Simulating Resilience: NASA's Agricultural Digital Twin (ADT) Approach to Drought Risk

1. Introduction: The Convergence of Planetary Science and Agronomy

The contemporary agricultural landscape is currently navigating a period of profound transformation, driven by the colliding forces of accelerating climate change, burgeoning global population demands, and the rapid digitization of environmental science. By the mid-21st century, the global demand for food, feed, and fiber is projected to increase significantly, yet the biophysical systems that underpin agricultural productivity—stable precipitation patterns, predictable growing seasons, and reliable soil moisture reserves—are becoming increasingly volatile. In this context, the traditional paradigms of agricultural management, which have historically relied on retrospective analysis, empirical intuition, and station-based weather averages, are proving insufficient to mitigate the risks posed by a destabilized climate. The emerging solution, poised at the frontier of Earth science and computational engineering, is the Agricultural Digital Twin.

This report offers a comprehensive deep-dive into the recent advances in NASA’s Agricultural Digital Twin initiatives, specifically examining their application in drought monitoring and risk assessment during the 2024-2025 period. Unlike static computer models or singular satellite maps, a Digital Twin represents a dynamic, living replica of a physical system. In the context of agriculture, it integrates continuous streams of data from diverse sources—including orbital sensors, ground-based flux towers, and biophysical crop models—to update a virtual representation of a farm, a watershed, or an entire continental breadbasket in near real-time.¹ This technological leap allows stakeholders to move beyond merely observing the current state of the Earth to simulating future outcomes under a vast array of hypothetical scenarios, answering the critical "what-if" questions that define modern risk management.³

The development of these systems represents a significant departure from previous generations of Earth system modeling. Historically, the domains of surface hydrology and crop physiology operated in relative isolation; hydrologists modeled the movement of water through the soil column, often treating vegetation as a simplified boundary condition, while agronomists modeled crop growth, frequently treating water availability as a static input. The Agricultural Digital Twin bridges this divide through the novel coupling of high-performance hydrologic frameworks, such as NASA’s Land Information System (LIS), with advanced crop growth models like the Decision Support System for Agrotechnology Transfer (DSSAT).⁴ This synthesis creates a unified system where the biological progression of the crop and the physical dynamics of the water cycle are inextricably linked, mirroring the complexity of the real world.

Furthermore, the integration of Artificial Intelligence (AI) and Machine Learning (ML) has revolutionized the operational capabilities of these twins. Through the development of "surrogate models" and emulators, researchers can now run thousands of simulations in the time it once took to run a single physics-based iteration, enabling the rigorous quantification of uncertainty and risk.³ As drought events increasingly manifest as rapid-onset "flash droughts," characterized by the swift depletion of soil moisture driven by high atmospheric demand, the capability of the digital twin to provide early warnings based on thermodynamic stress has become a critical asset for sustainable agriculture.⁷

Through the lens of pilot projects led by NASA Harvest, NASA Acres, and the Earth System Digital Twin (ESDT) program, this research articulates how these technologies are reshaping the epistemological foundations of agriculture. It explores the granular details of sensor capabilities—from the thermal precision of ECOSTRESS to the superspectral promise of Landsat Next—and evaluates the socio-economic implications of transitioning from reactive disaster response to proactive, data-driven resilience.

2. The Conceptual Evolution of the Digital Twin in Earth Science

To understand the magnitude of recent advances, it is necessary to contextualize the Digital Twin within the broader history of simulation technology. The concept originated in industrial engineering and aerospace, specifically within NASA’s Apollo program, where "living models" of spacecraft were created to simulate failures and test solutions on the ground before transmitting instructions to the crew in space.⁶ These industrial twins were closed systems, engineered by humans, where every component and variable was known and quantifiable.

Transposing this concept to the Earth system—and specifically to agriculture—requires a fundamental architectural shift. An agricultural ecosystem is an open, chaotic system characterized by non-linear interactions between the atmosphere, the pedosphere (soil), and the biosphere. Unlike a jet engine, which degrades in a predictable manner, a cornfield in Iowa interacts with dynamic weather patterns, heterogeneous soil properties that vary meter by meter, and complex human management decisions regarding irrigation, fertilization, and tillage.

2.1 The Earth System Digital Twin (ESDT) Framework

NASA’s approach to this complexity is anchored in the Earth System Digital Twin (ESDT) framework. An ESDT is defined not merely as a high-resolution model, but as a dynamic and interactive information system that fulfills three distinct functions: providing a continuous digital replica of the past and current state of the Earth; computing forecasts of future states under nominal assumptions; and offering the capability to investigate hypothetical scenarios under varying impact assumptions.²

This tripartite structure—Digital Replica, Forecast, and Impact Assessment—transforms the model from a scientific research tool into a decision-support engine. The "Digital Replica" component is fed by continuous observations from the Earth System Observatory (ESO), powered by data assimilation and fusion to ensure it remains a faithful mirror of reality. The "Forecast" component utilizes advanced computational capabilities to predict near-term evolution. The "Impact Assessment" component, perhaps the most critical for agriculture, utilizes machine learning and causality analysis to run large ensembles of "what-if" simulations, allowing users to explore the consequences of potential interventions.¹¹

