Stream sediment geochemistry in mineral exploration: a review of fine-fraction, clay-fraction, bulk leach gold, heavy mineral concentrate and indicator mineral chemistry

Stream sediment geochemistry is an important geochemical exploration tool used to characterize geology and geochemical variation, and to make discoveries in mineral exploration. This paper reviews the use of stream sediments as a sample medium for geochemical and mineralogical surveys in global terrains that may be conducted by the mineral exploration industry or government agencies. The review is one in a series of special papers published by this journal on sampling and analytical methods for various geochemical sample media that include waters, stream sediments, lake sediments, soils and vegetation globally.

Stream sediments have been used in mineral exploration for millennia, as one of the older mining and prospecting methods, advancing from gold or diamond panning to modern geochemical approaches (Ottesen and Theobald 1994; Stendal and Theobald 1994). Stream sediments have historically provided the most cost-effective sampling media to cover broad areas of appropriate terrain for baseline geochemical surveys to vector towards areas which warrant further investigation. Modern analytical techniques, improved analytical instrument capabilities and new research in indicator mineral chemistry have increased the contribution of drainage sediments to understanding regional geology and prospectivity (Averill 2009; Cooke et al. 2020). Lower analytical detection limits have defined geochemical background, improved geochemical contrast and allowed the use of more varied trace elements as pathfinders. The ability to map with those trace elements has improved the value, precision and accuracy of sediment surveys. New mineralogical instruments and software have allowed automation of mineralogical analysis, providing quantitative detail around mineralogy and mineral associations. Application of micro-analysis of resistate minerals now allows for a generation of new data for sediment grains, including age, trace element and isotopic compositions (O'Reilly et al. 2004; Chang et al. 2011).

This review summarizes the current best practices of stream sediment survey implementation and updates the reader with new developments in the discipline. A series of case histories demonstrate effective use of stream sediment programmes to define regional geochemical provenance and describe successful exploration discoveries. The review focuses on four sediment sample types: fine-fraction stream sediment samples; bulk leach extractive gold (BLEG); heavy mineral concentrate (HMC) samples; and indicator mineral chemistry of grains from stream sediment. Detailed background and historical application of stream sediment geochemistry are aptly summarized in previous publications: Rose et al. (1979), Levinson (1974), Meyer et al. (1979), Fletcher (1997), Ottesen and Theobald (1994), Stendal and Theobald (1994) and Lett and Rukhlov (2017). Whilst there are a variety of papers summarizing stream sediment and HMC surveys, there are fewer publications regarding gold-specific challenges and BLEG survey applications, some of which have been proprietary. These have been highly successful in early-stage exploration for Cu–Au porphyry deposits and various types of Au and Ag deposits. The ‘special case for gold’ section herein provides detail, and several successful discovery case histories are included. Finally, recommendations to optimize successful stream sediment geochemical outcomes and to integrate geochemical data into exploration and mapping programmes are provided.

Stream sediments are the product of weathered and eroded material accumulated within a drainage catchment. Weathering processes vary globally and alluvial material may result from the action of wind, water and/or ice, and moved by fluvial processes. Sediment may be derived from rock, soil, aeolian materials and/or glacial till. Active stream sediment includes clastic and hydromorphic contribution from waters, clastic material from stream banks, organic material, and material precipitated from stream water. Stream sediments are the most commonly used sample medium for reconnaissance exploration in cases for which the target mineralization and/or alteration are exposed to weathering and entrainment in drainage sediment and/or organic materials (Rose et al. 1979).

In some global geomorphological and climate regimes, there may be a significant influence from chemical erosion and hydromorphic dispersion of mobilized elements in ground and/or surface water and in a few environments precipitates or organic material may dominate. Chemically mobilized elements may precipitate within drainage sediments, in organic substrate, and/or in iron (Fe)–manganese (Mn)–aluminum (Al) oxides within the sediment. While clastic dispersal is the dominant influence on stream sediment geochemistry, chemically dispersed compounds and/or extremely fine-grained (<50 $μ$m) silt/clays/colloids and/or oxides may be important to the sediment chemistry and present isolated sample media themselves.

Stream sediment surveys involve the collection of some portion of the transported sediment for regional or detailed project-scale geochemical surveys. They are more commonly utilized at regional scales (1 sample per 10–100 km²). They may be valuable at more detailed scales (1 sample per 0.5–5 km²) in cases where terrain is steep, and stream sediment remains the most accessible geochemical sample medium (i.e. where rock outcrop or soil samples are sparse or difficult to obtain). Stream sediment surveys are a valuable first stage for mineral exploration programmes owing to the relatively low cost per square kilometre of ground being assessed and the ability to cover regional areas comparatively quickly.

Regional geochemical mapping programmes have been used to characterize environmental baselines, map regional geology in areas with little outcrop, and for direct targeting in mineral exploration. These surveys benefit from multi-element data. An example of using stream sediment data to map regional lithologies using advanced multi-element interpretation and imaging is provided in the Case history 6 section of this review. In some cases, mineral chemistry of grains may be useful to characterize catchment basin rock type, age of catchment rocks, and prospectivity for various types of alteration and mineral deposit styles. Examples of the growing field of mineral chemistry applied to sediment surveys are provided in Case history 5 of this review.

Sediment survey styles vary with physiography and climatic regime and include the collection of entire raw sediment, fine and extremely-fine sediment fractions, moss mats, coarse gold by panning, and HMCs. The last contains a range of semi-resistant minerals with densities >2.9 g cm⁻³. The sediment may be enriched in heavy minerals, and/or in chemically transported elements that may adhere on Fe–Al–Mn oxides or organic material in the stream bed. In these cases, a more generalized fine-grained stream sediment may be collected (100–300 g) of sand, silt, and/or clay collected in the field as a <2 mm (−10 mesh) and upgraded in a laboratory to a select fraction. In specific cases a finer fraction of <177 $μ$m (−80 mesh), finer <74 $μ$m (−200 mesh) or even <37 $μ$m (−400 mesh) sediment material may be collected. (Note that this review provides particle size comparison of sieve fractions in both $μ$m and mesh. The conversion of these is defined by ASDI Standard ASTM E11 for soil and sediment, wherein sieves of Tyler and American Standard −80 mesh = 177 $μ$m.) In some cases, a large raw sediment sample is preferable with a specific fraction later isolated in the lab. Many stream sediment samples have been sieved in the lab to <177 $μ$m (−80 mesh) and that historically was a mineral exploration standard sieve fraction (Garrett 2019). In the case of gold exploration, a sufficiently representative raw sediment sample can be processed either by gravity separation or by a BLEG method. Heavy indicator minerals may be useful as indicators of specific deposits or geological provenances, such as the collection of 1–5 critical grains of indicator minerals for kimberlite exploration work. The collection of zircon, ilmenite or monazite may be useful to characterize mineral chemistry and identify age, host rock chemistry and/or fertility for a domain to host specific types of rocks or mineral deposits.

Drainage systems around the globe vary with geology, geomorphology, weathering history (climate and tectonic history) and topographic relief. Very high-relief areas may have sparse sediment except at breaks in topography that allow sediment to collect. Very low-relief areas may have poorly defined drainage systems, or sheet-wash areas with alluvial fans of mixed and poorly sorted sediment material. Well-developed and naturally weathered areas may have drainages with well-sorted sediment, whereas some immature drainage systems contain an unsorted range of material from boulder to pebble–sand–silt to clay. Areas with temperate to tropical climates will generally have continuous water flow in the majority of the drainages. Rapid stream flow in steeper terrain may preclude the accumulation of sediment material, and in these cases the available sample may be the fine sediment trapped in collection traps and, in the case of cool temperate rain forest, by moss mats along the bank (e.g. British Columbia) (Matysek et al. 1988). Arid and hyper-arid climates may have drainages developed during rare episodic events, resulting in deposition of less sorted material. Dilution of a pre-existing geochemical signature may result from windblown sand (e.g. sub-Saharan Africa, northern China), windblown ash (e.g. Chile, Argentina) or fine glacial flour (e.g. Yukon). Equatorial regions with rainforests may have a greater mix of clay-sized material derived from deeply weathered regolith. Plate tectonic history of a domain will influence the current drainage pattern, such as in western Australia, where watersheds were first developed by equatorial tropical weathering followed by desiccation and arid climatic conditions. Glacially impacted areas of northern Russia, Europe, Canada and Alaska have glacial drift dispersed across a region with the current stream drainages reflecting sediment from both exposed geological outcrop and precursor glacial drift as much as 300 km from the source. It therefore becomes important to assess current topography and drainage networks, past ice-flow directions, and the tectonic/climatic history of an area to understand the stream sediment development and geochemistry. Stream sediment surveys may therefore benefit from an understanding of the regional and local geomorphology, and use of Quaternary or regolith maps to improve interpretation of the sediment geochemical response.

