spatial autocorrelation feature.