2.2 The Shift from Static Modeling to Dynamic Interaction

The transition to a Digital Twin architecture represents a move away from the static "run and publish" workflow of traditional academia. In a conventional workflow, a researcher might configure a crop model, feed it historical weather data, run a simulation, and publish the results months later. In the ADT framework, the model is "alive." It runs continuously, ingesting new satellite data as it becomes available (e.g., a new soil moisture map from SMAP every 2-3 days), correcting its internal states, and automatically updating its yield forecasts.⁹

This continuous loop creates a system that "learns" and adapts. If a forecasted rain event fails to materialize, the digital twin detects the discrepancy through satellite soil moisture observations, adjusts the water balance in the model, and immediately recalculates the stress impact on the crop. This dynamic responsiveness is essential for monitoring phenomena like flash drought, which can develop with a speed that outpaces traditional reporting mechanisms.⁷

3. Architectural Core: The Coupling of LIS and DSSAT

The scientific validity of the Agricultural Digital Twin rests on the rigor of its underlying physics engines. The architecture developed by researchers at NASA Goddard Space Flight Center (GSFC) relies on the bidirectional coupling of two premier modeling frameworks: the Land Information System (LIS) and the Decision Support System for Agrotechnology Transfer (DSSAT).

3.1 The Land Information System (LIS)

The Land Information System is a high-performance software framework designed to model land surface processes. It serves as the "hydrology engine" of the digital twin. LIS integrates satellite and ground-based observational data to solve the energy and water balance equations at the land surface.¹³ It calculates how precipitation is partitioned into runoff, infiltration, and evaporation, and how energy from the sun is partitioned into sensible and latent heat fluxes.

LIS allows for the use of an ensemble of land surface models (LSMs), such as Noah-MP, CLM, and VIC. This ensemble approach is crucial for characterizing uncertainty, as different models may parameterize soil physics differently. However, standard LSMs have historically possessed a limited representation of vegetation. They typically treat crops as a "green bucket"—a static layer of vegetation with fixed properties like root depth and leaf area index (LAI) that do not respond dynamically to environmental stress. This limitation makes them poor predictors of agricultural yield.⁴

3.2 The Decision Support System for Agrotechnology Transfer (DSSAT)

To address the biological deficiencies of LIS, the ADT integrates DSSAT, a suite of crop simulation models that encompasses over 42 different crops. DSSAT serves as the "crop engine." It simulates the phenological development of the plant day by day, from germination to harvest, based on genetic coefficients, soil conditions, and daily weather inputs.⁴

DSSAT excels at simulating the biological response of crops to management practices—such as the timing of nitrogen application or the choice of cultivar. However, standalone DSSAT simulations often rely on simplified "tipping bucket" water balance models that do not adequately capture the complex lateral and vertical movement of water in the soil, nor do they fully utilize the spatial richness of satellite remote sensing data.⁴

3.3 The Coupling Mechanism

The innovation of the NASA Agricultural Digital Twin lies in the coupling of these two systems. In this integrated framework, LIS provides the detailed soil hydrology and surface energy fluxes to DSSAT. It tells the crop model exactly how much water is available in the root zone and what the temperature of the canopy is. Conversely, DSSAT provides LIS with the evolving biological state of the crop. As the simulated corn plant grows, its roots deepen and its leaves expand; DSSAT feeds these updated Leaf Area Index (LAI) and root depth values back into LIS.⁴

This feedback loop ensures physical consistency. When the crop grows, it extracts more water from the soil in the LIS model, which in turn dries the soil, potentially inducing stress that DSSAT then registers, slowing the crop's growth. This non-linear interaction captures the complex feedbacks between vegetation and hydrology that characterize real-world drought impacts. The result is a system that leverages the superior hydrology of LIS and the superior biology of DSSAT.⁴

3.4 Data Assimilation: The Corrective Force

Even the best models drift from reality due to errors in input data (forcing) or imperfect parameterizations. Data Assimilation (DA) is the mathematical process used to correct the model trajectory using observations. The ADT utilizes advanced DA algorithms, such as the Ensemble Kalman Filter (EnKF), encapsulated within the LIS-DA subsystem.¹⁶

In operation, the ADT propagates a "cloud" of simulations (an ensemble), each with slightly different initial conditions or weather inputs to represent uncertainty. When a satellite observation becomes available—for example, a soil moisture retrieval from the SMAP satellite—the DA system compares the model's predicted soil moisture with the satellite's observed soil moisture. It then optimally adjusts the model's state variables (soil moisture, canopy temperature, biomass) to move the simulation closer to the observation, weighted by the relative uncertainty of the model and the observation.¹⁶

Recent advances in 2024 and 2025 have expanded the scope of assimilation to include vegetation optical depth (VOD) from microwave sensors and leaf area index (LAI) from optical sensors like MODIS and VIIRS. This multivariate assimilation allows the system to correct both the hydrological state (is the soil wet?) and the biological state (is the plant growing?) simultaneously, preventing logical inconsistencies in the digital twin.⁵

4. The Observational Nervous System: Remote Sensing Technologies

The "senses" of the Agricultural Digital Twin are the instruments aboard NASA’s fleet of Earth-observing satellites. These sensors operate across the electromagnetic spectrum, detecting signals that are invisible to the human eye but critical for diagnosing plant health and water status.