The usefulness of stream sediment geochemistry in exploration is enhanced with an understanding of the geochemistry of surface processes, mineral deposit formation and geochemical zonation around deposits. These factors should be kept in mind when designing a stream sediment geochemical survey and interpreting resulting data. Stream sediment survey design should begin with an assessment of the regolith conditions, topography, relative dominance of physical dispersal v. chemical dispersion, glacial history and geomorphology for the area. If available, satellite imagery is valuable in this assessment, including visible, near infrared and short-wave infrared wavelengths. Maps of the drainage network, glacial patterns and changes to geomorphology are valuable. At this stage, potential complexities in stream sediment survey design should be noted. These may include dramatic changes in topography and drainage network (natural catchment sizes), evidence of landslides, swampy areas, glacial drift, distinctive colluvial fans, fresh and saline lakes, the presence of gold- and other metal-robbing material such as carbon, recent wildfire activity, wind-blown material, volcanic ash and anthropogenic disturbance, all of which may contribute to the planning of an effective stream sediment programme.

Survey design should then consider purpose of the survey in terms of the appropriate sampling density, location of samples, anticipated type of stream sediment sample, grain size of interest for sample collection, and analytical protocol. Accessibility, the presence of wet or dry drainages and the nature of vegetation are important to note. Physiography and size of first- and second-order watersheds will play a role in sample planning. Site accessibility may play a practical role in site selection. The anticipated target feature size (geological or mineralogical) should dictate the sampling spacing. For instance, a target of exposed mineralization of 4 km² diameter is not likely to be identified with sampling broader than 1 sample/10 km², whereas a target of 20 km² may be identified with a drainage survey of 1 sample/50 km². Focus for an effective survey should be on the anticipated target size for a successful regional characterization.

It is not always clear which type of sediment sample may be preferable, what sample spacing is appropriate, or even the preferred sample preparation and analytical method. When there are no prior surveys or prior knowledge for a particular region, a geochemical orientation survey is recommended before embarking on the main survey. This may be a small survey of 30–50 samples taken for the purpose of defining: (1) the type of sample (stream sediment fraction, BLEG, HMC, indicator mineral), (2) preferable grain size, (3) laboratory sample preparation, (4) analytical method (digestion and instrumentation), (5) the most effective pathfinder elements or minerals and (6) sample spacing needed for a successful survey. Orientation surveys should include a known geological or mineralized area, with at least 20% of the samples collected in presumed background, non-mineralized areas. Orientation samples may be collected every 250–350 m upstream and downstream of known deposits to provide a sense of down-drainage geochemical dispersion and to help determine optimal survey sample density. A well-conducted orientation survey, incorporating appropriate sampling and quality control protocols, will increase the likelihood of success and/or discovery for the subsequent geochemical programme.

A fine-fraction stream sediment sample is the most commonly collected of the sample types described in this review and is useful for regional multi-element geochemical surveys. A successful application is described in Case history 1 below, describing a gold discovery in Colombia.

The density of stream sediment sampling depends on the scale of the mapping and stage of the exploration programme, as well as terrane variations. If a project is in the reconnaissance phase, spacing will be considerably broader (i.e. 1 sample/10 km² or up to 1 sample/20 km²) compared to follow-up surveys (c. 1 sample/1–2 km²). If known mineral deposits are present in the area, it is important to sample downstream from these, as well as distal to them to represent background. The Global Geochemical Reference Network project recommends collection of active stream sediment sample from the second-order catchment basins of <100 km² (Demetriades et al. 2022), which may be sufficient for regional mapping programmes but is generally too broad for mineral exploration. The stream sediment sample should be representative of all sediment in a stream basin, and therefore, if possible, several subsample increments of sediment within the stream (main channel, bars, eddies, etc.) should be collected over a distance of 5–100 m at each site. General guidance for selecting good sample locations on a given stream include the following: (1) sample well below a junction of two stream tributaries rather than immediately below to ensure that sediments from both sources have been mixed; (2) sample upstream of bridges, culverts or other human disturbances to avoid possible contamination; (3) avoid areas with obvious sources of contamination such as old fuel barrels or areas used for shooting target practice; and (4) in situations where small streams enter larger mainstream tributaries, sample far enough upstream in the smaller tributary to avoid the flood plain of the larger stream. Stream sediment samples should be taken across the active drainage channel to be as representative as possible.

The required sample size will vary depending on material available, whether or not field sieving is completed to remove coarse material, and on the amount of material required for analyses. In general, a 0.5 kg sample field screen to <2 mm is adequate except for programmes specifically designed for gold exploration, which require increased volumes of material (see the following section, special case for gold). A stainless steel field sieve at 2 mm is useful to eliminate larger pebble material and organics, and this sieve should be cleaned in the field between samples. Material collected in the field may then be further sieved to finer-size fractions in the laboratory. Stream sediment surveys utilize a variety of size fractions depending on terrane, targeted deposit type and survey purpose. In general, fine or ultrafine fractions (silt or <69 $μ$m; clay or <2 $μ$m, respectively) are useful because: (1) they tend to yield more reproducible results and higher-contrast anomalies than coarser fractions (Skeries et al. 2017; Noble et al. 2019); (2) finer fractions are less affected by the gold nugget effect; and (3) detection limit challenges become more prominent in coarser fractions due to the abundance of diluting quartz sand (Noble et al. 2019). A diagrammatic representation of grain sizes and scales is shown in Figure 1.

For dry or slightly damp sediment, field material of <2 mm material can be worked through a sieve. For sites with water, local stream water is used to wash sediment through the screen (carefully avoiding washing away the finer material). When at a site with very damp, clay-rich, or muddy sediment where there is not enough water to assist in screening, it may be necessary to eliminate the screen and simply hand pick out larger sticks and pebbles.

Although field sieving to <2 mm is common for many stream sediment programmes, field sieving at 590 $μ$m (−30 mesh) or finer fractions has been done in some circumstances. If the streams are wet, finer-mesh sieving can be effectively done in the field. However, if the streams are dry or damp, it is commonly difficult to adequately clean the sieve in the field, and an attempt at finer field sieving may result in contamination from one site to the next. As noted by Garrett (2019),

Over the last 70 years the use of the minus 80 mesh fraction for soil and stream sediment analysis has proven effective and led to the successful discovery of primarily base metal mineral resources. However, that does not mean the use of the minus 80 mesh should be taken for granted, as directed orientation work may indicate a more suitable size fraction for specific trace elements and mineralogical conditions, in particular precious metals (Garrett 2019, pp. 8–9).

A case study in Colombia included a collection of 24 000 samples in flowing drainages, wet sieved in the field to <74 µm (−200 mesh) (see Case history 1). Such screening requires longer periods of time in the field but eliminates the necessity to transport large samples and later laboratory sieving.

Laboratories may be instructed to analyse the sample as received with no further sample preparation if the sample is dry and fine-field sieving has been done. If sample drying is required, air drying, or oven drying at <60C, should be requested to avoid loss of volatile elements such as Hg and Sb. Sample sieving at the laboratory is commonly done at 177 $μ$m (−80 mesh) as a convenient sieve size, although this may not be suitable for any given survey. On request, labs may sieve to 149 $μ$m (−100 mesh) or 74 $μ$m (−200 mesh), if a finer fraction is of value and the collected sample has sufficient fine material. However, the finer the sieve fraction, the greater the opportunity for contamination between samples during preparation. Generally, the finer fractions will have higher concentration of trace elements, being less diluted by quartz, a potential problem with sand and silt size fractions. Most analytical work will require a prepared sample of 50–100 g, depending upon the analytical methods selected.

