Field Crop 2025, Vol.8, No.6, 293-300 http://cropscipublisher.com/index.php/fc 296 4.2 Calibration and validation using time-series remote observations No matter how advanced a model is, it still needs to be calibrated and verified to be reliable. Researchers usually fine-tune model parameters using continuous remote sensing time series, such as LAI, NDVI or soil moisture, and then compare the results with the real observed growth process and yield to see if they match (Zhuo et al., 2023). Especially during the critical periods such as jointing, panicle formation or grain filling, integrating the vegetation index into the model in stages yields more obvious results (Bouras et al., 2023). Nowadays, many teams combine remote sensing and ground measurement data to create grid-based or multi-model integrated calibration systems. This approach can significantly enhance spatial resolution and make predictions closer to reality. 4.3 Benefits: enhanced prediction accuracy, dynamic model adjustment Combining remote sensing with models has quite a few benefits. The most direct one is that the prediction is more accurate. Relevant studies show that the R² of wheat yield has increased and the RMSE has decreased (Zare et al., 2022). Meanwhile, the model can "update its status" at any time, promptly reflecting changes in crops or the environment, and is more sensitive to sudden situations such as drought and diseases. Moreover, this system is scalable and updated quickly. It can be of great use whether providing field advice to farmers or serving policy-making and food security assessment. 5 Case Study: Real-Time Wheat Yield Forecasting in the Indo-Gangetic Plains 5.1 Background: wheat production significance, climatic variability On the vast Ganges Plain of India, wheat is almost the lifeblood of farmers. It is not only one of the main production areas in the country but also closely related to the food supply of hundreds of millions of people. But this seemingly fertile land is not always stable. Frequent climate fluctuations, early arrival and early departure during the rainy season, droughts, heat waves, and sudden cold waves may all disrupt the planting rhythm (Zhao et al., 2020; Qader et al., 2023). Farmers often find it hard to judge the quality of their harvests, and the government is also prone to being "at a loss" when it comes to resource allocation and market regulation. Therefore, if the output can be predicted in advance, it can not only help farmers avoid risks, but also enable the policy level to be better prepared. 5.2 Methods: using Sentinel-2 data and WOFOST model with real-time assimilation To make the predictions more realistic, researchers attempted to combine the data from the Sentinel-2 satellite with the WOFOST crop model. Sentinel-2 takes fine shots and updates quickly, accurately capturing key parameters such as leaf area index (LAI) and soil moisture. Through methods such as Ensemble Kalman filtering (EnKF), these data were continuously input into the model, enabling it to be "corrected" in real time according to the field conditions throughout the growing season (Figure 2) (Wu et al., 2020). It is worth mentioning that the research also used Sentinel-1 radar data in combination, which can still obtain clear information in cloudy weather. This is particularly crucial for the Indian monsoon region. 5.3 Results: improved yield prediction accuracy and early warning benefits for farmers Practice has proved that the effect of this data assimilation method is quite remarkable. After integrating the LAI and soil moisture data of Sentinel-2 into the WOFOST model simultaneously, the prediction accuracy has been significantly improved. Research shows that when R² increases from 0.41 to 0.76, RMSE drops significantly, with the average relative error being as low as 3.17%. Compared with the traditional open-loop model, this joint assimilation can detect the trend of yield changes earlier, providing farmers with the opportunity to intervene in advance and allowing the government to be more composed when dealing with possible crop failures. Ultimately, the predictions not only became more accurate but also more "effective". 6 Benefits of the Integrated Approach 6.1 High spatial-temporal resolution for field-level management decisions Nowadays, farmland management emphasizes "seeing clearly and acting quickly". By integrating various remote sensing data such as satellite, unmanned aerial vehicle and ground sensors with crop models, researchers can estimate yields on a 20-meter square scale and even track subtle changes in crops at different stages (Yang et al., 2024). This improvement in resolution makes management more specific, and irrigation, fertilization and pesticide
RkJQdWJsaXNoZXIy MjQ4ODYzNA==