European Journal of Soil Science, October 2009, 60, 785–791 doi: 10.1111/j.1365-2389.2009.01170.x Near infrared spectroscopy for soil bulk density assessment C.S. M OREIRA a , D. B RUNET b , L. V ERNEYRE c , S.M.O. S ´ A d , M.V. G ALDOS a , C.C. C ERRI a & M. B ERNOUX b a Laborat´ orio de Biogeoqu´ ımica Ambiental, Centro de Energia Nuclear na Agricultura, Universidade de S˜ ao Paulo, 13416-000 Piracicaba, Brazil, b Institute de Recherche pour le D´ eveloppement- IRD, UMR Eco&Sols INRA-IRD-SupAgro, 2 place Viala, 34060 Montpellier, France, c DIREN Guyane, Service eaux et milieux aquatiques, risques et d´ echets, 97328 Cayenne, Guyane Fran¸ caise, and d Centro de Ciˆ encias Agr´ arias, Universidade Estadual do Maranh˜ ao, 65055-310 S˜ ao Lu´ ıs, Brazil Summary Soil bulk density values are needed to convert organic carbon content to mass of organic carbon per unit area. However, field sampling and measurement of soil bulk density are labour-intensive, costly and tedious. Near-infrared reflectance spectroscopy (NIRS) is a physically non-destructive, rapid, reproducible and low-cost method that characterizes materials according to their reflectance in the near-infrared spectral region. The aim of this paper was to investigate the ability of NIRS to predict soil bulk density and to compare its performance with published pedotransfer functions. The study was carried out on a dataset of 1184 soil samples originating from a reforestation area in the Brazilian Amazon basin, and conventional soil bulk density values were obtained with metallic “core cylinders”. The results indicate that the modified partial least squares regression used on spectral data is an alternative method for soil bulk density predictions to the published pedotransfer functions tested in this study. The NIRS method presented the closest-to-zero accuracy error (−0.002 g cm −3 ) and the lowest prediction error (0.13 g cm −3 ) and the coefficient of variation of the validation sets ranged from 8.1 to 8.9% of the mean reference values. Nevertheless, further research is required to assess the limits and specificities of the NIRS method, but it may have advantages for soil bulk density predictions, especially in environments such as the Amazon forest. Introduction In the past few decades, considerable efforts have been made to quantify carbon (C) dynamics and fluxes in the environment as affected by climate change, rising greenhouse gas levels and global warming. In this context, tropical rain forests such as the Brazilian Amazon represent significant sources/sinks of trace gases and are an important component within the global C cycle (Cerri et al., 2007). The soil is considered the largest terrestrial pool of carbon (excluding carbonate rocks), with 1500–2000 Pg C stored in the upper 100 cm of the soil profile (Batjes, 1996; Garc´ ıa- Oliva & Masera, 2004). Accurate estimates of soil carbon stocks (C t ) depend on the availability of soil carbon content (in g C kg −1 soil) and bulk density (D b ). In order to obtain accurate predictions of D b , labour-intensive and time-consuming field measurements are required (Manrique & Jones, 1991). In addition, field methods for measuring D b are limited in their ability to provide reliable, complete and uniform soil data (Bernoux et al., 1998b). Many conventional soil analytical techniques have been Correspondence: M. Bernoux. E-mail: martial.bernoux@ird.fr Received 22 August 2008; revised version accepted 28 May 2009 used in an attempt to establish the relationship between soil physical and chemical properties and individual soil components, often disregarding their complex, multi-component interactions (Viscarra Rossel et al., 2006). Pedotransfer functions (PTFs) for soil bulk density have been estimated from soil properties in the Amazon basin (clay, sand and silt contents, organic carbon and pH) using a stepwise multiple regression procedure (Bernoux et al., 1998b) and multiple linear regressions (Tomasella & Hodnett, 1998). Near-infrared reflectance spectroscopy (NIRS) offers an alter- native method to assess field soil variation. NIRS is a physi- cally non-destructive, rapid, reproducible and low-cost method that characterizes materials according to their reflectance in the wavelength range between 800 and 2500 nm (Stevens et al., 2006; Brunet et al., 2007). In this region of the electromagnetic spec- trum, each constituent of a complex organic mixture (C, N, H, O, P and S atoms) has unique absorption properties due to stretching and bending vibrations in molecular bonds (Odlare et al., 2005). The technique has been used in the past two decades to assess total carbon and nitrogen, nitrate-nitrogen (N-NO − 3 ), ammonia-nitrogen (N-NH + 4 ), respirometry, pH, cation exchange capacity, Ca, Mg, K,  2009 The Authors Journal compilation  2009 British Society of Soil Science 785