Canada has pledged to cut fertilizer-related emissions by 30% and overall greenhouse gas emissions by 45% by 2030, aiming for net-zero by 2050. Achieving these goals while maintaining or increasing crop yields is challenging because farmland is limited and must be used more efficiently. Stable and productive farmland is essential for meeting these targets, but detailed information about how yields change over time at the field or regional level has been lacking in Canada. This study addresses that gap by creating the first detailed mapping system to assess both the stability and productivity of crop yields over time. Using seven years of yield data (2017–2023) from about 15,000 field maps covering 14 million hectares in the Canadian Prairies, researchers combined on-farm yield data with satellite observations and environmental information. They integrated these datasets with artificial intelligence to categorize arable lands into stable-high, stable-low, and unstable yield areas at a fine 10‑meter resolution. Results show that nearly half of the studied land (45%) was unstable, with climate variability being a major cause. These findings provide valuable insights for precision agriculture, which helps farmers target high-yield areas for intensification and conserve less productive lands, supporting both food production and climate goals.