ASABE Technical Library - Abstract

Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, view access and subscription options

Yield Prediction from a Messy Yield Monitor Data

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  2026 ASABE Annual International Meeting  2601134.(doi:10.13031/aim.202601134)
Authors:   Xiaoyu Zhang, Sunoj Shajahan
Keywords:   Yield prediction, Satellite imagery, Machine learning, Vegetation indices

Abstract. Accurate yield prediction is critical for precision agriculture. It is strongly affected by the ground truth data quality and the satellite platform. This study evaluated the effects of satellite spatial resolution and yield monitor delay correction on field-scale yield prediction. We used multi-year data from 3 fields in Champaign, Illinois, including yield monitor data and multi-spectral imagery from Planet Labs (3 m), Sentinel-2 (10 m), and Landsat 8 (30 m). Yield data were cleaned using flow delay correction, and spatial interpolation was performed using Matérn kriging. Vegetation indices such as normalized difference vegetaion index (NDVI) and the green chlorophyll vegetation index (GCVI) were used as model inputs in a random forest framework. Results showed that higher-resolution imagery did not consistently improve prediction accuracy, and coarser resolution data achieved lower RMSE. NDVI saturated during peak growth, while GCVI remained more sensitive to yield variation. Flow delay correction had a strong impact on model performance, as even small time shifts (±4 seconds) introduced large errors in yield maps.

(Download PDF)    (Export to EndNotes)