Nickel laterites are Ni ± Co deposits formed by prolonged, intense weathering of peridotites, primarily in tropical latitudes but also at latitudes as high as ˜40° and probably higher in past warmer climates. The weathering process removes the principal components of the peridotite, MgO, and SiO₂, ultimately leaving a 5 to 7 percent residue of the original rock enriched in the other major components, such as Fe, Al, and Cr, and minor constituents, such as Ni, Mn, Cu, and Co. Generally, deposits developed in situ over the host ultramafic rocks are comprised of a more or less complete weathering profile, which may be complicated by the presence of an allochthonous cover or the absence of some weathering zones in the profile because of mechanical erosion. Erosion and sedimentation lead to the formation of transported deposits. Many such sedimentary deposits have been buried, lithified, and, in some cases, folded and metamorphosed.

The vast majority of deposits derive their Ni from igneous olivine and its alteration products, which are typically serpentine ± chlorite. Disseminated magmatic sulfide minerals in some host rocks can contribute to the Ni content of the laterite, but this is rarely the case for any economic nickel laterite deposits. This paper will concentrate on Ni laterites derived from olivine-rich rocks, reviewing advances in understanding of the evolution of in situ saprolitic deposits.

The general geology of nickel laterites has recently been comprehensively reviewed by Freyssinet et al. (2005). This paper begins with a general theory of nickel laterite evolution that relates different types and features of nickel laterites to the topographic setting of deposits on the one hand and to the long-term climate and geomorphic evolution of some lateritehosting regions on the other hand. Topographic setting has, historically, been important in nickel laterite exploration and it will be shown that three scenarios of climate change and tectonic setting help to explain many of the recognized relationships. Three assemblages of deposits exemplifying these scenarios, as well as their exploration potential, will be described and discussed in relationship to new information and developments in nickel laterite theories.

The zones of a complete nickel laterite profile and the terminology used here are summarized in Table 1. There is no widely accepted terminology in general use. Saprolite is defined regolith with inherited protolith textures. It is also commonly used to refer to the silicate-rich part of the profile, and the fine Fe oxide-rich units are often lumped as limonite or laterite in practice. The latter practice is followed here wherever literature sources do not permit more precise distinction. Quartz veinlets may have replaced specific minerals somewhere in the lower units where there is still adequate silica in solution. Quartz can persist, typically as a pervasive vein network into the upper levels of the profile, creating a siliceous version of each of the in situ units of Table 1.

Ores are commonly described as silicate, clay, or oxide (e.g., Freyssinet et al., 2005), depending on whether the saprolite, nontronite, or limonite zone, respectively, is economic. This is a simplification for several reasons. First, in some cases, ore is a mixture of two zones. For example, at the giant Goro deposit in New Caledonia, limonite and some of the underlying silicate saprolite constitute the ore. Second, the terms do not reflect the presence of quartz. In the case of two oxide deposits, the Goro limonite is very different from the siliceous limonite at the Ravensthorpe deposit in Western Australia. Finally, the presence of abundant chlorite and/or vermiculite can be important, such as in the Puma and Santa Fé deposits of Brazil where they cause development of a Ni-rich limonite zone containing Ni-enriched vermiculite. The economically important zone may not be the highest in Ni grade due to metallurgical considerations.

Nickeliferous forsterite-bearing igneous rocks of any kind and tectonic environment can be protoliths for nickel laterites if they have been exposed in humid tropical to Mediterranean weathering regimes since the mid-Paleozoic. In tectonic collision zones, ophiolitic harzburgites, dunites, and lherzolites, usually serpentinized, are the common Ni laterite hosts. Elsewhere, layered mafic complexes of any age, the cumulus olivine-rich channel facies of komatiites, or the intrusions related to them may be hosts. In Brazil, dunites and peridotites of alkalic complexes host nickel laterite deposits. The Ni content of protoliths ranges from 0.16 to 0.3 percent (Table 2), with localized exceptions where Ni has been transported for a few meters, from peridotite saprolite into the saprolites derived from normally barren pyroxenites and gabbros.

Some measure of mechanical transport may occur during the formation of reasonably complete laterite profiles. The duricrust and the typical red laterites at the top of the profile commonly show textures and intraformational breccia structures, and with the addition of nonultramafic components, suggest some very local movement of perhaps tens or hundreds of meters. These locally transported units typically have a low Co/Fe and Mn/Fe ratio compared to the in situ yellow or brown limonites or ferruginous saprolites that usually underlie them.

Movement by alluvial or mass wasting processes could involve longer distances of transport, yet still remain above the ultramafic protolith. An example is the Plaine des Lacs deposit, which lies under the lakes on the plain of that name on the southeastern massif of New Caledonia (Chen et al., 2004). There, the limonite must have been transported a few kilometers from elsewhere on the plateau into the lake basin. It contains plant material, some of which has been replaced by manganiferous siderite. Some of the siderite occurs as at least 1-cm-thick fossil beds. Classic bog-iron goethite concretions also formed in the lake bed.

In the Moa Baracoa ophiolite complex, Cuba, fossil evidence in the laterite of late Miocene foraminifera and corals shows that the laterite profile underwent a marine inundation to about the present 500-m elevation. It has been suggested that sorted pisolite beds in the overburden of the deposit represent beach bars formed during the transgression of the shoreline (Linchenat and Shirokova, 1964; Golightly et al., 2009a, b). In many of these deposits, saprolite is absent and rock boulders are encountered near the base of the limonite zone and an underlying silicate saprolite zone may be absent. The Cretaceous Çaldag deposit, Turkey, shows similar features (Thorne et al., 2009). However, the thicker zones of these Cuban deposits have saprolite roots and the overlying limonite has inherited protolith textures (Golightly et al., 2008a, 2009a, b). Linchenat and Shirokova (1964) suggested that much of the deposit at low elevations accumulated by mass flows containing serpentinite blocks and that most of the blocks were subsequently leached to form limonite blocks with inherited protolith textures.

Debris eroded from ultramafic rocks may have been deposited on other lithologic units. Some of the Mesozoic nickel laterites of the Balkans and Turkey accumulated on karstic limestones (Valeton et al., 1987; Economou-Eliopoulos and Michaelidis, E., 2000; Eliopoulos and Economou-Eliopoulos, 2000; Skarpelis, 2006; Laskou and Economou-Eliopoulos, 2007; Thorne et al., 2009). The mineralogy of such deposits is modified by diagenetic processes and in some cases metamorphism leading to the appearance of siderite, chamosite, and sulfide minerals. In the case of metamorphism, sodic amphiboles, chlorite, and magnetite may have formed. Because most of the paleolaterite deposits, whether in situ or transported, are covered, the original topographic distribution of the deposits is not well understood and thus is not discussed further in this paper.

Before discussing the development of nickel laterites in relationship to tectonism, topography, and climate change, it is necessary to review what is known about the solubility of the minerals, the controls that they imply, and methods for tracking movement of materials in a weathered profile.

In situ nickel laterites result from the dissolution and precipitation reactions in a more or less permeable host by relatively acidic soil water that is either seeping downward or fluctuating about a seasonally variable water table. The pH dependence of the equilibrium congruent solubility and dissolution kinetics of the minerals in Ni laterites are compared in Figure 1. The solubility of smectite clays, vermiculite, and chlorite is similar to that of the talc structure minerals.

Useful predictions that come from these solubility relationships are the following:

1. With the exception of quartz, the sequence of decreasing solubility accurately predicts the sequence in which minerals in laterites disappear upward in a typical profile and outward from corestones, which is olivine, orthopyroxene, clinopyroxene, serpentine, talc, chlorite, gibbsite, and goethite.

