Abstract
Sensed and soil sample data are used in two approaches for mapping soil properties in precision agriculture: management zone (MZs) and contour maps. This is the second paper in a two-part series that focuses on contour maps. Detailed and accurate contour maps of soil properties for precision agriculture are often costly to produce because of the large sampling effort required. Such maps or those of sensed ancillary data are often simplified to represent MZs. This research investigated the accuracy of detailed maps of soil properties produced inexpensively from sensed data by transforming them to z-scores. The z-scores of ancillary values are then transformed to values of soil variables using the mean and standard deviation of a small soil data set. The errors from this mapping approach are examined with historic soil data from three field sites with different scales of spatial variation in the United Kingdom. Errors from the conversion of z-scores of sensed data to soil variable ranges are compared with those from MZ averages (Paper I in this series). For soil properties with a moderate relation to ancillary data, the errors related to the z-score conversion were small irrespective of sample size. The root mean squared errors associated with the MZ mean rather than values from the digital map were generally smaller except when sample size was very small. The results suggest that when the scale of variation is small and more samples are required to define MZs, calibrating z-scores of sensed ancillary data may provide better MZ averages than sampling on a grid; it also provides a detailed map of spatial variation within the field. The z-score conversion approach is less sensitive to sample size and captures small features of the variation compared to the standard 100 m grid sampling to determine MZ averages.
Abstract
Sensed and soil sample data are used in two approaches for mapping soil properties in precision agriculture: management zone (MZs) and contour maps. This is the second paper in a two-part series that focuses on contour maps. Detailed and accurate contour maps of soil properties for precision agriculture are often costly to produce because of the large sampling effort required. Such maps or those of sensed ancillary data are often simplified to represent MZs. This research investigated the accuracy of detailed maps of soil properties produced inexpensively from sensed data by transforming them to z-scores. The z-scores of ancillary values are then transformed to values of soil variables using the mean and standard deviation of a small soil data set. The errors from this mapping approach are examined with historic soil data from three field sites with different scales of spatial variation in the United Kingdom. Errors from the conversion of z-scores of sensed data to soil variable ranges are compared with those from MZ averages (Paper I in this series). For soil properties with a moderate relation to ancillary data, the errors related to the z-score conversion were small irrespective of sample size. The root mean squared errors associated with the MZ mean rather than values from the digital map were generally smaller except when sample size was very small. The results suggest that when the scale of variation is small and more samples are required to define MZs, calibrating z-scores of sensed ancillary data may provide better MZ averages than sampling on a grid; it also provides a detailed map of spatial variation within the field. The z-score conversion approach is less sensitive to sample size and captures small features of the variation compared to the standard 100 m grid sampling to determine MZ averages. Leer más
Related posts:
Uncertainty resulting from constant bulk density assumption when interpreting soil nutrient concentrations
The misuse of permanganate as a quantitative measure of soil organic carbon
Field-scale digital mapping of top- and subsoil Chernozem properties
Predictive soil health indicators across a boreal forest to agricultural conversion gradient
Coherent Change Detection for High-Resolution Drone-Borne SAR at 24 GHz
Machine learning approach for satellite-based subfield canola yield prediction using floral phenology metrics and soil parameters