HYDRA-EO is a European Space Agency project that develops an operational framework to detect, quantify and differentiate abiotic and biotic crop stress. The project targets priority plant pathogens, including Flavescence Dorée in grapevine, Potato Virus Y in potato, and fungal diseases such as Septoria, while also assessing stress interactions under variable water and environmental conditions.
HYDRA-EO integrates Physically-based radiative transfer modelling approaches with machine learning algorithms to identify the spectral and spatial resolutions required to capture biotic stress signals effectively. By combining hyperspectral, thermal and multi-sensor Earth Observation data from field, airborne and satellite platforms, the project delivers early and reliable indicators to support agricultural resilience and evidence-based decision-making across Europe.
HYDRA-EO integrates UAV, satellite and RTM–ML processing for crop stress detection across studied European sites.
HYDRA-EO aims to develop and demonstrate a physics-informed, hybrid machine learning framework for early detection and attribution of multiple stressors in agricultural systems. By bridging leaf, canopy and landscape scales, the project supports ESA’s Digital Twin Earth vision for crops and prepares methodologies for upcoming missions such as FLEX, CHIME and LSTM.
The project addresses key Mediterranean and temperate crop systems (olive, citrus, pistachio, grapevine) and develops interpretable indicators of plant health based on hyperspectral, fluorescence and thermal data.
Multi-sensor datasets from field instruments, UAV platforms, airborne campaigns and satellite missions are combined with radiative transfer models to retrieve functional traits related to photosynthesis, structure and water status.
HYDRA-EO HYDRA-EO’s mission is to build an operational, multi-scale Earth Observation framework capable of detecting, quantifying and attributing crop stress across Europe. The project aims to bridge fundamental plant physiology with next-generation EO technologies, enabling early warning capabilities for agriculture and supporting ESA’s long-term vision for Digital Twin Earth.
By integrating radiative transfer modelling, hyperspectral and thermal sensing, and hybrid machine learning approaches, HYDRA-EO seeks to deliver interpretable, transferable stress indicators that can be deployed across different crops, environments and sensor platforms. Ultimately, the mission of HYDRA-EO is to accelerate the adoption of physics-guided ML in agricultural monitoring and contribute to more resilient, sustainable crop systems.
HYDRA-EO focuses on four strategic crops and a palette of biotic and abiotic stressors. Field campaigns and EO acquisitions are co-designed to capture contrasted responses across varieties, management and climate zones.
CIAG–IRIAF orchards with inoculated fungal and bacterial pathogens (e.g. Septoria spp., Pseudomonas savastanoi) under varying water regimes are monitored using UAV hyperspectral, thermal, and airborne-borne sensors to build a unique multi-sensor stress dataset.
Multi-stressor trials including Potato Virus Y and fungal infections in a dedicated phenotyping trial, combining leaf measurements, biochemical analysis, UAV imaging, and satellite observations. These experiments generate contrasted physiological responses across varieties under plant diseases.
Italian vineyard plots infected with Flavescence Dorée and additional fungal pathogens form the primary study area, enabling the development and validation of spectral–thermal indicators for early disease detection in perennial cropping systems.
Alfalfa experimental plots in Italy include controlled fertiliser treatments designed to capture a wide range of physiological responses. HYDRA-EO also investigates the spectral and thermal signatures associated with insect pressure in alfalfa, enabling early detection of pest-driven stress.
HYDRA-EO develops a hybrid modelling framework that couples radiative transfer simulations with machine learning to retrieve functional plant traits and disentangle overlapping stress signals (drought vs. disease) across multiple EO platforms and crops.
Retrieve traits such as chlorophyll content, anthocyanins, leaf area index (LAI), biomass, water status, canopy temperature, nitrogen- and red-edge-based indices, as well as SIF-related indicators. These traits are derived from hyperspectral, multispectral and thermal data using PROSAIL- and SCOPE-based simulations combined with machine learning.
Separate the spectral and thermal fingerprints of drought, fungal infections, bacterial and viral diseases, enabling early detection of biotic and abiotic stress. This HYDRA-EO analyses how these stressors interact, overlap and evolve over time, improving resilience assessments and supporting early warning systems in agricultural landscapes.
Quantify how plant trait signals degrade from UAV and airborne hyperspectral data to satellite missions such as Sentinel-2, PRISMA, EnMAP, FLEX, CHIME and LSTM, and design optimised bandsets and indices for operational monitoring. HYDRA-EO develops scaling strategies that link leaf, canopy and landscape processes, preparing methods for next-generation ESA missions.
Deliver open, ESA-oriented tools and Shiny interfaces that translate advanced radiative-transfer modelling, hyperspectral–thermal analysis and hybrid machine learning into practical decision-support applications for agronomists, advisors and policy stakeholders.
HYDRA-EO combines empirical datasets, radiative transfer models and advanced machine learning into a single workflow, aligned with ESA’s Digital Twin Earth vision.
HYDRA-EO activities, public presentations and scientific outputs are disseminated through ESA events, conferences and peer-reviewed papers. This section highlights key milestones and communication actions.
Across the Netherlands, Spain, and Italy, HYDRA-EO is combining field measurements with UAV, airborne hyperspectral and thermal imaging, fluorescence observations, and satellite acquisitions to understand how crop disease and stress become detectable across scales.
The 2026 campaign includes potato monitoring in Lelystad, olive and pistachio experiments at El Chaparrillo, vineyard observations in Italy, and collaborative activities with CERBERUS and regional plant-health partners.