4.1 SMAP: Monitoring the Soil Reservoir

The Soil Moisture Active Passive (SMAP) mission remains the cornerstone of hydrologic monitoring within the ADT. Operating in the L-band microwave frequency (1.41 GHz), SMAP is uniquely capable of penetrating cloud cover and moderate vegetation to sense the dielectric constant of the soil, which is directly related to its water content.¹⁸

While SMAP provides global coverage every 2-3 days, its native radiometer resolution of approximately 36 kilometers is too coarse for field-level agriculture. A major focus of recent research has been the development of downscaling algorithms. By fusing SMAP’s radiometer data with high-resolution radar data from the European Space Agency’s Sentinel-1 satellites (and potentially the upcoming NISAR mission), researchers have generated soil moisture products at 1-kilometer and even 400-meter resolutions.¹⁹

These high-resolution products are integrated into tools like Crop-CASMA (Crop Condition and Soil Moisture Analytics), a web-based geospatial application developed in partnership with the USDA. Crop-CASMA allows users to visualize soil moisture anomalies at the county and sub-county scale, providing a critical metric for assessing the "wetness" or "dryness" of the soil relative to historical norms.²² This data is essential for determining planting windows; if the soil is too wet (as seen in the 2019 Midwest floods) or too dry, farmers can adjust their schedules accordingly.

4.2 ECOSTRESS: Thermodynamics of Plant Thirst

Complementing the soil moisture data from SMAP is the thermal data from the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS). While SMAP measures the water supply in the soil, ECOSTRESS measures the water demand of the atmosphere and the physiological stress of the plant. It does this by measuring Land Surface Temperature (LST) with high precision.²⁴

The physics behind this measurement relies on the cooling effect of transpiration. A healthy, well-watered plant transpires water vapor through its stomata, a process that consumes energy and cools the leaf surface. When a plant is water-stressed, it closes its stomata to conserve water, causing the leaf temperature to rise. ECOSTRESS detects this rise in temperature before the plant shows any visible signs of wilting or browning.²⁴

A unique advantage of ECOSTRESS is its platform on the International Space Station (ISS). Unlike sun-synchronous satellites (like Landsat) that pass over a location at the same time every day (usually mid-morning), the ISS possesses a precessing orbit that allows ECOSTRESS to sample the Earth at different times of the day. This diurnal sampling capability is critical for agriculture because plant stress is often most acute in the afternoon when atmospheric demand is highest. A plant might appear healthy at 10:00 AM but shut down its photosynthesis by 2:00 PM. ECOSTRESS captures this afternoon stress signal, providing a more accurate assessment of daily water use.²⁵

The primary derived product from ECOSTRESS is the Evaporative Stress Index (ESI), which quantifies anomalies in evapotranspiration. Recent studies have shown that ESI can provide an early warning of flash drought up to several weeks before vegetation indices like NDVI show any decline, making it a vital component of the ADT’s early warning system.⁸

4.3 Landsat Next: The Superspectral Future

Looking toward the future horizon of the 2030s, the ADT architecture is being designed to ingest data from the upcoming Landsat Next mission. Landsat Next represents the most significant upgrade to the Landsat program in its 50-year history. It will replace the single-satellite operational model with a constellation of three identical satellites, improving the temporal revisit time from 8 days (with two satellites) to roughly 6 days.²⁸

Critically for agriculture, Landsat Next will feature a "superspectral" instrument with 26 spectral bands, compared to the 11 bands on Landsat 8/9. These new bands are strategically placed to measure specific biophysical and biochemical properties of crops.

• Red Edge Bands: Situated between the visible red and near-infrared, these bands are highly sensitive to chlorophyll content and leaf structure, enabling more precise detection of crop nitrogen status and early disease onset.²⁹

• Shortwave Infrared (SWIR) Bands: Additional SWIR bands will improve the discrimination of crop residues (dead vegetation) from bare soil, which is crucial for monitoring conservation tillage practices and soil health.²⁹

• Thermal Infrared Bands: Landsat Next will double the number of thermal bands, improving the accuracy of surface temperature retrieval and atmospheric correction, thereby refining field-scale evapotranspiration estimates.³⁰

• Water Quality Bands: New bands in the visible spectrum will allow for the detection of harmful algal blooms and suspended sediments in agricultural waterways, linking terrestrial management to aquatic health.³¹

The enhanced spectral resolution of Landsat Next will allow the ADT to move beyond simple biomass estimation to the detailed characterization of crop physiology, nutrient content, and water use efficiency at the scale of a single field (10-20 meters).³¹

4.4 OpenET: Democratizing Water Data

While high-end satellites provide the raw data, the translation of that data into actionable information often requires intermediate platforms. OpenET is a prime example of this, providing satellite-based estimates of evapotranspiration (ET) for the western United States. OpenET uses an ensemble of six different satellite-based models (including those used in the ADT) to calculate ET at the field scale.³³

The integration of OpenET data into the ADT provides a robust, transparent baseline for water accounting. By making this data publicly available and easily accessible via APIs, OpenET builds trust among stakeholders—farmers, water managers, and regulators—who can all refer to a "single source of truth" regarding water consumption. This is particularly vital in water-scarce basins like the Colorado River, where the ADT is used to simulate the impacts of water conservation strategies.³⁵

5. Computational Engines: AI, Emulators, and Cloud Infrastructure

The simulation of continental-scale agriculture at field-level resolution generates petabytes of data and requires massive computational resources. To make the Agricultural Digital Twin responsive and interactive, NASA has integrated cutting-edge developments in Artificial Intelligence (AI) and Machine Learning (ML).

5.1 Surrogate Modeling: Speeding Up the Physics

A significant bottleneck in traditional Earth system modeling is computational cost. Running a complex physics-based model like DSSAT for every field in the United States, under dozens of different climate scenarios, can take days or weeks on a supercomputer. This latency is unacceptable for real-time decision support.

