Estimating productivity of potatoes using digital agriculture tools

• Grant number: 23/14041-2
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: November 01, 2023
• End date: October 31, 2025
• Field of knowledge: Physical Sciences and Mathematics - Geosciences
• Principal Investigator: Rouverson Pereira da Silva
• Grantee: Samira Luns Hatum de Almeida
• Host Institution: Faculdade de Ciências Agrárias e Veterinárias (FCAV). Universidade Estadual Paulista (UNESP). Campus de Jaboticabal. Jaboticabal , SP, Brazil
• Associated research grant: 21/06029-7 - High resolution remote sensing for digital agriculture, AP.TEM

Abstract

The modernization of agriculture has provided strategies aimed at making the activity more sustainable and capable of meeting the growing demand for food. In the face of a scenario of population growth, ensuring food security is crucial for the future of upcoming generations. In this sense, it is of utmost importance to use technology to advance studies on foods that form the basis of human nutrition, such as potatoes (Solanum tuberosum L.), the fourth most consumed food in the world. Given the above, the use of spectral data from proximal sensors, coupled with unmanned aerial vehicles (UAVs) and satellites, can contribute to the generation of more robust models for estimating potato productivity. With this in mind, the aim of this work is to use different levels of remote sensing - orbital, aerial, and proximal - in conjunction with artificial intelligence algorithms to obtain more accurate and precise models for estimating potato productivity. The experiment will be conducted in two commercial areas cultivated with English potatoes (Solanum tuberosum) in the state of São Paulo. Climatic and soil information from the experimental fields will be collected for characterization purposes. The sample grid will be regular and composed of 25 points. Evaluations will be carried out from the tuberization phenological stage, with bi-weekly collection of biophysical parameters of the crop, both aerial and tuber, until the harvest date. Evaluations will begin approximately 45 days before harvest, totaling four temporal evaluations. Remote sensing data acquisition will occur at three collection levels: proximal, aerial, and orbital. At the proximal level, the ACS-430 Crop Circle sensor is intended to be used. This sensor is active, emitting electromagnetic light and capturing reflectance at central wavelengths of 670, 730, and 780 nm. Aerial flights will be conducted using the DJI Phantom P4 Multispectral UAV with RTK. This sensor captures reflectance in the blue (450 nm ± 16 nm), green (560 nm ± 16 nm), red (650 nm ± 16 nm), red edge (730 nm ± 16 nm), and near-infrared (840 nm ± 26 nm) bands. Orbital level will involve the use of images from the PlanetScope CubeSat platform sensor. This sensor generates surface reflectance images with a spatial resolution of 3 meters, composed of 4 spectral bands in the wavelength ranges of blue (455-515 nm), green (500590 nm), red (590-670 nm), and near-infrared (NIR - 780-860 nm). With the reflectance data from all platforms, vegetation indices will be calculated. The data will be analyzed through Pearson correlation analysis (p<0.01 or p<0.05) to identify whether spectral variables have any positive or negative relationship with productivity and quality variables. Subsequently, the vegetation indices will be used to generate prediction models using Artificial Neural Networks (ANNs). For this, two types of ANNs will be employed: Radial Basis Function (RBF) and Multilayer Perceptron (MLP), both of which are supervised learning models. For model training and testing, the database will be divided into 70% of the data for training and 30% for testing. The efficiency of the generated models will be analyzed in terms of accuracy and precision using Mean Absolute Error (MAE) and the coefficient of determination (R²). The knowledge generated in this research will enable: estimating and/or predicting potato productivity; indicating the optimal time for potato harvest based on quality; conducting intelligent harvesting of the crop, indicating zones where the potatoes have suitable tuberization for the consumer market; harvesting at the ideal moment will lead to increased productivity and quality of the harvested material. (AU)