Phytopathology • 2018 • 108:1420-1428 • https://doi.org/10.1094/PHYTO-03-18-0088-R Ecology and Epidemiology Regional Spatial-Temporal Spread of Citrus Huanglongbing Is Affected by Rain in Florida M. M. Shimwela, T. S. Schubert, M. Albritton, S. E. Halbert, D. J. Jones, X. Sun, P. D. Roberts, B. H. Singer, W. S. Lee, J. B. Jones, R. C. Ploetz, and A. H. C. van Bruggen† First, tenth, and twelfth authors: Department of Plant Pathology, IFAS, University of Florida, Gainesville 32611; first, eighth, and twelfth authors: Emerging Pathogens Institute, University of Florida, Gainesville 32610; second, third, fourth, fifth, and sixth authors: Florida Department of Agriculture and Consumer Services, Division of Plant Industry, Gainesville 33825; seventh author: Department of Plant Pathology, IFAS, SWFREC, University of Florida, Immokalee 34142; ninth author: Department of Agricultural and Biological Engineering, Gainesville, FL 32611; and eleventh author: University of Florida, Plant Pathology Department, TREC–Homestead, FL 33031. Accepted for publication 5 June 2018. ABSTRACT Citrus huanglongbing (HLB), associated with ‘Candidatus Liberibacter asiaticus’ (Las), disseminated by Asian citrus psyllid (ACP), has devastated citrus in Florida since 2005. Data on HLB occurrence were stored in databases (2005 to 2012). Cumulative HLB-positive citrus blocks were subjected to kernel density analysis and kriging. Relative disease incidence per county was calculated by dividing HLB numbers by relative tree numbers and maximum incidence. Spatiotemporal HLB distributions were correlated with weather. Relative HLB incidence correlated positively with Citrus huanglongbing (HLB) or greening is one of the most important citrus diseases worldwide (Gottwald 2010). HLB has been known in Asia and Africa since 1900 and 1920, respectively, but was discovered relatively recently in South (2004) and North America (2005) (da Graça et al. 2016; Gottwald et al. 2007; Halbert 2005). Major yield losses have been reported for all HLB-affected citrus production areas, including Florida (Farnsworth et al. 2014; Udell et al. 2017). In North America, HLB is associated with ‘Candidatus Liberibacter asiaticus’ (Las). Las is mainly restricted to the phloem of plants in the family Rutaceae, including species of Citrus and Murraya (Halbert et al. 2012). Las is naturally transmitted by the Asian citrus psyllid (ACP), Diaphorina citri Kuwayama (Hall et al. 2012). ACP feeds and reproduces on newly developing flushes of very young leaves on the expanding terminal ends of shoots (Chiyaka et al. 2012; Hall et al. 2012; Lee et al. 2015; Udell et al. 2017). Nymphs and adults obtain Las from the phloem and may inject the bacteria into the phloem of a hitherto healthy branch or tree (Ammar et al. 2016; Chiyaka et al. 2012). Flush colonization by nymphs enhances the chance of transmission (Lee et al. 2015). The †Corresponding author: A. H. C. van Bruggen; E-mail: ahcvanbruggen@ufl.edu Funding: This work was made possible by the generous support of the American people through the United States Agency for International Development (USAID)funded Innovative Agricultural Research Initiative (iAGRI) project (Award CA621-A-00-11-00009-00) that provided a PhD fellowship to M. Shimwela. We are grateful to the Esther B. O’Keeffe Foundation for contributing additional funding to this research. The opinions expressed herein are those of the authors and do not necessarily reflect the views of the U.S. Agency for International Development or the United States Government. *The e-Xtra logo stands for “electronic extra” and indicates that two supplementary figures, three supplementary tables, and one supplementary material file are published online. © 2018 The American Phytopathological Society 1420 PHYTOPATHOLOGY _ rainfall. The focus expansion rate was 1626 m month 1, similar to that in Brazil. Relative HLB incidence in counties with primarily large groves _ increased at a lower rate (0.24 year 1) than in counties with smaller _1 groves in hotspot areas (0.67 year ), confirming reports that large-scale HLB management may slow epidemic progress. Additional keywords: Diaphorina citri, citrus greening, ordinary kriging, area under the disease progress curve, frontal movement, weather effects. latency period in plant hosts (from infection to the first potential new infection) can range from 3 to 4 weeks, taking the psyllid generation time into account (Chiyaka et al. 