Estimation and Mapping of Vertisols Soil Nutrients by Geostatistics and Remote Sensing Approach

Abstract

he status of soil fertility is a concern, especially in the Deccan plateau vertisols of India. Vertisols are productive if they are managed well. Understanding the spatial variability of soil nutrients is necessary for agriculture to maintain sustainability. The objective of the present study is to characterize the status of soil nutrients, spatial variability of selected soil nutrients, and the estimation of the presence of these soil nutrients by spaceborne Hyperion data in scattered small-size fields of Gulbarga taluk, northern Karnataka, India. This region is known as the "pigeon pea vessel" of the state. The geostatistical analysis is carried out in SpaceStat 4.0® to find the spatial variability of all the selected nutrients. The coefficient of variation monitors the variation in the nutrients of the soil. The variogram analysis has shown that all the selected nutrients are the best fit for the spherical model except nitrogen, organic carbon, and phosphorus. The nugget/sill ratio is utilized to know the spatial dependence of soil nutrients. Using the best fit model, surface maps are generated by the ordinary kriging method. The estimation of soil nutrients from Hyperion data with statistical regression is measured as an alternative technique. The spectral information of the visible near infrared and short wave infrared range (400-2500 nm) is utilized to characterize soil nutrients. The potential of the Hyperion data has not yet been exploited completely due to noisy atmospheric components in spectral signatures especially in fields of smaller size. Sixty-eight random topsoil samples were collected from small farms, which are less than two acres in size. The systematic sampling of soil was conducted in the month (third week) of November 2016. This duration is also synchronized with the passage of the Hyperion satellite above the study area. The atmospheric (FLASSH) and geometric corrections is carried out and then the spectral reflectances are extracted. The PLS_Toolbox is used for filtering (Savitzky Golay), and the Partial Least square regression (PLSR) technique is applied for the estimation of soil nutrients by Hyperion data. The variable importance in projection (VIP) is identified, which reduces the non-significant wavelengths for the PLSR model. Two indices are ii used to assess the prediction accuracy, Coefficient of determination (R2), and root mean square error (RMSE). From analysis of soil nutrients, it is observed that the spatial variability maps exhibited a heterogeneous pattern of soil nutrients because of individual farming methods. The spatial variability maps are used as initial regulation by policymakers for site nutrient management, including fertilization in vertisols. This is essential for sustainable management of the fields, which are aimed at increasing the productivity of the crops; low productivity vertisols are to be used in cultivation on a global scale due to the current shortage of food supplies and agricultural resources land. The utilization of Hyperion data and PLSR technique showed that it has the low to moderate potential to estimate certain vertisols nutrients such as iron (R2=0.40), potassium (R2=0.45), and Copper (R2=0.41), and moderate estimation for nitrogen (R2=0.54) even though vertisols have less reflectance values compared to other soil types. The vertisols of India exhibit low reflectance, which are deficient in humus, nitrogen, phosphorus, and potassium due to low permeability and moisture stress throughout the drought. Hence the presence of soluble nutrients concentration is low compared to other soil. Generally, the white color contributes to higher reflectance in all wavelengths, so the grey-brown color is natural in the vertisols fields and along with less organic matter, which leads to the low reflectance. Hyperion data can be inventively utilized to estimate vertisols soil nutrients with reasonable accuracy in heterogeneous and small size fields.