Related Experiment Video
Updated: May 21, 2026
10:14
Reconstructing Terrestrial Paleoclimate and Paleoecology with Fossil Leaves Using Digital Leaf Physiognomy and Leaf Mass Per Area
Published on: October 25, 2024
Machine learning surrogate for the leaf PROSPECT-D model and its applications across plant species
Milad Rahimi-Majd1,2, Rudan Xu1,2, Stefan Bauermeister1
1Bioinformatics Department, Institute of Biochemistry and Biology, University of Potsdam, 14476, Potsdam, Germany.
Scientific Reports
|May 19, 2026
Summary
This study enhances leaf trait estimation from hyperspectral data by combining mechanistic (PROSPECT-D) and machine learning models. It introduces a framework to assess model transferability across species, improving accuracy for plant science applications.
Area of Science:
- Plant physiology and remote sensing.
- Utilizes leaf hyperspectral reflectance (HSR) for trait prediction.
Background:
- Leaf hyperspectral reflectance (HSR) is crucial for estimating leaf traits using machine learning (ML) and mechanistic models like PROSPECT.
- Limited ground truth data restricts comprehensive evaluation of PROSPECT model accuracy and transferability across diverse species.
Purpose of the Study:
- To evaluate and improve the accuracy and transferability of the PROSPECT-D model and ML models for leaf trait estimation.
- To develop a novel framework for analyzing ML model transferability across species using PROSPECT-D.
- To create a fast, accurate deep-learning-based surrogate for PROSPECT-D inversion.
Main Methods:
- Employed PROSPECT-D model inversion and forward simulation across multiple species.
- Introduced a partial least squares regression framework to analyze ML model transferability.
- Developed a deep-learning-based surrogate for PROSPECT-D inversion using neural networks.
Main Results:
- Identified four narrow wavebands influenced by environmental factors.
- Revealed trait-specific transferability patterns for ML models trained on PROSPECT-D forward simulations.
- Achieved fast and accurate leaf trait estimation using the developed deep-learning surrogate.
Conclusions:
- The data-driven framework enhances the accuracy and transferability of PROSPECT and similar mechanistic models.
- Deep learning offers a powerful approach for rapid and precise leaf trait estimation from HSR data.
- Improved mechanistic and ML model integration facilitates broader applications in plant science.
Related Concept Videos
02:16Light Acquisition
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
01:59Plant Breeding and Biotechnology
Crop cultivation has a long history in human civilization, with records showing the cultivation of cereal plants beginning at around 8000 BC. This early plant breeding was developed primarily to provide a steady supply of food.