Current Advances in Agricultural Sciences(An International Journal)

Soil-landscape modeling is one of the important researchable issues in soil science. Literature is scanty on such models especially using low intensity sampling and such models are unknown in rubber (Hevea brasiliensis) growing areas in Kerala. Prior visualisation of study area using a digital elevation model (DEM) with 90 mresolution was found useful in selecting the sites for soil profile excavation. Eleven soil profiles were cut and characterized and the resultant regression model i.e. actual soil volume (ASV) =-0.091 + 23.9 depth – 0.99 coarse fragments – 0.98 porespace (adjusted R² = 0.999) indicated a very good relation. Predictive maps were generated with data on 9 soil profiles using inverse distance weighting (IDW) method. The predictability, as tested by two samples i.e. soil profile 3 and 9 (which were excluded from model development), was 94, 89 and 82%, respectively for depth, coarse fragments and slope. Similarly, the error in prediction of ASV was 18 and 19% for soil profile 3 and 9, respectively. Nevertheless, the actual and predicted values placed the soils in the same class of slope and depth thus making the error in prediction ignorable. However, there are no such ranges identified to describe ASV and the coarse fragments with weighted means. The results of low intensity soil sampling (one soil profile per 4.4 ha) and interpolation with IDW are encouraging and this technique can help in generating predictive maps of similar such terrains at a lesser cost with acceptable error.