2012; Lee et al. 2015). The incubation period, from the time of infection until HLB symptom appearance, is much longer than the latency period and can vary from a few months to several years (Shen et al. 2013b). The incubation period also depends on the age and nutritional status of the tree (Shen et al. 2013a) and environmental conditions (Gottwald 2010). Management of HLB has been based primarily on the production of clean nursery stock, removal of symptomatic trees, and intensive use of insecticides (Gottwald 2010; Grafton-Cardwell et al. 2013). However, despite the production of healthy citrus trees in ACP-free nurseries (Halbert et al. 2012), rogueing of symptomatic trees by large citrus growers, and regular insecticide applications to control ACP in Florida, the disease has continued to spread (Shen et al. 2013b). Although regionally coordinated spray programs improved vector control in Florida (Udell et al. 2017), its complete control has not been possible thus far and ACP populations with insecticide resistance have developed (Grafton-Cardwell et al. 2013; Tiwari et al. 2011). Regular and frequent removal of symptomatic trees has slowed HLB spread in very large groves but has not been effective in smaller operations (Bassanezi et al. 2013a, b; Sisterson and Stenger 2013). HLB management could be improved with better understanding of the spatial patterns of its spread in relation to environmental variables (Narouei-Khandan et al. 2016). The spatial distribution of HLB within a grove generally is not random. Initially, affected trees often are more numerous at the edges of a grove (Gottwald et al. 2010; Leal et al. 2010; Luo et al. 2012; Shen et al. 2013b), and subsequent spread develops from these initial incursions (Bassanezi et al. 2005; Gottwald et al. 2007, 2010; Luo et al. 2012). Several methods have been used to describe the spatial distribution patterns of HLB-symptomatic trees (Gottwald et al. 2010; Leal et al. 2010; Luo et al. 2012; Parnell et al. 2011). Spatial dependence of disease occurrence has been analyzed with semivariograms, describing spatial variation in disease incidence as a function of distance (Leal et al. 2010; Shen et al. 2013b). These analyses have shown that the spatial dependence of HLB was greater within than across rows (Shen et al. 2013b). Geostatistical analyses, such as spatial autocorrelation and cluster analyses have been used to discover spatial patterns of disease incidence, and kriging interpolation has been employed to predict disease occurrence in unsampled locations (Bouwmeester et al. 2016; Leal et al. 2010; Luo et al. 2012; Nelson and Boots 2008). Kernel density estimation has been used to identify the location, spatial extent, and intensity of disease and vector hotspots (Bayles et al. 2017; Nelson and Boots 2008; Shimwela et al. 2016). At the regional and global scale, Narouei-Khandan et al. (2016) determined that the distributions of ACP and HLB were mainly related to rainfall, minimum temperature of the coldest month, and mean temperature during the driest quarter. The probability of ACP and HLB occurrence is higher in high rainfall areas, possibly due to the prevalence of young citrus flush during rainy periods (NaroueiKhandan et al. 2016). At a smaller scale, however, adult ACP counts were positively correlated with precipitation in one grove, but not in another, and adult ACP abundance was not correlated with flush counts at the same time (Udell et al. 2017). Las tolerates temperatures up to 35°C (Bové 2014), and the Las concentrations in ACP and citrus trees declined during hot summers with average maximum temperatures of 42°C in Pakistan (Razi et al. 2014). ACP is sensitive to temperatures above 40°C under controlled conditions (Hussain et al. 2016). Temperature relationships have been used to model flush abundance (Gutierrez and Ponti 2013), ACP dynamics and spread (Taylor et al. 2016), and potential HLB development using a combination of simulation modeling and GIS (Aurambout et al. 2009; Gutierrez and Ponti 2013; López-Collado et al. 2013). In another study, flush abundance was simulated with rainfall and temperature data, and ACP dynamics and potential distribution with temperature (Torres-Pacheco et al. 2013). However, rainfall has not been studied sufficiently in relation to the potential distribution of HLB in Florida. The spread of ACP and HLB (as evidenced by the presence of Las in citrus foliage) was estimated from diffusion