To overcome this, researchers are developing "surrogate models" or emulators. These are machine learning models—often Deep Neural Networks (DNNs) or Random Forests—that are trained on the inputs and outputs of the physics-based models. Once trained, the emulator learns the mathematical relationship between the inputs (weather, soil, management) and the outputs (yield, biomass).³

The CROMES (CROp Model Emulator Suite) is a pioneering example of this approach. CROMES provides a flexible pipeline for training emulators on Global Gridded Crop Model (GGCM) simulations. Recent validations have shown that CROMES-trained emulators can replicate the predictions of complex process-based models (like EPIC-IIASA) with an R² of 0.97 to 0.98, while running orders of magnitude faster.³⁸ This speed allows the ADT to perform "uncertainty quantification" by running thousands of Monte Carlo simulations in seconds, generating probability distributions of crop yield rather than single point estimates. This capability is essential for risk assessment, allowing the system to tell a farmer, "There is a 10% chance of yield failure, but a 60% chance of average yield," rather than a deterministic prediction that might be wrong.⁴

5.2 Physics-Informed Machine Learning

Pure data-driven AI models can sometimes produce physically impossible results (e.g., negative soil moisture) when faced with data outside their training set. To mitigate this, the ADT employs Physics-Informed Machine Learning (PIML). In this approach, the loss functions of the neural networks are constrained by physical laws (such as the conservation of mass and energy).²

For example, researchers are using Bayesian Neural Networks (BNN) coupled with LIS to predict crop yields. These models integrate the statistical power of deep learning with the physical constraints of the hydrological model. The BNNs are trained on historical yield data from USDA NASS and environmental variables from LIS. The "Bayesian" nature of these networks allows them to quantify the uncertainty in their own predictions, providing a confidence interval that is crucial for insurance and policy applications.¹⁴

5.3 Infrastructure and the "Last Mile"

The computational backbone of the ADT is supported by NASA’s Advanced Information Systems Technology (AIST) program. AIST develops the middleware that handles data ingestion, storage, and processing in the cloud. This includes the development of "Analysis Ready Data" (ARD) formats that remove the need for users to perform complex pre-processing on satellite imagery.¹⁰

A key objective is solving the "last mile" problem: delivering these sophisticated insights to users who may not be scientists. The CropSmart Digital Twin, a project funded by the NSF and USDA in collaboration with George Mason University, exemplifies this effort. CropSmart wraps the complex modeling (using the Noah-MP-Crop model) in a user-friendly interface accessible via web portals and smartphone apps. It translates the digital twin's output into specific, actionable recommendations, such as optimized irrigation schedules or planting dates, tailored to the user's specific geolocation and management goals.⁴¹

6. Drought Monitoring and Risk Assessment

Drought is the primary climatic adversary of agriculture. In the context of the ADT, drought is not viewed as a static event but as a dynamic process with distinct phases—meteorological, agricultural, hydrological, and socioeconomic. The ADT provides a unified platform for monitoring the evolution of drought across these phases.

6.1 The Phenomenon of Flash Drought

A critical focus of recent years has been "flash drought"—the rapid onset of drought conditions driven by extreme heat, low humidity, and high winds, which can deplete soil moisture in a matter of weeks. Traditional drought monitoring indices, often based on monthly precipitation deficits, are too slow to capture these events before significant crop damage occurs.⁷

The ADT addresses this by monitoring the "evaporative demand" of the atmosphere alongside soil moisture. In late 2024, a severe flash drought expanded across the central United States, impacting nearly 80% of the population. The ADT tracked this event using a combination of SMAP soil moisture anomalies (indicating a lack of supply) and ECOSTRESS Evaporative Stress Index (ESI) data (indicating high demand and plant stress).⁸ The integration of these variables allows the digital twin to issue early warnings, identifying regions where the "thirst" of the atmosphere is outpacing the soil's ability to provide water, often weeks before the vegetation turns brown.

6.2 Risk Assessment and "What-If" Scenarios

Beyond monitoring, the ADT enables proactive risk assessment through scenario analysis. Stakeholders can use the digital twin to simulate the impact of potential future weather patterns.

• Counterfactual Analysis: Researchers can replay historical drought events (like the devastating 2012 drought) using current crop genetics and management practices to see if modern resilience measures would mitigate the damage. This helps quantify the value of adaptation strategies.⁴

• Seasonal Forecasting: By driving the crop models with seasonal climate forecasts (S2S), the ADT can produce yield outlooks months in advance. While these forecasts have inherent uncertainty, the use of ensemble modeling allows for the generation of probabilistic risk maps, highlighting areas with a high likelihood of crop failure.⁴³

This capability transforms drought management from a reactive crisis response to a proactive risk management process. Farmers can use these insights to make decisions about purchasing crop insurance, selling grain futures, or altering their crop mix.

7. Institutional Frameworks: Harvest and Acres

To ensure these technologies reach their intended users, NASA has established two primary consortia: NASA Harvest and NASA Acres.

