Soil landscape characterization into landform elements for precision agriculture has become an important issue. As soil properties and crop yields change over the landscape, delineating landform elements as a basis for site-specific application of crop inputs has become a reality. Two different methods of delineating landform elements from agricultural fields were tested and compared. The first method delineated landform elements from digital elevation maps with the use of the LandMapR(tm) software, the second method delineated classes from IKONOS high resolution panchromatic images using an unsupervised classification algorithm. The LandMapR(tm) model delineated landform elements from true elevation data collected in the field and was considered the reference dataset to which the image classification maps were compared to. The IKONOS imagery was processed using a combination of one filtering algorithm and one unsupervised classification method prior to being compared to the classified DEM. A total of 20 filtering algorithms and two unsupervised methods were used for each of the five study sites. The study sites consisted of four agricultural fields covered with crop stubble and one field in summer fallow. Image classification accuracy assessment was reported as overall, producer’s and user’s accuracy as well as Kappa statistic. Results showed that filtering algorithms and classification methods had no effects on image classification accuracies. Highest classification accuracy of image map to landform element map comparison achieved for all study sites was 17.9 %. Classification accuracy was affected by the heterogeneity of the ground surface cover found in each field. However, the classification accuracy of the fallow field was not superior to the stubble fields. / May 2006

Soil landscape characterization into landform elements for precision agriculture has become an important issue. As soil properties and crop yields change over the landscape, delineating landform elements as a basis for site-specific application of crop inputs has become a reality. Two different methods of delineating landform elements from agricultural fields were tested and compared. The first method delineated landform elements from digital elevation maps with the use of the LandMapR(tm) software, the second method delineated classes from IKONOS high resolution panchromatic images using an unsupervised classification algorithm. The LandMapR(tm) model delineated landform elements from true elevation data collected in the field and was considered the reference dataset to which the image classification maps were compared to. The IKONOS imagery was processed using a combination of one filtering algorithm and one unsupervised classification method prior to being compared to the classified DEM. A total of 20 filtering algorithms and two unsupervised methods were used for each of the five study sites. The study sites consisted of four agricultural fields covered with crop stubble and one field in summer fallow. Image classification accuracy assessment was reported as overall, producer’s and user’s accuracy as well as Kappa statistic. Results showed that filtering algorithms and classification methods had no effects on image classification accuracies. Highest classification accuracy of image map to landform element map comparison achieved for all study sites was 17.9 %. Classification accuracy was affected by the heterogeneity of the ground surface cover found in each field. However, the classification accuracy of the fallow field was not superior to the stubble fields.

Soil landscape characterization into landform elements for precision agriculture has become an important issue. As soil properties and crop yields change over the landscape, delineating landform elements as a basis for site-specific application of crop inputs has become a reality. Two different methods of delineating landform elements from agricultural fields were tested and compared. The first method delineated landform elements from digital elevation maps with the use of the LandMapR(tm) software, the second method delineated classes from IKONOS high resolution panchromatic images using an unsupervised classification algorithm. The LandMapR(tm) model delineated landform elements from true elevation data collected in the field and was considered the reference dataset to which the image classification maps were compared to. The IKONOS imagery was processed using a combination of one filtering algorithm and one unsupervised classification method prior to being compared to the classified DEM. A total of 20 filtering algorithms and two unsupervised methods were used for each of the five study sites. The study sites consisted of four agricultural fields covered with crop stubble and one field in summer fallow. Image classification accuracy assessment was reported as overall, producer’s and user’s accuracy as well as Kappa statistic. Results showed that filtering algorithms and classification methods had no effects on image classification accuracies. Highest classification accuracy of image map to landform element map comparison achieved for all study sites was 17.9 %. Classification accuracy was affected by the heterogeneity of the ground surface cover found in each field. However, the classification accuracy of the fallow field was not superior to the stubble fields.