Non-allophanic Andosols of Trindade Island, south Atlantic: a new soil order in Brazil

The pedological studies carried out so far in Trindade Island (TI) have obtained patchy evidences of allophane, without detailed mineralogical and micromorphological studies to confirm the occurrence of Andosols in TI. Therefore, in this study, the mineralogical, micromorphological, physical and chemical characterization of four soil profiles from Vulcão do Paredão (P1) and Morro Vermelho formations (P2, P3, and P4) were carried on the latest volcanic events in Brazil from Trindade Island (TI) with the aim of to evaluate the presence of Andosols in this oceanic island. Profiles P1 and P2 are developed on pyroclastic bombs, and show, respectively, A-BiC and decapitated A-C horizons, whereas P3 and P4 are developed on lapillitic and bomb pyroclasts, show A-C horizons. The soil profiles have a reddish and brownish clayey matrix, are highly friable and show a plastic consistency. Their microstructures are granular, single grain and intergrain microaggregate, in which aggregates display an undifferentiated b-fabric. The mineralogical constituents of the bulk fraction are biotite, hematite, magnetite, ilmenite, pyroxene, olivine, halloysite, goethite, anatase, and rutile. The clay fraction is marked by the presence of halloysite, ferrihydrite, and little amounts of allophane. All soils presented andic properties and can be classified as non-allophanic Andosols. In addition, micromorphological features closely resemble those reported in Andosols from other volcanic islands from elsewhere. The predominance of halloysite in the clay fraction formed by alteration of sideromelane, suggests that allophane would be an intermediate phase of this rapid transformation favored by the wet climate conditions of the highest parts of TI.


INTRODUCTION
The soils formed from tephras or pyroclastic materials are known as Andisols (Soil Survey Staff, 2014), Andosols (IUSS Working Group WRB, 2015), and Kurobokudo (USCSJ, 2002). There are also some Andosols formed from non-tephritic materials, such as volcanic and sedimentary rocks, mixed materials of tephra and loess, etc. (Shoji et al., 1993). The Andosols are characterized by the presence of allophane or Al, Si, and humus compounds (Shoji et al., 1993). However, Shoji and Ono (1978) reported the occurrence of non-allophanic Andosols with Al/Fe humus characterized as Allophanic Andosols (Shoji et al., 1985). According to Soil Survey Staff (2014) an Andosol soil must have andic properties in at least 60 % of its profile. The andic properties of a soil are generally acquired during weathering of tephra or other parent materials which contain significant volcanic glass, leading to the formation of amorphous material as allophane, imogolite, ferrihydrite, or metal-humus complexes. The presence of amorphous materials is inferred from ammonium oxalate extraction of aluminum, iron, and silica (Soil Survey Staff, 2014). In the field, the andic soil properties may be inferred by using 1N sodium fluoride (NaF) (Soil Survey Staff, 2014). However, this test may not be specific for allophane as it may also reacts with A1 in Al-humus complexes (Parfitt, 1990).
The volcanic glass content is the percent of volcanic glass (by grain count) in the coarse silt and sand (0.02 to 2.0 mm) fraction (Soil Survey Staff, 2014). In addition, the pH(H 2 O) >5, low content of organic matter, and base-rich volcanic ash favor the formation of allophane (Dahlgren et al., 1993).
In Brazil, studies showing the occurrence of Andosols are rare and Schaefer et al. (2017) were the first to postulate their presence on the oceanic island. Dümig et al. (2008) were the first to detect non-allophanic Andosols in South America outside the radius of influence of the recent volcanism of the Andes region. These Andosols were described in the northeastern of Rio Grande do Sul State and were developed from rhyodacite (Jurassic -Cretaceous volcanism).