The choice of the most appropriate multi-element analytical method for stream sediment samples depends on a number of factors, including sediment composition and the expected pathfinder elements for the targeted terrain, the deposit type being sought, and whether especially low detection limits are needed in the survey area. For example, if the expected target is a porphyry Mo deposit, pathfinder elements such as Ag, Re or W may be present, but these elements may occur at very low concentrations, and therefore it is important to choose methods with appropriately low detection limits for these pathfinders. Because the sample size analysed for most multi-element methods is small (<1 g), default gold concentration results from the multi-element methods are rarely useful; gold should be determined separately from a larger and more representative sample (see the next section).

Contemporary multi-element analytical methods consist of acid digestion of the sample followed by a combination of inductively coupled plasma atomic emission spectroscopy (ICP-AES) and inductively coupled plasma mass spectrometry (ICP-MS) determination for a large number of elements (60 +). An aqua regia digestion is the most common digestion for stream sediment samples, which will dissolve oxide, sulfide and gold within the sample. However, most silicate and resistate minerals in the sample will not be dissolved by aqua regia, a potential problem if the elements of interest reside in these mineral phases. A four-acid digestion is considered a ‘near total’ method, in that most mineral phases will be digested. A few volatile elements such as Hg are not available from this stronger leach (they are lost during sample dissolution), and precision is generally slightly higher due to the more complex solution matrix caused by dissolution of silicate minerals. Some analytical methods such as instrumental neutron activation analysis (INAA) do not require any sample digestion and are considered ‘total’. Disadvantages of INAA include the smaller suite of elements determined and the long turn-around times typically required (Hoffman 1992).

Once digested, element concentrations may be determined by the more sensitive instrument, ICP-MS, or by ICP-AES which may be more economical, but has higher detection limits and a smaller suite of elements determined. Improved instrumentation in the past decade has resulted in lower detection limits for most pathfinder elements, in many cases well below natural background concentrations; for example lower detection limits for Au, Bi, Te and the rare earth elements (REEs) of 0.2 parts per trillion (ppt), 0.5 parts per billion (ppb), 3, and 2 ppb, respectively, are attainable. The resulting stream sediment data have improved precision and accuracy at lower concentrations, a greater suite of pathfinder elements, and the potential to acquire individual element isotopes (e.g. Pb²⁰⁷, Pb²⁰⁶). A discussion with the laboratory geochemist or chemist to select the best method is recommended for new programmes. Detailed description of various methods is included in the International Union of Geological Sciences Manual of Standard Methods (Demetriades et al. 2022).

A quality management system for a standard geochemical survey includes both quality assurance (QA) and quality control (QC) elements. The QA focus is mainly in the analytical laboratory environment, where the components of standard operating procedures, instrument logs, training records, data acceptance/rejection criteria and lab audits are covered. The QC element provides measures of the accuracy and precision of geochemical data produced by an analytical method. The accuracy and precision are established through the analysis of standard reference materials (SRMs), blank samples and field and analytical duplicate samples.

The accuracy of an analytical method is an attempt to quantify the ‘correctness’ of analysis, by including SRMs with the sample batch and comparing their analysis to the published value for those SRMs. Accuracy is calculated as the percent recovery by dividing the mean concentration by the target value of the SRM used and multiplying by 100. SRMs should be used that match the anticipated concentration of the stream sediment samples. If SRM concentrations are significantly higher, the lab may need to use a different method for analysis, and these higher-concentration reference samples may risk contaminating survey samples.

Blank samples are included for the purpose of identifying contamination within the sample preparation and analytical processes. If lower concentrations are anticipated, blanks are important to ensure that low concentrations are valid. A pure silica sand such as used for high-quality ceramic work provides a good pulp/analytical blank for sediment surveys, and these come in varying size fractions which may be selected to match the field samples.

Precision is a measure of the repeatability of results for duplicate pairs of samples. Field duplicate samples are taken at the same site, covering a similar section of the drainage. Precision may be calculated by one of several mathematical methods. Stanley and Lawie (2007) summarize the most commonly used precision variations and recommend calculation of percent coefficient of variation (%CV) calculation. If the outlier 5% samples are removed before calculation, then this measure is very similar to standard deviation calculation for accuracy measurements and is referred to as %CV^p95.

Measurements of precision and accuracy are optimal when elemental concentrations fall within the middle of the determination range for a specific analytical method and element. As a general rule, analytical determinations become less accurate and precise as data values approach the lower or upper reporting limits of determination. An example of the quality management system used for a wide variety of surficial sample types and analytical methods is provided by Anderson et al. (2011) and Friske and Hornbrook (1991).

If QC checks do not meet pre-established criteria, then the data should be investigated. Potential issues include sample inhomogeneity, analytical errors and field or laboratory cross-contamination.

Stream sediments are used in gold exploration and unique characteristics of gold require special considerations.

1. Gold occurs at three orders of magnitude lower concentration than other common exploration target elements.

2. Gold commonly occurs as a nearly pure nuggets.

3. Gold nuggets are heavy minerals with more than five times the density of other minerals found in stream sediments.

The first challenge to overcome is to obtain a representative sample at very low concentration levels. Clifton et al. (1969) calculated that a minimum of 20 grains of gold should be expected in an analytical aliquot if a precision of ±50% is to be expected. Several authors concluded that standard methods, which were adequate for base metal stream sediment exploration, were simply inadequate for stream sediments in gold exploration (Shelp and Nichol 1987). Early surveys of the 1970s demonstrated this by relatively poor precision for gold (Harris 1982; Arne and McFarlane 2014). To be fair, some of these earlier surveys were not originally designed to detect gold mineralization.

Gold explorers quickly realized that they needed to: (1) increase the gold grade, (2) decrease the grain size or (3) increase the analytical sample weight to achieve the precision that is required for effective use of stream sediments in gold exploration. The steady improvements in sampling and analytical procedures as well as the evolution of exploration practices is documented in Lett and Rukhlov (2017) for British Columbia regional programmes where they list the evolution of survey practices since 1976.

Fortunately, field duplicate data can be used retroactively to determine if the aggregate procedures, from sampling to final analysis, were adequate for the purposes of the survey. Unfortunately, these field duplicate data were either not collected or not published.

Because gold, in nugget form, is a heavy mineral, some explorers seek out sedimentary environments where gold may concentrate. Fletcher, together with his students, documented enhanced gold concentration in heavy mineral traps and other depositional environments (Day and Fletcher 1991; Fletcher et al. 1992; Hou and Fletcher 1996). They also reported that such traps yielded longer and more consistent dispersion trains in the heavy mineral concentrates. One concern with this method is that in regional surveys, where different and inconsistent depositional environments are found, the quality of depositional environments, unrelated to gold endowment, will have an undue influence on gold concentration at different sample sites.

Over the past several years, gold explorers have increasingly focused efforts on smaller and smaller grain sizes for the analytical aliquot. This is probably because of the empirical recognition that when a stream sediment is divided into different fractions, the finest fractions typically have the highest gold concentrations (Leduc and Itard 2003; Noble et al. 2019). This decrease in grain size also increases the number of grains analysed, for a given weight, as well as decreasing the maximum size of potential gold grains, all of which provide better precision and help to address the low precision reported in older surveys. Whereas a sieve fraction of less than 177 $μ$m was the norm for gold exploration 40 years ago (Rose et al. 1979), it is now more common to see <125 $μ$m, <63 $μ$m or even <50 $μ$m (Leduc and Itard 2003) in gold exploration (Fig. 1).

Obtaining these fine fractions from samples introduces challenges. Obtaining sufficient mass of the finer fractions normally requires significantly more effort at sieving. In addition, sieving a sample at sizes less than 177 $μ$m (−80 mesh) usually requires wet sieving. Mineral aggregates that form when a sample is dried will not pass the finer mesh sieves, such as less than 149 $μ$m (−100 mesh) or finer. Better sieving efficiency is achieved by wet sieving whereby a constant flow of water aids the breakdown of aggregates and passage of smaller mineral grains. This wet sieving can be performed at the collection site, or a bulk sample can be collected and wet-sieved in a camp or a lab. This introduces the concern that fine material will be washed away in the sieve wash water, and this fine material may be the target of the sieving. Alternatively, the wash water can be captured, and the fines collected with the use of a flocculant or filter press.

A <2 $μ$m fraction was isolated from Australian soils by Noble et al. (2019). The focus on finer and finer fractions is expected to continue as higher concentrations and improved precision are anticipated within finer fractions.