Explore the full campaign story and image gallery →
15–17 September 2026 · Lund, Sweden
José Luis Pancorbo will present two accepted contributions:
9–14 August 2026 · Washington, D.C., USA
The HYDRA-EO team presented two studies on trait-based hybrid modelling, ecosystem productivity, and forest-health monitoring:
Monitoring forest gross primary productivity with Sentinel-2 using plant trait-based hybrid radiative transfer and machine learning
Presenter: Dr. Carlos Camino
Read the conference paper (PDF) →
Early fungal detection in pine forest by fusing PRISMA and Sentinel-2 with biophysical models
Presenter: Dr. Carlos Camino
Read the conference paper (PDF) →
RTM-Suite brings leaf, canopy, soil, and atmosphere radiative transfer modelling together with sensor convolution and plant-trait inversion in one reproducible R and Python ecosystem. Explore 31 step-by-step tutorials, interactive applications, and verified examples for vegetation and Earth Observation research.
HYDRA-EO is coordinated by Wageningen University and brings together expertise in EO services, modelling, plant pathology, genetics and agronomy from the Netherlands, Italy and Spain.
HYDRA-EO builds on existing open-source ecosystems developed by the GRS group at Wageningen University and extends them with ESA-funded modules for plant trait retrieval and stress detection.
RTM-Suite connects leaf, canopy, soil, atmosphere, energy-balance and fluorescence models with sensor convolution, sensitivity analysis, look-up tables and plant-trait inversion. Its R and Python components are developed and numerically verified as one reproducible ecosystem.
ToolsRTMSCOPEinR
ToolsRTM.apptoolsrtmscopeinpython
R package for leaf and canopy optics, real-soil modelling, sensor convolution, spectral indices, look-up tables, and trait inversion using classical machine learning and deep learning.
GitLab repository ↗
R implementation of SCOPE 2.0 for coupled soil-canopy radiative transfer, photosynthesis, surface energy balance, thermal emission, and sun-induced chlorophyll fluorescence.
GitLab repository ↗
Nine interactive applications for PROSAIL, BRDF, MARMIT, SPART, SCOPE, look-up tables, model comparison, inversion, and STAC satellite workflows.
GitHub repository ↗
Python library for leaf and canopy simulation, MARMIT, sensor convolution, spectral indices, ML and deep-learning inversion, and STAC-based satellite data access.
GitHub repository ↗
Python implementation of SCOPE for optical, thermal, biochemical, energy-balance, photosynthesis, and fluorescence simulations within reproducible Python workflows.
GitHub repository ↗Short, reproducible snippets to explore the HYDRA-EO workflows with ToolsRTM and SCOPEinR. You can copy, paste and adapt them in your own R scripts or RStudio projects.
Function documentation and reproducible tutorials for leaf and canopy simulation, sensor convolution, inversion, and sensitivity analysis.
Python API documentation and a guided sensitivity-analysis chapter showing which traits dominate different spectral regions.
library(ToolsRTM)
# Launch Shiny simulators for different RTMs
ToolsRTM::get.simulator(app = "SCOPE")
ToolsRTM::get.simulator(app = "getLUT")
ToolsRTM::get.simulator(app = "PROSAIL-BRDF")
# Type ?get.simulator to see more options
library(ToolsRTM)
#Get default PROSAIL input list
inputs <- ToolsRTM::inputsPROSAIL
# 2.2 Generate a LUT with 100 samples
set.seed(1234)
LUT <- as.data.frame(ToolsRTM::getLUT(inputs = inputs, nLUT = 100, setseed = 1234))
# Summarise min–max per parameter
lut_ranges <- data.frame(
Parameter = names(LUT),
Min = round(sapply(LUT, min, na.rm = TRUE),3),
Max = round(sapply(LUT, max, na.rm = TRUE),3)
)
# Display in an interactive datatable
DT::datatable(
lut_ranges,
caption = "Ranges of sampled input parameters in the LUT",
options = list(pageLength = 10)
)
# DGet soil profile
rsoil.dry <- ToolsRTM::dataSpec_PDB$dry_soil
rsoil.wet <- ToolsRTM::dataSpec_PDB$wet_soil
psoil <- 0.5
rsoil.default<- c(psoil*rsoil.dry+(1-psoil)*rsoil.wet)
wl_grid <- 400:2500 # wavelength grid in dataSpec_PDB
#Simulate PROSAIL (foursail) for each LUT row
sims <- lapply(seq_len(nrow(LUT)), function(i) {
out <- suppressMessages(ToolsRTM::foursail(
inputLUT = LUT[i, ],
rsoil = rsoil.default,
LeafModel = "PROSPECT-PRO",
spectrum.all = TRUE
))
# out has rdot, rsot, rddt, rsdt; combine to BRF using solar zenith (tts)
brf <- ToolsRTM::Compute_BRF(
rdot = out$rdot,
rsot = out$rsot,
tts = LUT[i, "tts"],
data.light = ToolsRTM::dataSpec_PDB
)
# Return tidy tibble for this run
tibble(run = i, wl = wl_grid, rho = as.numeric(brf))
})
library(SCOPEinR)
# Read configuration and LUT tables
setopts <- read.table("input/setoptions.csv", header = TRUE, sep = ",")
inputLUT <- read.table("input/LUT_input.csv", header = TRUE, sep = ",")
# Execute the SCOPE model
db.sim <- SCOPEinR::get.SCOPE(LUT = inputLUT,
options.SCOPE = setopts, get.outputs = "ALL", get.plots = FALSE)
HYDRA-EO is currently in preparation for the ESA Kick-off meeting. For questions regarding the project, collaboration opportunities, datasets or technical details, please contact the project coordinator at Wageningen University & Research.
Dr. Carlos Luis Camino González
Geo-Information Science and Remote Sensing (GRS)
Wageningen University – Department of Environmental Sciences (WU-DES)
Wageningen, The Netherlands
Email:
carlos.caminogonzalez@wur.nl