Rev Bras Cienc Solo 2020;44:e0200007 Trindade Island (TI) is located in the mid-Atlantic with a surface of 13 km² and formed mostly by extrusive volcanic material and/or hyperabissal sodium-alkaline lava of Quaternary age (Almeida, 2002(Almeida, , 2006. Previously, Clemente et al. (2009) observed evidence of soils with andic properties and described a material similar to allophane at the base of a highland soil profile on tuffs, at the same location of the P1 of the present study. In soils developed from phonolitic and basic rocks of TI, Sá (2010) found pH values in NaF above 9.4, suggesting the presence minerals with a low degree of crystallinity, possibly allophane (Fields and Perrot, 1966;Brydon and Day, 1970). Consistently, phosphate retention values equal to or greater than 25 % and [Al (oxalate) + ½ Fe (oxalate)] > 0.4 were also observed (Sá, 2010). Clemente (2006) and Machado (2016) verified the presence of volcanic glass in a thin section in soils of TI, but a quantitative study was not performed, so they could not be clearly classified as Andosols.
Thus, the pedological studies carried out so far in TI have obtained patchy evidence of allophane, without detailed mineralogical and micromorphological studies to confirm the occurrence of Andosols in TI. To fill this gap, the present study aims to develop a combined mineralogical, micromorphological, physical, and chemical characterization of selected soils developed on TI pyroclasts to evaluate the presence of Andosols ( Figure 1a) and, assess the pedo-environmental condition of their occurrence.

Sample collected
Soil, rocks, and saprolites samples from four profiles were collected, one in the Vulcão do Paredão formation (VP) and three in the Morro Vermelho formation (MV) ( Table 1). The contact sections between rocks and their respective alterites (iso and alloalterite) and solum (pedogenetic horizons) were described and sampled in the field according to the standard guidelines (Santos et al., 2015).
The descriptions highlighted the main morphological variations at macroscopic scale, combined with the use of a magnifying glass (20X). Aspects such as color (Munsell, 2009), texture, and structure were described (Table 1).

Physical and chemical analyses
The physical and chemical analyses were performed in the Laboratory of Soil Analysis of the Federal University of Viçosa, following international standard procedures (Teixeira et al., 2017). The contents of sand, silt, and clay fractions were measured after slow agitation (16 h) using chemical dispersion with 10 mL of NaOH 1 mol L -1, and the proportions of each fraction were plotted in the triangular diagram to obtain the textural class. Coarse and fine sand were separated by sieving and, silt and clay were determined by the pipette method.
The bulk density was performed by volumetric cylinder method (Almeida et al., 2017). The samples collected in the field in metallic cylinders of 100 g cm -3 and were dried at 105 °C for 48 h, to obtain the soil density.
The quantification of volcanic glass was obtained from the coarse silt to sand fractions by separation by a magnifying glass Nikon C-POL.
The pH was measured in H 2 O (soil: solution 1:2.5) and in NaF (Pansu and Gautheyrou, 2006); the Melich-P concentrations were reported in mg dm -3 (Defelipo and Ribeiro, 1981); the exchangeable contents of K, Ca, Mg, and Al in cmol c dm -3 , allowed to sum of exchangeable bases (SB), the effective cation exchange capacity CEC (t), and total CEC (T) in cmol c dm -3 , base saturation (V) and aluminum saturation (m) in %. The organic matter (OM) content was determined by the Walkley-Black method, while the total organic carbon (TOC) content was estimated, in g kg -1 , by the equation (OM × 1.724). The P remaining (Prem) was calculated according to Alvarez et al. (2000) and P retention percentage (PR) as follows: P in the blank minus P in the equilibration solution (Prem).

Mineralogical analyses
The identification of the mineralogical phases was performed by X-ray diffractometry (XRD) in the fine-earth fraction and the clay fraction (<2 µm). The clay samples were saturated with Ca 2+ , deferrified with sodium dithionite-citrate-bicarbonate (DCB) according to the methodology of Mehra and Jackson (1960), treated with formamide and dehydrated at 110 °C for 6 h. The analyses were carried out in a Panalytical Empyrean  (CuKα, 45 KV and 40 mA) and in the Philips X ' Pert Pro Diffraction (CuKα, 40 KV, 40 mA) diffractometers. The scanning interval for the fine-earth fraction analysis was from 2 to 70 °, with a step of 0.02° (2θ min -1 ); and for the analysis of the clay fraction from 2 to 35°, with step 0.4° (2θ min -1 ). The diffractograms were interpreted by using the X'Pert HighScore Plus software and through the literature references (Brown and Brindley, 1980). Infrared spectroscopy was used to assist the mineralogical analyses. The analyses were performed in the IC2MP Institute, University of Poitiers using Fourrier transform spectrometers (Thermo Nicolet, Nexus 5700 series) for analyses of the bulk and clay fractions. The pressed discs were made with 1 mg of sample and 150 mg of KBr, which were analyzed with natural moisture and then dehydrated at 120 °C for 20 h and at 200 °C for 4 h to remove absorbed water. The mid-infrared spectra (4000-400 cm -1 ) were obtained using a source of white light, beamsplitter KBr, resolution of 4 cm -1 , the optical velocity of 0.4747 cm s -1 , and 100 scans. Clay fraction images were obtained by a Transmission Electron Microscope (TEM) JEOL 2011 Bruker Esprit in the CinaM-CNRS, Aix Marseille University. The suspended clay was deposited on carbon coated Formvar film Cu grids.

