Digital Twin‐Based Uncertain Weather Condition Monitoring for Enhanced Crop Yield Prediction

ABSTRACT Crop yield is impacted by uncertain weather conditions, including irregular rainfall, changing temperatures, and humidity variations. Existing crop yield prediction models fail to capture the dynamic nature of weather and mostly depend on limited data. Accurate modeling of weather variability is crucial for improving the reliability of crop yield prediction and climate‐aware agricultural decision‐making. Digital twin (DT) technology provides a solution by creating a virtual farm that replicates the real‐world farm conditions. This study develops a DT‐based crop yield prediction framework, CropTwin, to predict crop yield under uncertain weather conditions without relying on localized sensor networks. Meteorological data from the NASA POWER Data Access Viewer (DAV) were collected to predict crop yields in rice‐growing fields of the study location. We used a crop growth model to predict yields based on this original weather data. To validate how uncertain weather conditions affect yield, we created multiple scenarios with slightly varying weather conditions using a Monte Carlo simulation. We evaluated the CropTwin on all these scenarios to obtain a range of possible yields. The crop yield value is determined using various metrics, and the generalizability across different types of crops. Experimental results show that the proposed model accurately predicts the crop yields with a value of 5.90 t/ha under real‐time NASA data conditions and 5.62 t/ha under simulated uncertain weather conditions. Furthermore, the framework demonstrates good generalization performance across multiple crops and regions. This shows the efficient crop yield performance of the proposed DT‐based crop yield prediction framework.