Remote sensing data assimilation framework for improved crop yield estimation
A data assimilation framework integrating remotely sensed soil moisture and crop development data into the APSIM-Wheat crop model to improve yield prediction accuracy. Initial validation showed up to 30% improvement in yield estimation, with LAI assimilation reducing relative error from 38.3% to 7.6%.
Overview
Crop models forecast crop development and yield by simulating interactions among plants, atmosphere, and soil, but they depend on parameter data that are often costly and laborious to collect through traditional methods. This solution offers a data assimilation framework that integrates remotely sensed soil moisture and crop development observations into the APSIM-Wheat model to reduce prediction uncertainties and deliver more accurate yield estimates. By merging satellite-derived observations with process-based crop modelling, the framework addresses a critical gap in agricultural decision-making, enabling farmers, agribusinesses, and policymakers to make better-informed planting, irrigation, and harvest decisions.
Technical specifications
Key features:
- Data assimilation framework designed for the APSIM-Wheat crop prediction model
- Integration of remotely sensed soil moisture and crop development data, including Leaf Area Index (LAI) time series
- Assimilation of LAI observations at three-day intervals to refine model state estimates
- Tested using ground-based datasets collected in Victoria, Australia
- Demonstrated yield estimation improvement of up to 30% relative to uncalibrated model runs
- LAI assimilation reduced relative yield estimation error from 38.3% to 7.6% in validation experiments
- Designed to improve performance even when model parameters are not well-calibrated
Technology readiness level
The framework has been validated through controlled experiments comparing a reference APSIM-Wheat model run (meteorological data only, no assimilation) against a data assimilation experiment incorporating LAI time series. Results confirm significant improvements in yield estimation accuracy. The next phase of validation will extend testing to actual remotely sensed satellite data across a range of typical scenarios, assessing whether the approach can rescue poorly calibrated models and broaden applicability to real-world, large-scale agricultural monitoring contexts.
About Monash University, Melbourne
Monash University is a comprehensive public research university and one of Australia’s largest, known for scale, interdisciplinarity, and an applied orientation. Its Melbourne-based technology precinct brings together university laboratories, pilot-scale and prototyping suites, and company R&D groups alongside government research organizations to enable co-development and rapid iteration. Integration with a major hospital network supports clinical trials and translation, while structured industry placements and doctoral partnerships create a robust talent pipeline for corporate R&D. Research is backed by competitive funding from the Australian Research Council, the National Health and Medical Research Council, and state and federal programs that incentivize industry collaboration. A dedicated technology transfer office manages IP, licensing, and startup formation, with pathways to incubation and investment within the precinct.
delivered to your inbox.