La spectrométrie infrarouge : un outil pour la prédiction des constituants minéraux des sols tropicaux

In highly weathered tropical soils, the dominant residual minerals, i.e. kaolinite, gibbsite, hematite and goethite, play a key role with organic matter, in soil behaviour through their effect on the sorption of nutrients, such as phosphorus (Eberhardt et al., 2008) or exchangeable cations and exchangeable aluminium (Vendrame et al., 2013). Information on soil mineralogical properties is important for effective and sustainable soil fertility management, basis for sustainable productivity. It is of utmost importance in sub-Saharan Africa (SSA), as well as in other tropical regions, where agricultural intensification becomes critical and imperative to overcome the prevailing food insecurity (Nziguheba et al., 2015). There is a great and urgent need for cost and time effective and relatively simple soil analytical methods or procedures that enable local institutions and authorities to obtain relevant information, such as soil quality index for different soils or agroecological regions, that support decision making for sustainable management for agriculture. Soil analyses using traditional laboratory methodologies are generally expensive and often time consuming, restricting their use. The acquisition of large datasets on soil needs other technologies, such as sensing techniques (aircraft, satellite, on-the-ground spectroscopy etc.), to easily and accurately measure soil properties (Minasny and Hartemink, 2011). Infrared reflectance spectroscopy is accepted as a fast and nondestructive method to evaluate the components and properties of soils, and is considered as a possible alternative to improve or replace the conventional laboratory methods of soil analysis (Janik et al., 1998; Stenberg et al., 2010). Spectroscopy has also advantages over some of the conventional techniques, e.g. it is rapid, less expensive and more environmentally friendly (Viscarra Rossel et al., 2006b). Extensive literature exploits spectroscopy to predict soil components, soil organic matter and minerals, using spectroscopic techniques (Cécillon et al., 2009; Nocita et al., 2015; Shepherd and Walsh, 2002; Soriano-Disla et al., 2014; Viscarra Rossel et al., 2006b). For tropical soils, a large number of studies deal with the prediction of soil organic carbon as a result of the interest in carbon sequestration. Numerous studies have reported accurate predictions of soil organic matter content (Brunet et al., 2007; Madari et al., 2005; Móron and Cozzolino, 2003; Viscarra Rossel et al., 2006b). The quantitative prediction of the mineralogical composition of soils using spectroscopic techniques is, however, still limited. Nevertheless, spectral signatures for soil minerals in the visible, near- and mid-infrared regions, for clay and iron oxides, are Chapitre 2. La spectrométrie infrarouge : un outil pour la prédiction des constituants minéraux des sols tropicaux 49 numerous (Farmer and Russell, 1964; Soriano-Disla et al., 2014). The lack of useful prediction models is partly due to the fact that conventional analytical data are often unavailable due to the impeding cost of mineralogical analyses that require sophisticated equipment not available in most soil laboratories. However, the type, the proportion, and the concentration of soil minerals determine soil properties, such as cation exchange capacity or phosphorus sorption, and they are important for soil classification (Soriano-Disla et al., 2014), especially in the case of tropical soils. To date, only a limited number of studies have attempted to quantitatively predict the mineralogical composition of tropical soils with infrared spectrometry (Soriano-Disla et al., 2014; Vendrame et al., 2012). It has been shown that the quantification of iron oxides as well as the distinction between hematite and goethite are efficient (Ben-Dor et al., 2006; Viscarra Rossel et al., 2009) and successfully applied to soil surveys (BenDor et al., 2009). For kaolinite and gibbsite, the two most important clay particles of highly weathered tropical soils, Madeira et al. (1995) proposed a methodology using specific peaks of kaolinite and gibbsite, produced by vibrations of hydroxyl ions (OH- ) in their crystal lattice, situated between 2 200 and 2 300 nm. More recently, Vendrame et al. (2012) reported quantitative predictions for the mineralogical content of Brazilian soils using NIRS and chemometric methods. Coefficient of determination and ratio of performance to deviation (RPD) values as high as 0.86 and 2.5 (gibbsite) and 0.83 and 2.2 (kaolinite) were obtained from a diverse set of soils. More research about the use of infrared reflectance spectroscopy are still needed to provide quantitative analyses of the mineralogical composition of soils (Soriano-Disla et al., 2014), a key component for highly weathered tropical soils. The objective of this study was to investigate the use of near-infrared reflectance spectroscopy (NIRS) to estimate the mineralogical composition over a wide range of highly weathered soils in Madagascar. Two spectroscopic methods were compared: a method that use the entire near infrared spectra as input data and another that is based on diagnostic absorption peaks of minerals.