Many commercial laboratories offer procedures that will analyse up to 50 g of sample, including a 50 g fire assay/ICP or 50 g aqua regia digestion/ICP-MS. This value of 50 g is often the upper limit of scale and practicality. For these reasons, increasing the analytical weight is often not possible unless multiple aliquots are analysed.

The exception to analytical sample size limit are the cyanide leach (CN) procedures. Larger volumes of CN solution plus sample can be safely handled in a laboratory, provided the solution remains alkalic. Importantly, CN quickly and firmly bonds with gold and holds the gold tightly and securely in a liquid form. This principle is well known in gold processing plants and is often used in gold recovery operations. Equally important, CN solutions do not dissolve much else besides gold, silver and some copper (Rate et al. 2010). As a result, CN solutions are relatively clean and present very little interference or complication to analytical equipment, allowing for lower detection limits with higher precision.

CN leach techniques were first applied to large (up to 10 kg) stream sediments in Australia during the 1980s and the term BLEG was coined (Radford 1996). While the term BLEG lives on, the general technique has evolved with various versions including different sampling methods, size fractions, as well as pre- and post-leach treatments and methods of analytical determination. As a result, no two ‘BLEG’ methods are the same. Most commercial labs offer a ‘BLEG’ option but the procedures are not well standardized. Some mining companies have developed their own methods, although the details are often held closely by the mining company (Marshall 2022). BLEG surveys should specify additional details such as field sample size, grain size analysed, CN digestion method and instrumentation.

A disadvantage of CN leach techniques is that they are fine-tuned for gold determination, but do not allow for determination of other trace elements, which are often essential for interpreting the gold results and provide further exposure to different types of mineralization. For this reason, it is normally advised that trace elements be determined on a second sample or analysis collected at the same site, or on a second aliquot from the original sample. Multi-element determinations may be analysed by aqua regia/ICP-MS analysis on the fine-fraction sample.

Few comparative BLEG studies are available in the literature. Leduc and Itard (2003) compared results from Sudan, Guinea and Indonesia using a combination of two methods of BLEG analysing three sample fractions; <180, <125 and <63 $μ$m. While they reported that both BLEG and multiple-size fractions yielded similar results, they favoured the finer-grained fraction because the logistics were less complicated. Unfortunately they did not report precision for any of the methods.

BLEG gold results have, in fact, proven to be highly effective in many situations and can provide a robust, consistent and highly precise gold response within catchment basins. Two of the case histories (Case histories 2 and 3) provide examples of discoveries made resulting from customized regional BLEG geochemical programmes completed by geochemists (Radford 1996; O. Lavin [Newmont Mining], written communication 2014; Marshall 2022).

The current procedure for these customized BLEG surveys consist of collection of a finer-fraction sample of 1–10 kg in the field, proprietary processing at an in-house laboratory focused on upgrading the clay and fine silt sample fraction, followed by CN leach analysis for Au–Ag/ICP-MS determination and multi-element analysis by aqua regia/ICP-MS determination. The programme is specialized in terms of the field sample collected (field crews are carefully trained for each survey), use of an in-house processing laboratory, careful and detailed QAQC procedures, ultratrace–Au detection limits at the Au-processing laboratory, and specialized catchment-based multi-element data interpretation. The in-house ultra-clean laboratory immerses the sample in water, then stirs vigorously to create a slurry. The slurry is allowed to stand, allowing sand and coarser silt to settle. The remaining column of water, including suspended material standing above the settled coarser material, is decanted. In this way, only the finest-grained suspended sample material is separated for analysis (Marshall 2022). QAQC procedures include blank samples to ensure no contamination at very low concentrations of Au (sub-ppb), and abundant field duplicates (>10%). It is possible, with field duplicates, to determine retroactively if the procedures employed were adequate to provide reproducible results. An example of field duplicate data for a reproducible BLEG survey is provided in Figure 2, anticipated to have field duplicate precision of <15%CV^p95.

In summary, using stream sediments in gold exploration is difficult, primarily because of the very low concentration of gold, when compared to other elements of interest. Regional BLEG specialized programmes have proven successful at identifying Au, Ag–Au, and Cu–Au mineralization, despite challenges associated with gold geochemistry (see Case histories 2 and 3).

Heavy minerals are dense (>2.9 g cm⁻³) and resistant to weathering during alluvial, colluvial or glacial processes, thereby facilitating their preservation and identification in stream sediment samples. The HMC may include sulfide minerals, grains of iron and manganese oxides, and valuable indicator minerals. Prospectors have long used the presence of heavy minerals derived from alluvial sediments as an exploration tool. As an early example, a follow-up study of an HMC concentrate sample containing huebernite (a Mn-rich mineral of the wolframite solid solution series) in Golden, Colorado revealed the presence of a tungsten deposit in the catchment drainage 50 km to the west (Theobald and Thompson 1960). Further investigation led to discovery of the large Henderson Mo mine vertically below this tungsten occurrence by Stewart Wallace (SEG 2009).

Panning sediment in the field with a traditional gold pan, followed by heavy liquid and magnetic separations in the laboratory, effectively removes common rock-forming minerals such as quartz and feldspar, allowing for isolation and concentration of heavy resistate minerals. These minerals may be indicators of either mineralization or alteration, and/or may be indicative of important lithologies. Some HMC preparation laboratories have carefully refined separation procedures which allow for multiple fractions based on combinations of density and paramagnetic properties and allow separation of specific minerals (such as garnet), and the retention of every heavy mineral grain.

Because they are resistant to weathering, heavy indicator minerals will survive a great transport distance down drainages, and may show dispersal of tens of kilometres (depending upon drainage geomorphology and dilution from other sediment sources). The use of heavy mineral concentrates as an exploration sample medium may be particularly important in environments where alluvial sediments are contaminated by aeolian deposits, volcanic ash, or other surficial materials that are transported into stream valleys, thereby diluting and modifying the geochemical signature of stream sediments. The collection of heavy minerals is a means of screening and allowing focus on indicator mineral presence and chemistry.

Early studies focused on the presence of minerals that can be easily visually identified such as gold, gems, and tin and tungsten minerals, summarized in the Handbook of Exploration Geochemistry Drainage Volume (Stendal and Theobald 1994). The US Geological Survey routinely collected both HMC and stream sediments in mineral resource assessment studies from the 1970s through the 1990s (for example, Bell 1976; Miller et al. 1980). From the early 1970s, kimberlite indicator minerals have been used for diamond exploration globally, resulting in the discovery of many of the world's diamond deposits (Gurney et al. 1993). Geochemical surveys using Cr-pyrope in stream sediments led to a discovery in the Ural Mountains after recognition that these were similar to pyrope found in South African kimberlites (Lett and Rukhlov 2017). Identification of garnet, chromite, ilmenite and other minerals of a diagnostic chemical composition were indicative of diamondiferous kimberlites.

Over the past 10–20 years, research organizations, laboratories and exploration companies have conducted studies focused on heavy mineral fractions of till or sediment to determine specific mineral indicators of mineralization or alteration, and the indicator mineral chemistry. Results from these studies show that there are diagnostic minerals indicative of porphyry copper deposits, magmatic nickel–copper deposits, metamorphosed massive sulfide deposits, Broken Hill-type deposits, and a variety of other deposit types (McClenaghan et al. 2020).

Analyses of HMC by multi-element chemical techniques were historically incorporated in exploration programmes because doing so facilitated the detection of many elements that would otherwise have been too low to be detected in stream sediments (Stendal and Theobald 1994). Most recently, automated scanning electron microscope (SEM) techniques have replaced the visual identification of minerals, and there are now options to analyse the bulk HMC sample, or utilize microanalytical techniques such as electron micro-probe (EMP) and laser ablation ICP-MS (LA-ICP-MS). Chemistry and zonation of specific minerals in an HMC sample can add significantly to exploration tools.