2. Nickel-rich serpentine, Ni-rich talc (kerolite), and Nirich chlorite (nimite) are less soluble than their Mg-rich counterparts and an important reaction is Ni exchange for Mg in more soluble Mg-rich serpentine and Mg-rich chlorite and vermiculite as the profile evolves.

3. At a particular point in a profile, in open spaces, as the alteration front moves downward, precipitates of garnierite minerals in veins become increasingly Ni rich over time and eventually, as the pH drops, these are followed or cut by quartz precipitates.

4. Because soil water normally increases in pH from slightly acidic at the surface to alkaline at the weathering front, where soil water is predicted to approach equilibrium with serpentine at a pH of 9.2, primary silicate minerals should be replaced by talc or smectite minerals at depth or quartz nearer the surface if saturation is approached.

Important conclusions derived from the last three points include:

1. The Ni/Mg ratio increases upward in the profile from bed rock to the magnesium discontinuity, the point at which Mg silicates disappear in the profile at the top of the serpentine or nontronite saprolite zones.

2. The leaching of soil water from more decomposed material into more Mg rich, less decomposed material is necessary for true supergene Ni enrichment. Typically this means downward leaching above the water table, but could mean subhorizontal movement of water on the water table if it lies in the saprolite zone. As a corollary, supergene enrichment is not expected to occur if the water table lies within the limonite zone. Ni released there should be lost by lateral movement in solution.

3. Therefore, a low-lying water table caused by strong fracturing of the underlying rocks and or relatively elevated topography at a hill top or the edge of a plateau is expected to favor supergene enrichment of Ni in the saprolite.

Mathematical modeling of the leaching process predicts an exponential increase of insoluble elements such as Fe, Al, Cr, and Ti upward and a complementary decrease in soluble elements, such as alkalis and alkaline earths, under isovolumetric weathering (Golightly, 1981; Brantley et al., 2008). In Ni laterites concentrated near the top of the saprolite, the collapse of the saprolite under overburden pressure causes the profile to be shortened vertically and causes a departure from the exponential trend.

The lower solubility of talc or smectite relative to serpentine predicts that protolith Mg silicates will be entirely replaced by smectites or quartz, providing the percolating water reaches saturation. In many profiles occurring in constantly wet rain-forest environments, this is only true for olivine alteration. Orthopyroxene in corestone rinds, in many cases, has dissolved to form goethite or smectite-rimmed, but otherwise empty cavities, rather than being entirely replaced by talc or smectite (Golightly, 1981; Nahon and Colin, 1982; Pelletier, 1996). This suggests that such profiles are so efficiently flushed by meteoric water that the water has not attained quartz or smectite saturation within the saprolite zone. Where a nontronite zone has formed, smectite saturation has occurred due either to restricted circulation, as in the lowland profiles of New Caledonia (Trecases, 1975) or Nicaro, Cuba (Linchenat and Shirokova, 1964), and/or because of a drier climate, as in the Murrin Murrin, Bulong, Brolga, and Greenvale deposits in Australia (Burger, 1979; Golightly, 1981; Parianos and Rivers, 1996; Elias, 2006), as well as the deposits at Tiebaghi on the dry, leeward side of New Caledonia (Latham, 1986).

Studies of biologic effects in weathering profiles on quartz diorite bed rock in the Rio Icacos watershed in Puerto Rico give some results that may be relevant to weathering profiles on ultramafic rocks (Buss et al., 2005). Core stones weather spheroidally (Fletcher et al., 2006) to form concentric shells of partially weathered rock layers, referred to as rindlets, which slowly transform to saprolite. The rind zone in their study (0.2−2 m thick) is overlain by saprolite (2−8 m thick) topped by soil (0.5−1 m) thick. Direct microscopic counts of total cell densities were higher than 10⁸g⁻¹ at three depths: in the upper 1 m, at 2.1 m, and between 3.7 and 4.9 m, just above the rindlet zone. Data were collected from enumeration of culturable aerobic heterotrophs, extraction of microbial DNA for yield calculations, and biochemical tests for iron-oxidizing bacteria. Total cell densities, which ranged from 2.5 × 10⁶ to 1.6 × 10¹⁰ g⁻¹ regolith, and total and culturable cells and DNA yields at lower depths correlate with moisture and HCl-extractable iron. Biochemical tests for aerobic iron oxidizers were also positive at 0.15- to 0.6-, at 2.1- to 2.4-, and at 4.9-m depths. It appears that Fe⁺⁺ released in the initial stages of weathering supported Fe-oxidizing bacteria, causing ferrihydrite and goethite precipitation.

There are many studies of gains and losses of elements in laterite profiles relative to their protolith (e.g., Trescases, 1975; Colin et al., 1985, 1990; Schellmann, 1989; Bandyayera, 1997). A fundamental assumption in most of these studies is that an element R may be sufficiently insoluble such that the variation of its ratio to a relatively mobile element X, (X/R)/(X_o/R_o) can measure the relative movement of mobile element X in the profile. If good estimates of bulk density of the material are available, then the movement can be quantified. A second assumption is that the concentration of the mobile and immobile elements, Xo and Ro, of the protolith is sufficiently well know from assays of fresh rock and is sufficiently homogeneous, so that interelement mobility is the most important source of variation in the profile. Generally, Ir, Zr, Ti, Cr, Fe. and Al are thought to be the best choices of an immobile element in order of increasing solubility. Commonly, however, the best choices are rarely available on sufficiently large geochemical data sets to come to reliable, deposit-wide conclusions. If the Ni to immobile element ratio is greater than that of the protolith, then supergene enrichment has likely occurred. As a general rule, this is confined to the silicate parts of the profile, whether saprolite or nontronite. In contrast, where ratios in the profile are less than those for the protolith, residual enrichment only has taken place. This is typical of limonite and siliceous limonite zones.

Until approximately 30 years ago, routine exploration data for laterites included only Ni, Co, and Fe, but many other important elements were lacking. In recent decades, routine exploration data have also typically included analyses for elements such as Mg, Si, Al, Cr, and Mn and, in some cases, Ti, Ca, Na, P, and LOI. Work with the full body of exploration data for a deposit, commonly comprising tens of thousands of analyses, has the potential to provide some more definite answers to mass-balance problems. For example, the scatter of Al and Fe analyses in several thousand assays in approximately 1-m drill core samples from the La Gloria deposit in Guatemala is shown in Figure 2A. The serpentinite saprolites form a broad, linear trend of positive correlation between Fe and Al, extending from fresh rock assays at the lower left to limonite, L, where the trend ends at about 62 percent Fe₂O₃. This trend is terminated by a trend of limonite samples showing a negative correlation between Fe and Al, from 50 percent Fe₂O₃ and 17 percent Al₂O₃ to 75 percent Fe₂O₃ and 2 percent Al₂O₃. For Al₂O₃ >17 percent, there is an additional trend of very high Al samples (labeled C, Fig. 2A), which are samples from a near-surface layer of partially exotic material derived from a more felsic protolith that includes tuffs with small quartz phenocrysts and accessory zircon and xenotime. Some samples with high Al and Ti are saprolitic material derived from minor gabbroic intervals in the protolith.

