When precision fanning management zones （MZs） are delineated in an agricultural field for precision nutrient management, unsupervised classification and cluster analysis procedures using remote sensing image analysi...When precision fanning management zones （MZs） are delineated in an agricultural field for precision nutrient management, unsupervised classification and cluster analysis procedures using remote sensing image analysis software are performed. These unsupervised classification and cluster analysis procedures are performed on the basis of the assumption that grouping of data points into naturally occurring clusters reduces within zone variability. The problem is that, there are small patches of different soil types within each management zone that are regarded as insignificant by the farmer, and are assimilated within larger MZs. These will consequently make soils within a management zone to be inhomogeneous. The objective of this study was to determine the probability of soil sampling occurrences on patches assimilated during delineation of MZs after a cluster analysis was performed. The study was conducted on a 5.0 ha （25°05′34.46″ S and 28°18′30.01″ E） and a 24.4 ha （23°59′04.61″ S and 28°52′29.43″ E） fields in the Waterberg District of the Limpopo Province in South Africa. A bare-soil high resolution Quickbird satellite imagery of a conventionally tilled agricultural field was used to develop MZs in the field. Soils were sampled using systematic unaligned sampling on a 35.0 m and 30.0 m grids for the 24.4 ha and 5.0 ha fields, respectively. Probabilities were calculated based on percentage area assimilated during the cluster analysis procedure that was performed using remote sensing image analysis software. The results indicated that in the 24.4 ha field there were 2.5 ha patches of high and medium zones that were assimilated within the low zone, and thus making low zones non-homogeneous. After cluster analysis and assimilation of patches, the low zone in the 24.4 ha field increased by 45.5% （2.5 ha） while the high zone was 16.4% （2.4 ha） smaller in size. In the smaller field of 5.0 ha, the high zone, which was originally 3.20 ha, lost 0.37 ha （11.6%）, which was assimilated in either low or medium zone. The study indicates that unequal probability proportional to size sampling could be used to minimize error when sampling across precision farming MZs because typically the low, medium and high MZs are not of equal size and do not contribute equally towards the mean values of soil samples.展开更多
文摘When precision fanning management zones （MZs） are delineated in an agricultural field for precision nutrient management, unsupervised classification and cluster analysis procedures using remote sensing image analysis software are performed. These unsupervised classification and cluster analysis procedures are performed on the basis of the assumption that grouping of data points into naturally occurring clusters reduces within zone variability. The problem is that, there are small patches of different soil types within each management zone that are regarded as insignificant by the farmer, and are assimilated within larger MZs. These will consequently make soils within a management zone to be inhomogeneous. The objective of this study was to determine the probability of soil sampling occurrences on patches assimilated during delineation of MZs after a cluster analysis was performed. The study was conducted on a 5.0 ha （25°05′34.46″ S and 28°18′30.01″ E） and a 24.4 ha （23°59′04.61″ S and 28°52′29.43″ E） fields in the Waterberg District of the Limpopo Province in South Africa. A bare-soil high resolution Quickbird satellite imagery of a conventionally tilled agricultural field was used to develop MZs in the field. Soils were sampled using systematic unaligned sampling on a 35.0 m and 30.0 m grids for the 24.4 ha and 5.0 ha fields, respectively. Probabilities were calculated based on percentage area assimilated during the cluster analysis procedure that was performed using remote sensing image analysis software. The results indicated that in the 24.4 ha field there were 2.5 ha patches of high and medium zones that were assimilated within the low zone, and thus making low zones non-homogeneous. After cluster analysis and assimilation of patches, the low zone in the 24.4 ha field increased by 45.5% （2.5 ha） while the high zone was 16.4% （2.4 ha） smaller in size. In the smaller field of 5.0 ha, the high zone, which was originally 3.20 ha, lost 0.37 ha （11.6%）, which was assimilated in either low or medium zone. The study indicates that unequal probability proportional to size sampling could be used to minimize error when sampling across precision farming MZs because typically the low, medium and high MZs are not of equal size and do not contribute equally towards the mean values of soil samples.