Minerals including apatite, rutile, titanite, apatite and magnetite have recently been shown to have chemical characteristics that distinguish barren v. mineralized rocks (Bouzari et al. 2016; Mao et al. 2016; Kelley et al. 2022). Age dating of zircon in HMC has been used to infer age of intrusives within a catchment basin (O'Reilly et al. 2004). Common resistate minerals sourced from porphyry copper or molybdenite copper deposits may include jarosite, chalcopyrite, gold, garnet, epidote, wolframite, tourmaline and apatite. Resistate minerals near kimberlite may include garnet, ilmenite, chromite, spinel, diopside, olivine and diamond; near magmatic copper–nickel deposits spinel, ilmenite; near metamorphosed massive sulfides gahnite; near gold deposits: gold, rutile, tourmaline, and scheelite. Other resistant minerals of interest to the geological provenance may include zircon, monazite, apatite, xenotime, spinel and columbite and are described in the next section of this paper.

The HMC sample should be collected specifically from stream locations that are most likely to contain minerals with densities greater than those of quartz and feldspar. These can include depositional environments where the water velocity slows, such as point bars, gravel bars, behind and under large boulders, or between and under cobbles. In streams without cobbles or coarse gravel, material may be collected from bars, or from roots of plants growing along the drainage. Any area with heavy mineral streaks of black sand (magnetite) or garnet are likely to contain other heavy minerals. Maximum dispersal trains for most heavy ore minerals were initially reported to be 5 km (Stendal and Theobald 1994) but more recent studies have shown that with collection of a 10–20 kg bulk sample and careful laboratory processing, sediment dispersal trains can be tens of kilometres (Averill 2001).

Sediment at the sample site is sieved to <2 mm size fraction (−10 mesh). The bulk HMC sample may then be sent for detailed laboratory processing, or it may be further refined in the field. Field panning with a standard gold pan can remove up to two thirds of the material and may be done until the first signs of heavy minerals (black sand or garnets, etc.) are visible in the pan. A gravity separator, sluice or jig may be used in the field in lieu of manual panning. Surveys that seek only a few diagnostic mineral grains, such as kimberlite surveys, are not processed in the field and the bulk sample of 10–20 kg is sent to a specialized HMC preparation laboratory.

Techniques for indicator mineral studies involve incremental separation of mineral grains using sieving, density and electromagnetic properties of the individual stream sediment mineral grains. Some labs begin with sieving the sample, to allow focus on the 0.25–0.5 mm fraction for consistency in subsequent stages. The sample may then be further separated by gravity (such as using a riffle table) where gold is easily separated. Separation by heavy liquids serves to remove the majority of quartz and lighter silicate minerals. Liquid densities of 2.9 up to 4.1 g cm⁻³ are available (Stendal and Theobald 1994). Magnetic separations are then completed, at which stage strongly magnetic minerals (magnetite) are separated from paramagnetic minerals (e.g. garnet) and non-magnetic minerals (e.g. zircon). Numerous published studies provide details of sample preparation methods used in reconnaissance surveys (Averill 2001; De Cariat and Cooper 2011; Lee et al. 2016; Lett and Rukhlov 2017; Wilton et al. 2017; Kelley et al. 2022).

Mineral identification in HMC samples was historically done optically using a hand lens or binocular microscope. In the laboratory, minerals may be identified using an SEM with polished epoxy mineral grain mounts. More recently, commercial software programs such as Quantitative Evaluation of Materials by Scanning Electron Microscopy (QEMSCAN; Pirrie et al. 2004), Mineral Liberation Analyzer (MLA), or TIMA (TESCAN Integrated Mineral Analyser) and INCA Mineral of Oxford Instruments are designed to provide automated identification of tens of thousands of grains in short time intervals. The methods use back-scattered electron (BSE) imaging to map grain mounts and to identify the minerals based on energy dispersive spectrometry (EDS). The EDS spectra are compared with spectra held in a look-up table, allowing a mineral or phase assignment to be made at each acquisition point. The assignment makes no distinction between mineral species and amorphous grains of similar composition. Software data are summarized in a spreadsheet giving the area percent, mass percent, number of mineral grains or number percent of mineral grains of each composition in a look-up table. Particle distribution maps allow for identification of mineral associations and presence of mineral inclusions.

Automated mineralogy has a major advantage in the quality and quantity of data acquired compared to visual determination (Lougheed et al. 2021). A large number of minerals can be rapidly identified, with the time required for analyses dependent on the desired resolution. Rare and low-abundance indicator minerals may be missed during traditional visual observation, or may occur as inclusions within other resistant minerals. The presence of a few key indicator mineral grains in a sample may be the key to diagnose prospectivity and suggest additional detailed examination or sampling. Indicator mineral grains mounted with epoxy in pucks can be analysed chemically with streamlined methods so that rapid throughput can be attained without hand-picking of grains. Automated SEM techniques allow for the identification of textures, mineral associations and inclusions that might be specific to a deposit type (Plouffe et al. 2022). Finally, automated SEM techniques remove bias or error that is likely inherent during visual inspection (Lougheed et al. 2021).

Chemical and isotopic compositions of resistate minerals may be used in exploration as indicators of the deposit type, the rock type, age and prospectivity of the host rock, and the proximity to a mineral deposit. Major elements of minerals are typically measured using an electron micro-probe (EMP, also known as electron probe microanalysis, EPMA) and the trace elements measured using LA-ICP-MS. Elemental mapping of minerals may be obtained through BSE imaging on an SEM or EMP (major elements, qualitative), microXRF (micro-X-ray fluorescence; major and minor elements; qualitative), PIXE (proton induced X-ray emission; major to trace elements, quantitative), synchrotron (minor and trace elements; quantitative), and LA-ICP-MS (minor and trace elements; quantitative). Isotopic compositions, mainly Hf isotopes of zircons are used in TerraneChron (O'Reilly et al. 2004), measured using LA-MC (Multi-Collector)-ICP-MS instruments.

Historically, major element compositions of resistate minerals were used to determine mineral species which may be sufficient to indicate the source rock types (e.g. gahnite). Garnet and chromite of specific compositions suggest a kimberlite source (Sobolev 1971, 1977; Gurney and Switzer 1973; Grutter and Menzies 2003). More recently, with the development of LA-ICP-MS techniques, trace element chemistry of many minerals can be used as indicators, discriminators or vectors in mineral exploration (Chang et al. 2011; Cooke et al. 2014, 2017, 2020; Wilkinson et al. 2015; Mao et al. 2016; Kelley et al. 2011). Some of these indicator minerals may survive weathering and be preserved in stream sediment HMC samples. It is beyond the scope of this review to summarize all of the indicator mineral chemistry research which is currently one of the major research topics of the economic geology community. Here we summarize a few minerals that are being used, or have the potential to be used, in sediment HMC surveys.

Zircon occurs in the heavy non-magnetic HMC sediment fraction. Some trace element compositions of zircon (Ce⁴⁺/Ce³⁺ ratio) are recognized to be related to the oxidation state of magmas, suggesting magma fertility potential. High values indicate a more oxidized state (Ballard et al. 2002). Shen et al. (2015) found that larger porphyry Cu deposits (>1.5 Mt Cu metal) are associated with more oxidized intrusions with higher Ce⁴⁺/Ce³⁺ ratios (>120). Shu et al. (2019) reported a similar relationship between this ratio (and other zircon trace element parameters) and the metal tonnage of porphyry and skarn Mo deposits. Loucks et al. (2020) proposed another redox state indicator using the concentrations of Ce, Ti and age-corrected initial U in zircons. Detrital zircon in stream sediments was applied by O'Reilly et al. (2004) to infer the age, crustal evolution and fertility of the corresponding catchment lithologies via their method, TerraneChron, which integrates zircon U–Pb age dates, Hf isotopic composition and trace element composition.

Garnet occurs in the paramagnetic heavy fraction of HMC and its chemistry may indicate both lithological information and mineral prospectivity. Andraditic garnet is derived from skarns, whereas grossularitic garnet, with or without spessartine components, may be from either skarn or marble (Meinert et al. 2005; Chang et al. 2019). Metamorphic garnet is likely rich in almandine endmember pyrope garnet. Cr-pyrope (including G9/G10 compositions) is an accepted indicator mineral for diamond-bearing kimberlites (Nixon 1980; Grutter and Menzies 2003; Grutter et al. 2004). Two examples of successful diamond exploration using stream sediment HMC are summarized in Case history 5.