and gradient models using first-occurrence data from four transects in Mexico (FloresSánchez et al. 2017). Similar first-occurrence data in Florida was monitored by the Florida Department of Agriculture and Consumer Services (FDACS), Division of Plant Industry (DPI), starting with the very first detection of ACP and HLB in 1998 and 2005, respectively. Although this information helped describe the spread of the vector and the disease in general terms (Gottwald et al. 2007; Halbert 2005), the Florida database has not been used to analyze the spatiotemporal distribution of HLB over time. Georeferenced HLB incidence in southern Florida was also monitored by United States Sugar Inc. in Clewiston, Florida. These data were used to validate simulation models (Gottwald et al. 2010; Irey et al. 2011; Parry et al. 2014) but did not include the FDACS-DPI data that cover the whole State. In models, the front of a disease epidemic can move at a constant speed or at increasing speeds depending on the model assumptions about the size and the shape of the distribution kernel, in particular the thickness of the tail (Cunniffe et al. 2015a; Mundt et al. 2009; Parry et al. 2014). The rate of the frontal movement has not been determined for HLB. Yet, spatial and temporal models with specific distribution kernels have been used to examine the potential effects of rogueing of infected, symptomatic trees and psyllid control on HLB management (Cunniffe et al. 2015b; Taylor et al. 2016; Xie et al. 2016). Under the assumption that the epidemic could be slowed down by these practices, millions of citrus trees have been rogued throughout Florida (but not in all groves) and 285,000 tons of insecticides have been applied in citrus groves annually primarily to manage HLB in this State (https://quickstats.nass.usda.gov/). Although these intensive management practices may have slowed the HLB epidemic in large-scale farms (Bassanezi et al. 2013b), additional study of their impact is needed. Considering the above mentioned knowledge gaps about the spatial-temporal spread of HLB in Florida, this study had the following objectives: (i) determine the spatial distribution of HLB occurrence in citrus blocks over time, (ii) estimate the rate of expansion of HLB-positive foci, (iii) relate the distribution of HLBpositive blocks and the relative disease incidence in citrus producing counties to long-term rainfall, relative humidity and temperature patterns, and (iv) compare HLB disease progress in counties with primarily large versus smaller citrus groves in Florida. Using the FDACS-DPI HLB database, we examined the HLB distribution in citrus production areas in Florida (about 150 km × 400 km) over a period of 6 years. Spatiotemporal distributions of HLB data were related to weather data obtained from 27 weather stations of the Florida Automated Weather Network (FAWN) evenly spread over the Florida peninsula. MATERIALS AND METHODS Data collection. The main data used in this study were collected between 2007 and 2012 by staff of the Citrus Health Response Program (CHRP) of FDACS-DPI and of the U.S. Department of Agriculture, Animal and Plant Health Inspection Service (USDAAPHIS), Plant Protection and Quarantine (PPQ) program. The data included information on HLB occurrence from systematic statewide surveys, DPI and USDA inspectors, and occasionally individual homeowners and growers in 19 Florida counties. Ten mature leaf samples from all symptomatic citrus trees per block were sent by express mail to the DPI diagnostic clinic in Gainesville, Florida, to be tested for the presence of Las. The results were pooled per block or per homestead; thus, a block or homestead was considered positive when at least one tree was positive at that location. HLB presence data per citrus block and GPS coordinates of the center of each block with Las-positive trees and to a limited extent (<1%) of GPS units of individual backyard trees were stored in the DPI Plant Pathology Specimen Tracker (PPST) system. This database was handed over to the first author by the second and third authors for geostatistical analysis. Details of the surveys and sample collection methods are presented in the Supplementary Material File S1. Although the PPST database does not contain information about HLB cases in Miami-Dade and Broward counties in the early part of the epidemic (2005 to June 2007), selected HLB cases in those counties in the beginning of 2007, documented