Materials and methods

Study sites

The study areas were located in different sites in Madagascar, corresponding mainly to the crystalline basement of the island (Fig. 1.). This basement is formed of strongly metamorphosed Precambrian meta-sedimentary units and is intruded by various 0 granitic, mafic and mafic-ultramafic rocks (Collins and Windley, 2002). The climate of the island is subtropical, characterized by a mean annual rainfall of more than 3 000 mm along the East Coast to less than 1 000 mm in the south-east region, and varying from 1 000 to 1 800 mm in the highlands. The mean annual temperature depends largely on the altitude, being above 25°C on the coasts and below 20°C in the highlands. The studied soils were classified as Ferralsols, Cambisols and Nitisols (IUSS Working Group WRB, 2006) and were mostly classified as ferralitic soils in the former French classification system (Commission de Pédologie et de Cartographie des Sols; CPCS, 1967). The selection of the soils was based on field observations, i.e. the presence of a ferralitic horizon, resulting from long and intense weathering. According to Bond et al. (2008), the central plateau of Madagascar is dominated by grasslands and savannas with, for example, Aristida rufescens, Loudetia simplex, Trachypogon spicatus, Hyperthelia dissoluta, Ctenium concinnum, whereas the west is dominated by herbaceous savannas represented by species like Heteropogon contortus, Hyparrhenia spp., Loudetia spp., Themeda quadrivalvis, Panicum spp. 1 Figure 1. Sampling sites localization in Madagascar. The island is predominantly formed by strongly metamorphosed Precambrian meta-sedimentary units (crystalline basement) and Mesozoic sedimentary rocks. 2 2.2 Soil sampling and reference analyses Soil samples were collected at 120 sites close to the main roads (Fig. 1.) at the top or upper third part of the hills. The sites selected were at least 100 m away from the road, and were relatively undisturbed by human activities, i.e. grasslands used by farmers for extensive grazing. At each of the 120 sites (Fig. 1.), composite samples were taken at 0- 0.1, 0.1-0.2, 0.2-0.3, 0.5-0.6 and 0.8-0.9 m depth, using an auger (Edelman auger), resulting in 600 soil samples. The spectra of the 600 samples were acquired, and 148 of those were selected from all the five soil horizons, according to their spectral representativeness, and analyzed by reference methods. Iron oxides were determined by the citrate-bicarbonate-dithionite (CBD) deferrification method (Mehra and Jackson, 1960). One gram (1 g) of crushed (200 µm) soil was placed in a centrifuge tube and mixed with 1 g of dithionite and 50 mL of sodium citrate (78.43 g L-1) and sodium bicarbonate (9.82 g L-1) solutions. The mixture was warmed up and kept at 40°C for two hours in a water bath. Then, the samples have been being mixed continuously in an end-over-end shaker for 16 hours at 25°C. After that the supernatant was separated by centrifugation at 4 000 rpm for 15 minutes and reserved to determine iron (Fe) and aluminium (Al) by atomic absorption spectroscopy (AAS) (Thermo Scientific ICE 3000 series). The Fe2O3 content determined by the CBD extraction method (Fe2O3_CBD) represents the amount of free iron oxides and the Al2O3 content (Al2O3_CBD) corresponds to the amount of Al substituting Fe in iron oxides. The ratio of Al substitution (Alsub, %) was calculated from the amount of Fe2O3_CBD and Al2O3_CBD expressed in moles (Jeanroy et al., 1991). The goethite (Gt) and hematite (Hm) contents were computed by combining the two equations as follows (see Reatto et al., 2008, 2009, for more details): Fe2O3_CBD = 0.8989 × Gt + Hm (1) Hm / (Hm + Gt) = (RI – 3.50) / 8.33 (2) where Gt and Hm are the goethite and hematite contents (g kg-1) of the sample, respectively, Fe2O3_CBD is the Fe2O3 content of the sample determined with CBD extraction (in g kg−1), 0.8989 the specific proportion of Fe2O3 in a goethite (for goethite assumed to be not Al substituted), RI is the red index according to (Santana, 1984) and equalled to RI = M + V/C, with M a parameter related to the hue, C the chroma and V the value of the Munsell notation.