Apatite is semi-resistate and also of marginal density (3.15–3.20 g cm⁻³), occurring in the intermediate-density and non-magnetic HMC fraction. Because it is difficult to separate from other rock-forming minerals in the indicator mineral processing, identification of apatite grains is particularly enhanced by use of automated SEM techniques. Apatite is a common mineral in igneous, metamorphic and sedimentary rocks, and in various types of deposits including magmatic magnetite–apatite deposits, carbonatite deposits, sedimentary apatite deposits and hydrothermal deposits including iron–oxide–copper–gold (IOCG), iron–oxide–apatite Kiruna deposits, porphyry, skarn, epithermal, diamond and orogenic gold deposits. There have been extensive studies of apatite chemistry (summarized in Belousova et al. 2002). The research relevant to mineral exploration includes: (1) distinguishing different types of hydrothermal deposits (Mao et al. 2016); (2) discriminating magmatic apatite and hydrothermal apatite from different alteration zones in some porphyry deposits using apatite trace element composition and cathodoluminescence colours (Bouzari et al. 2016); (3) inferring the F, Br and I concentrations of the fluids if an apatite is known to be hydrothermal (Kusebauch et al. 2015); and (4) inferring magma fertility using apatite S and Cl contents (Peng et al. 1997; Chelle-Michou and Chiaradia 2017; Meng et al. 2021). Magma fertility may also be inferred by magma redox state and its evolution based on Mn in apatite (Miles et al. 2014) or S speciation in apatite (Meng et al. 2021).

Apatite in stream sediments has been used for provenance studies, using Sr, Y, Mn and total REEs of apatite to distinguish its rock sources (Belousova et al. 2002), sometimes coupled with U–Pb geochronology (O'Sullivan et al. 2018). Magma fertility indicators may be applied to assess the occurrence of potential fertile intrusions in a drainage catchment. A broad estimation of potential deposit type(s) may also be possible based on conclusions in Mao et al. (2016) although more research would assist in reducing uncertainties of the discrimination. Kelley et al. (2022) showed that, although not definitive, the Eu anomaly and Mn concentrations of apatite from sediment samples matched that from mineralized core samples at a porphyry deposit in eastern Alaska.

Spinel (specifically magnetite) is ubiquitous in HMC samples. Magnetite is easily separated by its density and strong magnetic response. It occurs in igneous, sedimentary and metamorphic rocks, and in a wide range of mineral deposit types (e.g. magmatic Fe–Ti–V–P, magmatic Ni–Cu–PGE, porphyry, skarn, iron–oxide–copper–gold (IOCG), volcanic-hosted massive sulfide (VHMS) and banded-iron formation (BIF)-hosted Fe ores). Magnetite trace elements have been applied to distinguish magnetite from different types of deposits. A set of discrimination diagrams from Dupuis and Beaudoin (2011) use Ni + Cr v. Si + Mg to identify Ni–Cu–PGE deposits, Al/(Zn + Ca) v. Cu/(Si + Ca) to identify VHMS deposits, and Ca + Al + Mn v. Ti + V to distinguish magmatic V–Ti–Fe, porphyry, skarn, Kiruna, IOCG and BIF deposits. Although these are popularly used and effective in distinguishing magmatic and hydrothermal magnetite, Cooke et al. (2017) voiced caution about classification uncertainty, potential issues in LA-ICP-MS analysis of magnetite, and the challenges in discriminating different types of hydrothermal deposits. Efforts are continuing with more and more data being reported (e.g. Dare et al. 2014; Nadoll et al. 2014, 2015; Huang et al. 2019a, b). The potential of magnetite trace elements to provide vectors towards ores continues to be examined. Zhang et al. (2017) report that the magnetite Mn content at the Ernest Henry IOCG deposit in Australia decreases upwards and outwards from the core of the system. The scatter of the Ti, Mn, Al and V concentrations, expressed as the standard deviation of the element abundances, shows a similar trend.

Both McCurdy (2021) and McClenaghan et al. (2022) found compositions of magnetite in streams draining the Casino porphyry Cu–Au–Mo deposit in Yukon, Canada that suggest a magmatic hydrothermal source. The sediment magnetite is similar in composition to hydrothermal magnetite in the potassic alteration zone of the porphyry deposit.

Spinel studies found unique gahnite (Zn-spinel) chemistry near Broken Hill-type (BHT) deposits (Walters et al. 2002). Rare Cr-enriched spinel is diagnostic of kimberlite deposits along with Cr-diopside; only a few grains in a stream sediment sample are sufficient to indicate the presence of a kimberlite deposit (Averill 2009).

The effective display and interpretation of stream sediment data are integral to realizing full value of drainage geochemistry. Stream sediment data represent all material input from the upstream catchment basin and not a localized point source, such as would be the case for lithogeochemical or soil geochemical data. There are several methods used to visualize stream sediment data in 2D or 3D, and selection of the best approach is based on the scale and density of the survey.

Interpretation approaches start with consideration of the geomorphology, geology, dominant structures and type of sediment. Data-driven interpretations may be utilized to interpret geochemical maps and anomalies in an exploration survey that warrant follow-up study. The potential for zonation in a mineralized system requires a careful interpretation and offers an opportunity to interpret vectors towards a mineralized centre. Grunsky and Caritat (2019) advocate the use of a systematic data interpretation approach to allow proper classification, prediction, mapping and prospectivity evaluation.

Single-point plots are shown as ‘bubble plot maps’ wherein the point of sample collection is shown as a symbol with gradational sizes and/or varying colour schemes to represent elemental concentrations (Fig. 3). These are useful in areas with high sample density and when stream data are viewed as overlays on image base maps (geology, geophysics, etc.). These maps, however, do not identify the upstream catchment area that is being sampled. Displays of the sediment data using catchment polygons are useful for regional surveys, to identify gaps in survey coverage, and to identify large differences in the catchment sizes between samples. Several geographic information system (GIS) software programs create catchment areas from a digital elevation model (DEM) and sample point data. Shuttle radar topography mission (SRTM) data are available for most of the globe (at 30 m resolution). Higher-resolution light detection and ranging (LIDAR) is preferable if available.

Accurate catchment polygons that represent the upstream area of each stream sediment sample allow calculation of ‘local area’ (catchment area from the sample point up to the next upstream sample) and ‘total area’ (catchment area from the sample point to the ridgeline of the headwaters regardless of other upstream samples). The ‘total area’ may be used to level data between catchments, useful in surveys with dramatically different catchment sizes and variable sediment load per sample. Catchment maps may then be attributed by colour or patterns by element concentration (Fig. 4).

‘Worm’ maps show stream drainage coloured upstream of each sample point (Fig. 5). The drainage lines can then be symbolized by element concentrations. These worm plots may also be enhanced by sizing the catchment by stream order, so that lower-order drainages are shown with a thicker line as demonstrated in Case history 3.

In areas with high-density sampling, or for regional surveys, interpolated raster maps may be useful, particularly when using the sediment data to map geological features. Case history 6 demonstrates an example of raster imaging of sediment data to interpret regional geology.

Any of these maps may then be incorporated into a 3D software program by draping the image over topography. Especially in areas with extreme topographic relief, it may be important to drape these catchment features on topography, particularly in cases for which the drainage intersects both lateral and vertical geochemical dispersion.

Stream sediment samples have an elemental concentration related to size of the catchment with a given background concentration, affected by outcropping mineralization (Fig. 6; Hawkes 1976). The calculation of area and concentration implies that the sediment concentration will decrease exponentially based on the size of the catchment and dilution by barren sediment. Given the same size of outcropping mineralization, a larger catchment will have a lower concentration of the metal of interest than a smaller catchment. A plot of elements of interest v. the upstream catchment area may show a clear downstream decay pattern (Fig. 7). Depending on the element of interest, outcropping mineralization, grade of mineralization, climatic environment, geomorphology and variable amounts of clastic dispersal v. chemical dispersion, the transport distance of an anomaly can vary greatly. Drainages in areas with cover or thick alluvial basins may have a more subtle response than those free flowing across lithological units and mineralization. Some leveling by catchment size and/or type of cover may be useful. Chemical effects on the dispersion of a geochemical response may include the presence and impact of organics in the stream bed, carbon-bearing bed load, and the development of Fe-, Mn- and/or Al-oxides. These may serve to enhance a geochemical response by scavenging charged ions, creating apparent anomalies. These may be interpreted by ratioing the element to Fe-, Mn- and Al-data, and may also be clear from field sample notes.

