Predicting Maize Theoretical Methane Yield in Combination with Ground and UAV Remote Data Using Machine Learning.

Area of Science:

• Agricultural remote sensing
• Precision agriculture
• Biomass estimation

Background:

• Accurate estimation of maize total-above ground biomass (TAB) and theoretical biochemical methane potential (TBMP) is crucial for agriculture.
• Unmanned aerial vehicle (UAV) and machine learning (ML) offer potential for non-destructive estimation but are underutilized in the Nordic-Baltic region.

Purpose of the Study:

• To assess the efficacy of UAV-based multispectral data and ML models for estimating maize TAB and TBMP.
• To determine the optimal phenological stages for accurate prediction in the Nordic-Baltic region.

Main Methods:

• UAV multispectral data (blue, red, green, red edge, infrared) were collected during key maize phenological stages.
• UAV-derived vegetation indices were integrated with field measurements and various ML models (generalized linear, random forest, support vector machines).

Main Results:

• ML models achieved high prediction accuracy (88-95% for TAB, 88-97% for TBMP) between the blister (R2) and Dough (R4) stages.
• The V7-V10 growth stages were identified as the earliest practical timing for adequate TAB and TBMP prediction for farmers.

Conclusions:

• UAV technology combined with ML models provides a successful approach for maize TAB and TBMP estimation.
• Further research is recommended to refine these techniques for broader agricultural applications.