Diffuse reflectance spectroscopy for estimating soil properties: A technology for the 21st century

Spectroscopic measurements of soil samples are reliable because they highly repeatable and reproducible. They characterise the samples' mineral–organic composition. Estimates concentrations constituents inevitably less precise than estimates obtained conventionally by chemical analysis. But cost each spectroscopic estimate is at most one-tenth a determination. Spectroscopy cost-effective when we need many data, despite costs errors calibration. Soil spectroscopists understand risks over-fitting models to dimensional multivariate spectra have command mathematical statistical methods avoid them. Machine learning has fast become an algorithmic alternative analysis for estimating from reflectance spectra. As with any modelling, judicious implementation machine as it also carries risk predictions irrelevant elements To use confidently, validate outcomes appropriately sampled, independent data sets. Not all should be considered ‘black boxes’. Their interpretability depends on algorithm, some interpretable explainable. Some difficult interpret complex transformations or their huge complicated network parameters. there rapidly advancing research explainable learning, these finding applications in science spectroscopy. In parts world, environmental scientists recognise merits building spectral libraries which can draw localise modelling derive information new projects within domains. We hope our article gives readers more balanced optimistic perspective spectroscopy its future. Highlights reproducible analytical technique. Spectra calibrated properties known error. properties. becoming ever powerful extracting accurate spectra, interpreting exist. Large provide that used locally aid samples.