Manual interpretation of potential mineralization styles involves evaluating anomalies of every element of interest, either in the context of global anomalous values, or in the context of the local geology of the survey. Based on the anomalous elements present in each sample, a potential mineralization style can be assigned (Fig. 8).

Exploratory data analysis (EDA) involves an inquisitive investigation of the sediment multi-element geochemical data, using summary statistics and data visualization methods such as probability plots, histograms, correlation plots, matrix correlations, etc. These approaches are amply described in the literature (Till 1974; Rollinson 1993). The purpose of EDA is to identify what the visualization methods tell us about the survey data, to explore subtleties within the data, and to formulate hypotheses about root causes of the geochemical dispersion. This may involve a simple plotting of data distributions and correlations, or it may involve a detailed review of the data variation. Good quality multi-element stream sediment data should be amenable to a detailed evaluation in the context of regional lithologies, alteration, mineralization, direct indicator of ore, and prospectivity analysis. The workflow for EDA includes pre-processing of data, careful quality check of the accuracy and precision of the data, correlation and population evaluation, cluster analysis, and then selection of appropriate multi-element interpretation methods.

Following basic EDA, a multivariate statistical method such as principal component analysis (PCA) or factor analysis is useful for interpretation of various subtle aspects of the multi-element data. A summary of both the factor analysis method and various calculated parameters is provided by Statistica software (Statistica 2014a). Multivariate methods are referred to as ‘guided’ because some EDA and interpretation is involved in application of the statistical methods. A number of groupings may be selected based on interesting geochemical associations rather than simply based on the percent of data variation explained, thereby both reducing the number of variables and classifying them, providing quantitative data summarizing correlation structure. Note, however, that variable element zonation around a potential mineral deposit may not be well represented by statistical summaries, and zonation may require additional scrutiny of key individual element data. Examples of multivariate element groupings may generally include: (1) high concentrations of Fe–Co–Ni–Cr–Cu that may suggest mafic lithologies; (2) a Al–Li–Rb–REE–Zr suite that may suggest felsic lithologies within the catchment basin, as shown in Case history 6; (3) a strong association of Fe–Mn–Al with pathfinder elements that may suggest that sediment chemistry is impacted by oxide development, possible scavenging by these elements, and producing anomalies unrelated to geology or mineralization; (4) a suite of the pathfinders As–Sb–Hg that may suggest Au mineralization; and (5) a Cu–Mo–Re–Bi–Te suite that may suggest a higher-temperature, hydrothermal-style of mineralization.

Machine-learning approaches may be non-guided methods, for which software manipulations evaluate and summarize data structure independent of any user selection. These have no a priori bias and may provide information within the data not evident from a simpler user-driven interpretation. They are valuable for extremely large and complex datasets. These include, but are not limited to, neural networks, random forest, boosted tree analysis and self-organizing maps (Parsa et al. 2022).

Indicator mineral data interpretations are commonly completed based on key ratios of a few specific elements, and correlation plots of those ratios, such as the classification of G10 kimberlite garnets based on composition of Cr, Ca, Mg and Fe (Grutter et al. 2004). If data populations are simple, i.e. groupings of ‘mineralized’, ‘distal to mineralization’ and ‘background’ then a tool such as multiple discriminant analysis is useful to separate unknown samples into probable affiliation with these simple bins allowing a numerical calculation of probability of affinity to each group. A description of discriminant analysis and the mathematical calculations is provided by Statistica software (Statistica 2014b).

Regional stream sediment programmes are useful both for creating regional geochemical maps, and for geochemical exploration programmes at greenfields and near-mine stages. Numerous global mineral deposit discoveries are attributed in part to favourable multi-element stream sediment geochemical anomalies. Mineral discoveries have involved stream sediment surveys at scales of 100–1000 km² at early stages of exploration which are then followed by integrated multi-disciplinary exploration approaches at a detailed scale.

A stream sediment survey was part of an integrated greenfields exploration programme in the tropical environment of central Colombia, completed from 2003 to 2007. The sediment geochemistry contributed to the discovery of La Colasa, a major Au deposit announced in 2007, with 22 Moz Au in grades up to 0.8 g t^–1 (Lodder et al. 2010; Naranjo et al. 2017).

The exploration geologists embarked on a programme to cover the prospective area with stream sediment samples collected at spacing of 1 sample per 1–2 km² catchment basin size. The area is a tropical rainforest in lower reaches, but rises to high elevation mountains with few outcrops. The drainages provide access and outcrop exposure. The area has a uniform coverage of streams with ample sediment. The programme collected c. 23 000 stream sediment samples from flowing streams and samples were wet sieved in the field to a 74 µm (−200 mesh) fraction. The samples were then analysed by 50 g fire assay – ICP Au determination and a separate multi-element analysis with aqua regia digestion and ICP-MS determination (P. Allen, [Independent Consulting Geochemist], written communication 2013). The programme outlined 600 geochemical anomalies of which further geological work generated 34 drill-ready targets. La Colosa was one of several deposit discoveries in the programme (Lodder et al. 2010).

A combined stream sediment, BLEG and HMC survey was part of a greenfields exploration programme in the high-relief, tropical environment around prospective islands in Indonesia, completed by geologists from 1987 to 1989 (Maula and Levet 1996). The BLEG survey data contributed to the discovery of a major Cu–Au porphyry, the Batu Hijau deposit, with an initial resource of 914 Mt averaging 0.53% Cu (4.8 Mt Cu) and 0.40 g t^–1 Au (366 t Au) and an estimated 45-year mine life (Clode et al. 1999).

Sites were initially accessed by boat around the islands with geologists trekking up drainages to a suitable sample collection site, at a targeted density of 1 sample/10 km². The original survey included 900 BLEG samples. A bulk sample was collected in the field, processed at the customized company laboratory to a <50 $μ$m fraction, digested with a 50 g CN leach/Atomic Absorption (AA) determination for Au. The Batu Hijau discovery sample was a single BLEG sample with 7.4 ppb Au collected 9 km down drainage from the deposit, within a 20 km² catchment basin, shown in Figure 9 (Maula and Levet 1996). A follow-up sampling campaign demonstrated a 196 ppb Au anomaly 1–2 km from the deposit. It was only at this location that a <177 $μ$m (−80 mesh) stream sediment sample analysed by aqua regia digestion/AA determination) reflected the deposit with a multi-element anomaly in Au–Cu–Mo–As (Marshall 2013, 2022). A follow-up BLEG sampling campaign further defined the geochemical anomalies, with the deposit clearly shown in numerous catchment basins (Fig. 10; Marshall 2022). The anomaly at Batu Hijau was one of an initial 36 anomalies generated by the survey, and one of several deposit discoveries made by the programme (Maula and Levet 1996).

A greenfields BLEG Au survey in the hyper-arid environment of southern Argentina was completed by geologists in 2001. Drainages in the area are poorly developed due to low rainfall and relief. Further, the survey was complicated by stream sediment dilution by fine-grained ash from past volcanic eruptions to the west. The BLEG survey data contributed to discovery of the Navidad Ag–Pb–Cu–Zn district, initially reported with a resource of 630 Moz Ag with significant Pb, Cu and Zn, in eight intermediate sulfidation epithermal stratiform deposits (Williams 2010).

Geologists collected 1200 bulk samples at a density of 1 sample per 5 km² catchment basin. The samples were processed at a customized in-house laboratory to a <50 $μ$m fraction, digested with two methods: a 50 g CN leach/AA determination for Au, and an aqua regia digest/ICP determination for pathfinder elements. While the programme initially targeted gold deposits, it resulted in the discovery of this significant polymetallic Ag–Pb district, demonstrating the utility of BLEG for Ag as well as Au exploration. The initial anomalies are robust, including a very strong anomaly in Ag–Au–Pb–Zn–As–Bi–Sb–Mo with 11 samples outlining a 9 km long zone (Fig. 11). The highest BLEG Au anomaly was 2.96 ppb, while the Ag anomaly in nine samples was >150 ppb. The difference in threshold for Au between this case history and Case history 2 described above may be related to several factors: this is a base metal–Ag district rather than Cu–Au mineralization; it is an arid environment with possible tuffaceous ash dilution compared to Case history 2.

… the Navidad discovery was the end result of simple prospecting follow up of a regional geochemical stream sediment survey (BLEG). Hence, it can be clearly classified as a great geochemical success … (Lhotka 2010, p. 183).