The ultramafic protoliths are >70 percent serpentinized harzburgite and minor dunite. The distribution of data within the positive correlation trend exhibited by the saprolite samples shows two narrow, relatively highly populated trends of data (indicated by the arrows in Fig. 2A). Because the alumina in the saprolite is mainly derived from the orthopyroxene in the harzburgite protoliths, the simplest explanation of this pattern is that these two populations are a clear indication of two distinct harzburgite protoliths with two distinctly different, sharply defined modal proportions of pyroxene. Microprobe analyses of the orthopyroxene indicate 2.82 percent Al₂O₃ and 57.75 percent FeO. The only other source of Al is the accessory spinel in the dunite protolith with 27.65 percent Al₂O₃, 40.41 percent Cr₂O₃, and 19.26 percent FeO (Tav -chandjian and Golightly, 2010). The deduced proportions of primary igneous minerals are shown in Figure 2B. The diffuse, less populated band lying between these two trends and the Fe₂O₃ axis represent intervals with mixed harzburgite and short widths of dunite protoliths, a pattern commonly seen in ophiolites. The coherence of the two harzburgite populations with Al/Fe of 0.22 and 0.155, respectively, from about 15 percent Fe₂O₃ up to the limonite trend, clearly demonstrates that there is no relative movement of Al and Fe during saprolite development and it shows this with far higher precision than data from a single profile could ever do. At <15 percent Fe₂O₃ down to the bed-rock values near 8 percent, the two populations show a convex trajectory that suggests a small loss of Fe relative to Al during the initial stages of weathering of the protolith. This is most likely to be a loss in solution of Fe⁺⁺, the predominant form of Fe in the host rock. This is likely a loss from serpentine and olivine dissolution because pyroxene survives the initial stages of leaching in these saprolites.

The broad, steep negative covariation in the limonite population is a typical result of mobilization of Al and Fe due to solution in relatively acid, near-surface soil water, a process that led, in the La Gloria deposit, to formation of ˜1- to 3-cmwide goethite veinlets. The trend lies on a tie line between goethite and gibbsite. The high Al phase in the exotic cover is kaolinite and that trend lies on a tie line with a smaller slope between Al-rich nickeliferous limonite and pure kaolinite labeled K in Figure 2A.

Despite the fact that Fe is likely more mobile than Cr and Al, the latter two are present in much smaller absolute concentrations and Cr is particularly more subject to nugget variations due to its occurrence, in part, in chromian spinel. As a result, the precision with which patterns can be defined by Fe assays is superior. Therefore, Ni and Fe are used below to calculate the amount of rock represented by those metals in laterite profiles, and Ni, Co, and Fe are used to distinguish between supergene and residual enrichment in different zones in the weathering profile.

Given the concentration, C_i, of element, i, in a laterite profile of thickness, T, and the dry bulk density of the laterite, d, the thickness of parent rock, T_o, needed to supply the profile is given by:

T_o = Σ(TC_i d)/C_iod_o, where C_io and d_o are the concentration and density values for the protolith, respectively.

The sum is carried over the units of the profile or over depth. Estimates for Ni and Fe provenance in a number of deposits for which reasonably adequate data are publicly available are given in Table 2. Most are based on global average compositions of the different stratigraphic units of the profile without application of a Ni cut-off grade. A 1-m thickness of typical limonite, the most leached product in a profile, grading 68.6 percent Fe₂O₃ and 1.26 percent Ni, requires 3.76 m of protolith to supply the Fe and only 2 m to supply the Ni. The concentration of 3.76 m of rock into 1 m of limonite involves a 73 percent volume loss, much more than for the weathering of any common rock type except for limestones and dolomites. This volume loss can result in the development of karst topography on peridotites.

The estimates in Table 2 are minimal for two reasons. First, the global estimates may include incomplete profiles near the margins of the deposits, and second, in general, losses can be expected. Preferential loss of iron may potentially be accounted for by erosion from the top of the profile or loss of Fe⁺⁺ in solution from the bottom of the profile. Erosion of limonite may remove more Fe than Ni, which may already have been mainly leached from the limonite and concentrated in the preserved saprolite. If the water table lay high in a limonitic profile, then Ni may be selectively lost in ground water that has not filtered through the saprolite. Alternatively, a transported limonite may have an excess of Fe due to the Ni depletion. As shown in Figure 3, the ratio of rock thickness required to supply Ni and Fe in deposit-hosting profiles differs in the three scenarios of climate discussed in detail below.

Topography has played an important role in the exploration for and discovery of nickel laterite deposits. A main objective of this paper is to understand the origin of this control. The position and flux of water through the water table and its interaction with an evolving landscape drives the movement of material within in situ laterite deposits. Three common regional scenarios of climate and tectonics are used to explain much of the variation:

1. Wet scenarios characterize tectonically active areas with high relief and permanently wet profiles under a rain forest. These profiles are usually located on hilly terrane or plateaus above subduction zones along active continental margins. Saprolite zones have strong, true supergene enrichment of Ni, and nontronite and/or siliceous zones are rare.

2. Wet-to-dry scenarios are defined by peneplains with a history of increasing aridity. These have no characteristic topographic expression on the host peneplain. The saprolite zone has minor supergene Ni enrichment and the transition zone has supergene Co and Mn enrichment at a redox front in a siliceous limonite and/or nontronite zone.

3. Dry-to-wet scenarios are located on dissected peneplains in a formerly arid setting that has evolved into a permanently wet one. This may typically be on an elevated peneplain remnant. The saprolite zone has true supergene Ni enrichment and may have inherited silica or nontronite and supergene Co and Mn enrichment zones.

The overall distribution of Fe, Ni, and Co indicates the processes involved. The variation of Ni and Fe in profile zone averages is shown in Figure 4. Saprolites from the permanently wet category have a high Ni/Fe ratio, which is much greater than the protolith value and indicates strong supergene enrichment. The Ni/Fe ratio for the wet-to-dry class of saprolites is roughly the same as the protolith, indicating minor supergene enrichment and that for the dry-to-wet category is variable from minor to strong supergene enrichment.

Most deposits in the permanently wet setting have Co/Fe ratios close to the protolith value (Fig. 5), but the Goro deposit is a higher exception. Some deposits in the other two settings have higher Co/Fe ratios than their protoliths, at least in part of the profile, indicating supergene Co enrichment at a redox front. Generally, the red limonite and duricrust zones show a decrease of Co with Fe. A problem with broad application of the Co/Fe ratio is the poor definition of the bed-rock Co values because they are usually present in concentrations near to the detection limit of routine assays. It would be better to use Mn, but that is not commonly available in published averages.

The wet rain-forest settings mainly host the deposits developed on Tertiary ophiolite complexes in tectonically active areas such as New Caledonia, Sulawesi, Indonesia, and Guatemala. Ground water is generally Mg and Si rich, each in concentrations of about 25 mg/l, and has an alkaline pH >8 (e.g., Trescases, 1975; Gleeson et al., 2003). Rainfall is typically ≥2 m per year and pore space is nearly saturated with water, even above the water table. Vegetation is rain forest with local savannah. Some features common to nickel laterite profiles in this environment are illustrated in Figure 6. Peridotite saprolite is comprised of joint blocks or fragments in breccias with an unweathered core and a porous saprolite crust. The saprolite crust increases in thickness and the corestone diameter typically decreases upward in the profile. Serpentine increases in Ni and Fe and decreases in MgO concentrations from the core to the bounding fracture. Olivine decomposes first, generally being replaced by smectites, by quartz, or by an amorphous ferric-silicic phase that, in turn, varies gradually in composition outward (Golightly, 1979a, 1981). Pyroxenes may be replaced by talc or smectite or may weather to voids lined by limonite and a small amount of smectite. Fractures and open space in breccia zones are lined by zoned, colloform talc-kerolite garnierite that shows a progressive increase in Ni/Mg ratio during precipitation (Golightly, 1981; Cluzel and Vigier, 2008; Wells et al., 2009). Garnierite is crosscut by veinlets of colloform chalcedonic quartz. Quartz veinlets are generally minor and tend to be concentrated in breccia zones with a large amount of open space or in deposits developed from unserpentinized peridotite. The overlying limonite is a ferruginous saprolite in which the inherited protolith structures and textures are typically compressed to form a foliation parallel to the upper contact of the peridotite saprolite. The ferruginous saprolite hosts asbolite veinlets and concretions, in many cases nucleated on magnetite originally formed during serpentinization of the protolith. Rarely a smectite clay zone forms a transition zone between the peridotite saprolite and ferruginous saprolite.