Micromorphological and microchemical analyses
Thin sections of soils and rocks previously impregnated were described in a Zeiss microscope fitted with a digital camera. The micromorphological description of soils was based on Stoops (2003) and (Stoops et al., 2018).
The microchemical analyses were performed with electron microprobe JEOL JXA-8230 at the Microanalysis Laboratory of Geology Department/Federal University of Ouro Preto. The acceleration and current conditions were 15 kV and 20 nA, and corrections of the common ZAF matrix were applied. Counting times on the peaks/background were 10/5 s for all elements (Si, Al, Fe, Ca, Na, and K), except for Ba (30/15 s). The point-source analyses were calibrated with quartz-TAP, coridon-TAP, almandine-LIFH, augite-PETJ, anorthoclase-TAPH, microcline-PETL, and barite-PETH. The FeO was considered as the amount of total iron obtained through the microprobe.
The A1:Si ratio of the allophane was calculated from [A1(o)-Al(p)]/Si(o) and the amount of allophane in the sample was estimated by multiplying of Si(o) by the factor proposed by Parfitt (1990): for Al:Si atomic ratios close to 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, is multiplied Si(o) by the factors 5, 6, 7, 10, 12, 16, respectively. Ferrihydrite content was determined by multiplying Fe(o) values by l.7 (Parfitt and Childs, 1988). Table 1 shows the macromorphological and physical characteristics of the four soil profiles. The profile P1 is 1.30 m thick and is located at an elevation of 460 m, with A, Bi, C1, and C2 horizons and developed from pyroclastic bombs ( Figure 1b, Table 1). Profile 2 (P2), is 1.20 m thick, located at 351 m on deposit of pyroclastic bombs, and has lost most of its A horizon; P3, at an elevation of 258 m, is approximately 1.00 m thick and shows A, C1 and C2 horizons, with 0.10 m layer of lapillitic pyroclasts between C1 and C2; P4 has a thickness of 0.57 m, located at an elevation of 72 m, with a sequence of A, C1, and C2 horizons. Bulk density is low (<0.90 g cm -3 ), consistently with soils developed from volcanic ash (Soil Survey Staff, 2014).  Figure 2). In P1, the glass is stained (reddish brown to yellowish), showing alteration to palagonite. However, this alteration is only partial in P2, P3, and P4 ( Figure 2).

Physical and chemical properties
The bases saturation of P1 (VP) is lower than P2, P3, and P4 (MV) decreasing to the A horizon (Table 2). Profile 1 (P1) is located at the highest altitude; hence it is exposed to more intense weathering, which promotes greater leaching of K + , Na + , Ca 2+ , and Mg 2+ , resulting in lower pH(H 2 0) -between 4.82 and 6.01. The pH in NaF is greater than 9.5 in all horizons, except A and C2 of P3. The sum of exchangeable bases (EB), effective CEC [CEC(t)] and potential CEC [CEC(T)] are high, which suggests little weathered young soils. A slight increase of EB is seen in the A horizon highlighting the influence of organic surface. The base saturation (V) is lower than 50 % in A, Bi, and C1 of P1, and above 50 % in all others (P2, P3, and P4) ( Table 2). This characterizes the P1 as a dystrophic and the other profiles as eutrophic. The content of exchangeable aluminum (Al 3+ ) is high in A, Bi, C1 horizon of P1, with a marked value in A horizon (4.2 cmol c kg -1 ).