Sediment and HMC in a till-covered, lower-relief area of lakes and bogs of the James Bay lowlands, west central Ontario were sampled by exploration companies, initially for diamond exploration. Chrome–clinopyroxene and Mg–ilmenite in HMC samples were the main indicators of kimberlite intrusions in this region. Regional background counts of kimberlite indicator minerals in HMC samples identified a large kimberlite field, and led to the discovery of the Attawapiskat kimberlite field and diamond deposit in 1987 (Crabtree 2003; Kjarsgaard et al. 2019). These surveys also identified a region with anomalous chromite (>55 grains) in sediment HMC samples which led to the discovery in 2007 of the Eagle's Nest Ni–Cu deposit in what is now known as the Ring of Fire Cr–Ni–Cu–PGE–V district. The Eagle's Nest deposit contains proven and probable reserves of 11.1 Mt grading 1.68% Ni, 0.87% Cu, 0.87 g t^–1 Pt, 3.09 g t^–1 Pd and 0.18 g t^–1 Au. Also within the Ring of Fire, the Blackbird deposit contains measured and indicated resources of 20.5 Mt averaging 35.76% chromite (Scales 2017). This example highlights the success of using HMC, resulting in the discovery of multiple mineralization styles in diverse geological terranes.

Indicator minerals in stream sediments were used in an exploration programme from 1971 to 1979, leading to the discovery of the Argyle diamond lamproite pipe in the Kimberley region, western Australia. More than 750 million carats of diamond have been produced from Argyle since 1983. The first-round reconnaissance stream heavy mineral survey covered an area of c. 200 000 km², with a sample size of 8 kg and a spacing of c. 15 km. Sample site quality was given higher priority than the exact sample interval. The follow-up in-fill survey was carried out at c. 5 km spacing with a 40 kg sample. Exploration follow-up samples were loam samples collected at intervals of 10–20 m and processed for indicator minerals, paired ground magnetic and radiometric surveys (Smith et al. 2018). The most useful heavy mineral was chromite (composition and quantity), with pyrope garnet (G10), picroilmenite, megacrystal zircon, and diamond also identified. The diamond-indicating compositional criteria for garnet and chromite were initially from Sobolev (1977) and then refined with local data (Smith et al. 2018), based on EMP analyses.

The discovery of the Bunder diamond pipes in India in only 18 months (2003–04) was another success, in which rare G9/G10 garnets and chromite from HMC were the most effective indicators (Krishna et al. 2018). Similarly, at the Anatapur District, India, stream sediment samples of 15 l volume were collected from second- and third-order drainages at density of 1 sample/2.4 km², and sieved to <2 mm. The samples were then passed through a Gerrytz jig, processed for HMC, and mineral grains manually picked focusing on collection of garnet, Cr–diopside, chromite, ilmenite and zircon grains. The HMC survey was followed by a soil (loam) sampling grid survey and a ground magnetic survey, all of which contributed to the identification of two new kimberlites in the district (Mukherjee et al. 2007).

A subset of British Geological Survey multi-element stream sediment data in SW England effectively mapped four key geological provenances and 16 specific lithological domains (Johnson 2011; Kirkwood et al. 2016a, b). The sample subset included 3745 stream sediment samples collected at a density of 1 sample/5 km² (Fig. 12). Kirkwood et al. (2016a) utilized domain-weighted compositional PCA paired with kriging to map litho-tectonic boundaries. The first three associations within the PCA analysis describe 70% of data variation related to four major geological domains. Granitic intrusions in metasedimentary palaeo-basins were clearly identified in the data interpretation and imaging (Fig. 12), characterized by high concentrations of incompatible elements U, Ta, Hf, Th and Rb. Further, sediment geochemistry identified distinct lithologies characterized by high concentrations of elements Cr, Ni, Ca and Mg (mafic rocks), the REEs (felsic rocks) and high Se, Ca and Na, and associated low Si, U and Nb related to the metasedimentary package. Comparison with the known geology of SW England confirmed these stream sediment geochemical interpretations, which supports the value of using stream sediment multi-element data as an aid to geological mapping and creating environmental geochemical baseline maps.

Stream sediment samples are an effective sampling medium applied to regional geochemical surveys for the purposes of geochemical mapping and/or geochemical exploration. A variety of surveys are described in this review, from fine-grained stream sediment samples of various size fractions to heavy mineral concentrates paired with indicator mineral chemistry, to BLEG samples. These sediment samples may be collected in wet or dry drainages, and are effective at providing regional geological and geochemical characterization of the catchment drainage basins. Stream sediment samples rely on active erosion from the area sample by a catchment basin; if the lithologies and/or potential mineralization are entirely covered with no exposure, then in such cases the survey may need to shift to groundwater, soil or other types of sampling programmes. So long as the lithology and any potential mineralization are impacted by either physical dispersal or chemical dispersion captured within a drainage, there is potential to identify protolith outcrop chemistry and mineralization.

Improvements in the analytical instruments are providing high-quality data below regional background concentration for many trace elements, with data now available below crustal abundance for many elements including Tl, Te, Bi, W and REEs. Elemental mapping of indicator minerals may be obtained through BSE imaging on an SEM or EMP, microXRF, PIXE and LA-ICP-MS-MS allow for detailed diagnostic mineral chemistry. Advancement of automated mineralogy allows the study of mineral chemistry to diagnose source of the mineral, lithology, age and exposure to hydrothermal fluids. A single grain in a stream sediment sample may suggest fertile lithologies (e.g. presence of intrusives or kimberlites), permissive host rock ages and/or proximity to mineralization (e.g. massive sulfide, porphyry, gold deposits).

Improved interpretation of data contributes to understanding stream sediment geochemistry with a broader perspective. Application of improved statistical machine-learning algorithms may assist the geochemist to extract value from these more complex datasets and incorporate both multi-variate and geospatial aspects to interpretation.

Selected examples of these applications have been provided in the Case histories section of this review. Geological mapping has been done successfully using detailed and systematic sediment sampling in SW England (Kirkwood et al. 2016a). A 23 000 sample fine-fraction stream sediment (<74 $μ$m, −200 mesh) programme in Colombia resulted in identification of several deposits, the most notable being the La Colosa Au–Cu deposit (Case history 1). Application of heavy mineral concentrates now includes both the identification of diagnostic minerals, and diagnostic mineral chemistry for specific minerals. Case history 4 describes discovery of chromite deposits in the James Bay lowland, Ontario. Case history 5 describes indicator mineral surveys applied to exploration for kimberlite occurrences in Australia and India. Finally, BLEG surveys of +1000 samples using a customized company method contributed to many discoveries by greenfields geologists including those described in Case histories 2 and 3: the Batu Hijau Cu–Au porphyry, Indonesia and the silver-base metal district at Navidad, Argentina (O. Lavin [Newmont Mining], written communication 2014; Marshall 2022). These are simply a few examples of opportunities for discovery provided by large-scale, early-stage reconnaissance application of stream sediment geochemistry and indicator mineral chemistry globally. In regions where stream sediments actively disperse mineral grains and induce chemical dispersion, it is anticipated that geological mapping and the discovery of mineral deposits will continue from stream sediment, HMC, indicator mineral chemistry and BLEG geochemical programmes.

The authors describe stream sediment drainage geochemistry on the heels of a century of work amply summarized by A. Rose, A. Levinson, J. Plant, M. Hale, R. Ottesen, P. Theobald, A. Bjorklund, R. Mazzucchelli, R. Lett and K. Fletcher. BLEG work acknowledges the development work of W. Griffin, N. Radford, O. Lavin, K. Arndt and S. Marshall. We thank Newmont for permission to publish some details of the Newmont customized BLEG procedures which transitioned from Normandy to work by the Newmont geochemistry team. Thank you as well for advice for this article from B. McClenaghan, R. Eppinger and to the GEEA reviewers. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government.

MED: writing – original draft (lead); KA: writing – original draft (equal); ZC: writing – original draft (equal); KK: writing – original draft (equal); OL: writing – original draft (equal).

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The datasets generated during and/or analysed during the current study are not publicly available because confidential company data were used to generate representative map styles. Data sharing is not applicable to this article as this is a review article and no new datasets were generated or analysed during the current study.