Biosystems Engineering (2005) 91 (3), 305–312 doi:10.1016/j.biosystemseng.2005.04.015 PA—Precision Agriculture Spectral Phosphorus Mapping using Diffuse Reflectance of Soils and Grass I. Bogrekci; W.S. Lee Agricultural and Biological Engineering, Frazier Rogers Hall, PO Box 110570, IFAS, University of Florida, Gainesville, FL 32611-0570, USA; e-mail of corresponding author: bogrekci@ufl.edu (Received 22 October 2004; accepted in revised form 21 April 2005; published online 17 June 2005) Phosphorus (P) concentration was determined from reflectance spectra of grass and soils in a total of 150 samples of each collected from three different sites in the Lake Okeechobee drainage basin. The reflectance spectra of both fresh and dried samples for grass and soil were measured in the ultraviolet (UV), visible (VIS), and near infrared (NIR) regions from 225 to 2550nm with an interval of 1nm. Phosphorus concentrations of the samples were correlated with the absorbance of the same samples. Two-thirds of both vegetation and soil samples were used for calibration and the remaining one-third of vegetation and soils were used for validation. Stepwise multiple linear regressions (SMLR), and partial least-squares (PLS) analyses were applied to the data sets in order to predict P concentrations for soil and grass. Actual and predicted P concentration maps of the fields for vegetation and soil were plotted. Strong relationships (coefficient of determination R 2 ¼ 0778; 0914, and 0922) in PLS for the validation data sets were obtained between absorbance and P concentrations in soils. However, a weak relationship ðR 2 ¼ 0425Þ in PLS for the validation data set was produced from absorbance and P concentrations in vegetation samples. Spatial variation in actual and predicted maps showed that P variability could be represented using diffuse reflectance spectroscopy in the UV, VIS, and NIR regions. r 2005 Silsoe Research Institute. All rights reserved Published by Elsevier Ltd 1. Introduction Reflected light from soil and vegetation carries physical and chemical information about the material interacted. This nature of electromagnetic radiation has been studied by many researchers. Most spectroscopy studies on vegetation and soils have focused on visible and near infrared regions of the electromagnetic spectrum due to many reasons such as cost, availability, and sensitivity. However, in this study, the ultraviolet (UV) region was added to explore more possibilities of developing better P prediction models. Nitrogen (N) and phosphorus (P) content in spring barley was investigated using hyperspectral line scanning in visible range (Christensen et al., 2004). Near infrared (NIR) reflec- tance spectroscopy was used for the determination of P concentrations in sugarcane leaves (Chen et al., 2002). Soil reflectance measurements have been used to predict different soil properties: P and potassium (K) for different soil orders (Lee et al., 2003), soil moisture and organic matter (Varvel et al., 1999; Hummel et al., 2001), and soil mineral N (Ehsani et al., 1999). Ehsani et al. (1999) studied an NIR technique for the rapid determination of soil mineral nitrogen. The conclusion drawn from their study was that the NIR response of soil in the wavelength range from 1800 to 2300nm could be used to determine nitrate content of soil successfully. Varvel et al. (1999) worked on the relationships between spectral data from an aerial image and soil organic matter as well as P levels. Few image acquisition problems were reported from their study and similarities were found between surface maps of organic matter and Bray-1 P. A correlation coefficient of 057 was reported between soil organic matter and Bray-1 P. Bray 1 is a soil testing method that extracts ‘plant available’ P from soils using acid-fluoride reagent. Lee et al. (2003) conducted a research to estimate chemical properties of Florida soils using visible (VIS) and NIR spectroscopy. They produced prediction models to estimate pH, organic matter (OM), P, K, calcium (Ca), and magnesium (Mg) concentrations of three representative soil orders in Florida, referred to Alfisol, Entisol, and Ultisol. ARTICLE IN PRESS 1537-5110/$30.00 305 r 2005 Silsoe Research Institute. All rights reserved Published by Elsevier Ltd