Accurate crop modeling at the field-level is important for yield prediction and agricultural risk mitigation, but is often hindered by the lack of information on field management as well as crop phenology of different cultivars. This study aims to develop a data assimilation framework for field-level crop modeling without management or crop phenology information for potential remote sensing applications. To do this, we first present a Monte Carlo simulation-based approach to estimating planting date and quasi-calibrated phenological parameters. Second, a simplified fertility stress scheme is developed for the AquaCrop model. The aim here is not necessarily to improve the AquaCrop model but to facilitate ensemble simulation when the field-level fertility stress condition is unknown. Finally, in situ soil moisture, canopy cover and biomass measurements are assimilated into the model to estimate crop yield, with the potential for yield prediction also explored. The experiments were performed for a rainfed maize field over 9 growing seasons, with each using a different maize cultivar. Results suggest that the planting dates can be accurately estimated (RMSE = 7.1 days, MAE = 5.4 days), and that the simplified fertility stress scheme adequately approximates the biomass and yield estimates from the original AquaCrop model under different fertility stress conditions. Data assimilation improves yield estimation, with an RMSE of 0.97 Mg/ha compared to 2.14 Mg/ha from the no-assimilation case. Yield prediction experiments reveal that the method is able to predict yield within 15% of the observed values up to 3 months before harvest. The proposed methodology does not rely on field-based information (e.g., planting date, plant density, crop phenology, fertility condition), and illustrates the potential for field-level crop modeling and yield forecasting using remote sensing data.
Bibliographical noteKAUST Repository Item: Exported on 2022-03-17
Acknowledged KAUST grant number(s): OSR-2017-CRG6
Acknowledgements: The work undertaken herein was partly funded through the ’A new paradigm in precision agriculture: assimilation of ultra-fine resolution data into a crop-yield forecasting model’ project, supported by the King Abdullah University of Science and Technology, grant number OSR-2017-CRG6, and through the ’Building REsearch Capacity for sustainable water and food security In drylands of sub-saharan Africa (BRECcIA)’ project, which is supported by UK Research and Innovation as part of the Global Challenges Research Fund, grant number NE/P021093/1, and by the National Natural Science Foundation of China (Grant Nos. 52179029, 51879289) and the Guangdong Basic and Applied Basic Research Foundation (2019B1515120052). The authors thank David Scoby from University of Nebraska-Lincoln for providing data.