The bulk composition of a profile through a silicate saprolite follows an exponential upward increase of Fe, Cr, and Al, and complementary decrease of Mg and Si. The Ni/Mg ratio also increases exponentially upward (Golightly, 1981). As can be gauged by the variation of the Ni/Fe ratio of the typical serpentinite profile (Fig. 4), Ni is leached from the limonite zone and the top of the peridotite saprolite and precipitated deeper in the saprolite zone. Therefore, Ni grades are generally highest in the saprolite zone decreasing upward toward very low values of about 0.5 percent in the most evolved duricrusts and downward to 0.2 to 0.3 percent in the parent rock. The top of the saprolite zone forms a discontinuity in Mg and Si distribution, where the silicate minerals have been completely exhausted.

The upper and lower contacts of the peridotite saprolite are highly variable. The local vertical relief of the lower contact of the saprolite can have slopes of about 25 to 50 percent over horizontal distances of tens of meters. Reflecting this topographic variability, the geostatistical range of intersample grade correlation, in variograms along the saprolite zone, is also short. The range of Ni-grade variograms is typically 12 to 50 m (Horn and Bacon, 2002; Murphy et al., 2004; Golightly, 2005).

Deposits in the Massif Sud of New Caledonia provide classic examples of topographic controls on laterite development under the wet scenario (Fig. 7). The topography of the ultramafic massif is the result of the progressive uplift and dismantling of a Tertiary plain (Trescases, 1975). In the mountainous northwestern part of the massif, the uplift is to >900-m elevation and terrane forms forested V-shaped mountain valleys with slopes of >20°. Nickel laterites are preserved on fragments of the uplifted and dismantled peneplain. Deposits, primarily of very high Ni grade peridotite saprolite, are found on plateau remnants, on ridge crests, and on topographic spurs with slopes of as much as 20°. Locally small karstic basins occur on these surfaces (Latham, 1986). Examples of the classic relationship of saprolite deposits to topography from the northern mountainous parts of the New Caledonian southern massif can also be seen at Koniambo on the west side of New Caledonia (Latham, 1986; Murphy et al., 2004), in the Philippines at Acoje (Daniels, 2008) and Agata (Cox, 2008, 2009), and in the Soroako district, Sulawesi, Indonesia (Golightly, 1979b).

In the southeast, the Tertiary plain is only slightly uplifted so that it forms a plateau covered by thick laterite at 200- to 250-m elevation. It is divided into 3- to 5-km-wide basins that are divided by rocky ridges rising typically 100 to 200 m. The basins have been interpreted as peridotite karst polje (Trescases, 1975; Genna et al., 2005) and the terrane commonly described as karstic. Many have an elevated water table and many ponds and lakes fill sinkholes in the thick (5−10 m) duricrust. The most southeastern basin, Goro, is well drained and hosts the giant Goro nickel laterite deposit. According to the metal balance in Table 2, the Goro deposit developed from 86 to 127 m of harzburgite protolith to provide the Ni and Fe, respectively, in the complete laterite profile (based on data in Horn and Bacon, 2002). This is a rock column that would rise 45 to 87 m above the current surface of the plateau, the latter number coming close to the elevation of the rocky ridges bounding the basin on the south.

Karst topography results from the large volume loss associated with the weathering of the peridotite. Although they are best developed here, depressions associated with nickel laterite are not confined to New Caledonia. They also are found in the Soroako district, Sulawesi, where they form large closed marshy depressions filled with transported laterite and organic detritus (Hope, 2001) and in deposits on plateaus above 500-m elevation in the Moa-Baracoa ophiolite in Cuba (Golightly et al., 2008b, 2009a).

The La Gloria nickel laterite deposits in Guatemala, (Harju, 1979; Tavchandjian and Golightly, 2010) are located on an undulating plateau at 600- to 1,000-m elevation and on terraces along the linear southeastern edge of the plateau facing the Lago de Izabal. The terraces and topographic spurs hosting the individual deposits of the La Gloria area (Fig. 8) are thought to represent graben-bounding faults. Laterite preserved on the plateau requires approximately equal amounts of protolith weathering to supply the Ni and Fe. On the terraces, however, where the Ni grades are highest, the amount of rock needed to supply the Ni exceeds the amount required for the Fe. An exotic, partially tuffaceous cover unit is preserved on the plateau and the parts of the terraces most protected from erosion. The pattern is consistent with increasing erosion of the in situ limonite toward the southeast.

The mined-out footwall of the saprolite orebodies of the historic Exmibal deposit (Fig. 9A) shows the typical variability of the lower contact of the saprolite zone. The water table underlying the terrace that hosted the deposit lies at 50-m depth below the premining surface.

Nickel laterite deposits of the Yilgarn craton (Fig. 10) occur in the cumulate olivine-rich facies of Archean komatiites or olivine-rich intrusions. The mined deposits, including Murrin Murrin, Cawse, and Bulong, are located in the central to northeastern part of the craton, with the exception of Ravensthorpe, which lies near the southern edge.

The deposits are characterized by strong horizontal layering of siliceous zones and precipitation of magnesite in the saprolite zone. They are dominantly either siliceous limonite (Cawse and Ravensthorpe) or nontronite (Murrin Murrin and Bulong), but both features are present in each. The Ni-enriched zone at Bulong is comprised of saprolite derived from serpentinite with centimetre-scale subhorizontal chalcedony veinlets, overlain by a nontronite zone and above that by a goethite + kaolinite ± quartz (limonite and ferricrete) zone (Burger, 1996). Although zones of Mn oxides and silica veinlets tend to be very flat-lying, the zonation in the saprolite and/or nontronite is quite irregular. Nontronite zones are not continuous but commonly pinch out on saprolite pinnacles. At Cawse, a series of northeast-trending ˜45°-dipping shears has a spatial relationship to the location of Ni enrichment in the deposits.

Supergene Ni enrichment in the saprolite or nontronite zones is comparatively minor (Fig. 4). A zone of supergene enrichment of Co + Mn oxides occurs at the base of the limonite or in the nontronite zones. Cawse, in particular, has some very high Co/Fe ratios that are much greater than of any feasible protolith values (Fig. 5). At Cawse, a strong correlation of Ni to Co and/or Mn suggests that some true supergene Ni enrichment may be due to coprecipitation as oxides, with little to do directly with Ni enrichment in silicate minerals.

The Yilgarn craton is an area of generally very low relief that has experienced increasing aridity as the Australian fragment of Gondwana moved northeast through the Tertiary into the tropical arid latitudes. The present climate in the northeast part of the craton is arid, with a sporadic annual rainfall of about 0.2 m and the water table at about 35-m depth. The region is characterized by a network of dry river channels draining into major dry lakes. A low-temperature (<300°C) paleomagnetic component at Murrin Murrin is consistent with late Miocene to Pliocene weathering (Wells and Butt, 2006). The δ¹⁸O for kaolinite, ranging between 13 and 16 per mil (Wells and Butt, 2006), is consistent with the more southerly pre-middle Tertiary latitude of Australia (Bird and Chivas, 1989).

In situ deposits in the craton do not show a consistent association with topographic highs or with any particularly distinctive topographic feature. In the Murrin Murrin area, the maximum relief of 215 m is provided by a granite. The Murrin Murrin South deposit lies under gently undulating terrane where weathered ultramafic rocks are exposed as subparallel, elongate, northeast-trending ridges and isolated mesas and buttes (Wells and Butt, 2006). The Murrin Murrin North and Cawse deposits lie under the edge of a dry, Tertiary river channel that deposited a layer of alluvium derived from a felsic igneous source. The surface expression of Bulong is a 10- to 20-m-deep “karstic soup-plate” depression (Burger 1996). Ravensthorpe lies at 200-m elevation under the top of a very gradual rise, about 50 to 100 m above the general level of the surrounding terrane.