In all other soils, the content of Al 3+ is null. The available phosphorous showed similar values to soils on tuffs described by Clemente et al. (2009); the phosphorous may be attributed to sideromelane (0.8 % in pyroclasts of P2 and P3), and 2.3 % in palagonitized sideromelane of P1 (Mateus et al., accepted article). The sodium saturation in the C2 horizon of P1 and P4 represents a solodic character (sodium saturation between 6 and 15 %). The phosphate retention (PR) is high in all horizons with values between 65 and 97 %; in contrast, the total organic carbon contents (TOC) are low (Table 2).
The c/f related distribution is coarse monic in C horizons (Figure 3b), changing to chitonic and eunalic in Bi and A horizons, showing a single grain and intergrain microaggregate microstructure, respectively at the bottom and top of the profiles (A and Bi horizons), (Figures 3b, 3f, 3h, and 3i). In Bi and A horizons, the pyroclasts fragments are split into smaller fragments characterizing the transition from simple grain to intergrain microaggregate microstructure (Figures 3d, 3f, and 3i). The typical granular microstructure also occurs in the A horizon (Figures 3f and 3l). The pyroclasts fragments and olivine and magnetite crystals decrease in size and quantities towards to the surface. At the bottom (C2 horizon), pyroclasts fragments of 15.0 mm   Aggregates occur in Bi and A horizons and show undifferentiated b-fabric micromass ( Figure 3i). As pedological features, reddish and yellowish clay coatings occur, evidencing illuviation of fine colloidal materials (Figures 3e and 3f) and infillings of clay due to the bioturbation (Figure 3i).

DISCUSSION
The presence of andic properties on TI soils was evaluated using the criteria in the Soil Taxonomy (Soil Survey Staff, 2014), for moderate or weakly weathered soils, and in the IUSS Working Group WRB (2015). Recently, in 2018, the Brazilian Soil Classification System -SiBCS (Santos et al., 2018) incorporated this property, using the same criteria used by the IUSS Working Group WRB (2015) ( Table 5).
According to Soil Survey Staff (2014), soil with andic properties must have less than 25 % of total organic carbon (TOC), with one of the following criteria that must be met: 1) for moderately weathered soil: the values of [Al(o) + ½Fe(o)] >2 %; the bulk density of soil equal or minor than 0.90 g cm -3 and phosphate retention equal or minor than 85 %; or 2) for weakly weathered soil: ratio [Al(o) (2015) and SiBCS (Santos et al., 2018), only the parameters shown in 1 are taken into account.
Considering the Soil Taxonomy criteria 1, WRB and SiBCS, only three horizons (P1A, P1C2, and P3A) met all criteria. The main attribute not met was the bulk density, since only the A horizons of P1, P3, and P4, and C2 horizon of P1 showed density lower than 9.00 × 10 7 mg m -3 . However, except for the P4C1 horizon (Table 2), the values are not much greater than 9.00 × 10 7 mg m -3 , varying between 9.30 × 10 7 and 10.00 × 10 7 mg m -3 , so very close to the limit.
The soils of TI have met criterion 2 of the Soil Survey Staff (2014) because they contain less than 25 % of TOC (Table 2) and are weakly weathered. The presence of primary minerals in A and B horizons (glass, olivine, pyroxenes, and feldspar), inside and outside of altered pyroclast fragments, including in the clay fraction, corroborate the low weathering degree of these soils (Figures 3d, 3e, and 3i). The P1, P2, and P3 have andic properties by this criterion in 60 % of their profiles (Table 5). Only the Bi horizon of P1 showed a value lower than of the sum [(Al (O) + 1/2 Fe (o)] × 15.625 + volcanic glass content (%) ≥36.25, while the C2 horizon of P3 presented less than 5 % of volcanic glass, and  (Yeomans and Bremner, 1988); PR: phosphate retention (Alvarez et al., 2000); (1) Coding of horizon according to Brazilian Soil Classification System (Santos et al., 2018).
