Frequently Asked Questions

Everything you want to know

UltraFine+ FAQ’s

Open allClose all

UltraFine+® is a novel method for soil analysis designed for 0-30 m of transported cover developed by CSIRO in collaboration with LabWest. This method is based on separating and analysing only the ultrafine fraction (< 2 µm) of a given sample, as most metals in transported cover tend to adsorb preferentially to clay particles and other fine “scavenging” phases with large surface areas. By removing the bulk of the coarse-grained, “barren” portion of the sample, the signal to background ratio is increased and nugget effects (for gold) are removed. In addition, the analysis requires smaller sample volumes due to the enhanced sensitivity with up to 100 – 300 % increased concentrations of Au, Cu and Zn. Since the commencement of the UltraFine+® Next Gen Analytics research project, additional soil properties, including spectral mineralogy, EC, pH and particle size analysis have been added to the standard multi-element analysis. The UltraFine+® soil analysis method is commercially available exclusively through LabWest.

The UltraFine+® Next Gen Analytics for Discovery research project is taking the UltraFine+® method further, lowering detection limits for the 50+ elements, developing valuable additional soil property analyses, and creating an automated QA/QC “traffic light” system for all analyses. Above all, the Next Gen Analytics provides landscape context for your samples using machine learning methods to integrate spatial data and soil properties in several derived outputs. These additional first-pass interpretations and results add value to the standard commercial output provided by LabWest.

Next Gen Analytics is provided by CSIRO and includes the following outputs: proxy regolith landscape clusters, maps and boxplots of elemental outliers by landscape type, Principal Component Analysis and exploration ratios, soil texture diagrams, source and dispersion directions, and regolith geochemical indices. Outputs are available in a number of easy-to-use files from a web viewer platform, GeoTIFFs, PNGs and CSVs to shapefiles to make it very easy to fast-track your interpretation.

Yes, through our commercial partner, LabWest. You can simply submit your samples like you would for standard soil analysis at a commercial laboratory. Fill out the LabWest sample submission form and request UltraFine+®. The lab may have a couple of different packages depending on whether you need chemistry only, additional REEs, or the spectral mineralogy and sizing data.

As for any soil sampling method, this really depends on your exploration needs and the size of your tenement. However, for the machine learning outputs, ideally, models perform better with hundreds to thousands of samples. In general, we consider the ideal minimum number of samples for sound interrogation to be 50 samples for each landscape type (we currently generate outputs for 4, 8 and 12 landscape clusters). The sample density and number of landscape clusters to use is based on experience and judgement, with the UltraFine+® Next Gen Analytics providing the first-pass assessment requiring the end user to consider the outputs.

Whereas the UltraFine+® method eliminates a lot of uncertainty around sample depth by only analysing the <2 µm fraction, the technique, like most soil sampling, is best conducted on uniform/standardised sample materials. Sometimes this can be a specific horizon or depth. We commonly advise to do some vertical and horizontal orientation studies when first setting up a large survey. Traditionally, many samples have been sampled at approximately 5 to 15 cm below the surface – excluding the organics at the surface. A sampling guide is available here. These are only general guidelines and depend on your exploration context.

This is unlikely. Rare surface geochemistry predictions through deep cover have been observed (Cameron and Leybourne 2005) but many of these are only weakly tested and rarely are statistical checks applied to these results. The UltraFine+® soil analysis method and Next Gen Analytics project are based on scientific principles and processes and can only detect elements that have made their way into the surface soil. However, the UltraFine+® method can detect very subtle anomalies in transported cover that other detection methods may miss. The additional soil properties analysed with the UltraFine+® method will also give you more information on your soil composition, and the integration of spatial data provides landscape context which enables the detection of subtle anomalies in transported cover. Where metals can reasonably migrate to the surface in great enough concentrations, the UltraFine+® method will work as well or better than other methods (Anand et al. 2016).

If you plan to submit samples from overseas, please contact LabWest as any soil samples submitted from outside of Australia are subject to strict quarantine laws. LabWest have recently acquired certification for international quarantine and will be able to assist (international sample submission). There are options for gamma irradiation and heat treatment to release samples from quarantine upon arrival in Australia. The CSIRO team have completed test work on heat treatment of soils from New Zealand and other test samples in Australia, and no appreciable difference has been detected for geochemical assays with respect to target and pathfinder metals. Initial test work on the effects of heat treatment on mineralogy also shows no appreciable differences. Initial results will soon be published here.

