Accurately identifying what crops are grown, and when, is very important for managing farms effectively—especially in places like Yuma County, Arizona, where farmers grow many crops each year. In this study, we used daily, high‑resolution satellite images (Planet Fusion, 3 m) from 2019–2023 and applied machine learning to figure out which crops were in each field. We trained and checked our model using field data collected by the U.S. Bureau of Reclamation.

Among several machine learning methods, XGBoost worked best, reaching about 98% accuracy in training and 94% in testing. Adding information about plant growth stages and how each field’s satellite signals vary over time helped the model tell similar crops apart. The model worked especially well for lettuce, wheat, cotton, and mature date palms, though it had more trouble with Bermuda–Klein grass and young citrus trees.

Overall, this approach can quickly and accurately track crops multiple times a year, helping with better farm and water management.