(2) Coding of horizon according to Soil Taxonomy (Soil Survey Staff, 2014). therefore only, these two horizons do not follow criterion 2. The P4 soil did not have 5 % volcanic glass in 60 % of the profile, and therefore does not meet criterion 2, either (Table 5). It is important to emphasize that the C2 horizon of P4 consists of fragments of a mineralogical composition different from the other profiles, suggesting a different volcanic event of pyroclastic bomb deposition.
In addition, based on criterion 2, the andic properties detected were: ratio Al(o) + ½Fe(o) higher than 0.4 % ( Table 5); phosphate retention (PR) greater than 25 % in all horizons, with values between 65 and 97 % (Table 2); the sand and silt contents exceeding 30 % in all profiles (Table 1), corroborating less weathered soils; also, all horizons, excluding the C2 and C1 of P4 and C2 of P3, showed at least 5 % of glass volcanic in the coarse sand fraction in part of the profile (Table 1; Figure 2).
Pyroclasts fragments consisting of palagonized glass corroborate the content of more than 5 % of volcanic glass in the 2 mm fraction (Figures 3b, 3d, 3g, and 3h).
The NaF values greater than 9.4 also indicate the presence of non-crystalline minerals. However, all profiles presented little amounts of allophane (between 0.36 and 3.19 %), and the Si(o) content ≥0.6 % further suggests the presence of these amorphous materials.
The pH(H 2 O) >5 in all profiles and low content of organic material would contribute to the allophane formation (Dahlgen et al., 1993). The values of the Fe(o)/Fe(d) ratio are also high in all profiles, indicating the presence of no crystalline minerals and a poorly weathered soil (Table 5).
Although the selective extractions indicated the presence of amorphous minerals, the mineralogical analyses (XRD, Infrared, and TEM) could not detect their presence, although halloysite was found. Consistently, in Cabo Verde Island, halloysitic soils with high amounts Al-humus complexes were classified as non-allophanic Andosols (Madeira et al., 1994). Accordingly, soils of the present study have an undifferentiated b-fabric characteristic of the gradual replacement of allophane by halloysite, and illuviation features in all horizons, particularly more pronounced in the Bi horizon of P1, shows that they are Andosols intermediate to soils with more crystalline clays (Stoops et al., 2018).
The classification of non-allophanic horizons requires some properties, such as Alp/Alo of 0.5 or more, and/or Si(o) content of less than 6 g kg -1 ; and thickness of 0.25 m or more (Tamura, 2012;USCS, 2002). All profiles of this study have Si(o) content less than 6 g kg -1 and thickness of 0.25 m or more, so, they can be classified as non-allophanic Andosols. Alternative names for allophane-like materials which could be used are defective kaolin allophane or halloysite-like allophane, when Al:Si ratio value is near 1 (Yoshinaga, 1986) or proto-imogolite allophane and imogolite-like, when this ratio is between 2 and 4 (Farmer and Fraser, 1979;Parfitt and Wilson, 1985). In this case, the soils of P1 and P3 would present halloysite-like allophane, whereas P4 a proto-imogolite allophane and imogolite-like. According to the IUSS Working Group WRB (2015)  We suggest that sideromelane is the likely precursor of halloysite in TI soils. The study of Mateus et al. (2020) on the alteration of pyroclasts of TI showed the presence of little quantities of halloysite in zones where occur altered sideromelane (palagonite) occurs. In addition, the halloysite of P4 also has feldspars as possible precursors (oligoclase and anorthoclase), besides volcanic glass.
Soils with andic properties, also considered non-allophanic, have already been recognized in the Brazilian territory. In a pioneer report, Ker (1988) suggested the possibility of soils with andic properties in southern Brazil. Later, Gama et al. (1992)  Recently, Santos Junior (2017) confirmed the presence of non-allophanic Andosols in the escarpment of Serra Geral Formation, in the states of Santa Catarina and Rio Grande do Sul. This author identified eight Aluandic Andosols, and two Histosols with andic properties.
The organometallic complexes formed in the cold-humid-high elevation areas explained the formation of soils with andic properties associated with non-volcanoclastic rocks.
