A soft computing approach for epidemiological studies of coffee and soybean rusts

Abstract

Solutions of complex problems require intelligent systems that combine knowledge, techniques and methodologies, from different sources, considering environmental changes, for decision support improvement. Thus, it became necessary to apply robust methodologies to characterize the interaction among climatic variables related to epidemic progress. The objective of the present work was to evaluate the effects of temperature and leaf wetness in asian soybean (Glycine max L.) rust (Phakopsora pachyrhizi H. Sydow & P. Sydow) intensity in Suprema cultivar and coffee (Coffea arabica L.) rust (Hemileia vastatrix Berkeley & Broome) intensity in Mundo Novo and Catuaí cultivars using linear regression (LR), nonlinear regression (NLR), fuzzy logic systems (LFS) and neuro-fuzzy systems (NFS). Comparing observed and estimated values for both diseases, NFS increased the precision and accuracy of the estimated values, following in decrease order by LFS, NLR and LR. NFS enabled to explain 85% and 99% of asian soybean rust and coffee rust monocyclic process, respectively.