The nickel laterite deposits on the Brazilian shield, in and adjacent to the Araguaia-Tocantins watershed (Fig. 11), are examples of deposits formed in environments that have changed from arid to wet climates. Today the natural cover of this region is rain forest with annual rainfall of 1.5 to 2 m. However, the Brazilian shield must have experienced increasing erosion and rainfall as the South Atlantic Ocean opened ca. 120 m.y. Immediately before the initiation of sea-floor spreading, the Brazilian shield was covered by sand dunes (Scherer, 2000). During the initial stages of opening, extrusion of the Parana and Parnaiba flood basalts to the south and northeast of the Araguaia-Tocantins basin, was followed by deposition of a variety of Cretaceous formations. Features such as calcrete deposits, playa lake sediments, desert ventifacts, and eolian sands are preserved in the Sao Francisco basin east of the Araguaia-Tocantins basin (Verner Inda et al., 1984) and are evidence of an arid climate. The Mata de Corda Formation includes flows and pyroclastic rocks believed to be the extrusive equivalents of a province of alkalic intrusions dated at 87± 1.5 Ma (Sonoki and Garda, 1988). The ultramafic members of the alkaline complexes host nickeliferous laterites in Western Goias and São Paulo states, including the Santa Fe deposit described below. Cretaceous formations are preserved in some regional basins and as caps to relatively isolated mesas. This late Cretaceous to Eocene surface, known as the South American surface (King, 1956), forms isolated mesas and plateaus in many locations. In the absence of Cretaceous cover, such as in the Carajas highlands of Para, weathering profiles have been dated by Ar⁴⁰/Ar³⁹ on Mn oxides at 69 to 65, 56 to 51, and 43 to 40 Ma (Vasconcelos et al., 1994). At elevations lower than this surface, the more extensive Velhas surface is similarly dated as mid-Tertiary. Both the South American and Velhas surfaces host many Ni laterite deposits. Although the deposits have not been directly dated, the surface ages are indicative at least of latest Cretaceous to mid-Tertiary ore formation.

The region hosting the deposits lies in the basins of the Tocantins and Araguaia Rivers, which drain northward to the vicinity of the mouth of the Amazon River near Belem. The Tocantins-Araguaia basin is mainly underlain by Archean to Late Proterozoic rocks of the Amazonas and Sao Francisco cratons and the intervening Araguaia fold belt. The chemistry of sediments preserved on the Ceara Rise in the Atlantic Ocean derived from the Amazon has been interpreted (Dobson et al., 1997) as comprising three terrigenous components. These include two early periods of Archean provenance, a late Tertiary period derived from Phanerozoic sediments of the Amazon basin, and a post-6 Ma influx of Andean-sourced sediments. The two periods of Archean-sourced sediments center on 50 and 30 Ma and thus coincide with the ages of the South American and Velhas surfaces, respectively. These surfaces and their nickel laterite deposits originated during a period of relatively rapid erosion of the Archean shield in the Amazon and Araguaia basins, which was followed by a period of relatively slow erosion.

Generally, the South American peneplain remnants lie 100 to 400 m above the surrounding terrane. Deposits on the peneplain remnants are characterized by major siliceous saprolite and/or siliceous limonite zones. The siliceous zones occur either as an extensive, subhorizontal layer, or silica cap, such as at Onça, Puma, Vermelho, Jacare, Morro do Engenho, Agua Branca, and parts of Ipora, or in and bordering steep faults or fracture zones as at Barro Alto, Niquelândia, Serra do Tapa, Vale dos Sonhos, and Quatipuru. (Baeta Júnior, 1986; Berbert, 1986; Heim and Werneck, 1986; Pedrozo and Schmaltz, 1986; Werneck and Mattos, 1986; Silva and Oliveira, 1995; Carlon et al., 2006; Falconbridge Brasil, 2006; Dreisinger et al., 2008). Deposits on the younger lowlying surface have less relief and relatively little siliceous material, albeit still more than most wet scenario deposits. The Onça-Puma, Santa Fé, and recently discovered Serra do Tapa and Vale dos Sonhos deposits are presented as examples.

The four nickel laterites hosted on the 2380 Ma Catete suite of layered ultramafic rocks (Macambira and Ferreira-Filho, 2002) intrude into rocks of the Archean Xingu Complex of the Amazon craton in the Carajas region of Para, are examples of the silica cap deposit type. The following description of the Onça and Puma deposits is based on Heim and Werneck (1986), Grant and Majorkiewicz (2003a, b), and Sucharda (2005).

The Onça and Puma deposits and their lithologic hosts, although they are being developed by Vale as a single minesmelter complex, are not identical. In both cases, the ultramafic host rocks are south-dipping layered intrusions. The basal units are strongly serpentinized cumulus dunite with an orthopyroxenite layer near the top of the dunite. At Onça, a second pyroxenite is followed upward by anorthositic gabbro and gabbros. The dunite at Onça has a variable, high chromite content, whereas Puma does not. There are a variety of minor occurrences of wehrlite, harzburgite, and rhyolite only at Puma. Also, importantly, at Puma, pyroxene has been replaced by chlorite, which also forms pegmatitic veins. The presence of abundant chlorite is a feature in common with the Vermelho deposit (Schwab et al., 1985; Silva and Oliveira, 1995) and the Santa Fé deposit described below.

Igneous layering in the Onça and Puma intrusions dips 45° to 60° south. At Onça, jointing is related to dip parallel foliation, but at Puma some of the foliation appears to be a local exfoliation surface. The Onça ridge is a cuesta landform, with the steep surface dipping north, whereas the Puma ridge and mesa generally slope steeply in both directions.

A Landsat image composite (Fig. 12) shows the distinctive topographic landforms of the Onça and Puma deposits. The deposits are not uniformly covered by siliceous limonite. Onça has a major siliceous limonite cap covering the eastern one-third and a small area on the west-central part of the deposit. At Puma, there is one extensive siliceous limonite mesa covering the central two-thirds of the deposit and two smaller areas west of a gap in the ridge. Elsewhere, the laterite profile is incomplete, mainly comprised of serpentinite saprolite, with or without a minor amount of limonite that is exposed primarily in the western and central parts of Onça and the eastern and western extremities of Puma. In detail, the silica cap is a rounded mesa. The siliceous zones, presumed to be originally flat, are partially broken up by erosion and tend to comprise a jumble of blocks of siliceous limonite, capping the ultimate mesa landform. The cap at Onça is highest in the east-central part of the deposit, at nearly 500-m elevation (Fig. 13). The elevation of the edges of the cap is near 400 m. The ridge line is inclined toward the east, but the silicacapped landform drops very sharply to the north and east to 260 to 330 m above the base of the slope.

The overall composition of each of the visually logged weathering zones at Onça and Puma is given in Table 4 (Grant and Majorkiewicz, 2003a, b). Based on this average profile, the Onça laterite required 34 m of dunite to supply both the Ni and Fe, which is roughly 12 m of additional dunite above the current ground surface. Puma West required a similar amount to supply Ni but less for the Fe, probably because the limonite was eroded. Because serpentinite saprolite crops out extensively in the western parts of Onça and Puma, erosion may make these thicknesses a considerable underestimate. The base of the saprolite profile has a similar relief to that observed in modern wet profiles and a similar short range of correlation of Ni grades along the layers. However, unlike deposits in the wet scenario, the core stones in the saprolite are altered to the core. Fresh rock cores are not seen in the ore zone, not even in the exceptional boulders that show spheroidal weathering joints (Fig. 9A). Supergene Ni enrichment is restricted to serpentinite saprolite; pyroxenite saprolite shows no significant enrichment.

