Plant species richness prediction from DESIS hyperspectral data: A comparison study on feature extraction procedures and regression models

TL;DR

This study demonstrates that spaceborne DESIS hyperspectral data can predict ground-measured plant species richness in two Australian habitats with region-specific models, leveraging dimensionality reduction (PCA, CCA, PLS) and regression (KRR, GPR, RFR). The best results reached r = 0.76, RMSE = 5.89 in the Southern Tablelands and r = 0.68, RMSE = 5.95 in the Snowy Mountains, with the red-edge, red, and blue bands providing the strongest predictive signal. A two-component reduction consistently yielded optimal performance across feature extractors, and region-specific modelling outperformed pooled modelling, highlighting the importance of local spectral–biodiversity relationships. Compared to Sentinel-2, DESIS offered higher predictive accuracy, underscoring the value of hyperspectral information for biodiversity mapping, while also identifying practical considerations and limitations for scaling to larger areas and future SWIR-enabled missions.

Abstract

The diversity of terrestrial vascular plants plays a key role in maintaining the stability and productivity of ecosystems. Airborne hyperspectral imaging has shown promise for measuring plant diversity remotely, but to operationalise these efforts over large regions we need to advance satellite-based alternatives. The advanced spectral and spatial specification of the recently launched DESIS (the DLR Earth Sensing Imaging Spectrometer) instrument provides a unique opportunity to test the potential for monitoring plant species diversity with spaceborne hyperspectral data. This study provides a quantitative assessment on the ability of DESIS hyperspectral data for predicting plant species richness in two different habitat types in southeast Australia. Spectral features were first extracted from the DESIS spectra, then regressed against on-ground estimates of plant species richness, with a two-fold cross validation scheme to assess the predictive performance. We tested and compared the effectiveness of Principal Component Analysis (PCA), Canonical Correlation Analysis (CCA), and Partial Least Squares analysis (PLS) for feature extraction, and Kernel Ridge Regression (KRR), Gaussian Process Regression (GPR), and Random Forest Regression (RFR) for species richness prediction. The best prediction results were $r=0.76$ and $\text{RMSE}=5.89$ for the Southern Tablelands region, and $r=0.68$ and $\text{RMSE}=5.95$ for the Snowy Mountains region. Relative importance analysis for the DESIS spectral bands showed that the red-edge, red, and blue spectral regions were more important for predicting plant species richness than the green bands and the near-infrared bands beyond red-edge. We also found that the DESIS hyperspectral data performed better than Sentinel-2 multispectral data in the prediction of plant species richness.

Figures (12)

• Figure 1: Locations of in-situ plant species richness samples collected in field experiments.
• Figure 2: Histograms of species richness distribution for sampling plots in the (a) Southern Tablelands, and (b) Snowy Mountains regions.
• Figure 3: Comparison of the DESIS (before and after pre-processing) and Sentinel-2 spectra at one of the ground sampling plots. An EnMAP spectrum simulated for a random location is also shown for reference.
• Figure 4: Average DESIS reflectance spectra calculated from field sample plots with low, intermediate, and high species richness for the (a) Southern Tablelands and (b) Snowy Mountains regions.
• Figure 5: Impact of number of components on the estimation accuracy of plant species richness with (a) Principal Component Analysis (PCA), (b) Canonical Correlation Analysis (CCA), and (c) Partial Least Squares analysis (PLS).