Our work was presented at Living Planet Symposium 2025 (23-27 June, 2025 Vienna) during the poster session.
Enhancing Sugarcane Stress Detection with Hyperspectral and Thermal Data: Insights from the PRISMA4AFRICA Project.
The PRISMA4AFRICA project aims to establish a partnership between African and European organizations to advance the adoption and use of Earth Observation (EO) technologies for precision farming and food security. This initiative is designed to address user needs while leveraging the opportunities and challenges offered by recent EO data processing and modelling. In the framework of this project, we develop and disseminate tools based on thermal and hyperspectral EO data to detect plant stress, with a particular focus on stresses impacting sugarcane plantations. Sugarcane is widely cultivated in all the collaborating countries—Gabon, Mozambique, and South Africa—which are represented by AGEOS, INIR, IIAM, and SASRI, namely African Early Adopters (AEA). The SASRI team has identified key stress factors affecting sugarcane, including yellow sugarcane aphid infestations and Eldana damages, while INIR is particularly interested in water stress. This collaboration is fundamental for validating the products using in-situ data collected quasi real-time with the hyperspectral acquisitions. To this end, an online training session was organized to share the theory and practice of data collection with African collaborators.
Hyperspectral and thermal data, generally less exploited compared to multispectral data due to their complexity and current limitations in terms of revisit time, offer instead significant opportunities because enabling the retrieval of (a) crop biochemical and biophysical vegetation parameters (e.g., LAI, FAPAR, LCC/CCC and CWC), (b) soil properties (e.g., soil organic carbon – SOC), and (c) evapotranspiration (ET) and Evaporative Stress Index (ESI). The goal of the joint activity is to generate stress maps by combining outputs from different processing and modelling used as input PRISMA (ASI Italian mission) and EnMAP (DLR German program) for the hyperspectral (0.4-2.5 µm) dataset and the ECOSTRESS (Ecosystem Spaceborne Thermal Radiometer Experiment on Space Station on the ISS) and Landsat-8/9 for the thermal ones (8-12 µm).
Crop biochemical and biophysical parameters retrieval was achieved through a hybrid approach. The radiative transfer model PROSAIL was used to generate a training dataset encompassing different illumination and geometry configurations, which is then applied to train Machine Learning Regression models (tree-based models and Gaussian Process Regression – GPR models, depending on the target variables). These models have been validated in different countries, achieving promising results: RMSE = 0.38 m²/m², R² = 0.82 for LAI; RMSE = 0.093, R² = 0.805 for FAPAR; RMSE = 0.019 g/cm², R² = 0.77 for CWC; and RMSE = 0.38 µg/cm², R² = 0.695 for Chlorophyll. Whereas SOC was estimated using a 1-Dimensional Convolutional Neural Network (1D-CNN), trained on an extensive global PRISMA dataset, combined with SOC values from the ICRAF and KKSL (https://soilspectroscopy.org/) spectral libraries. Transfer learning was subsequently applied to refine retrieval processes for the specific areas of interest. A preliminary test of the methodology was conducted in South Africa, yielding an R² value of 0.47.
ECOSTRESS L3/L4 standard products were exploited to extract information about the ET and water stress. Unfortunately, these data are not always produced because of the lack of ancillary layers required by the ECOSTRESS processing chain for the ET calculation. For this purpose, an ad hoc workflow was set up to derive both albedo and LAI directly using PRISMA images. To improve the spatial resolution, the Data Mining Sharpener (DMS) algorithm was successfully applied to the sharpening of the LST products using PRISMA-derived 30m-NDVI.
While crop vegetation parameters and soil properties will be validated using in situ data collected by the AEA, the absence of Eddy Covariance or ET stations within the study areas will limit the evaluation of the retrieved ET/ESI to a qualitative assessment. So, this evaluation will involve comparisons with WaPor FAO portal values (https://data.apps.fao.org/wapor) or cross-validation against data from better-instrumented reference sites. Preliminary results clearly show that the 30m ET products derived by combining PRISMA and ECOSTRESS using the Priestley-Taylor Jet Propulsion Laboratory (PT-JPL) algorithm are of good quality in terms of dynamic range and spatial pattern. The PRISMA-ECOSTRESS ET product shows a high correlation with ESA-STIC and NASA-PT-JPL products and a RMSE of 32 and 19 W/m2, respectively.
To conclude, this study evaluates the potential of thermal and hyperspectral data for detecting stress and damages in crop fields like sugarcane. By integrating these advanced technologies, it becomes possible to provide critical insights to enhance the resilience of plantations against various stressors and to contribute to food security efforts. Furthermore, with the upcoming hyperspectral missions like CHIME (ESA) and SBG (NASA), as well as thermal missions such as LSTM (ESA), SBG-TIR (NASA), and TRISHNA (ISRO), these tools pave the way for the development of an operational monitoring system in the framework of precision farming and food security.
Authors: Roberta Bruno (1), Raffaele Casa (2), Francesca Fratarcangeli (1), Saham Mirzaei (3), Francesco Palazzo (4), Simone Pascucci (3), Stefano Pignatti (3), Nitesh Poona (5), Chiara Pratola (1), Zoltan Szantoi (4), Alessia Tricomi (1)
Organisations: (1): e-GEOS S.p.A.; (2): Department of Agriculture and Forestry Sciences (DAFNE), University of Tuscia, Via San Camillo de Lellis, 01100 Viterbo, Italy; (3): Institute of Methodologies for Environmental Analysis (IMAA)- Italian National Research Council (CNR), C. da S.Loja, 85050 Tito Scalo, Italy; Via Tiburtina, 965 – 00156 Rome – Italy; (4): European Space Agency (ESA), Frascati, Italy; (5) South African Sugarcane Research Institute (SASRI), Mount Edgecombe, 4302, South Africa