The low-lying Santa Fé deposits and the similar Ipora deposits are hosted on alkalic complexes in southwestern Goias state. The Santa Fé deposits protolith is an 87± 1.5 Ma alkalic dunite and/or clinopyroxenite complex (Sonoki and Garda, 1988), about 4.5 × 6 km in extent (Fig. 14). Clinopyroxenite forms a discontinuous, ≤500-m-wide zone surrounding the dunite, accompanied by minor 500- × 100-m gabbro bodies (Dreisinger et al., 2008). The dunite, apart from a small central zone several hundred meters in diameter, is nearly totally serpentinized and is cut by numerous small, very coarse grained vermiculite bodies. Jointing of the igneous rocks is predominantly flat-lying.

The weathering zones, including locations outside the identified resources, have the global compositions in Table 4, which were used to calculate the rock column needed to supply the laterite (Table 2). Serpentinite saprolite is comprised of Ni- and Mg-rich serpentine, Ni vermiculite, and talc or chlorite after pyroxene, overlain by pisolitic limonite.

The Santa Fé deposits lie on a slight rise of 40 to 60 m emerging through the Tertiary unconsolidated, alluvial Araguaia Formation. The laterite-covered terrane is comprised of a plain that gently slopes away from tors or inselbergs of dunite. Their Ni and Fe were concentrated from an average of about 14 m of dunite and peridotite. This thickness of dunite is only about a quarter of the 50- to 100-m height of the inselbergs (Fig. 14). The position of the water table was monitored for more than 1 year in boreholes distributed at 200-m intervals across the Santa Fé and Ipora deposits. The Ni-enriched saprolite lies entirely within the vertical range of fluctuation of the water level. The relief on the contacts of the saprolite zone is very small, <10 percent slope, and the geostatistical range of Ni grade correlation in the mineralized zone is about 120 m, much longer than the range in most saprolite deposits. At Santa Fé, saprolite derived from pyroxenite in the vicinity of dunite protoliths is strongly enriched in Ni compared to the protolith.

The Serra do Tapa and Vale dos Sonhos deposits (Falconbridge Brasil, 2006) are recent discoveries in the Araguaia fold-and-thrust belt of the Late Proterozoic. The two deposits are hosted in harzburgites in a mafic volcanic rock complex of oceanic origin. The Serra do Tapa deposit is about 6 × 1 km in north-south extent and lies on both sides of a ridge, with adjacent zones of moderately east dipping silexite interpreted to represent a fault zone, rather than a part of the nickel laterite profile. It is a very thick saprolite (Fig. 15), as much as 80 m, capped by as much as 10 m of red limonite and duricrust and extends down to the water table. A very unusual feature, a zone of anomalously FeO-rich “orange Tapa” located in the deepest part of the saprolite may be related to the water table. The Vale dos Sohnos deposit is about 3 × 1 km in extent, with thinner saprolite. It lies under relatively low terrane on the east side of a silexite ridge and close to the regional base level. The SiO₂/MgO variation of graded intercepts indicates that the Vale dos Sohnos saprolite with ˜33 percent SiO₂ and ˜18 percent MgO, at ˜29 percent Fe₂O₃, is close to a pure serpentine saprolite profile, in contrast to Serra dos Tapa with ˜39 percent SiO2 and ˜18 percent MgO, at ˜24 percent Fe₂O₃, which has an SiO₂/MgO ratio that is more indicative of a mixture of serpentine saprolite and a smectite zone.

For the wet scenario, several facts indicate rapid leaching by undersaturated meteoric water. Strong supergene Ni enrichment occurs within topographically elevated locations. Occurrence of quartz is minor and, in New Caledonia, smectite zones are restricted to lowland profiles in the humid Massif Sud (Trescase, 1975) or on plateaus on the relatively dry leeward southwest side of the island, such as at Tiebaghi, (Latham, 1986). In Cuba, smectite zones are found in the low-lying deposits of Nicaro (Linchenat and Shirakova, 1964).

The classical concept of saprolite formation in joint systems that antedate weathering is shown in Figure 6. The joint blocks were computer generated using a probabilistic recursive fracturing algorithm and a saprolite crust with thickness exponentially decreasing with depth superimposed on each block. The scheme produces valleys in the top and bottom of the saprolite, where the joint blocks are small leave ridges between them. The model predicts that the contacts of the saprolite zone show rapid lateral variation in vertical position, the lower contact more variable than the upper. They are effectively etched surfaces. If the bed-rock surface lies well above the water table, then the joints allow meteoric water at a slightly acid pH to seep down though the system of joint blocks, leaching Mg and Si from the blocks and forming a porous saprolite crust on the blocks. At the base of the system, the water finally becomes saturated in serpentine and attains a high pH of about 9. Because Ni serpentine is less soluble than Mg serpentine, Ni/Mg exchange leads to a progressive Ni enrichment outward in the crusts and upward in the profile. Trescases (1975) showed that, on the basis of the balance of solutes in drainage basins in the highland areas of the Massif Sud of New Caledonia, the rate of advance of the weathering front to be 29 to 48 m/m.y. Similar terrane in areas underlain by intermediate and felsic rocks in the Rio Icacos basin of Puerto Rico gives a similar rate of 58 m/m.y. (White et al., 1998) on the same basis or 43 m/m.y. (Brown et al., 1995; Riebe et al., 2003) based on accumulation of cosmogenic Be10. Apparently, the weathering rate in such terrain in New Caledonia is not unusually high because of the higher solubility of peridotites; the equal rate may result from processes dependent on the similar rapidly evolving topography.

The development of the Icacos weathering profile was modeled quantitatively with a steady-state, “conveyor-belt” model in which the rate of removal of soil at the top of the profile exactly equals the rate of advance of the weathering front at depth (Turner et al., 2003; Fletcher et al., 2006). In their models, expansion of biotite accompanying initial oxidation causes the formation of exfoliation cracks, speeding access of water and oxygen to the fresh rock cores by an order of magnitude. In nickel laterite hosts, formation of serpentine or smectites from olivine or pyroxenes, as part of the initial weathering process, could provide the expansion required for exfoliation, but there is no process to do this in a serpentinite. Exfoliation structures are known, but rarely reported, in nickel laterite saprolites. They are reported in diorites associated with the nickel laterites of the Sarçimen area in eastern Anatolia (Çolakog˘lu, 2010). Exfoliation shell structures are present in gabbros and gneisses in the Onça and Puma deposits (Fig. 9B). They are seen in coarsely jointed serpentinite in these deposits, at Soroako, and at the Koniambo deposit in New Caledonia. However, they are exceptional, so it seems unlikely that the exfoliation fracturing process contributes strongly or is essential to rapid nickel laterite development.

In some cases, blocks are created by tectonic processes resulting in breccia zones with high fractions of open space available as a site for garnierite and quartz precipitation. Such breccias host very high grade garnierite ores in New Caledonia, at Soroako in Sulawesi, and in a few other locations. Structural evidence (Genna et al., 2005; Cluzel and Vigier, 2008) indicates that the breccia zone formation underlying saprolite deposits in New Caledonia is synchronous with laterite formation. The brecciation is caused by hydrofracturing at or below the base of the profile, where meteoric water drains through the weathering profile into karstic aquifers and is evident by the formation of karstic depressions, or dolines, in the overlying topography. On the other hand, there are no clear karst features at the La Gloria deposit, Guatemala, Locally waterlaid sediments comprised of kaolin clay and sapropelic peat indicate the presence of a former pond, perhaps in a sinkhole. On the higher fault terraces (Fig. 8), narrow crevasses filled with water-sorted saprolite and sequences of in situ laterite in fault contact above transported limonite are seen in drill cores. All these features indicate a complex history of fault movement contemporaneous with the laterite development in New Caledonia and Guatemala.

