EOAgriTwin

DestinE DestinationEarth ESA

EOAgriTwin is an innovative project under the European Space Agency’s Digital Twin Earth (ESA DTE) initiative, serving as one of the Earth Observation Lead Digital Twin Components with a focus on agriculture. The aim of EOAgriTwin is to create a comprehensive virtual replica of agricultural systems, at multiple scales, with a focus on agriculture under multiple stressors, and to deliver functional Digital Twin to support monitoring of crop condition, simulation of growth dynamics production under different conditions and stress factors.

Objective:

EOAgriTwin develops an Earth observation–based Digital Twin of agricultural systems to monitor crop conditions, assess risks, and support sustainable decision-making.


Project duration:

24 Months (11/2024-11/2026)


Target audience:

Target users include researchers, policymakers, and agricultural stakeholders.


Consortium & partner role:

The project is implemented by a multidisciplinary consortium of research institutions and industry partners with expertise in Earth observation, agricultural modelling, and digital platforms.


ZALF – Leibniz Centre for Agricultural Landscape Research (Coordinator):

Overall project coordination and scientific leadership; integration of agricultural models, data, and systems for the EO-based Digital Twin.


Humboldt-Universität zu Berlin (HU Berlin):

Satellite-based monitoring of crops and land use, providing EO analytics and time-series products for agricultural assessment.


Remote Sensing Solutions GmbH (RSS):

Technical development of Digital Twin and Dashboard; Operational Earth observation data processing, automated workflows, and scalable EO services.


Università Cattolica del Sacro Cuore (UCSC):

Crop and agro-ecosystem modelling, focusing on crop growth, weather impacts, and stress response simulations.


International Centre of Insect Physiology and Ecology (ICIPE):

Expertise on biotic stressors, particularly insect pests and management systems (Push-Pull Technology) supporting resilience analysis in agricultural systems.


Funding:

Funded by the European Space Agency (ESA) under the Digital Twin Earth (DTE) programme. (Contract No.: 4000146541/24/I-KE)


Contact:

Location Leibniz-Zentrum für Agrarlandschaftsforschung (ZALF) e.V. Eberswalder-Str. 84 15374 Müncheberg, Germany
Phone +49 (0) 3343282-411
Email contact@eoagritwin.eu

Contact point: Dr. Gohar Ghazaryan (ZALF)

EOAgriTwin Project Manager


Disclaimer:

This platform is provided for research and demonstration purposes. Outputs do not replace official advisory or decision-support services.

Drought Hazard:


This layer shows the probability of agricultural drought occurrence on a monthly basis. It represents the likelihood that current vegetation and climate conditions correspond to drought effects. The product is derived using a logistic regression model for drought hazard model that links satellite-based indicators of vegetation condition, surface temperature, and precipitation anomalies with historical crop yield losses. The model produces spatially explicit drought probability values ranging from low to high hazard.


Data sources include Copernicus Sentinel-3 for land surface temperature and vegetation indicators since 2020, MODIS data for land surface temperature and vegetation indicators from 2001 to 2019, TAMSAT Precipitation data to derive the Standardized Precipitation Index (SPI), FAOSTAT national crop yield statistics for model training and calibration and the Copernicus Landcover as a mask for agricultural land.


The layer is intended for near–real-time drought monitoring and comparative assessment of drought conditions across regions.


The algorithm is based on:

Maximilian Schwarz, Tobias Landmann, Natalie Cornish, Karl-Friedrich Wetzel, Stefan Siebert, Jonas Franke (2020) A Spatially Transferable Drought Hazard and Drought Risk modeling Approach Based on Remote Sensing Data. Remote Sensing 12:237, doi:10.3390/rs12020237

NDVI – Normalized Difference Vegetation Index:


This layer shows monthly composites of vegetation greenness and photosynthetic activity, which are key indicators of crop growth, biomass development, and seasonal vegetation dynamics.


NDVI is calculated from optical satellite observations using the contrast between red and near-infrared reflectance, which changes with vegetation density and condition. In this dashboard, NDVI is derived from MODIS (2001-2019) and Copernicus Sentinel-3 (since 2020) data and is used to monitor crop development, detect anomalies, and support drought and crop condition assessments.

NDII – Normalized Difference Infrared Index:


This layer represents vegetation water content and provides information on crop and canopy moisture conditions. It is provided as monthly composites.


NDII is calculated from near-infrared and shortwave-infrared satellite bands, which are sensitive to water content in vegetation. The product is derived from MODIS (2001-2019) and Copernicus Sentinel-3 (since 2020) observations. Lower NDII values indicate reduced canopy moisture and increasing water stress, making this layer particularly relevant for drought monitoring and irrigation analysis.

LST – Land Surface Temperature:


This layer shows the temperature of the land surface as monthly composites, an important indicator of surface energy balance, evapotranspiration, and heat stress in agricultural systems.