in a separate database, were included in the analysis. The disease was firmly established in those counties by 2007 and omitting this information would give a biased impression of the HLB distribution in Florida. The selection procedure for these occurrence data are given in the supplement. Monthly weather data (rainfall, relative humidity, and minimum and maximum temperatures) were downloaded from the FAWN database for the period 2007 to 2012. The FAWN database contains data from 28 weather stations and their GPS units, evenly spread over the Florida peninsula. DNA extraction and PCR protocols. DNA from leaf samples was extracted following USDA protocols (Li et al. 2006). Purified DNA was stored at _20°C for further analysis. Las was detected by conventional PCR using the primer sets OI1/OI2c (Jagoueix et al. 1996) until 2009. Thereafter, qPCR was performed using the USDA protocols of Li et al. (2006). Ct values less than 34 were considered positive. Details of the HLB identification protocols are given in the supplement. Preparation of HLB data for geostatistical and temporal analyses. All data in the PPST database and the additional database for Miami-Dade and Broward counties were checked manually; duplicate and triplicate data were deleted and GPS units were sometimes estimated from physical addresses using Google maps. The original databases together consisted of approximately 31,000 entries, but almost half of them had to be removed during this preparatory work. The remaining HLB-positive cases (mostly citrus Vol. 108, No. 12, 2018 1421 blocks and some individual trees) and GPS units of both databases were combined in one Excel spreadsheet. The number of new cases per year and county are given in Supplementary Table S1. Locations that were found to be HLB-positive in 1 year were considered to be positive in following years, resulting in cumulative data over time. For spatial analysis, these data were preprocessed in ArcGIS 10.2.1, and included projection and coordinate transformation, clipping, masking, feature selection, buffering, grid resampling, and reclassification (Ormsby et al. 2010). Cumulative HLB cases in each of the years from 2007 to 2012 were mapped in ArcGIS 10.2.1 (Fig. 1). Ten foci were identified visually on these maps of cumulative HLB-positive cases (Fig. 2). The visually identified foci coincided with foci identified in GIS by other researchers (S. E. Halbert, unpublished data). The longest diameter of each focus (front) was measured on printed maps and plotted over time (years). For temporal analyses and relational analyses with weather data all HLB-positive cases were summed up per county for each separate year (Supplementary Table S2). In addition to these cumulative HLB cases (mostly blocks), the number of HLB-positive trees per county per year was corrected for the number of citrus trees in a county as obtained from the USDA National Agricultural Statistics (NASS) website (https://www.nass. usda.gov/fl) for 2008, 2009 to 2010, and 2011 to 2012. Since it was not known how many trees were infected per county, the number of HLB cases could not be simply divided by the number of citrus trees in each county. First, the relative number of citrus trees per county was calculated by dividing the number of trees by the highest number of trees (in Hendry County); then, the number of HLB cases in each county and year was adjusted by dividing the cases by the relative tree number in the county. Finally, this adjusted number of HLB cases in each county was divided by the highest adjusted number of HLB cases (in Palm Beach County). Since tree densities were missing for Broward and Miami-Dade counties in the USDANASS database, the same and 1/3 the tree numbers as in Palm Beach County were used for these two counties, respectively. These calculations resulted in a “relative HLB incidence” with a value between zero and one per county and year (Supplementary Table S3). The area under the disease progress curve (AUDPC) was calculated using the midpoint (trapezoidal) integration method for each county from this relative HLB incidence over time (2007 to 2012). The AUDPC values for all counties were subjected to kriging, and the resulting map was compared with kriging maps of various weather data. Geostatistical analysis. Kernel density estimation and ordinary kriging were employed to analyze the pattern of HLB spread in each year of observation. Kernel density estimation was used