Integrating Remote Sensing and Mechanistic Modeling for Crop Monitoring and Yield Estimation
Georgios Ntakos is a PhD student in the department of Water Resources. (Co)Promotors are dr.ir. C. van der Tol from the faculty of Geo-Information Science and Earth Observation, University of Twente.
Agricultural crop production is essential for global food security, yet the ability to accurately monitor and estimate yields remains a persistent challenge. Traditional methods often depend on empirical relationships, which, while offering valuable insights, fail to fully capture the complex and dynamic interactions between genetics, environment, and management that drive crop growth and development. However, with the growing availability of advanced remote and proximal sensing technologies, we now have an opportunity to move beyond empirical techniques and build methodologies based on mechanistic and physical models. This shift may allow the use of detailed knowledge about crop processes, resulting in more reliable and explanatory insights into agricultural productivity. Such developments are important for optimizing resource use, improving crop management strategies, and addressing global challenges caused by climate change and increasing global food demand.
This research seeks to help address these issues by fusing remote sensing data with mechanistic and physical modeling frameworks to provide a more robust and explanatory approach to yield estimation. As part of this effort, a starting point involves combining retrieved biophysical variables with Gross Primary Production (GPP) fluxes simulated using the Soil Canopy Observation of Photochemistry and Energy fluxes (SCOPE) model, along with a Harvest Index. This step translates photosynthetic and energy flux processes into yield predictions, enabling a baseline assessment of crop productivity. By moving beyond empirical approximations, such as fixed Harvest Index values, the research aims to demonstrate the value of integrating physiological processes with data-driven models to gain a clearer understanding of yield variability across a range of conditions.
Building on this foundation, the study introduces a coupling of SCOPE’s Radiative Transfer Model (RTMo) with the WOFOST Crop Growth Model (CGM). This integration connects reflectance-based biophysical retrievals and yield simulations by embedding mechanistic knowledge directly into the Radiative Transfer modeling framework. The result is a reduction in dependence on empirical factors, such as the Harvest Index and an increase in the explanatory power of the models, by directly linking reflectance to plant traits and yield through mechanistic physiological knowledge.
Furthermore, this research explores the potential of incorporating fluorescence sensor measurements into mechanistic models. Fluorescence serves as an indicator of photosynthetic efficiency and has proven to be a valuable proxy for assessing plant health and biomass accumulation. By integrating fluorescence data, the study further reduces the reliance on empirically derived parameters and establishes a direct connection between plant function and biomass production. This approach not only improves the accuracy of crop modeling but also provides insights into dynamic processes such as the photosynthesis.
Finally, the research evaluates the different techniques for simulating yields at a regional scale, comparing and investigating empirical, mechanistic, and physical modeling techniques. This assessment highlights the strengths and limitations of each approach and discusses how their combination can lead to more reliable predictions across various regions and growing seasons. By using these techniques, the study underscores the scalability and adaptability of the proposed methodologies for real-world agricultural monitoring.
In conclusion, the integration of remote sensing and observation with empirical, mechanistic, and physical modeling at different scales, has the potential to benefit academic research, through improved modeling approaches, but also provides tools, insights and knowledge that can help farmers and policymakers make better decisions. In the future, Artificial Intelligence (AI) could support these approaches by enabling more efficient data processing and boosting the accuracy of yield predictions, while also creating new opportunities to investigate and understand crop systems and their dynamics.
