Relações entre atributos físico-químicos do solo e de Pratylenchus brachyurus na cultura da soja

• Autor(a) principal: Kashiwaqui , Marcos Massuo
• Data de Publicação: 2021
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações do UNIOESTE
• Texto Completo: http://tede.unioeste.br/handle/tede/5426
• topic: Nematoides; Geoestatística; Árvore de decisão; Weka; Mineração de dados; CIÊNCIAS AGRÁRIAS:AGRONOMIA

Brazil is one of the major world producers of soy, however some pests and diseases are of concern in the production environment of this oilseed. The root lesion nematode is an example of this, since the main control methods used have limitations for the management of these organisms. Proper management of soil chemical attributes can positively contribute to the management of Pratylenchus brachyurus, by reducing its population and damage, as well as improving crop development. Some recent advances, such as precision agriculture and the discovery of knowledge in databases (KDD), can contribute to the identification of factors that act for the spread of nematodes. The objective of this study was to obtain relations between physicalchemical attributes and the population density of P. brachyurus, using the decision tree technique, in an area cultivated with soy in Mundo Novo/MS. Data collection was carried out in the 2018/19 harvest. Geostatistical analysis of the attributes, pH, phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), potential acidity (H + Al), aluminum (Al), cation exchange capacity (CTC) was performed, base saturation (V), content of sand, silt, clay and number of individuals of P. brachyurus. The mapping was performed using the ordinary kriging method. The decision tree was modeled using the Weka software. The nematode density, class attribute, was categorized into two classes, “High” and “Low”, using the EM (Expectation-Maximization) algorithm. The imbalance in the data was minimized with the use of the resample filter, and for the selection of attributes, the Wrapper method was adopted. The decision tree was modeled using the J48 classifier and evaluated for accuracy, Kappa index, precision and recall. In the geostatistical analysis, a predominance of the gaussian and spherical models was observed, for 5 and 4 attributes, respectively. The Matérn model was observed in 2 attributes, while for the silt and the density of P. brachyurus, no spatial dependence, pure nugget effect was observed. Regarding the degree of spatial dependence, the physical attributes of sand and clay were highlighted, in addition to the chemical attributes H + Al and CTC, with strong spatial dependence, while the other chemical attributes were presented in a moderate way. The decision tree generated using the J48 classifier had 91.99% accuracy and 0.81 Kappa index, whereas the “High” and “Low” classes showed 0.96 and 0.84 accuracy, as well as 0, 93 and 0.90 for recall, respectively. The decision tree obtained enabled the identification of 5 classification rules, with the participation of the attributes pH, P and K, in 5, 3 and 2 rules, respectively. In the geostatistical analysis, the attributes of greater spatial continuity were sand and clay, while most of the chemical attributes were identified with moderate dependence. The population density of P. brachyurus could not be mapped due to the pure nugget effect. The decision tree technique proved to be promising and made it possible to identify the pH, P and K attributes, which were related to the population density of P. brachyurus.