Development and assessment of a warning system for coffee rust management and its use for disease risk evaluation

Author(s)

Fernando Dill Hinnah

• Document type: Doctoral Thesis
• Press: Piracicaba.
• Institution: Universidade de São Paulo (USP). Escola Superior de Agricultura Luiz de Queiroz (ESALA/BC)
• Defense date: 2018-03-29

Examining board members

Paulo Cesar Sentelhas; Luiz Roberto Angelocci; Carlos Alberto Alves Meira

• Advisor: Paulo Cesar Sentelhas

Abstract

Coffee crop is of major importance to Brazil, being cultivated on more than 2 million hectares. It is a strategic commodity for the country, which is the main world producer. Several factors influence the yields, mainly a disease known as coffee leaf rust (CLR), caused by the fungus Hemileia vastatarix. This disease can reduce yield up to 35% and the most common strategy for CLR controlling is by spraying fungicides, with interval based on the residual period and according to regional CLR intensity. This traditional way does not consider the climate influence on disease development. With the aim of developing a forecast system (FS) for CLR management, employing weather data from CLR field assessments since 1998, several steps were performed: a) CLR epidemiology analysis; b) correlation between disease progress rates and weather variables; c) development of a forecast system, in order to rationalize chemical control; d) assessment of the FS performance on field trials; e) generation of an agro-climatic index for CLR risk assessment in Brazilian coffee areas; and f) evaluation of possible El Niño Southern Oscillation (ENSO) influence on CLR epidemics. Analising 88 siteseason CLR epidemics, from Varginha, Boa Esperança and Carmo de Minas, MG, the best fit was obtained by Gompertz model. Using stepwise method, CLR infection rates were estimated with multiple linear regressions, using minimum temperature and relative humidity as inputs. The model performed well, presenting less than 9.5% of false negatives in the months assessed. To evaluate CLR forecast system, two field trials were performed during 2015-16 season (Varginha and Boa Esperança), and five during 2016-17 season (Varginha, Boa Esperança, Uberlândia, Buritizal, and Campinas). The FS treatments performed better than the calendar spray system in six trials, with the exception for Campinas. The poor FS performance in Campinas evidenced the necessity of FS threshold calibration at sites different from the region where the FS was developed, once it is empirical. In order to assess the risk, the estimated CLR infection rate was evaluated for 46 different sites in Brazilian coffee producing region. Historical weather data since 1961 to 2015 for each site was used to estimate daily values of cumulative infection rate (CIR). Each site and season were classified into five CIR scores from Very Low (score 0) to Very High (score 4). The risk was spatialized using multiple linear regression based on geographical coordinates and altitude. The Brazilian coffee region was classified into four risk classes, being most of them between Medium to High risks in the area currently cultivated with coffee. For the same historical serie, CIR was estimated for 45 locations and then classified by ENSO phases: El Niño (EN); Neutral (NT); and La Niña (LN). A predominant absence of ENSO effect on CLR in Brazil was observed. Only in Paraná and São Paulo states there was ENSO effect, with higher CIR during EN seasons. (AU)

• FAPESP's process: 14/17781-8 - Development and application of phytosanitary warning systems for the management of coffee diseases
• Grantee: Fernando Dill Hinnah
• Support Opportunities: Scholarships in Brazil - Doctorate