Spectroscopy

Predicts soil properties from spectral fingerprints using ML calibration, enabling fast, low-cost soil characterization.

This project uses soil spectroscopy to estimate soil properties rapidly from Vis–NIR and FTIR spectral signatures. We pair spectra with laboratory reference measurements to calibrate machine learning models that translate spectral fingerprints into soil property estimates. Preprocessing improves signal quality, and validation tests robustness across soil types and value ranges. Deliverables include calibrated prediction workflows that accelerate soil characterization and strengthen mapping datasets.

Main focus

  • Fast estimation of soil properties from Vis–NIR and FTIR spectra.
  • ML calibration models linking spectral features to reference measurements.
  • Scalable workflows for monitoring, mapping, and sampling efficiency.

Objectives

  1. Build spectral libraries linked to laboratory soil property measurements.
  2. Train and validate ML models for accurate, transferable spectral prediction.
  3. Deploy calibrated tools to support high-throughput soil assessment and mapping.

Graphical abstract

:::: {.row .justify-content-sm-center} ::: {.col-sm-10 .mt-3 .mt-md-0}

::: ::::

::: caption Graphical abstract summarizing the workflow and key results. :::

Papers (selected)

Paper 1 — Predicting Soil Weathering Indices Using FTIR

DOI




Paper 2 — Predicting Soil Organic Carbon Using FTIR

DOI




Paper 3 — Remote Sensing and Vis–NIR for Soil Mapping

DOI



References