Previsão da dinâmica sazonal de Frankliniella schultzei (Thysanoptera: Thripidae) em cultivos de tabaco usando redes neurais artificiais
Loading...
Date
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Viçosa
Abstract
Os cultivos de tabaco (Nicotiana tabacum) têm grande importância econômica e social devido a renda e aos empregos gerados. O tripes Frankliniella schultzei (Thysanoptera, Thripidae) é uma das principais pragas nesses cultivos. A compreensão da dinâmica populacional das pragas é importante para os programas de Manejo Integrado de Pragas, pois isso possibilita se identificar os fatores reguladores e prever períodos de maior ocorrência desses organismos. As Redes Neurais Artificiais (RNAs) destacam-se como ferramentas de inteligência artificial com elevado potencial para modelar e prever fenômenos complexos, como a dinâmica sazonal de populações de pragas. Assim, esse trabalho teve como objetivo determinar um modelo de previsão da dinâmica sazonal de F. schultzei em cultivos de tabaco usando RNAs. Para tanto, durante três anos em 87 cultivos conduzidos em pivôs centrais de cerca de 150 ha foram monitoradas a cada cinco dias as densidades de F. schultzei, a idade das plantas de tabaco e os elementos climáticos. Foram determinadas 1320 RNAs e foi selecionada aquela com maior poder preditivo. Essa RNA foi validada e posteriormente, o seu desempenho foi comparado com modelos de Floresta aleatória, regressão linear múltipla e ARIMA. A RNA selecionada utilizou como preditores a idade das plantas de tabaco, a temperatura média do ar, a velocidade média dos ventos e a precipitação pluviométrica. Nesse modelo foi usada uma defasagem de 30 dias para os elementos climáticos e ele possui dez neurônios na camada oculta, função de ativação tangente hiperbólica e algoritmo de aprendizado resiliente backpropagation. Essa RNA apresentou o melhor desempenho preditivo entre as RNAs avaliadas e do que os modelos de Floresta Aleatória, Regressão Linear Múltipla e ARIMA. Verificou-se que o aumento da densidade populacional desse tripes nos cultivos de tabaco esteve associado ao aumento da idade das plantas e com períodos de temperaturas mais amenas, precipitação pluviométrica intermediária e menores velocidades dos ventos. Portanto, o modelo de RNA estabelecido neste trabalho é apropriado para determinar a dinâmica sazonal de F. schultzei nos cultivos de tabaco. Palavras-chave: Nicotiana tabacum; Tripes; Inteligência artificial; Elementos climáticos; Idade das plantas
Tobacco (Nicotiana tabacum) crops are of great economic and social importance due to the income and jobs they generate. The thrips Frankliniella schultzei (Thysanoptera, Thripidae) is one of the major pests affecting these crops. Understanding pest population dynamics is crucial for Integrated Pest Management programs, as it enables the identification of regulatory factors and the prediction of periods when these organisms are most prevalent. Artificial Neural Networks (ANNs) stand out as artificial intelligence tools with high potential for modeling and predicting complex phenomena, such as the seasonal dynamics of pest populations. Thus, this study aimed to develop a predictive model for the seasonal dynamics of F. schultzei in tobacco crops using ANNs. To this end, F. schultzei densities, tobacco plant age, and climatic variables were monitored every five days over a three-year period across 87 crops managed under center-pivot irrigation systems (approximately 150 hectares each). A total of 1,320 ANNs were developed, and the one with the highest predictive power was selected. This ANN was validated, and its performance was subsequently compared with Random Forest, multiple linear regression, and ARIMA models. The selected ANN used tobacco plant age, average air temperature, average wind speed, and rainfall as predictors. The model incorporated a 30-day lag for climatic variables and featured ten neurons in the hidden layer, a hyperbolic tangent activation function, and the resilient backpropagation learning algorithm. This ANN demonstrated superior predictive performance compared to the other evaluated ANNs and to the Random Forest, Multiple Linear Regression, and ARIMA models. The study found that increases in the population density of this thrips species in tobacco crops were associated with increasing plant age and with periods characterized by milder temperatures, moderate rainfall, and lower wind speeds. Therefore, the ANN model established in this study is suitable for determining the seasonal dynamics of F. schultzei in tobacco crops. Keywords: Nicotiana tabacum; Thrips; Artificial intelligence; Climatic elements; Plant age
Tobacco (Nicotiana tabacum) crops are of great economic and social importance due to the income and jobs they generate. The thrips Frankliniella schultzei (Thysanoptera, Thripidae) is one of the major pests affecting these crops. Understanding pest population dynamics is crucial for Integrated Pest Management programs, as it enables the identification of regulatory factors and the prediction of periods when these organisms are most prevalent. Artificial Neural Networks (ANNs) stand out as artificial intelligence tools with high potential for modeling and predicting complex phenomena, such as the seasonal dynamics of pest populations. Thus, this study aimed to develop a predictive model for the seasonal dynamics of F. schultzei in tobacco crops using ANNs. To this end, F. schultzei densities, tobacco plant age, and climatic variables were monitored every five days over a three-year period across 87 crops managed under center-pivot irrigation systems (approximately 150 hectares each). A total of 1,320 ANNs were developed, and the one with the highest predictive power was selected. This ANN was validated, and its performance was subsequently compared with Random Forest, multiple linear regression, and ARIMA models. The selected ANN used tobacco plant age, average air temperature, average wind speed, and rainfall as predictors. The model incorporated a 30-day lag for climatic variables and featured ten neurons in the hidden layer, a hyperbolic tangent activation function, and the resilient backpropagation learning algorithm. This ANN demonstrated superior predictive performance compared to the other evaluated ANNs and to the Random Forest, Multiple Linear Regression, and ARIMA models. The study found that increases in the population density of this thrips species in tobacco crops were associated with increasing plant age and with periods characterized by milder temperatures, moderate rainfall, and lower wind speeds. Therefore, the ANN model established in this study is suitable for determining the seasonal dynamics of F. schultzei in tobacco crops. Keywords: Nicotiana tabacum; Thrips; Artificial intelligence; Climatic elements; Plant age
Description
Citation
DONATO, Conceição Aparecida da Silva. Previsão da dinâmica sazonal de Frankliniella schultzei (Thysanoptera: Thripidae) em cultivos de tabaco usando redes neurais artificiais. 2026. 26 f. Dissertação (Mestrado Profissional em Defesa Sanitária Vegetal) - Universidade Federal de Viçosa, Viçosa. 2026.
