Abebe et al. (2026) Assimilating leaf area index and soil moisture into the WOFOST model for improved maize (Zea mays L.) yield estimation in Ethiopia

Identification

• Journal: International Journal of Applied Earth Observation and Geoinformation
• Year: 2026
• Date: 2026-03-25
• Authors: Gebeyehu Abebe, Odunayo David Adeniyi, Amazirh Abdelhakim, Tesfaye Balemi, Tsegaye Tadesse, Berhan Gessesse, Beatrice Asenso Barnieh, Zoltan Szantoi
• DOI: 10.1016/j.jag.2026.105257

Research Groups

• African Research Fellow, Directorate of Earth Observation Programmes, European Space Agency (ESA)/ESRIN, Frascati, Italy
• Department of Natural Resources Management, Debre Berhan University, Debre Berhan, Ethiopia
• Centre for Remote Sensing Applications (CRSA), Mohammed VI Polytechnic University (UM6P), Ben Guerir, Morocco
• FAO-Ethiopia, Semera Field Coordination Office, Semera, Ethiopia
• National Drought Mitigation Centre, University of Nebraska-Lincoln, Lincoln, NE, USA
• Department of Remote Sensing, Space Science and Geospatial Institute, Addis Ababa, Ethiopia
• Department of Geography and Environmental Studies, Koteb University of Education, Addis Ababa, Ethiopia
• Department of Geography and Environmental Studies, Stellenbosch University, Stellenbosch, South Africa

Short Summary

This study developed a data assimilation framework using the Ensemble Kalman Filter to jointly assimilate satellite-derived Leaf Area Index (LAI) and Soil Moisture (SM) into the WOFOST crop model, significantly improving maize yield estimation accuracy in Ethiopia compared to univariate assimilation or open-loop simulations.

Objective

• To estimate maize (Zea mays L.) yield and improve its accuracy in Ethiopia by jointly assimilating satellite-derived Leaf Area Index (LAI) and Soil Moisture (SM) from Earth Observation (EO) data into the WOFOST crop growth model.

Study Configuration

• Spatial Scale: Main maize-growing administrative zones in Ethiopia, covering an area of 410,149 km², across twenty-eight administrative zones in seven regional states.
• Temporal Scale: Main cropping seasons of 2015/2016 and 2016/2017 for model calibration, and the 2021/2022 growing season for validation and assimilation. Maize is typically grown from May to September.

Methodology and Data

• Models used:
  • WOrld FOod STudies (WOFOST) crop growth model (water-limited mode).
  • Ensemble Kalman Filter (EnKF) algorithm for data assimilation.
  • Sobol method for sensitivity analysis.
  • SUBPLEX optimization algorithm for parameter calibration.
• Data sources:
  • Earth Observation (EO) Data:
    • Leaf Area Index (LAI): MODIS product (MOD15A2H.006), smoothed using Savitzky-Golay filtering.
    • Soil Moisture (SM): Soil Moisture Active Passive (SMAP) L3SMP Level 3 product (top 5 cm soil, 9 km spatial resolution).
    • Crop Mask: World Cereal Crop Maps.
  • Ground Observation Data:
    • Maize yield and Above Ground Biomass (AGB): Taking Maize Agronomy to Scale in Africa (TAMASA) project (2015/2016 and 2016/2017 seasons).
    • Official yield data: Agricultural Sample Survey of 2021/2022, Central Statistical Agency (CSA) of Ethiopia (administrative zone level).
  • Ancillary Data:
    • Crop management data (sowing dates, fertilizer rates, harvesting): TAMASA field observation data.
    • Soil data (field capacity, saturated soil, wilting point): International Soil Reference and Information Centre-World Soil Information (ISRIC-WISE) databases.
    • Daily weather data: NASA Langley Research Center (LaRC) Prediction of Worldwide Energy Resource (POWER) Project.

Main Results

• Joint assimilation of LAI and SM significantly improved maize yield estimation accuracy and reduced discrepancies between simulated and observed LAI and SM values.
• Performance metrics for maize yield estimation (2021/2022 growing season):
  • Joint DA (LAI & SM): R² = 0.69, RMSE = 0.048 kg/m², NRMSE = 13.22%.
  • DA with LAI only: R² = 0.63, RMSE = 0.062 kg/m², NRMSE = 16.44%.
  • DA with SM only: R² = 0.54, RMSE = 0.070 kg/m², NRMSE = 18.21%.
  • Open-loop (original WOFOST): R² = 0.31, RMSE = 0.072 kg/m², NRMSE = 19.5%.
• The original WOFOST model (open-loop) tended to overestimate LAI and AGB, and underestimate SM in the root zone under water-limited conditions.
• Assimilating SM alone showed less significant improvement than LAI assimilation, partly attributed to the spatial resolution mismatch between MODIS LAI (500 m) and SMAP SM (9 km).
• All three assimilation strategies (LAI, SM, joint LAI & SM) produced maize yield maps with improved spatial variability compared to the open-loop simulation.

Contributions

• Developed and implemented an innovative multivariate data assimilation (DA) approach that jointly integrates satellite-derived Leaf Area Index (LAI) and Soil Moisture (SM) into the WOFOST crop growth model using the Ensemble Kalman Filter (EnKF) for maize yield estimation in Ethiopia.
• Demonstrated that this multivariate DA significantly improves maize yield estimation accuracy and captures spatial variability better than univariate DA or open-loop simulations, which is crucial for rainfed agriculture and food security in Ethiopia.
• Provided valuable insights for improving agricultural productivity and decision-making processes in a region vital for national food security.
• Utilized the EO AFRICA Innovation Lab cloud computing platform for implementing the WOFOST-EnKF DA framework, showcasing state-of-the-art cloud computing for geospatial big data processing and analysis.

Funding

• Financial support from the European Space Agency (ESA) through the African Research Fellowship (ARF) Initiative.

Citation

@article{Abebe2026Assimilating,
  author = {Abebe, Gebeyehu and Adeniyi, Odunayo David and Abdelhakim, Amazirh and Balemi, Tesfaye and Tadesse, Tsegaye and Gessesse, Berhan and Barnieh, Beatrice Asenso and Szantoi, Zoltan},
  title = {Assimilating leaf area index and soil moisture into the WOFOST model for improved maize (Zea mays L.) yield estimation in Ethiopia},
  journal = {International Journal of Applied Earth Observation and Geoinformation},
  year = {2026},
  doi = {10.1016/j.jag.2026.105257},
  url = {https://doi.org/10.1016/j.jag.2026.105257}
}