In summary, for the tectonically active wet scenario, the strong supergene enrichment of Ni in peridotite saprolite (Fig. 4), the relatively low Co/Fe ratios (Fig. 5), and the common lack of a smectite zone are consistent with rapid leaching of the profiles by downward-percolating meteoric water, which does not attain saturation with metals until the water table is reached in the lower part of the saprolite zone. The hilly topography and the local underlying karst channels and fault zones provide this drainage.

For the peneplain-hosted wet-to dry-deposits, rainfall is low and runoff is typically high. In the Yilgarn craton, rainfall is 0.2 m annually, an order of magnitude less than the wet scenario described above. Nickel laterite development is attributed to a more humid climate during the Tertiary in a more southerly latitude (C. Butt, writ. commun. 2010). Current water tables lie well below the Yilgarn deposits, at an average depth of 20 to 50 m (Gray and Noble, 2007). The reason for the minor amount of supergene Ni enrichment is likely the evolving relationship of the water table to the terrain surface. In order for supergene enrichment of Ni to occur, Ni has to be released by decomposition of olivine, serpentine, chlorite, and/or nontronite. Ni can also be concentrated by dissolving and reprecipitating in limonite or precipitating in exchange for Mg that is released by the evolution of the saprolite. This cannot happen below the water table if it lies above the silicate saprolite contact because the Ni released in saprolite formation has to move upward into a zone of lower pH and Mg, preventing the normal silicate precipitation mechanism described above.

There are two possibilities leading to enrichment of Ni in the saprolite: Ni enrichment when the terrane was hilly at an earlier stage of topographic development of the peneplain, or downward leaching of Ni to a water table that was starting to lower as the climate began to evolve toward the present arid state. Preexisting hills are possible at Ravensthorpe and perhaps Murrin Murrin, but if they had existed at Cawse or Bulong, then they would have been overprinted by the alluvial channel overlying Cawse and the “karstic soup plate” hollows at Bulong. In these two cases, it is possible that the negative topographic features focused the waning water supply into the deposit (Freyssinet et al., 2005). The complexity of the saprolite and/or nontronite contacts at Murrin Murrin is thought to result from leaching in early, much wetter conditions. The relatively weak supergene Ni enrichment in saprolite in this scenario (Fig. 4) is consistent with downward leaching through the saprolite being a relatively minor phenomenon. Some profiles from this scenario show increased supergene ratios of Co/Fe relative to protolith values (Fig. 5). Upward movement of Ni from the saprolite zone, along the inclined shear zones, to precipitate at the water table may explain Ni grade variation in a Co- and Mn-rich zone at Cawse (Brand et al., 1996). A Co-Mn enrichment horizon at Ravensthorpe has been labeled as a redox front (Miller et al., 2004). It seems likely that Co, Mn, and Ni, present in the aquifer in the saprolite zone, diffused upward to precipitate at the redox front at the water table. The horizontal zones of siliceous saprolite at Bulong, Cawse, and Ravensthorpe, and the nontronite clay zones at Murrin Murrin and Bulong are consistent with deposition from saturated solutions either at the water table or under a waning water supply.

For the dry-to-wet dissected peneplain deposits, the importance of silica deposition and the general replacement of pyroxenes by smectites, talc, and chlorite-vermiculite are consistent with leaching by saturated waters. The importance of quartz deposition is consistent with precipitation at a shallowwater table on the peneplain at the beginning of deposit formation. The strong supergene Ni enrichment results from downward leaching in the hills and mesas left by stream erosion and the lowering of base level. The evolution of the deposits (Fig. 16) may take place either with or without development of a siliceous zone.

The formation of the siliceous deposits was interpreted in the Tocantins-Araguaia drainage and elsewhere on the planalto of Brazil (Melfi et al., 1988) as beginning with the near-surface deposition of chalcedonic quartz at the water table in the weathering profile of serpentinites. The stability fields in the phase diagram (Fig. 1) suggest that the pH must have been relatively low, and thus unlike the alkaline conditions caused by the release of Mg by the weathering of serpentine and olivine near the base of the saprolite zone.

The Onça and Puma deposits are close enough to one another to obtain a better idea of the original plain. The deposits lie close to the Catete River, which probably was also located within a few kilometers of its present course on the original peneplain (Fig. 12). The river now flows through a water gap in the Serra Arqueada, a ridge of iron formation and volcanic rocks that lies between the Onça and Puma deposits, with a crest that is generally higher in elevation than the ridges that host the deposits. Possibly, the ancient Catete River flowed across the current silcrete mesas at the two deposits and then, as downcutting proceeded, it was deflected toward its present course by the siliceous limonite and saprolite; this seems most likely at the northeast end of the Puma deposit. The elevation difference between the Catete River valley and highest points on the Onça and Puma siliceous limonite mesas is about 260 to 285 m. If the laterite started developing at 60 Ma, then the average rate of downcutting must have been about 4.5 to 5 m/m.y and likely greater at the beginning of erosion during both the Late Cretaceous-Eocene and Miocene surfaceforming events.

If silcrete did not form and protect the weathering profile, perhaps in initially more elevated parts of the peneplain, then erosion could eventually reduce the ground level close to the water table with the depth of the water table being limited by the base level of erosion on nearby streams. If erosion is sufficiently slow, as would be expected in the later stages of downcutting, then the saprolite will not be directly eroded and Ni will continue to be enriched from the decomposition of the overlying limonite. If the evolution of the laterite attains base level, so that the base of the saprolite zone reaches the water table, then further downward advance of the weathering front is slowed. Subsequently, the bed-rock pinnacles are completely altered and the saprolite zone flattens out conforming to the relatively smooth geometry of the water table. The Santa Fé and Ipora deposits in Goias, Brazil, provide examples of these features.

A second consequence of the confinement of the Ni supergene enrichment zone at Santa Fé to the water table is the enrichment of pyroxenite-derived saprolite in Ni by lateral movement along the water table from dunite-derived saprolites. This is a feature in common with the Niquelândia deposit (Pedrozo and Schmaltz., 1986; Colin et al., 1990) and a difference from the Onça and Vermelho deposits, where pyroxenite-derived laterite is barren and relatively unweathered. The lateral movement of Ni at Niquelândia is not so clear. Because the pyroxenite units dip about 60° to the west, there may have been overlying peridotite contributing Ni to pyroxenite-derived laterite as well.

The deposits in the Araguaia Tocantins watershed lie on topographic remnants of the ˜70 Ma South American surface elevated by about 200 m above their immediate surroundings. If this elevation was produced during the time when the landscape was lowered to the level of the 25 Ma Velhas surface, then this took place at a rate of about 6 m/m.y. Current rates of erosion determined using ¹⁰Be accumulation in various areas in Brazil range from 2.5 to 9 m/m.y., the same range as suggested by the topographic expression (Braucher et al., 1998).

Nickel deposits in laterites were first discovered by Jules Garnier in New Caledonia, in 1864, in very high grade garnierite veins. Mining started in 1875. Nickel began to be recognized in laterites elsewhere in the world in the early 20^th century, typically developed in large ophiolite complexes in New Caledonia, Cuba, and the Philippines, or over layered mafic complexes such as Niquelandia in Brazil, but interest in these as significant Ni resources only started in the mid-1930s. In some cases, Ni was discovered in limonite used as iron ore; the brittle, undesirable qualities of Mayari pig iron from Cuban limonite were found in 1907 to result from the high Ni content. The development of the Nicaro mine and plant in Cuba 1942 to 1943 was a wartime project brought to production in about 1 year (McMillan and Davis, 1955) and brought the importance of nickel laterites as major resources to the forefront after the war.

