215-9 Transferability of Visible/Near-Infrared Diffuse Reflectance Spectroscopic Models of Critical Soil Fertility Properties Between Two Small-Holder Agricultural Villages in Southern India.

See more from this Division: ASA Section: Agronomic Production Systems
See more from this Session: Technologies for Determining Nutrient Needs and Improving Nutrient Use Efficiency: Graduate Student Competition
Tuesday, November 4, 2014: 3:15 PM
Long Beach Convention Center, Seaside Ballroom A
Share |

Christopher M Clingensmith1, Sabine Grunwald1 and suhas wani2, (1)Soil and Water Science Department, University of Florida, Gainesville, FL
(2)ANDHRA PRADESH, INDIA, ICRISAT, PATANCHERU, INDIA
The demand for quantitative soil data has increased with the need for land resource assessments, which is especially important in food production. Soil maps prepared traditionally require the collection and analysis of numerous soil samples, which is time consuming, labor intensive, and costly. For small-holder farmers in developing nations, this type of information is far out of reach but urgently needed to sustain soil quality, food security and livelihoods. Diffuse reflectance spectroscopy offers a rapid and cheap alternative to conventional soil analysis that could be employed in developing countries. We assert that once a stable spectral soil model has been built, it can be transferred to another similar region to produce soil property predictions with little analysis and low cost.

The objectives were to (i) apply chemometrics to infer the concentrations of a suite of soil properties and (ii) assess the transferability of locally derived chemometric models. We utilized VIS-NIR diffuse reflectance spectroscopy as a means of rapid soil analysis. Soil sampling was performed over two small-holder agricultural villages in southern India, Kothapally and Masuti, where 255 and 257 soil samples were collected, respectively. The soil samples were analyzed for texture, soil organic carbon, pH, electrical conductivity, and soil macro- and micronutrients. Sieved soil subsamples were scanned in the VIS-NIR range (350 – 2,500 nm) at 1 nm resolution. Partial least squared regression was employed to produce prediction models that were subjected to rigorous validation analysis to identify the best performing model for each soil property. The models selected for one village were then used to predict properties in the second village using its spectral data. The results not only confirm that VIS-NIR soil prediction models are a viable alternative to traditional soil analysis but that locally constructed models can be transferred to areas with similar conditions and produce satisfactory predictions.

See more from this Division: ASA Section: Agronomic Production Systems
See more from this Session: Technologies for Determining Nutrient Needs and Improving Nutrient Use Efficiency: Graduate Student Competition