Establishment of soil management zones based on spatial variability of soil properties for precision agriculture using Gis in Katumani, Machakos district of Kenya

Abstract

Analysis and interpretation of spatial variability of land/soil properties is a key-stone in precision agriculture. Spatial variability is a problem in soil testing because mixing soil cores from high- and low-fertility areas creates a soil sample that does not represent either area because each point in a field is unique. Precision agriculture needs to be employed in solving this problem, which involves the use of Global Positioning System (GPS) and Geographic Information Systems (GIS). However, reliable soil information required for precision farming planning is scarce, particularly in Kenya. Therefore a more detailed soil study of a part of Kenya Agricultural Research Institute (KARl) Katumani Research Centre was conducted with the principal objectives of (a) characterizing spatial variability of soil properties (b) determining relationships between spatial variability of soil fertility and their determining factors and (c) evaluating Gridpoint versus Grid-cell soil sampling schemes for precision farming. Spatial variability and soil management zone maps were presented each at scale 1: 6,000 and the results of soil properties and their determining factors relationship were presented in bar charts. GPS was used to geo-reference the sampled locations, which were based on the gridpoint and grid-cell soil sampling schemes. Physical and chemical properties of the soil samples were analyzed using the standard procedures. Spatial analyst extension of Arcview GIS software was used for the GIS analysis. Genstat was used for statistical analysis, where one-way analysis of variance to test the significance of the studied factors. Spatial analysis by calculating variogram statistics was used to test spatial variability of soil properties. Generally soil fertility data showed variability where, soil phosphorus and percentage clay content were highly variable with 888.0292 and 115.3625 general vanances respectively. Total nitrogen was the least variable with general variance of 0.0022. Present land use, vegetation cover and soil texture were the major factors influencing soil phosphorus, total nitrogen and soil pH distributions in the study area respectively. This is because they were all significant at P<0.05. Micro-relief had no effect on the soil fertility within the study area by depicting unusual relationship with the exception of phosphorus, which did not show significance at P< 0.05. Organic matter was influenced by vegetation cover and not by slope due to its unusual relationship, although was significant at P< 0.05. Phosphorus was high in cultivated area than in grazing area. This was probably due to application of phosphatic fertilizers. Soil reaction (pH) was found to be low in loamy sand soils than in the sandy clay, due to high rate of leaching of basic nutrients and low organic matter hence low buffering capacity in loamy sand soils. Generally, the study area had relatively acid soils due to the presence of gneiss parent material. Vegetation cover played a major role in the distribution of soil nitrogen in the area hence high levels were found in areas with thick herbs and shrubs. By observation, grid-point sampling scheme was a better strategy in characterizing spatial variability of soil properties in the area than the Grid-cell sampling scheme. This is because variability was precisely defined within very short distances than in Grid-cell sampling scheme, which generalized variability in each cell. This was also supported statistically where grid-point sampling scheme gave higher general variances for all land/soil properties than Grid-cell sampling scheme. Soil management decisions would therefore be based on the developed soil management zones for precision agriculture where inputs should be matched with the site potential to maximize crop yields while minimizing excess usage of crop inputs.