The spore of the beans: Spatially explicit models predict coffee rust spread in fragmented landscapes

Landscape structure influences the spread of plant pathogens, primarily by affecting pathogen dispersal. Coffee leaf rust (Hemileia vastatrix), a fungal disease that causes heavy economic losses in the coffee industry, is likely to be affected by landscape structure via dispersal of its wind-borne spores. Previous studies have found positive associations between leaf rust incidence and the proportion of pasture cover, suggesting that deforestation may facilitate rust spore dispersal. We explored this idea by modeling the spread of rust transmission in simulated landscapes. Specifically, we modeled within-patch transmission using a probabilistic cellular automata model, and between-patch transmission using a random walk with spore movement inhibited by forest canopy cover. We used this model to understand how the spread of coffee rust is affected by: 1) clustering of coffee plants, 2) clustering of deforestation, and 3) proportion of landscape deforestation. We found that clustering of coffee plants is the primary driver of rust transmission, affecting the likelihood and severity of rust outbreak. Deforestation is important in landscapes with high clustering of coffee: rust outbreaks are more severe in landscapes with a higher proportion of deforested areas, and more variable in landscapes where deforested areas are more evenly dispersed throughout the landscape.


Introduction 1
Many ecological systems are characterized by their landscape configuration, including natural 2 or anthropogenic habitat fragmentation (Levin 1992;Fahrig 2003). Landscape structure 3 includes the distribution and quality of habitat, which in turn affects the connectivity of habitat 4 patches. Furthermore, landscape structure is known to affect various species through its effect 5 on connectivity of high quality habitat patches, patch size and extinction risk, as well as edge pathogens. Yet, there have been limited efforts to understand mechanistically how landscape 18 structure affects plant populations (Cunniffe et al. 2015a).

19
Here we use the plant disease, coffee leaf rust (Hemileia vastatrix), as a model system for 20 understanding issues of fragmentation and disease spread. First recorded in 1879, coffee leaf 21 rust is a fungal disease notably recognized for completely destroying the coffee industry in Sri 22 Lanka, which used to be one of the largest coffee-producer countries in the world. Since the 23 1970s, coffee rust has spread to the largest coffee-producing regions in the world including 24 Brazil, Mexico, and Colombia. Reports of up to 30-50% losses due to coffee rust in Brazil and 25 Costa Rica, 31% in Colombia, and 16% in Central America make this disease an urgent priority 26 for the coffee-growing industry ( underneath leaves that release between 300,000 to 400,000 spores into the environment 33 (Kushalappa and Eskes 1989). The infection process is composed of three principal steps: 34 germination, penetration, and colonization of spores, with germination and penetration 35 requiring the presence of running water (Kushalappa and Eskes 1989). Other factors influencing 36 germination and penetration include leaf age and spore concentration and distribution at the 37 infection site. Colonization of spores into the stomatal opening of the leaves is followed by the 38 production of spores (sporulation). Within one to three weeks, pale yellow spots on the 39 underside of leaves are the first signs of infection. Depending on the age of the leaves, spore 40 production can begin anywhere from two weeks to months following infection (Kushalappa and 41 Eskes 1989). At maturation, spores are released into the air and can travel as high 1000 meters 42 above coffee canopies (Martinez et al. 1975). Disease severity has been associated with the 43 number of visible spores per leaf, the duration of leaf wetness, rainfall, and minimum 44 temperatures (Zambolim 2016). At the canopy level, rainfall can facilitate the spread of spores 45 via raindrops that splash between leaves (Kushalappa and Chaves 1980;Waller 1982). Thus, the 46 spread and severity of coffee leaf rust are heavily determined by climatic conditions.

47
In addition to rain, spore dispersal is primarily facilitated by strong wind patterns that 48 potentially carry spores hundreds of miles from the origin (Waller, 1982). Earlier studies have 49 provided evidence for the dispersal of airborne spores through trapping techniques above and 50 within coffee tree canopies (Martinez et al. 1975, Becker et al. 1975). Under outbreak 51 conditions, maximum wind speed and low relative humidity have been associated with 52 increases in spore dispersal during late morning and afternoon (Becker et al. 1977a).

53
Furthermore, wind speeds between 12-20 km/hr have been linked to high numbers of spores 54 with moderate numbers at wind speeds as low as 7 km/hr (Martinez et al. 1977). Wind gusts, in    infection rates of coffee rust. Therefore, here we use simulation models to investigate how 67 landscape composition and configuration influence the spread of the coffee rust. 68 We hypothesize that the windborne dispersal of rust spores is facilitated by landscape 69 composition and configuration. Specifically, we examine the effects on disease transmission 70 from the clustering of coffee plants, proportion of deforestation within the landscape, and the 71 degree to which deforested areas are scattered in space. We predict that rust spores will 72 disperse more readily through landscapes with high coffee clustering and deforestation levels; 73 resulting in a higher incidence of coffee rust in landscapes that exhibit these characteristics.   (Table 1).

86
The second landscape aspect represents the surrounding matrix and "mirrored" the coffee  (Table 1).     Figure 4A).

178
Maximum infection tends to decrease with dispersion, but this association is weak (ρ = -0.255, 179 Figure 4B). Deforestation was also strongly associated with skew, with a greater degree of right 180 skew at low values of deforestation (ρ = -0.514, Figure 4C). Finally, the concentration parameter 181 tended to be highest at high values of dispersion (ρ = 0.455, Figure 4D). The concentration  Using a spatially-explicit model, we found that the spread of coffee rust is primarily affected by 353 clustering of coffee plants (Figure 3). In addition, clustering interacts with deforestation and 354 fragmentation to influence disease spread (Figures 4, 5). Overall, the models predict a general 355 pattern of highly clustered coffee plots and deforestation leading to greater coffee rust 356 prevalence, with dispersed deforestation increasing individual outbreaks' predictability.

358
Our results have important implications for local and regional management practices.