7.1 NASA Harvest: A Global Perspective

NASA Harvest, led by the University of Maryland, operates with a global mandate to improve food security through Earth observation. It functions as a boundary organization, connecting the scientific capabilities of NASA with the operational needs of humanitarian agencies, ministries of agriculture, and international markets.⁴⁶

In 2024, Harvest demonstrated the power of remote sensing in crisis response. Monitoring the El Niño-induced drought in Southern Africa, Harvest utilized satellite data to predict maize crop failures months before harvest. These early warnings allowed governments and aid organizations to mobilize resources and plan imports to avert famine.⁴⁷ Additionally, Harvest has been pivotal in monitoring the impact of the conflict in Ukraine, using satellite imagery to map unharvested fields, quantify the impact of artillery craters on arable land, and estimate global grain supply disruptions.⁴⁹

7.2 NASA Acres: Domestic Precision

NASA Acres, established more recently, focuses on the United States. Its mission is to bridge the gap between space-based data and the American farmer. Acres works through a network of pilot projects and partnerships with land-grant universities, agricultural extension services, and private industry.⁵¹

Acres addresses the specific challenges of high-input, high-tech U.S. agriculture. Key themes include optimizing nitrogen management to reduce runoff, verifying carbon sequestration practices for emerging ecosystem service markets, and improving water use efficiency in the water-limited West. Acres also prioritizes the "trust infrastructure," working to ensure that data sharing arrangements respect the privacy and commercial interests of farmers, a critical barrier to the adoption of digital technologies.⁵²

8. Applied Case Studies in Sustainable Agriculture

The theoretical power of the Agricultural Digital Twin is best illustrated through specific case studies from the 2024-2025 period.

8.1 Corn Yield Prediction in Iowa

Iowa, the heart of the U.S. Corn Belt, serves as a primary testbed for the ADT. In 2024, researchers utilized the LIS-DSSAT coupled system to simulate corn yields under varying climatic conditions. By assimilating high-resolution SMAP soil moisture data, the digital twin was able to capture the subtle onset of water stress during the critical silking stage of corn development, a period when the plant is most sensitive to moisture deficits.⁴

The results showed that the assimilation of satellite data significantly reduced the error in yield predictions compared to open-loop model runs (simulations without satellite correction). Furthermore, the digital twin allowed for sub-field level analysis, identifying specific zones within fields that were underperforming due to soil compaction or poor drainage. This level of detail empowers farmers to practice precision agriculture, applying inputs only where they are needed.⁵⁵

8.2 Soybean Management in Kansas

In Kansas, situated on the drier western edge of the Corn Belt, the focus has been on optimizing management to mitigate drought risk. NASA Acres pilots have used the ADT to analyze the interactions between planting dates, cultivar maturity groups, and seasonal weather patterns.⁴

The digital twin simulated thousands of "virtual seasons" to determine optimal planting windows. The results indicated that for certain soybean varieties, shifting the planting date could allow the crop to avoid the peak heat stress of late summer during its reproductive phase. These "in-silico" experiments provide farmers with evidence-based recommendations that would be too risky and expensive to test in the real world.⁵⁶

8.3 Water Conservation in the Colorado River Basin

In the arid Western U.S., the integration of OpenET data with the ADT is supporting critical water conservation efforts. As states negotiate reductions in water usage from the Colorado River, accurate measurement of agricultural water consumption is paramount.

The ADT provides field-scale estimates of evapotranspiration, allowing water managers to audit water use and verify compliance with conservation programs. It also enables the simulation of "fallowing scenarios"—estimating how much water could be saved if specific fields were taken out of production for a season. This data-driven approach is essential for designing equitable and effective water policy in a drying climate.³⁴

9. Future Horizons and Challenges

As the Agricultural Digital Twin matures, several challenges and opportunities lie ahead.

Interoperability and Federation: The vision is for a "system of systems" where the Agricultural Digital Twin interacts seamlessly with Digital Twins of the atmosphere, the ocean, and urban systems. Achieving this requires rigorous standards for data interoperability and model coupling, a task being spearheaded by the AIST program.¹

The Data Deluge: The launch of next-generation missions like Landsat Next, SBG (Surface Biology and Geology), and NISAR will generate data at a scale previously unimagined. The ADT infrastructure must scale to ingest, process, and assimilate this flood of information without latency.

Global Scaling: While the ADT works well in data-rich environments like the U.S., scaling it to smallholder systems in Africa or Asia, where ground truth data is scarce, remains a challenge. Innovations in transfer learning and the use of crowd-sourced data are being explored to bridge this gap.⁵⁷

Adoption and Trust: Ultimately, the success of the ADT depends on its adoption by the agricultural community. This requires not just better algorithms, but better user interfaces and stronger guarantees of data privacy. Farmers must trust that the "Digital Replica" of their farm is working for them, not just observing them.⁵³

10. Conclusion

The Agricultural Digital Twin represents a paradigm shift in our relationship with the land. It moves agriculture from a reactive industry, forever responding to the caprices of weather, to a proactive one, capable of anticipating risks and optimizing resilience through simulation. The integration of NASA’s space-based sensor web with advanced biophysical modeling and artificial intelligence has created a tool of unprecedented predictive power.

From the soil moisture radiometers of SMAP to the thermal sensors of ECOSTRESS, and the computational logic of the LIS-DSSAT coupler, the components of this system are now operational and delivering value. The pilot projects in Iowa, Kansas, and globally have demonstrated that these tools can accurately predict yields, provide early warnings of flash drought, and support critical management decisions.

As climate change continues to rewrite the rules of agriculture, the "digital replica" provided by the ADT offers a safe harbor where farmers, policymakers, and scientists can test their strategies against the future. It is a necessary adaptation for a sustainable civilization, ensuring that even as the environment becomes more uncertain, our food systems remain secure.