As a sponsor of the UltraFine+® Next Gen Analytics project you will receive a Data Package with many different outputs in various formats (GeoTIFFs, PNGs, CSVs and shapefiles) and a web browser-based data viewer, the DSO (Digital Sample Observations). While the DSO is still under development, we will soon publish a short overview video.

The data package will include your soil analysis data (chemical assays, pH, EC, sizing and spectral mineralogy) and QA/QC of your standards and duplicates in addition to first-pass interpretations generated by machine learning. These include landscape clusters (regolith proxies), outliers by landscape type, soil texture, exploration indices and dispersion and source direction for your samples. Data is available as CSV, shapefiles and PNG or GeoTIFF files.

Some of the data we generate may be unfamiliar at first glance. However, in most cases you can treat additional soil property data similarly to the way you would geochemical data. We want to make it as easy as possible to work with the final outputs, so our team is working on developing a range of How-to Guides. They will all be available here, so please check back if the guide you are looking for is not yet available. If you have any other questions (or the guide you need is missing) please contact

As an UltraFine+® Next Gen sponsor you will receive additional information pertaining to 17 soil property parameters from VNIR and 5 parameters from FTIR analysis. These parameters provide information concerning the composition and chemical properties of your ultrafine soil samples and can be used to understand landscape settings and mobile element uptake/adsorption processes. In addition, the spectral data may help identify false positive geochemical anomalies in relation to mineral exploration. The VNIR data can also be directly related to parameters obtained from optical remote sensing imagery (i.e., hyperspectral and multispectral satellites like PRISMA, EnMap, ASTER, Sentinel 2 and Landsat). Tables of the measured parameters and their potential applications along with limitations can be found in our How-to Guides for VNIR and FTIR.

The main difference between the VNIR and FTIR spectroscopy data is that the VNIR uses spectral information which is obtained in the wavelength region between 350 to 2500 nm whereas in the case of FTIR, the region between 4000 to 400 cm-1 (or 2500 to 25,000 nm) is used for determining the spectral properties of the ultrafine soils.

Not just yet… but the team is working towards making the Next Gen Analytics automated and commercialised. In the short-term, contact We are looking to make this a simple addition for industry in the future. More options will be available soon through the next R&D project that will look to advance these outcomes further. For more information on project developments head to UltraFine+® The Future tab on this website.

In general, we can only compare selected elemental data to other methods, so there is not a comparable technique for all the other soil properties reported. However, we have compared the technique to other extractions. For most elements of interest, the UltraFine+® method extracts greater concentrations. Comparison studies were conducted on 4-acid digestion, various aqua regia extractions, hydroxylamine hydrochloride, 24h cyanide leach (Au only), MMI, water leach and other commercially available clay separation techniques. Head to the Publications and Case Studies tab for publicly available comparisons.

If you are looking for mineralisation through great depths of transported cover (>30 m) and have no likely mechanism in place to transport a geochemical signature, because these settings are not conducive to any surface geochemical sampling. UltraFine+® is also not optimised for working with resistate minerals and associated elements (W, Te), especially in residual terrains. In these settings, more traditional litho-geochemical approaches will work well, and we would recommend 4-acid and/or fusion XRF to produce good results. We do favour UltraFine+® over these approaches when the cover is mixed (transported and residual) or there is considerable variance, as the landscape analytics add much more value here than traditional methods. Another reason not to use UltraFine+® might be pre-existing orientation work or historical research that identified an approach that successfully works in your area (don’t change a good thing!) even though UltraFine+® will likely also work in these environments.

We developed an easy “traffic light” system for first pass QA/QC of your duplicates and the UltraFine+® Standard for all analyses. Like a real traffic light, the rules are very simple: Green – you are all good to go. Yellow/Orange – be aware. Red – stop and evaluate. For more information head to our How-to Guides.

Since the UltraFine+® method uses only the <2 µm soil fraction, you cannot currently use CRMs for the whole process (sieving these to <2 µm would change the results rendering them incomparable to the certified reference values). Hence, we have developed the UltraFine+® Standard (QC_320_UFF). This standard has been generated by blending representative soil types from across Australia and we have tested this standard with over 300 analyses for ICP and spectral analyses. Other CRMs can be incorporated into the geochemistry analytical steps only. These can be used, but expect the UltraFine+® method to report higher concentrations of many metals (due to the more aggressive closed vessel aqua regia approach compared to open block). A summary comparison of some commonly used CRMs for soil analyses has been conducted by LabWest and can be found here.

