Please login first
Estimating photosynthetic and non-photosynthetic vegetation fractional cover and traits in semi-arid tree-grass ecosystems using Sentinel 2 images
1 , 2 , 2 , 3 , 4 , 1 , 1 , 1 , * 1
1  Environmental Remote Sensing and Spectroscopy Laboratory (SpecLab), Spanish National Research Council (CSIC), Madrid 28037, Spain
2  Institute of Agricultural Sciences (CSIC), Tec4AGRO Group, Serrano, 115b, 28006, Madrid, Spain
3  Department of Environment, National Institute for Agriculture and Food Research and Technology (INIA, CSIC), Madrid 28040, Spain
4  Department for Biogeochemical Integration, Max-Planck-Institute for Biogeochemistry, Jena 07745, Germany
Academic Editor: Riccardo Buccolieri


Monitoring photosynthetic (PV) and non-photosynthetic vegetation (NPV) fractions over time can help to detect and assess changes in ecosystem function and services which are critical for resource management and the mitigation of climate change effects. Mixed tree-grass ecosystems are one of the most prevalent forms of terrestrial vegetation across the globe. They provide the basis for livestock production and significantly impact regional and global productivity and food quality. However, these ecosystems represent a challenge for remote sensing applications due to the spectral heterogeneity stemming from the multiple landscape features such as grasses, trees, and shadows. Such a mixture makes it difficult to accurately identify and characterize the different vegetation layers at pixel level in most satellite imagery.

The aim of this study was to use Sentinel 2 (S2) images to estimate PV and NPV fractions in the herbaceous layer of a tree-grass ecosystem and develop empirical models to relate estimated NPV fractions and key vegetation traits associated to senesced vegetation such as Aboveground biomass (AGB). This information will allow to analyze the impact of NPV spatio-temporal variability on grassland productivity as standing senescent plants can compete with green vegetation and difficult seed germination [1].

Spectral mixture analysis (SMA) [2] is a powerful technique that can separate mixed spectral signals to estimate the fractional cover of individual landscape features. SMA and similar unmixing techniques are often applied without considering the spectral phenology of the vegetation layers which may reduce the accuracy and reliability of the results. In this work, we addressed these limitations by using a seasonal endmember spectral mixture analysis (SESMA) approach that accounts for temporal endmember variability. We downloaded and processed a S2 time series of 449 images from 2016 to 2022. Atmospherically corrected Surface Reflectance sub-scenes of about 1.1 x 1.3 km centered on the study area at the Majadas de Tiétar research station ( have been analyzed in combination with spectral and biophysical in-situ data acquired at the site to estimate PV and NPV grass fractions and predict AGB of the NPV vegetation. The SESMA S2 endmembers were derived from field spectra measured with an ASD Fieldspec 3 spectroradiometer (400 to 2500 nm) including PV, NPV and soil measurements acquired in different phenological periods. A partially constrained unmixing with a sum-to-unity constraint on the abundance fractions was applied using ENVI 5.7 image analysis software. Linear regression was further applied to resulting S2 image fractions to relate NPV fraction within-situ NPV AGB as derived from the senescent vegetation samples acquired in 5 to 12 permanent plots in 34 campaigns from 2017 to 2022. Best results (R2 =0.46) were achieved during the grass decay season, where a mixture of green and senescent species is usually found in semi-arid grasslands. An important finding of our study was the effect of flowering, that caused the shifting from direct to inverse relationship between NPV fraction and NPV AGB, which confirms the need to consider phenology, including flowering, in the development of NPV estimation models in highly dynamic herbaceous covers.

[1] Xu, D., Liu, Y., Xu, W., & Guo, X. (2022). The Impact of NPV on the Spectral Parameters in the Yellow-Edge, Red-Edge and NIR Shoulder Wavelength Regions in Grasslands. Remote Sensing, 14(13), 3031.

[2] Shimabukuro, Y. E., & Ponzoni, F. J. (2019). The Linear Spectral Mixture Model. In Spectral Mixture for Remote Sensing: Linear Model and Applications (pp. 23-41). Cham, Switzerland: Springer International Publishing.

Keywords: Sentinel-2; Tree-grass ecosystem; Non-photosynthetic vegetation; Spectral mixture analysis
Comments on this paper
Currently there are no comments available.