Selected Data Tables

Table 1: Key NASA Satellites and Instruments Supporting the Agricultural Digital Twin

Satellite / Instrument	Primary Measurement	Agricultural Application	Revisit Time	Key Advantage
SMAP (Soil Moisture Active Passive)	Soil Moisture (top 5cm)	Drought monitoring, planting conditions	2-3 Days	Global coverage, L-band microwave penetrates clouds
ECOSTRESS (ISS Instrument)	Land Surface Temperature (LST)	Evapotranspiration (ET), Water stress detection	Variable (Diurnal)	Detects plant stress before visible signs; diurnal sampling
Landsat 8 / 9	Optical / Thermal Imagery	Crop health (NDVI), Crop type mapping	8 Days (combined)	High spatial resolution (30m), long historical record
GPM (Global Precipitation Measurement)	Precipitation	Rainfall inputs for hydrologic models	< 3 Hours	Accurate global rainfall data, essential for crop models
Landsat Next (Future - Launch ~2030)	Superspectral Imagery (26 bands)	Residue cover, Snow/Water quality, Nutrient content	6 Days (triplet)	Improved spectral and temporal resolution for precision ag
GRACE-FO	Terrestrial Water Storage	Groundwater depletion monitoring	Monthly	Tracks deep aquifer trends critical for irrigation

Table 2: Components of the NASA Agricultural Digital Twin Architecture

Component	Function	Description
LIS (Land Information System)	Hydrology Engine	Models land surface states (water, energy) by assimilating satellite data. Acts as the boundary condition provider.
DSSAT (Decision Support System for Agrotechnology Transfer)	Crop Engine	Simulates crop growth, development, and yield based on genetics, management, and environment.
Data Assimilation (DA)	Correction Mechanism	Uses algorithms like Ensemble Kalman Filter (EnKF) to update model states with real-time satellite observations.
CROMES (CROp Model Emulator Suite)	AI/ML Accelerator	A machine learning emulator that mimics physics-based models to enable rapid, large-scale simulations.
CropSmart (User Interface)	Decision Support	NSF/USDA funded interface delivering actionable insights (e.g., irrigation advice) directly to farmers via apps.

Table 3: Comparison of Drought Monitoring Indices

Index	Primary Input	Response Speed	Application in ADT
PDSI (Palmer Drought Severity Index)	Precipitation, Temperature	Slow (Weeks/Months)	Long-term climatological context
NDVI (Normalized Difference Vegetation Index)	Vegetation Greenness	Medium (Lagged)	Assessing visible crop damage
ESI (Evaporative Stress Index)	Land Surface Temperature (ET)	Fast (Days)	Flash drought early warning; detects stress before browning
Crop-CASMA	Soil Moisture (SMAP)	Fast (Days)	Assessing root-zone water availability for planting/growth

Citations used in this report: ¹

Works cited

1. NASA Earth System Digital Twins (ESDT) Prototypes, accessed January 17, 2026, https://destination-earth.eu/wp-content/uploads/2024/05/NASA-Earth-System-Digital-Twins-ESDT-Prototypes.pdf

2. NASA Earth System Digital Twins (ESDT) For a Sustainable Future, accessed January 17, 2026, https://ntrs.nasa.gov/api/citations/20250000171/downloads/2025-01_AMS25_NASA-ESDT_LeMoigne.pdf

3. Earth Systems Digital Twins (ESDT) - NASA Technical Reports Server, accessed January 17, 2026, https://ntrs.nasa.gov/api/citations/20230010074/downloads/2023-06-15_Keynote-STC-DT-Webinars_Final.pdf

4. NASA Agricultural Digital Twin will help farmers improve crop yield forecasts, accessed January 17, 2026, https://esto.nasa.gov/nasa-agricultural-digital-twin-will-help-farmers-improve-crop-yield-forecasts/

5. Genetics and Sustainable Agriculture Research: Mississippi State, MS - Project : USDA ARS, accessed January 17, 2026, https://www.ars.usda.gov/research/project/?accnNo=445679

6. Digital Twins - eoPortal, accessed January 17, 2026, https://www.eoportal.org/other-space-activities/digital-twins

7. The SMCE-Built NASA Earth Information System (EIS) Cloud Computing Platform Supports the Analysis of Flash Droughts, accessed January 17, 2026, https://www.nccs.nasa.gov/news-events/nccs-highlights/Flash-Drought

8. Drought Expands Across the U.S. - NASA Science, accessed January 17, 2026, https://science.nasa.gov/earth/earth-observatory/drought-expands-across-the-us-153526/

9. Engineers using 'digital twins' to improve agriculture, health, manufacturing and more, accessed January 17, 2026, https://trac-ai.iastate.edu/2025/04/09/engineers-using-digital-twins-to-improve-agriculture-health-manufacturing-and-more/uncategorized/

10. Earth System Digital Twins - HPC User Forum, accessed January 17, 2026, https://www.hpcuserforum.com/wp-content/uploads/2024/04/Jacqueline-Le-Moigne_NASA_Earth-System-Digital-Twins.pdf

11. Earth System Digital Twins - NASA Earth Science and Technology Office, accessed January 17, 2026, https://esto.nasa.gov/earth-system-digital-twin/

12. Drought and Soil Moisture Data - Drought.gov, accessed January 17, 2026, https://www.drought.gov/topics/soil-moisture

