Nov 13 – 15, 2024
Leipziger KUBUS Helmholtz-Zentrum für Umweltforschung – UFZ
Europe/Berlin timezone
Welcome to the iDiv Conference 2024! Registration is now closed

Using machine learning for PROSAIL inversion to retrieve functional plant traits of intensively and extensively managed grasslands

Nov 15, 2024, 9:00 AM
15m
Leipziger KUBUS/1-B - Hall 1 B (Leipziger KUBUS)

Leipziger KUBUS/1-B - Hall 1 B

Leipziger KUBUS

150
Talk Flexpool Talk Session

Speakers

Daniel Doktor (Helmholtzzentrum für Umweltforschung - UFZ) Maksim Iakunin (Helmholtz Center for Environmental Research (UFZ))

Description

Grasslands, representing the most extensive terrestrial biome, are increasingly subjected to management intensification, particularly in Europe, where they play an essential role in agricultural systems. The ecological and environmental functions of these grasslands are affected by management practices, which vary in intensity according to environmental conditions. While intensified management enhances agricultural productivity, it often results in negative environmental impacts such as reduced biodiversity and soil nutrient depletion. Traditional biodiversity monitoring methods, despite their precision, are resource-intensive and inadequate in isolation for addressing these challenges. Consequently, remote sensing has emerged as an indispensable tool, providing comprehensive and continuous data for monitoring grassland management and biodiversity.
This study explores the relationship between grassland management regimes and remote sensing signals through the inversion of the PROSAIL radiative transfer model. The aim is to retrieve community-averaged plant traits from canopy hyperspectral reflectance observations. A machine learning model trained on a look-up table generated from various combinations of plant traits and reflectance spectra, facilitates the estimation of management intensity and biodiversity richness. Fieldwork was conducted in experimental grasslands sites (GCEF) with different management regimes to collect functional plant traits and in-situ canopy reflectance spectra throughout a complete vegetation period. The findings illustrate PROSAIL limitations when retrieving plant functional traits (leaf pigment contents, LAI, leaf mass to area ratio) especially for extensively managed grasslands and advanced phenological stages. The purely empirical categorisation of management types based only on observed spectral signals worked well. This indicates i) that the hyperspectral measurements capture substantial information on traits and the underlying management, but also ii) that standard in-situ sampling schemes and radiative transfer modelling approaches lack components to more exactly capture and quantify the increased ratio of ripe/senescent plant parts in diverse grasslands.

Status Group Postdoctoral Researcher

Primary author

Maksim Iakunin (Helmholtz Center for Environmental Research (UFZ))

Co-authors

Daniel Doktor (Helmholtzzentrum für Umweltforschung - UFZ) Franziska Taubert (Helmholtz Centre for Environmental Research GmbH - UFZ) Hannes Feilhauer (Remote Sensing Center for Earth System Research, Department of Remote Sensing in Geo- Ecosystem research, Leipzig University) Dr Reimund Goss (Institute of Biology, Department of Plant Physiology, University of Leipzig) Dr Severin Sasso (German Center for Integrative Biodiversity Research (iDiv), Halle-Jena-Leipzig)

Presentation materials

There are no materials yet.