Methodologies based on Remote Sensing for detecting Risks not Covered in Agricultural Insurance

• Grant number: 23/17866-2
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Duration: August 01, 2024 - July 31, 2026
• Field of knowledge: Interdisciplinary Subjects
• Principal Investigator: Daniel Lima Miquelluti
• Grantee: Daniel Lima Miquelluti
• Host Company: Agririsk Soluções em Gerenciamento de Risco Ltda
• CNAE: Tratamento de dados, provedores de serviços de aplicação e serviços de hospedagem na internet
• City: Piracicaba
• Pesquisadores principais: Vitor Augusto Ozaki
• Associated researchers: Lucas Barbosa Cavalcante ; Luis Henrique Andia ; Pedro Henrique Batista de Barros
• Associated scholarship(s): 24/14121-9 - Full Stack Developer, BP.TT; 24/14127-7 - Support in data processing and analysis applied to Agricultural Insurance, BP.TT

Abstract

Weather, market and logistical fluctuations are among the main risks that affect agribusiness. However, a very common risk, but little understood by the insurance market, is related to losses due to pests, diseases and weeds, and the interaction with climatic factors. There is high complexity in separating the damage caused by climatic factors, for example, drought, from those resulting from pests, diseases and/or weeds. Each of these factors affects crop productivity differently, and the problem is that only losses caused by adverse weather conditions are subject to compensation by insurers. In 2021 and 2022, rural producers suffered heavy losses, and the amount compensated by insurance companies reached a record amount of R$15.9 billion. The market estimates that between 10 and 15% of this total is related to measurement errors, that is, losses caused by phytosanitary problems (not covered by insurance), and also fraud. Therefore, we seek to improve the management of agricultural claims, using remote sensing techniques. The purpose of this project is to collect and process high-resolution satellite images of agricultural areas (initially for soybeans, first and second harvest corn and wheat), using pre-processing techniques to improve the quality and clarity of images. images; and, develop algorithms for image processing and machine learning, capable of identifying and differentiating areas affected by weeds, pests and diseases. In other words, a classification model will be developed that categorizes the affected areas and maps the spatial distribution of damage caused by these phytosanitary problems in the monitored areas, integrating the information extracted from satellite images to identify patterns and areas of infestation. used to correct the survey of crop losses, and consequently, the amount to be compensated. The results of the project will be important as they will increase the effectiveness and precision in quantifying crop losses, thereby reducing moral risk and improving the financial result of the market as a whole. (AU)