LST is derived from thermal infrared satellite measurements from MODIS (2001-2019) and Copernicus Sentinel-3 (since 2020). Elevated land surface temperatures often indicate reduced evapotranspiration and limited water availability, and therefore provide complementary information to vegetation indices for assessing drought and heat stress impacts on crops.

Land Cover:


This layer shows aggregated annual land cover classes and cover fractions at 100 m spatial resolution, describing the dominant land cover type and the proportional presence of key land cover classes such as cropland, forest, grassland, water, and built-up areas.


The product is based on the Global Land Cover dataset provided by the Copernicus Global Land Service using multi-year satellite time series and supervised classification. It is primarily based on PROBA-V surface reflectance data, complemented by ancillary datasets and expert rules to ensure temporal consistency. Land cover classes follow the FAO Land Cover Classification System (LCCS) and are provided together with fractional cover layers and quality indicators.


Citation:

Buchhorn et al. (2020): Copernicus Global Land Service – Land Cover 100 m: Version 3 Globe 2015–2019, Product User Manual. Zenodo.

Irrigated Systems:


This product shows a classification of irrigated and rainfed agricultural systems by identifying differences in seasonal vegetation behavior.


The product is derived from multi-year Sentinel-2 vegetation time series (2022-2024) using harmonic analysis to capture characteristic crop phenology patterns. Irrigated fields typically maintain higher and more stable vegetation signals during dry periods. Thermal and evapotranspiration information is used as supporting data to improve the discrimination between irrigated and rainfed systems, particularly in water-limited environments. Post-processing included a slope mask derived from the Copernices DEM, an agricultural mask (Digital Earth Cropland Mask), and a mask for protected areas (WDPA).


The algorithm is based on:

Tobias Landmann, David Eidmann, Natalie Cornish, Jonas Franke & Stefan Siebert (2019) Optimizing harmonics from Landsat time series data: the case of mapping rainfed and irrigated agriculture in Zimbabwe, Remote Sensing Letters, 10:11, 1038-1046, doi:10.1080/2150704X.2019.1648901

Trend Analysis:


Example text

Crop Yield under Different Management Practices


This What-If Scenario explores how crop yield responds to different water management practices during the growing season.

Simulations are based on the MONICA crop model, which represents crop growth, soil water balance, and yield formation. Earth Observation–based information is used to identify irrigated areas and to parameterize management conditions.

Two management options are considered:

The resulting maps show spatially explicit crop yield outcomes under the selected management assumptions.


How to Use:

  1. Select an area.
  2. Select a crop type.
  3. Select a year.
  4. Choose the management practice (rainfed or irrigated).
  5. View and compare yield maps to assess the impact of water management on crop production.

Results show relative differences between scenarios and years rather than absolute yield forecasts.

Drought Impact during Crop Growth Stages


This What-If Scenario assesses drought stress impacts on crops during different growth stages under alternative water management practices.

The scenario allows users to explore how rainfed and irrigated management influence drought stress severity during specific crop growth phases.

The following growth stages are considered for each crop type:

Stage Winter Wheat Silage Maize
2 Emergence to Double ridge Emergence to Shooting
3 Double ridge to Flowering Shooting to Tasselling
4 Flowering to Grain filling Tasselling to Flowering
5 Grain filling Flowering to Grain filling
6 Senescence Grain filling
7 — Senescence

The resulting maps show spatially explicit crop yield outcomes under the selected management assumptions.


How to Use:

  1. Select an area.
  2. Select a crop type.
  3. Select a year.
  4. Choose the management practice (rainfed or irrigated).
  5. Select a phenological growth state.
  6. Explore stage-specific drought stress maps, ranging from no stress to extreme stress.

Results show relative differences between scenarios and years rather than absolute yield forecasts.

Drought Hazard under Alternative Rainfall Conditions


This What-If Scenario explores how drought hazard changes under different rainfall assumptions relative to the local climatology in the upcoming months based on current conditions.

Rainfall conditions are systematically modified into five scenarios based on local (1° grid cells) rainfall climatology time series analysis:

For each rainfall scenario, vegetation and surface condition indicators (NDVI, NDII, and LST) are modelled consistently with the altered precipitation input. These indicators are then used as inputs to a drought hazard model, which estimates the probability of drought occurrence.


How to Use:

  1. Select the current month (base month from which the scenarios will be calculated into the future).
  2. Select a rainfall scenario.
  3. Select the time horizon (1 month/3 months ahead).
  4. Select output/Scenario Variable (Drought Probability, Land Surface Temperature, NDVI, NDII, SPI3*).
  5. View the corresponding drought hazard map or index to assess changes in changing rainfall conditions.

*The SPI3 scenario output layer is not a prediction. The basis for the SPI3 are the five rainfall scenarios based on the climatology.

Results show relative differences between scenarios and years rather than absolute yield forecasts.