Trindade Island soils show an unusual situation, with non-allophanic soils derived from typic pyroclastic materials. As previously described, the other non-allophanic soils in Brazil occur because of cold climate conditions and the role of the organic matter. Sieffermann and Millot (1969) studied the soil genesis on Late Quaternary pyroclastic rocks, scorias, tuffs, and ashes of basaltic composition from the Federal Republic of Cameroon. Sieffermann and Millot (1969) showed that in tropical climates, with average annual precipitation between 1.5 and 1.6 m, and a marked dry season and in less leached zones, there is a rapid neoformation of halloysite through allophane as an intermediate phase, maintaining lower amounts of allophane. In our study at TI, we have found similar characteristics: pyroclasts of basaltic composition and tropical oceanic climate with average annual rainfall between 926 mm and a marked dry season, between January and March. Therefore, the climatic conditions of the island could contribute to the rapid formation of halloysite from an allophane precursor, helping to explain the small amounts of allophane in these soils.
Andosols show a well-developed medium granular microstructure at the top which changes to coalesced microstructures towards the bottom of profile (Sedov et al., 2010), very similar to those reported here, but less stable than microgranular microstructures of Latossolos (Oxisols). The A horizons showed a granular microstructure, with an undifferentiated b-fabric typical of Andosols on pyroclasts from oceanic islands Pico and Fail in Azores (Gérard, 2007), Terceira (Pinheiro et al., 2003), Canarias (Sanchez Diaz, 1978;Benayas et al., 1980), and Santa Fé (Morrás, 1978).
Despite the criteria of the IUSS Working Group WRB (2015) be considered satisfactory for the SiBCS (Santos et al., 2018), they fail to meet the specificities of TI soils. Soil Taxonomy will offer better conditions to incorporate soils with andic properties with different degrees of development. Thus, we believe that the criteria currently adopted by Soil Survey Staff (2014) are more adequate for keying out soils like those of Trindade Island, in the SiBCS.
There is no place for Andosols in the SiBCS. Using the current system (Santos et al., 2018), the soils of this study would be classified as Cambissolo Háplico Sódico típico (P1), with A moderate, presence of Bi horizon, and lower base saturation (<50 %) ( Table 2); Neossolo Regolítico Eutrófico típico (P3), and Neossolo Regolítico Eutrófico solódico (P4). The P3 has A moderate, absence of B, lithic contact at a depth greater than 0.50 m, with semi-altered pyroclasts fragments within the C; P4 is different from P3 because it presents a solodic character (sodium saturation between 6 and 15 % in C2 horizon) ( Table 2). Using the IUSS Working Group WRB (2015), only pedon (P1) can be classified as Andosol (Table 6). The others are classified as Regosols, but this class does not admit the andic suffixes as supplementary qualifiers. The reason why some pedons, such as P3, have not been classified as Andosols is because even though they have the andic properties on the A horizon, they are 0.20 m thick, whereas a minimum of 0.30 m is required by the WRB/FAO System. According to Soil Survey Staff (2014), P1, P2, and P3 are all Andisols, respectively Alic and Typic Hapludand.
We recommend two possibilities to improve the SiBCS in view of the presence of Andosols in TI: 1) the expansion of research in other regions of Brazil, such as in Brazilian oceanic islands and in places where occur similar pedoenvironmental characteristics found in these islands for inclusion of the Andosols order in the SiBCS and/or; 2) the adoption of Soil Taxonomy criteria for the definition of andic properties for weakly weathered soils, maintaining the WRB/FAO criteria already included in 2018.

CONCLUSIONS
The soils of Trindade Island (TI) present andic properties and should be classified as Andosols, and meeting the criteria to be classified as non-allophanic Andosols. In addition, the micromorphological features are like the Andosols described from elsewhere.
The Andosols in TI are non-allophanic for a different reason, compared with other non-allophanic Andosols already reported in Brazil. In Trindade Island, soils are derived from pyroclastic materials, and the silandic conditions are weakened by the rapid weathering under warmer climates. Other non-allophanic soils in the southern highlands of Brazil have andic properties because were formed under cold-mesothermic climate conditions.
The predominance of halloysite in the clay fraction is consistent with the non-allophanic character of TI soils. Halloysite was formed by alteration of sideromelane, in which allophane is the intermediate phase.
These unique TI soils should be recognized in the SiBCS as special soils, and the adoption of Soil Taxonomy criteria for weakly weathered soils allows us to classify them out as non-allophanic Andosols.