13. LIS Framework - NASA, accessed January 17, 2026, https://lis.gsfc.nasa.gov/

14. Project Selections for AIST-21 - NASA Earth Science and Technology Office, accessed January 17, 2026, https://esto.nasa.gov/project-selections-for-aist-21/

15. Advanced Information Systems Technology (AIST) 2023 Annual Reviews Earth System Digital Twins (ESDT) - NASA Technical Reports Server, accessed January 17, 2026, https://ntrs.nasa.gov/api/citations/20230013564/downloads/ESDT-2023GroupedReviews.pdf

16. Land Information System | LIS Framework - NASA, accessed January 17, 2026, https://lis.gsfc.nasa.gov/software/lis

17. Land Data Assimilation System | LDAS - NASA, accessed January 17, 2026, https://ldas.gsfc.nasa.gov/

18. NASA Soil Moisture Data Advances Global Crop Forecasts, accessed January 17, 2026, https://science.nasa.gov/earth/earth-observatory/nasa-soil-moisture-data-advances-global-crop-forecasts-92228/

19. A 1 km soil moisture dataset over eastern CONUS generated by assimilating SMAP data into the Noah-MP land surface model - ESSD Copernicus, accessed January 17, 2026, https://essd.copernicus.org/articles/17/4587/2025/

20. The Future of Soil Moisture Sensing - Texas Water Resources Institute, accessed January 17, 2026, http://twri.tamu.edu/blog/2024/08/10/the-future-of-soil-moisture-sensing/

21. High Spatial Resolution Soil Moisture Mapping over Agricultural Field Integrating SMAP, IMERG, and Sentinel-1 Data in Machine Learning Models - MDPI, accessed January 17, 2026, https://www.mdpi.com/2072-4292/17/13/2129

22. Sensing soil moisture for better crops | National Snow and Ice Data Center, accessed January 17, 2026, https://nsidc.org/news-analyses/news-stories/sensing-soil-moisture-better-crops

23. Tracking Deluge and Drought through Soil Moisture - NASA Science, accessed January 17, 2026, https://science.nasa.gov/earth/earth-observatory/tracking-deluge-and-drought-through-soil-moisture-150138/

24. Research — ECOSTRESS - NASA, accessed January 17, 2026, https://ecostress.jpl.nasa.gov/science

25. At Which Overpass Time Do ECOSTRESS Observations Best Align with Crop Health and Water Rights? - MDPI, accessed January 17, 2026, https://www.mdpi.com/2072-4292/16/17/3174

26. Agricultural monitoring from the International Space Station: linking ECOSTRESS at different overpass times to actual field conditions, accessed January 17, 2026, https://agu.confex.com/agu/agu24/meetingapp.cgi/Paper/1615341

27. Development of a Global Evaporative Stress Index Based on - Western Water Applications Office (WWAO), accessed January 17, 2026, https://wwao.jpl.nasa.gov/documents/35/Factsheet-NASA-Drought-Early-Warning.pdf

28. Landsat Next Satellites offer improved data to Earth science - USGS Publications Warehouse, accessed January 17, 2026, https://pubs.usgs.gov/fs/2024/3005/fs20243005.pdf

29. Beyond the Visible: Landsat Next's New Spectral Bands - NASA Science, accessed January 17, 2026, https://science.nasa.gov/missions/landsat/beyond-the-visible-landsat-nexts-new-spectral-bands/

30. Landsat Next Update - Western States Water Council, accessed January 17, 2026, https://westernstateswater.org/wp-content/uploads/2023/03/Landsat-Next-Update-to-WSWC-Fall-2023-Meeting-USGS_Stryker_Final.pdf

31. Landsat Next - eoPortal, accessed January 17, 2026, https://www.eoportal.org/satellite-missions/landsat-next

32. Landsat Next Defined - NASA Science, accessed January 17, 2026, https://science.nasa.gov/missions/landsat/ls-landsat-next-defined/

33. FARMing with Data: OpenET Launches New Tool for Farmers and Ranchers - DRI, accessed January 17, 2026, https://www.dri.edu/farming-with-data-openet-launches-new-tool/

34. OpenET Expands Satellite-based Water Data Across 48 States, accessed January 17, 2026, https://etdata.org/openet-expands-satellite-based-water-use-data-48-states-farmers-water-managers-across-nation/

35. OpenET: Open-Source Transparent Water Management Data, accessed January 17, 2026, https://etdata.org/

36. FARMing with Data: OpenET Launches new Tool for Farmers and Ranchers, accessed January 17, 2026, https://science.nasa.gov/missions/landsat/farming-with-data-openet-launches-new-tool-for-farmers-and-ranchers/

37. Digital Twins of the Earth System via Hybrid Physics-AI Models - SIAM.org, accessed January 17, 2026, https://www.siam.org/publications/siam-news/articles/digital-twins-of-the-earth-system-via-hybrid-physics-ai-models/

38. CROMES v1.0: a flexible CROp Model Emulator Suite for climate impact assessment - GMD, accessed January 17, 2026, https://gmd.copernicus.org/articles/18/5759/2025/gmd-18-5759-2025.pdf

39. (PDF) CROMES v1.0: a flexible CROp Model Emulator Suite for climate impact assessment, accessed January 17, 2026, https://www.researchgate.net/publication/395350711_CROMES_v10_a_flexible_CROp_Model_Emulator_Suite_for_climate_impact_assessment