to identify the location, spatial extent and intensity of HLB hotspots (Nelson and Boots 2008). Details of kernel density estimation and its results are given in the supplement. Ordinary kriging was used to obtain the best linear unbiased estimates of the variables at unsampled locations, s0 (Smith et al. ^ 0) for each variable at any 2007). In ordinary kriging, estimates Z(s location were derived from linear combinations of the neighboring measured values (si) according to the following equation: ^ 0 Þ = åni=1 liZðsiÞ Zðs (1) where Z (si) is the sample value at location i, li is a weight, n is the number of samples, s0 is a prediction location, and n is the number of measured values. This estimation is unbiased and has minimum variance (Johnston et al. 2001). Ordinary kriging assumes a constant but unknown mean and fits a mathematical function to a specified number of points to determine the output values for all surrounding locations as follows: ZðsÞ = µ + eðsÞ (2) where e(s) is a zero mean second-order stationary random field with covariogram function C(h) and variogram g (h) (Johnston et al. 2001). Fig. 1. Distribution of cumulative huanglongbing (HLB) cases identified by the diagnostic clinic of the Florida Department of Agriculture and Consumer Services, Division of Plant Industry at Gainesville, Florida, from 2007 to 2012. New HLB findings in Miami-Dade and Broward counties were relatively low, because most trees were already infected (and removed) in 2005 and 2006. 1422 PHYTOPATHOLOGY Fig. 2. Foci of cumulative huanglongbing (HLB) cases visually identified from the HLB distribution maps from 2007 to 2010 (Fig. 1), projected on the map of 2008. Expanding foci were not detected in Miami-Dade County (southeastern Florida), because HLB was already firmly established in 2007. Ordinary kriging was performed with the geostatistical Analyst tool in ArcGIS 10.3.1. Semivariograms were prepared by fitting the spherical model, which was selected based on an average error close to 0 and a root mean square error standardized (RMSS) close to 1 (Johnston et al. 2001). The anisotropy effect (differences in autocorrelation resulting in different slopes of the spherical model in different directions), also was checked within the GIS program. As ordinary kriging is an exact interpolator, meaning that the predicted value is equal to the measured value at the sampled locations, the predicted spatial distribution could be highly clustered (Johnston et al. 2001). To limit the importance of individual cumulative HLBpositive cases the maximum neighborhood was set to search a radius of 20 points with a minimum of five points. AUDCP values per county (GPS units for centroids) were also subjected to kriging with maximum neighborhood restrictions with a radius of two to five points. In addition to cumulative HLB cases and AUDPC values, statewide average rainfall, and minimum and maximum temperature from 2007 to 2012, calculated from FAWN data, were subjected to kriging using a radius of five points with a minimum of two points around each weather station. Temporal statistical analysis. The longest diameter of each of 10 foci was calculated, plotted, and regressed over time after the Gompertz transformation, Y = _ln[_ln(y)], resulting in values between 0 and 1, assuming that the maximum asymptotic value is 1. Regressing the transformed data over time (year) resulted in a straight line of values, which were then back-transformed to obtain a sigmoid curve with plateau 1. The back-transformed predicted values of the sigmoid curve were multiplied by the actual maximum value to be comparable with the original data. The final predicted values were regressed on the observed values by linear regression and the R2 and MSE were used to evaluate the goodness of fit. The slope of the straight line of transformed data were back-transformed as well to obtain a relative rate of spread, and this relative rate was multiplied by the maximal asymptotic distance to obtain an absolute rate. These calculations were first carried out for individual foci, then for groups of foci (1 to 5 and 6 to 10), and finally for all foci together. The rate of spread could not be calculated for foci in Miami-Dade and Broward counties as the disease was already firmly established there in 2007. The relative HLB incidence was used to calculate the AUDPC per county. The AUDPC was regressed on the estimated mean rainfall, minimum and maximum temperature