Estimating grainsize using the thermal infrared spectral features of quartz and carbonate

Hyperspectral logging of diamond drill core has become a routine technique in state drill core repositories across Australia. Manually logging grainsize can be difficult and is a subjective and labour-intensive process that can lead to inconsistencies. This study investigates the application of spectral calculations to automate grainsize estimation using the primary and secondary Reststrahlen features of quartz and carbonate within the 6500–14 400 nm range. Several spectral methods were tested, including automated relative abundance of quartz, feature extraction, polynomial fitting and arithmetic expressions. Data processing was performed using The Spectral Geologist™ 8 software, which automatically generated results for the relative abundance of quartz. This metric proved valuable for identifying lithological boundaries and tracking changes in quartz content, aiding interpretation of depositional environments. Although the feature-extraction method initially produced unexpected spectral outputs, it uncovered variability within lithological units that could be used for further classification. Polynomial fitting at 8190 nm did not correlate well with geological logging, but the spectral fitting at 8630 nm showed excellent agreement with geological observations. Two arithmetic methods for calculating quartz and carbonate grainsizes both performed accurately; however, the quartz grainsize scalar did not provide additional value compared with the other approaches. This paper aims to: (1) showcase the use of thermal infrared spectral features for rapid and reliable grainsize estimation and (2) describe simple spectral calculations geologists can apply during core logging. The results indicate that grainsize estimation can now be automated across a variety of platforms using the spectral algorithms detailed herein. This approach shows potential for the ability to create or improve existing grainsize logs where they exist in historic datasets. These methods, particularly the 8630 nm trough for quartz, also offer significant potential for accurate lithological differentiation and can be applied across different mineral systems, including those associated with economic mineralisation.