40. DSFAS-AI: Developing an Integrated Deep Learning Model Framework for County-Level Crop Yield Prediction in support of USDA NASS Operation - UNIV OF WISCONSIN, accessed January 17, 2026, https://portal.nifa.usda.gov/web/crisprojectpages/1028199-dsfas-ai-developing-an-integrated-deep-learning-model-framework-for-county-level-crop-yield-prediction-in-support-of-usda-nass-operation.html

41. The CropSmart Digital Twin | George Mason University, accessed January 17, 2026, https://www.gmu.edu/news/2024-12/cropsmart-digital-twin

42. CropSmart - a digital twin for making wiser cropping decisions nationwide, accessed January 17, 2026, https://nsf.elsevierpure.com/en/projects/cropsmart-a-digital-twin-for-making-wiser-cropping-decisions-nati/

43. Drought Monitoring, Prediction, and Projection using NASA Earth System Data, accessed January 17, 2026, https://www.earthdata.nasa.gov/learn/trainings/drought-monitoring-prediction-projection-using-nasa-earth-system-data

44. New Warning System Identifies Flash Drought Quickly | U.S. Geological Survey - USGS.gov, accessed January 17, 2026, https://www.usgs.gov/news/technical-announcement/new-warning-system-identifies-flash-drought-quickly

45. Machine Learning Emulators and Empirical Models Combining Climate and Global Crop Models for Seasonal Agricultural Production - NASA Technical Reports Server (NTRS), accessed January 17, 2026, https://ntrs.nasa.gov/citations/20230012472

46. NASA Harvest, accessed January 17, 2026, https://www.nasaharvest.org/

47. Using El Niño Forecasts to Give Early Warning of Crop Failures - NASA Harvest, accessed January 17, 2026, https://www.nasaharvest.org/news/using-el-nino-forecasts-give-early-warning-crop-failures

48. GEOGLAM Crop Monitor for Early Warning Releases 100th Issue - NASA Harvest, accessed January 17, 2026, https://www.nasaharvest.org/news/geoglam-crop-monitor-for-early-warning-releases-100th-issue

49. NASA Harvest to Tackle Food Security as Part of Google.org AI Collaborative, accessed January 17, 2026, https://www.nasaharvest.org/news/nasa-harvest-to-tackle-food-security-as-part-of-googleorg-ai-collaborative

50. NASA Harvest at AGU 2024, accessed January 17, 2026, https://www.nasaharvest.org/news/nasa-harvest-at-agu-2024

51. NASA Acres, accessed January 17, 2026, https://www.nasaacres.org/

52. Our Projects - NASA Acres, accessed January 17, 2026, https://www.nasaacres.org/our-projects

53. Getting Farmers the Most From Their Data: Ag Data Transparent and Farm Journal's Trust In Food Join Forces with NASA Acres to Move the Needle on Farm Research, accessed January 17, 2026, https://www.nasaacres.org/news/agdatatransparent

54. Iowa Cropland 2001-2020 - NASA Scientific Visualization Studio, accessed January 17, 2026, https://svs.gsfc.nasa.gov/4968/

55. Improving Crop Yield Predictions with Satellite Assist - NASA Science, accessed January 17, 2026, https://science.nasa.gov/missions/landsat/improving-crop-yield-predictions-with-satellite-assist/

56. Digital Twin Models for Crop Yield Prediction | Anvil Labs, accessed January 17, 2026, https://anvil.so/post/digital-twin-models-for-crop-yield-prediction/

57. From Orbit to Farmland: How Space Technology—and NASA Harvest—Are Reshaping Global Agriculture, accessed January 17, 2026, https://www.nasaharvest.org/news/from-orbit-to-farmland-how-space-technologyand-nasa-harvestare-reshaping-global-agriculture

58. 2025 Entries Make History in Kansas Soybean Yield and Value Contests, accessed January 17, 2026, https://kansassoybeans.org/press_release/2025-entries-make-history-in-kansas-soybean-yield-and-value-contests/

59. Earth Science in Action, accessed January 17, 2026, https://science.nasa.gov/earth/in-action%20/

60. Evaporative Stress Index (ESI) - Drought.gov, accessed January 17, 2026, https://www.drought.gov/data-maps-tools/evaporative-stress-index-esi

61. ECOSTRESS Gridded Evaporative Stress Index disALEXI 24-Hour L4 CONUS 70 m V002, accessed January 17, 2026, https://www.earthdata.nasa.gov/data/catalog/lpcloud-eco-l4g-esi-alexi-002

62. MF3154 Kansas Soybean Management 2025 - K-State Extension Midway District, accessed January 17, 2026, https://www.midway.k-state.edu/crops/docs/kansas-soybean-management-2025_MF3154.pdf

63. ECOSTRESS Tiled Evaporative Stress Index disALEXI 24-Hour L4 CONUS 70 m V002 | NASA Earthdata, accessed January 17, 2026, https://www.earthdata.nasa.gov/data/catalog/lpcloud-eco-l4t-esi-alexi-002

64. Crop/Plant Yields - NASA SVS | Search, accessed January 17, 2026, https://svs.gsfc.nasa.gov/search/?keywords=Crop/Plant%20Yields

65. Reaping the benefits of a digital twin | NCAR & UCAR News, accessed January 17, 2026, https://news.ucar.edu/132980/reaping-benefits-digital-twin

66. George Mason University - CropSmart Digital Twin - YouTube, accessed January 17, 2026, https://www.youtube.com/watch?v=IqtVmHO8vD8