pedometricians have come here
The urssa code provides basic unsupervised functionalities for the use of spectra in laboratory routines.
This repository provides a csv file conainting 350-2500 nm soil spectra and codes (divided into modules) following a reproducible example.
- Unsupervised spectral clustering of samples
- Internal quality control (outlier detection and removal) of analytical results
- Correlation analysis between soil data and spectra
- Samples selection for traditional analysis
- Samples selection for soil prediction (cost reduction)
The Soil_data_spectra.csv file contains:
- Soil Attributes data:
- Clay (g kg-1)
- Sand (g kg-1)
- Organic Matter - OM (g kg-1)
- Cation Exchange Capacity - CEC (mmolc kg-1)
- Base Saturation - V (%)
- Soil reflectance spectra:
- From 350 to 2500 nm at 1 nm resolution
urssa_01.R:
- MODULE 1: Importing and pre-processing data
- MODULE 2: Unsupervised spectral clustering of samples
- MODULE 3: Variable importance
- MODULE 4: Identification of soil attribute outliers
- MODULE 5: Correlation analysis
- MODULE 6: Plotting boxplot and spectra by cluster and Laboratory
urssa_02.R:
- MODULE 1: Quantification of outliers
urssa_03.R:
- MODULE 1: Splitting data into training and test subsets
- MODULE 2: Assessment of subsets
- MODULE 3: Soil attributes modeling with CUBIST
- MODULE 4: Assessment of prediction results
Please, cite the following paper when using urssa:
Poppiel, R.R., Paiva, A.F. da S., Demattê, J.A.M., 2022. Bridging the gap between soil spectroscopy and traditional laboratory: Insights for routine implementation. Geoderma 425, 116029. https://doi.org/10.1016/j.geoderma.2022.116029