Remote Sensing

• Satellite and drone imagery, and the resolution each provides
• NDVI, NDRE and their saturation at high canopy cover
• Cloud cover, atmospheric correction and revisit interval as practical limits

Modelling

• Yield prediction and the small number of seasons available as training data
• Disease and pest detection from images, and field conditions against clean datasets
• Spatial and temporal autocorrelation, and why random splits overstate accuracy

Systems

• Edge against cloud processing when bandwidth and power are constrained
• Data storage, gateway design and handling intermittent connectivity
• Dashboards and alerts that a farmer will actually act on

Adoption

• Smallholder economics and the cost ceiling for any deployed system
• Agronomic validation with trials rather than model accuracy alone
• Data ownership, advisory liability and trust in automated recommendations