SpecTraits

SpecTraits is a Shiny application designed to predict and build models for inferring leaf traits using spectroscopy

SpecTraits offers four main functionalities:

Predict

Predict leaf traits using user-provided Partial Least Squares Regression coefficients or Radiative Transfer Models

Build

Build Partial Least Squares Regression (PLSR) models from a user-defined dataset of leaf traits and spectra

Pre-process

Apply basic pre-processing methods on leaf spectra (e.g., resampling, transformations).

Data

Download curated datasets that integrate leaf traits and spectra

Predict leaf traits using leaf spectra

Predict leaf traits by uploading a .csv file that contains leaf spectra.

SpecTraits provides two ways for predicting leaf traits: i) by importing PLSR coefficients, and ii) by using radiative transfer models (RTM).

The spectra file most contain wavelengths (nm) as columns and samples as rows, a first column should be named ID.

For PLRS coefficients predictions, both the spectra and coefficients files should share the same columns name. For RTM predictions, the spectra file should contain columns between 400 and 2400 nm at 1 nm spacing.

Predicted traits can be validate by uploading a .csv file that contains columns of leaf traits and the exact number of samples as the spectra file.

An example of files containing leaf spectra and traits can be downloaded here.

Step 1 - Import spectra file

Choose file:

Browse...

Step 2 - Select method

Choose a method:

Upload PLSR coefficients:

Browse...

RTM model selection (cctm package; Visser 2021):

PROSPECT-D

PROSPECT-5B

Step 3 - Apply method

Step 4 - External validation (optional)

Choose file:

Browse...

Select trait:

Step 5 - Export predicted traits

Download

Build Partial Least Squares Regression (PLSR) models

Build PLSR models by uploading .csv files that contains leaf spectra and leaf trait.

The spectra file most contain wavelengths (nm) as columns and samples as rows. The first column should be named ID.

The leaf trait file most contain traits as columns and samples as rows, The first column should also be named ID.

SpecTraits provide a way to estimate the optimal number of components using machine learning frameworks.

SpecTraits also provide a way to build and export PLSR models while assessing their performance.

An example of files containing leaf spectra and traits can be downloaded here

Step 1 - Import files

Choose a spectra file:

Browse...

Choose a trait file:

Browse...

Step 2 - Define data split approach

Choose a data split approach:

Data for training:

Step 3 - Evaluate the optimal number of components

ML framework to use:

Leave-one-out CV

10-fold CV

Permutation

Enter a maximun number of components to start:

Number of permutation:

Proportion of data for each permutation:

Step 4 - Run final PLSR models

Enter the optimal number of components to use:

ML framework to use:

Leave-one-out CV

10-fold CV

Permutation

Number of permutation:

Proportion of data for each permutation:

Step 5 - Export models

Include report in ZIP

Report Author:

Report Format:

Download

PLSR Coefficients

Variable of Importance of Projection

Training

Testing

Pre-process leaf spectra

Upload a spectra file and select a preprocessing method to apply.

The spectra file (.csv) must contain wavelengths (nm) as columns and samples as rows, with a first column named ID.

Step 1 - Import spectra file

Choose file:

Browse...

Step 2 - Select preprocessing method

Preprocessing method:

Spectral resampling

Spectral smoothing

Spectral transformation

Step 3 - Configure settings

Define the target spectral spacing and range for FWHM-based resampling

Spectra will be resampled using a Gaussian model that uses the Full Width at Half Maximum (FWHM) of the instrument bands (spectrolab; Meireles et al. 2022).

Target start wavelength (nm)

Target end wavelength (nm)

Target step (Δλ, nm)

Define parameters for Savitzky-Golay smoothing filter

The filter smooths spectra using a polynomial fit within a moving window.

Window size (must be odd)

Derivative order

Window size must be an odd number ≥ 3. Derivative order: 0 = smoothing only, 1 = first derivative, etc.

Select a transformation to apply to the spectra

Transformation type

Scales (comma-separated, e.g., 1,2,3)

Variance

Summed-wavelet spectra using continuous wavelet transform (CWT package)

Band window

Scale order

Derivative transformation with specified band window and scale order

Vector normalization using Euclidean norm (2-Norm). No parameters needed.

Step 4 - Run preprocessing

Apply FWHM-based spectral resampling

Apply Savitzky-Golay smoothing filter to spectra

Apply spectral transformation

Step 5 - Export processed spectra

Download the processed spectra as a .csv file.

Download

Original Spectra

Processed Spectra

About SpecTraits

Version 0.1 | GitHub

Leaf spectroscopy has emerged as a powerful tool for the rapid, non-destructive estimation of leaf traits. However, deriving trait information from spectral data is often not standardized across research studies and can be technically demanding, creating substantial barriers for users without expertise.

SpecTraits was developed to bridge this gap by providing an integrated, user-friendly platform that consolidates tools for deriving leaf traits from spectral data. Our goal is to make leaf spectroscopy more accessible to researchers and plant scientists who use leaves for phenotyping (e.g., ecologists, agronomists, and evolutionary biologists), enabling them to fully leverage the potential of spectroscopy to advance the understanding of plant evolution and function.

If you use SpecTraits in your research, please cite:

@software{SpecTraits,
  author = {Guzmán, J. Antonio and Cavender-Bares, Jeannine},
  title = {SpecTraits: A Shiny Application for Leaf Trait Prediction Using Spectroscopy},
  year = {2026},
  version = {0.1},
  url = {https://github.com/ASCEND-BII/SpecTraits}
}

The development of SpecTraits was supported by:

NSF DBI: 2021898

For questions, bug reports, or feature requests, please visit our GitHub Issues page or contact the development team.

Last updated: 2026-04-01