for each county based on the maps obtained by kriging of the FAWN data. PROC GLM was used in SAS version 9.4 to test effects of individual weather variables and their interactions on AUDPC. Studentized residual values were tested for normality using the Shapiro-Wilk test in PROC Univariate Normal Plot. Residual values were plotted versus observed values and inspected for trends. Stepwise regression was used to determine the relative effects of the individual weather variables. Mean relative HLB incidence per year was calculated for (i) Hendry and Collier counties (with primarily large citrus farms of >1,000 ha); (ii) Hardee, Glades, Highlands, and Desoto counties (HLB hotspot counties as determined by kernel density analysis, Supplementary Fig. S1; with small-medium size groves of 40 to 500 ha); and (iii) remaining counties (with a wide array of grove sizes). The classification was made prehoc, because differences in HLB development between large-scale and small-scale citrus groves had been published (Bassanezi et al. 2013a, b; Sisterson and Stenger 2013). Disease progress curves were determined for each group of counties using the Gompertz transformation and linear regression in SAS version 9.4. The parameter estimates were compared using t tests, and the relative rate parameters were back-transformed and multiplied with the observed asymptotic value to enable comparisons of the relative and absolute rates with those in the literature. RESULTS Since its discovery in Miami-Dade County in 2005, HLB had spread to 12 counties in southern and middle Florida by 2007 (Fig. 1). In the first 2 years of our study (2007 to 2008) six foci (1 to 6) could be distinguished (Fig. 2). Thereafter, these foci became larger and new foci (7 to 10) appeared (Fig. 2). The diameter of each focus increased over time, mostly in north-to-south direction, seemingly along roads (Fig. 1). Disease progress varied slightly among foci and continued expanding until foci merged. On average, the focus diameter increased asymptotically according to a Gompertz function (Fig. 3), with a relative rate of 0.56 year_1 and an absolute rate of 19.5 km year_1 (1,626 m month_1). There were no significant Fig. 3. Averages and standard errors of the longest diameter of foci (n = 10) of huanglongbing (HLB)-affected trees in commercial citrus groves in Florida between 2007 and 2012. The diameter of the foci was assumed to be 0 km in 2005, when the disease was first detected in residential properties close to Miami, FL. Average diameters are indicated by dots and back-transformed diameters predicted from the Gompertz model by a solid line. Fig. 4. Kriging-predicted distribution of cumulative huanglongbing (HLB) cases in the main commercial citrus production areas in Florida from 2007 to 2012 (Fig. 1 provides original data). The probability of HLB findings in Miami-Dade and Broward counties was relatively low, because most trees were already infected (and removed) in 2005 and 2006. Vol. 108, No. 12, 2018 1423 differences between the mean expansion rates (± standard errors) of foci 1 to 5 (17.9 ± 10.4 km year_1) and foci 6 to 10 (21.3 ± 13.5 km year_1). A paired t test for the two groups of foci on the predicted values in 7 years showed that the values were not significantly different (P = 0.84). Kriging indicated that in 2007 the main concentrations of HLBaffected trees in Florida were on the southwestern and mideastern coasts but shifted to the south-center in 2008 and expanded around that area (Fig. 4). Kernel density analysis also showed that hotspots of HLB cases were at the western and eastern coasts at first but moved to the south middle of the peninsula later. The concentrations of HLB-diseased trees were seemingly low in the southeast (Miami-Dade and Broward counties) in that period (2007 to 2012), because most citrus trees had been removed already in 2005 and 2006. The distribution patterns of HLB cases (as determined by kriging or kernel density analysis) did not coincide with the pattern of maximum temperature and only slightly with that of minimum temperature (Fig. 5). The distribution pattern of HLB cases did not coincide with the rainfall distribution pattern either (Fig. 6B). The results from kriging of HLB cases (Fig. 4) may reflect the number of citrus trees in an area rather than HLB incidence, because more samples were collected in areas with many citrus groves than in areas with fewer groves. To avoid the confounding