Exploration boomed during the post-war decades until about 1980, with large programs leading to the discovery of most nickel laterites known today. A variety of methods successfully led to discoveries. Prospecting and mapping found many deposits, such as the Wingellina nickel laterite developed on the mafic-ultramafic Giles Complex in central Australia and Ambatovy in Madagascar. The United Nations Development Program stream-sediment geochemical surveys in Africa found the Sipilou and Moyango nickel laterites in Cote d’Ivoire, and the Musongati deposit in Burundi, in the early 1970s (Argosy, 1999), and the Co-rich Nkamounah (Cameroon) Ni laterite district about 1985. Air photo interpretation was used by the International Nickel Company (Allum, 1981) to select targets on the basis of topographic signature (discussed above) in large areas of the virgin Amazon rain forest in the early 1970s, leading to the discovery, in quick succession, of the Onça, Puma, Jacare, and the Vermelho laterite deposits. Most exploration programs concentrated on the major ophiolite complexes where laterite potential had been identified in the previous decades. In some cases, this was the result of following-up on samples collected by farmers, such as in the Lago de Izabal region of Guatemala (Harju, 1979) and the Brolga and Marlborough districts in Queensland, Australia. Komatiite-hosted nickel laterites started to be recognized in Western Australia after the discovery of the Kambalda nickel sulfide deposits in the late 1960s.

After low levels of activity through the 1980s, much new exploration was carried out in areas of Ni-rich laterites that were identified earlier and at projects that had been suspended during bad economic times. In many cases, exploration has consisted of raising inferred resources to indicated or measured status. However, in some examples, the renewed activity led to grassroots discoveries, in smaller ultramafic bodies of which the Serra do Tapa and Vale dos Sonhos deposits in the Araguaia belt of Para, Brazil, are the most important. Other grassroots discoveries, such as the Lontra and Vila Oito (Horizonte Minerals, 2009), are still being evaluated in the same belt, and the Dutwa deposit (African Eagle Resources, 2008) has been recently discovered in Tanzania. These demonstrate that, despite the relative ease of exploration for surficial deposits, exploration for this class of deposit continues to be productive.

The problem of target selection for nickel laterite exploration falls naturally into two categories dependent upon the size of the ultramafic bodies: major known ultramafic complexes or relatively small, hitherto unidentified ultramafic bodies. The former usually are large ophiolite complexes (e.g., in Sulawesi [Indonesia], Philippines, New Caledonia, Cuba, Oregon [USA], Guatemala) or major layered ultramafics (e.g., Musongati [Burundi], Giles Complex [Australia], Niquelandia and Barro Alto [Brazil]), which have been known for some time. If they had to be located today, most would be easily identified using satellite imagery. Finding the deposits in these was mainly a matter of recognizing the structural and topographic settings of known deposits elsewhere in the complexes. In the latter case, finding an ultramafic body and assessing its potential is the initial problem.

In general, bodies of ultramafic rocks with a high concentration of olivine may make suitable hosts. Forsteritic olivine in very primitive intrusions and volcanic rock sequences with Fo₉₀ to Fo₉₃ tend to contain 0.30 to 0.35 percent Ni and about 8 to 10 percent total Fe₂O₃. The approximately 16 percent H₂O that it takes to alter olivine to serpentine ± brucite lowers this grade to 0.25 to 0.30 percent Ni in derived serpentinites. However, although ophiolitic ultramafic rocks normally meet this criterion, some other ultramafic rocks may not and can be eliminated as permissive exploration targets if their Ni content is too low. The potential laterite Ni grade can be estimated as the ratio of Ni to the other insoluble components of the parent rock: potential Ni grade = Ni/(Ni + Al₂O₃ + Cr₂O₃ + Fe₂O_3(total)). Potential Ni grade decreases rapidly as a function of MgO. Ni = 0.000004852 MgO ^2.25 for a broad collection of komatiite data (Fig. 17). Actual grade for supergeneenriched saprolite would typically exceed this and for limonite it would be slightly less. A potential grade of 1 percent Ni corresponds to 25 to 35 percent MgO; in typical komatiites, this is the olivine cumulus part of flows or intrusions. They are typically serpentinized and the associated magnetite provides a magnetic anomaly.

In the past 10 years, some other geophysical methods, most notably ground-penetrating radar, have been used at the advanced exploration stage to define the thickness and depth of the silicate saprolite zone (Francke and Nobes, 2000). Not all ultramafic rocks are serpentinized and thus lack a magnetic anomaly; for example, large parts of Soroako, Indonesia (Golightly, 1979b) and Poro, New Caledonia (Besset and Coudray, 1978) are not serpentinized. Small ultramafic bodies in granulite terranes may not be serpentinized. Detailed magnetic surveys have been used at Soroako to map serpentinization.

Falconbridge’s (now Xstrata Nickel) Araguaia project (Falconbridge Brasil, 2006) lies in the Araguaia tectonic belt, a Late Proterozoic collisional zone between the Archean Amazon craton to the west and the Sao Francisco craton (Paixão et al., 2008). The presence of ophiolitic rocks in this belt has been known for many decades. There is widespread deep lateritic weathering in the region and the Quatipuru Ni laterite deposit, known since the 1970s, occurs in the southern part of the belt about 145 km from the new Serra do Tapa/Vale dos Sonhos discovery. The essential components of the program used by Falconbridge Brasil (2006) included a geologic interpretation of a regional Landsat image followed by detailed air photo interpretation to produce an enhanced geologic map. A 49,000 line km high-resolution airborne geophysical survey, including magnetics and radiometrics, was carried out. The high MgO + Ni ophiolitic peridotites, permissive for hosting Ni laterite, normally have a strong magnetic anomaly due to the formation of accessory magnetite accompanying serpentinization and would also be expected to have very low radiometric responses due to the lack of U, Th, and K in the ultramafic rocks. The high-resolution survey was important for accurately locating the joint magnetic high-radiometric low targets. Diamond drilling tested combined geophysical and geochemical anomalies and this led quickly to the discovery hole at Serra do Tapa, on October 7, 2004, in the second year of the program. Subsequent drilling has demonstrated two deposits, Serra do Tapa (5 × 1 km) and Vale dos Sonhos (3 × 1 km), with a combined resource of about 104 million tons (Mt) at 1.33 percent Ni (Xstrata , 2009).

Formation of the deposits listed in Table 2 required the consumption of approximately 41 m of protolith. At the rate of 6m/m.y., indicated by the 45-m.y. time gap between the South American and Velhas surfaces, the formation of the laterite deposits could have taken ⁠. This is a small part of the time interval available for the formation of the wet-to-dry and dry-to-wet categories of deposits. It suggests that the formation of the deposits may have happened rapidly following a major tectonic event that led to peneplain uplift followed by a relatively steady state of landscape evolution, with increasing rainfall or the approach to the base level of erosion. Saprolite deposits formed in wet, tectonically active, mountainous environments, where the observed basin-wide rates of weathering are almost an order of magnitude faster than rates in stable continental settings. This suggests that they are in a dynamic steady state, losing as much off the top by erosion as forms at the base of the profile, which is just slow enough to allow supergene enrichment of Ni to accumulate in the saprolite zone, while keeping limonite accumulation to a minimum.

In all of the climate scenarios, deposits fall roughly into two types: those that form above the water table with consequent strong topographic and fracture control and variability, and those that form down to and are controlled by the water table with consequent weak topographic control and uniformity. Deposits such as Santa Fé and Ipora show that the second is an attainable stage of development. In the absence of inselbergs or protective siliceous facies, the latter may have little or no topographic expression. Other low-relief nickel laterites are already being found in such terrane in the Araguaia belt. Exposed deposits that remain to be discovered seem likely to fall into this category and geophysical methods, such as the described combined magnetic and radiometric approach and electromagnetic measurement of weathering profile thicknesses are likely to be more important in exploration than the traditional methods of prospecting, mapping, and geochemical soil surveys.

This paper has benefited from very thorough reviews and suggestions by Charles Butt and Sarah Gleeson.