Author(s): Vinay HT, Mallikarjun B Hanji, Jagadeesh MS, V Ramamurthy, Mohan Kumar TL and KN Krishnamurthy

Abstract: The set of management practices and soil behavior highly depend on soil characteristics which are most variable in nature. Soil behaviour helps in measuring soil performance for growing crops, which in turn will help the farmers to make decisions about the crops to be grown. In order to predict soil behaviour and to know the appropriate multivariate statistical methods to classify soil samples of varied kind, the knowledge of soil grouping based on similar characteristics is necessary. To this end, the data on soil samples was collected from ICAR-NBSS & LUP Regional Center, Bengaluru, Karnataka. Gower’s distance metric was employed in the present study to get distance matrix of the data which are both quantitative and qualitative in nature. By Elbow method, the ideal number of clusters was found to be three. Clustering of soil samples was done by employing five techniques namely Single linkage, Complete linkage, Average linkage, Ward’s method and K-Medoid methods. The classified soil samples were validated using Dunn, CH and Silhouette Index. In the present study among clustering techniques, K-Medoid method had the highest value for Dunn Index (1.29), CH Index (42.98) and Silhouette Index (0.13). Similarly, soil classification based on this technique was found to be realistic in taking decisions for land use planning. Hence, this method was considered to be the better method for classifying varied nature of soil samples for efficient agricultural land use planning.

DOI: 10.22271/maths.2024.v9.i4a.1770

Pages: 61-68 | Views: 114 | Downloads: 5

Download Full Article: Click Here