problem with the number of citrus trees in an area, the relative HLB incidence (relative to citrus tree number per county and the highest HLB incidence) was calculated for each county and each year. The AUDPC of relative HLB incidence was highest in Miami-Dade, Broward, and Palm Beach counties along the southeastern coast and declined inland (Fig. 6A). The spatial distribution of AUDPC largely coincided with the rainfall distribution in the Florida peninsula, except for low predicted AUDPC values along the Gulf of Mexico, where commercial citrus is not produced (Fig. 6B). The AUDPCs in 21 counties were positively correlated with mean annual rainfall over the 6 years of this study (r = 0.83, P = 0.001) and with the annual mean minimum temperature (r = 0.75, P = 0.001) (Supplementary Fig. S2). General linear model analysis of AUDPC versus rainfall and maximum and minimum temperatures and their interactions indicated that rainfall, minimum temperature and their interaction contributed significantly to explaining the variability (P = 0.0001; residuals normally distributed). In stepwise multiple regression rainfall was the only variable selected that contributed significantly to explaining the variation in AUDPC (P = 0.005; residuals normally distributed). The relative disease progress (relative HLB incidence over time) was highest in counties with consistent hotspots during the study (Hardee, Glades, Highlands, and Desoto counties, which had mostly medium size groves) compared with two large-grove counties (Hendry and Collier) and the rest of the counties with smaller groves (Fig. 7). The relative disease progression rate, derived from the Gompertz equation, was significantly (P = 0.05) higher in the hotspot counties (0.67 year_1) than in the large-grove counties (0.24 year_1), as listed above. However, the relative disease progress rate in the rest of the counties with smaller groves (0.42 year_1) did not differ significantly from those in the other two groups of counties. Fig. 5. Ranges in A, maximum and B, minimum temperatures obtained from kriging of average weather data (2007 to 2012) of the Florida Agricultural Weather Network. The locations of weather stations are indicated by dots. 1424 PHYTOPATHOLOGY This lack of significance was due to the large variation among the counties with the smaller groves. DISCUSSION In this paper, survey data for regulatory purposes were used to estimate regional spread of HLB in Florida. Most data were collected according to a risk-based sampling protocol. These data may not be ideal for the analysis of spatial spread of HLB, because they do not reflect random samples and do not include negative data. However, other regional survey data are not freely available for Florida and would be very difficult to collect using random sampling, as about 30,000 samples were needed to obtain reasonable estimates of spatial distributions over time. Similar survey data meant for regulatory purposes were used for spatial analyses in Brazil (Bassanezi et al. 2005; Gottwald et al. 2006) and Mexico (Flores-Sánchez et al. 2017). The very large numbers of samples in these and in our data sets, as well as the high intensity and widespread occurrence of HLB toward the end of the sampling period, minimizes the risk of bias in the data sets. Moreover, it is now clear that the whole of Florida is suitable for ACP and HLB development, but regional differences in intensity due to small differences in climate remain despite the widespread occurrence of HLB throughout the Florida peninsula (Narouei-Khandan et al. 2016, https://www.freshfromflorida.com/Divisions-Offices/PlantIndustry/Agriculture-Industry/Citrus-Health-Response-Program/ Citrus-Quarantine-and-Disease-Detection-Maps). Finally, the change in diagnostic protocol in 2009 was unlikely to affect the detection efficiency of Las in HLB symptomatic tissues, as conventional and real-time PCR methods are equally reliable to confirm the presence of Las in samples from symptomatic trees (Li et al. 2007), although real-time qPCR was 10- to 100-fold more sensitive than conventional PCR (Li et al. 2007). Thus, later samples tested by qPCR may have overestimated HLB presence compared with earlier samples tested by conventional PCR. Despite these limitations, this research resulted in three major findings: (i) confirmation of a positive relationship between HLB relative incidence and rainfall distribution patterns, (ii) the Fig. 