Unfortunately, due to supply limits, the UltraFine+® Standard UFF320 is only available to project sponsors. However, as UFF320 is being phased out, LabWest in collaboration with CSIRO is currently developing additional standards, which will be available to all LabWest clients who use the UltraFine+® analysis.

Yes, the outliers provided by the UltraFine+® Next Gen Analytics workflow are “Tukey” outliers, i.e. points outside k * IQR (interquartile range), where k=1.5. All concentrations are treated as lognormal. The outliers on a log-transformed distribution will be the same as the outliers for the CLR-transformed distribution pertaining to the same element.

The Uniform Manifold Projection and Approximation is a dimensionality reduction algorithm that is applied to transform the spatial input data for each project area to a three-dimensional representation of data. The method does not explicitly include any location information, spatial relationships, or spatial features (e.g., textures) as only the per-pixel values of each input layer are considered.

After UMAP has been applied to the spatial input data, “agg12”, “agg8” and” k-means4” are the algorithms we use in the UltraFine+® Next Gen Analytics workflow to cluster locations with similar spatial data signatures. If you would like more information on these algorithms head to the Publications tab.

The Multi-resolution Valley Bottom Flatness is an index developed over Australia by CSIRO (Gallant and Dowling 2003) that can be used as an indication of depth of transported cover. We use the MrVBF as one of the spatial input layers to generate landscape cluster proxies. The MrVBF is computed from digital elevation data and is publicly available through Geoscience Australia. The UltraFine+ Team in collaboration with John Gallant have also used the algorithm for parts of New Zealand. This will enable us to run our models in most parts of the world.

Orientation work over known deposits and adjacent to known deposits in Western Australia using UltraFine+® successfully located the DeGrussa Cu-Au deposit (2.8Mt grading 5.8% Cu and 1.9g/t Au) and the Mt Eureka Au-Ag deposits. Other deposits located using UltraFine+® include the Federation Pb-Zn-Ag and the Wagga Tank Au-Cu prospects in NSW and the Gruyere (Au) deposit in WA (Bonwick 2020; hons thesis, available on request). Early test work was also conducted near the Hemi Au deposit in the Pilbara and the Lake Rebecca Au deposit, but did not cover the main mineralisation well. Hence results, though positive, were less conclusive and it is unclear from this data how effective the detection was.

We have also seen many positive results in greenfield environments and in areas where sub-economic mineralisation has been detected, e.g., at Blue Billy in the Kalgoorlie region. While not the million-ounce deposits people are looking for, the important consideration is that it is working where we expect it to work: in transported cover. The above examples all have a component of transported cover from 1 to 30 m. A growing list of very positive outcomes are also reported in a number of company ASX reports compiled by our commercial partner .

We are also comfortable to say that it has not always worked. We do still see false positives and Encounter Resources drilled a solid gold anomaly that was enhanced from traditional test work only to get very weak Au in quartz veins at depth. These results do link into the Lake Rebecca study that now shows more promising results. However, the full analytics approach has not been applied in these settings.

The UltraFine+® Next Gen workflow produces clusters of data with similar spatial properties and gives each cluster a unique colour. The machine learning does not take any information on the geographical location (spatial context) or soil properties of these data points into account. We also do not carry out ground-truthing for each site (we will soon publish some preliminary results of our ongoing ground-truthing studies). Hence, the landscape clusters, or landscape types, are considered proxies for the physical regolith types in your project area.

Not as part of this project, but if you have a specific question for your exploration program or tenement package that you would like to explore, you can contact CSIRO for collaboration opportunities on how machine learning might help you answer that question. Email or head to the Contact tab.

The research project started in April 2020 and the final report will be produced in April 2023.

As we are now in the final months of the UltraFine+® Next Gen Analytics research project, spaces for sponsorship are closed. A main constraint is the timing of your analyses so they can be incorporated into the research project. Contact to learn more. The project team is currently looking for the next developments to support industry related to the UltraFine+® technology into the future. New project proposals and collaborations will be coming. You can find more information here: UltraFine+® The Future.

The colours in your outputs (for your 4, 8 or 12 landscape clusters) range from dark red-brown to dark blue. The machine learning workflow assigns a colour to each cluster based on the mean value of the MrVBF calculated for each cluster. The lowest mean MrVBF values are assigned dark red-brown, and often indicate outcrop and subcrop. The highest mean MrVBF values are assigned dark blue, and often indicate areas of deep transported cover. However, the scale is relative (based on the lowest and highest mean MrVBF values in your particular area) and occasionally, the presence of, e.g., elevated lowlands or outcrops in otherwise deep cover can influence the order of landscape colours you might expect.