Published March 13, 2026
| Version 0.3.0
Software
Open
autoSpectra: Soil Spectral Modeling, Visualization and Prediction
Authors/Creators
Description
An R package and Shiny application for predicting soil physical and chemical properties from diffuse reflectance spectra. Supports VisNIR (350–2500 nm) and MIR (600–4000 cm⁻¹) spectral domains. Uses soilVAE asymmetric autoencoders trained on the Open Soil Spectral Library (OSSL) v1.2 with a canonical two-step Savitzky-Golay preprocessing pipeline. Supports 34 OSSL Level-1 harmonized soil properties and sensor-agnostic prediction across all OSSL-contributing instruments.
Notes
Data used by this package: Open Soil Spectral Library (OSSL) v1.2, available under CC-BY 4.0 at https://docs.soilspectroscopy.org. Spectral preprocessing uses the prospectr R package (Stevens & Ramirez-Lopez, 2014). Predictive models are soilVAE asymmetric autoencoders implemented via Keras/TensorFlow.
Files
HugoMachadoRodrigues/autoSpectra-v0.3.0.zip
Files
(1.3 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:1cfca733860f029f5b63dc065611ba3b
|
1.3 MB | Preview Download |
Additional details
Related works
- Is supplement to
- Software: https://github.com/HugoMachadoRodrigues/autoSpectra (URL)
- References
- Journal article: 10.5194/essd-15-3829-2023 (DOI)
Software
- Repository URL
- https://github.com/HugoMachadoRodrigues/autoSpectra