Recôncavo Baiano is an area favourable for the occurrence of citrus greasy spot (CGS) (Mycosphaerella citri), but there has been no study of this pathosystem in Brazil. This work aimed to characterise the temporal patterns of CGS-induced defoliation in sweet orange cultivars 'Bahia' (Washington Navel) and 'Pêra'. Temperature, rainfall and relative humidity were recorded, as well as weekly defoliation (fallen leaves/canopy m 2 or m 3 ). Considering the mean of fallen leaves per canopy m 2 , and mean canopy area, the total annual defoliation was estimated to be around 32 000 leaves per plant for 'Bahia' and 18 500 for 'Pêra' sweet orange. Spectral density analysis showed that defoliation has a 5-week-long main cycle for both cultivars. The proportion of symptomatic fallen leaves was never below 0.97. The monthly number of fallen leaves per canopy area was positively correlated with the mean CGS incidence on leaves. Defoliation was significant, resulting in a low leaf density throughout the year. Many defoliation cycles and the very high proportion of symptomatic fallen leaves assure a constant inoculum supply. Based on these results, CGS cannot be considered a minor disease, at least in Recôncavo Baiano.

13 mathematical model, meta-population model. 14 15 Abstract 16 Citrus greasy spot (CGS), caused by Zasmidium citri, induces premature defoliation and yield 17 loss in Citrus spp. CGS epidemiology is well understood in areas of high humidity such as 18 Florida (USA), but remains unaddressed in Brazil, despite differing climatic conditions and 19 disease management practices. We characterize the spatiotemporal dynamics of CGS in the 20 Recôncavo of Bahia, Brazil, focusing on four hierarchical levels (quadrant, plant, grove and 21 region). A survey conducted in 19 municipalities showed that disease is found throughout the 22 entire region with a prevalence (i.e. proportion of affected sampling units) of 100% in groves 23 and plants, and never lower than 70% on leaves. Index of dispersion (D) values suggest the 24 spatial pattern of symptomatic units lies somewhere between random and regular. This was 25 confirmed by the parameters of the binary power law for plants and their quadrants (log(A)<0 26 and b<1). Variability in disease severity at different plant heights (0.7 m, 1.3 m and 2.0 m) 27 was tested, but no consistent differences were observed. We introduce a simple 28 compartmental model synthesising the epidemiology of the disease, in order to motivate and 29 guide further research. The data we have collected allow such a model to be parameterised, 30 albeit with some ambiguity over the proportion of new infections that result from inoculum 31 produced within the grove vs. external sources of infection. By extending our model to 32 include two populations of growersthose who control and those who do notcoupled by 33 the spread of airborne inoculum, we investigate likely performance of the type of cultural 34 controls that would be accessible to citrus growers in Northeastern Brazil. Our model shows 35 that control via removal of the key source of inoculumi.e.