Detecting Sensitive Spectral Bands and Vegetation Indices for Potato Yield Using Handheld Spectroradiometer Data.

Area of Science:

• Agricultural Science
• Remote Sensing Technology
• Crop Physiology

Background:

• Precision agriculture leverages remote sensing for efficient crop management.
• Canopy reflectance offers non-destructive insights into crop health and yield potential.
• Potato (Solanum tuberosum L.) yield prediction requires optimized data acquisition strategies.

Purpose of the Study:

• Determine the optimal temporal window for remote sensing measurements of potato canopy signals related to yield.
• Identify the most effective spectral bands (350-2500 nm) and vegetation indices for predicting potato tuber yield.
• Develop a predictive model for potato yield using selected spectral features.

Main Methods:

• Conducted a two-year study (2020-2021) measuring plant-by-plant canopy reflectance of six potato varieties.
• Utilized correlation analysis and dimensionality reduction to identify yield-correlated spectral features.
• Employed multiple linear regression and Leave-One-Out Cross-Validation (LOOCV) to build and validate the yield prediction model.

Main Results:

• Identified 23 independent spectral features significantly correlated with potato tuber yield.
• The Gitelson2 and Vogelmann indices were among the most significant predictors.
• The optimal measurement period for yield prediction was found to be 56 to 100 days after planting.
• The developed model achieved an RMSE of 702 g with a %RMSE of 29.2%.

Conclusions:

• Remote sensing, specifically canopy reflectance measured between 56-100 days post-planting, is effective for predicting potato yield.
• Specific spectral bands and vegetation indices can be tailored for sensor applications in precision potato farming.
• This research supports the advancement of precision agriculture, potentially enhancing food security through improved crop management.