6. Kriging-estimated ranges of A, areas under the disease progress curves of relative huanglongbing (HLB) incidence from 2007 to 2012 and of B, rainfall over the same period obtained from the Florida Agricultural Weather Network (locations of weather stations indicated by dots). The relative disease incidence was calculated from the number of HLB cases in each county detected by Florida Department of Agriculture and Consumer Services, Division of Plant Industry divided by the relative number of trees in those counties according to the USDA-NASS website and related to the highest incidence calculated (in Palm Beach County). The number of reported cases in Miami-Dade and Broward counties had already declined in 2007 compared with Palm Beach County. Vol. 108, No. 12, 2018 1425 identification of foci of HLB cases from where the disease expanded into surrounding citrus groves, and estimation of the expansion rates of the foci, and (iii) demonstration of differences in relative disease progress rates among counties with large versus smaller citrus groves. Some of these findings were possible because of the approach that was used to analyze spatial occurrence data that had not been used previously. Spatial regional occurrence data of HLB cases in Florida were analyzed after calculating cumulative cases over time and adjusting these data for local citrus tree densities. The use of cumulative data is common for the calculation of disease progress curves, where dead plants continue to be considered as diseased at the highest severity level (Forbes et al. 2014). Cumulative data were also used by Flores-Sánchez et al. (2017) for HLB occurrence in Mexico. Although HLB-affected trees may have been rogued, they could have contributed to disease progress before their removal since the latent period for HLB is as short as 3 weeks and its incubation period can span several years (Lee et al. 2015; Shen et al. 2013b). Thus, the use of cumulative data is justified. Without adjustment for the number of citrus trees, the apparent clustering of HLB cases partly reflects the spatial distribution and sampling of citrus trees. The distribution of HLB cases regardless of number of citrus trees in an area was not related to regional differences in weather. However, the distribution of AUDPC, based on HLB cases adjusted for tree number (relative HLB incidence per county), reflected regional rainfall patterns with higher rainfall in the southeastern than in the middle and northern parts of the Florida peninsula. The relationship with rainfall was obtained despite the use of the less sensitive PCR method for earlier samples in the southeast, possibly resulting in an underestimation of the real regional differences in HLB incidence due to differences in rainfall. Narouei-Khandan et al. (2016) reported relationships between the occurrence of HLB and ACP and rainfall at a global scale, even though their HLB and ACP data were not adjusted for citrus tree density. However, since the range of rainfall intensity is greater at the global scale than at smaller regional scales, relationships are more easily found at the global scale. Fig. 7. Mean relative incidence (and standard errors) of huanglongbing (HLB) in counties with commercial citrus groves from 2007 to 2012. The relative incidence was calculated from the number of HLB cases in each county detected by Florida Department of Agriculture and Consumer Services, Division of Plant Industry divided by the relative number of trees in those counties according to the U.S. Department of Agriculture-National Agricultural Statistics website and related to the highest incidence calculated (in Palm Beach county). Hendry and Collier counties have primarily large groves (>1,000 ha); hotspot counties (Hardee, Glades, Highlands, and Desoto counties) were identified by kernel density analysis of HLB cases; and other counties consisted of the rest of the counties with commercial citrus groves in Florida. The Gompertz transformation and back-transformation were used to create the disease progress curves. The estimated relative epidemic rates were 0.67 ± _ 0.33, 0.42 ± 0.26, and 0.24 ± 0.14 year 1 for hotspot counties, residual counties, and large-grove counties, respectively. The slopes of the regression equations differed significantly (P = 0.05) between the large-grove counties and the hotspot counties, but not between the rest of the counties and either hotspot or large-grove counties according to t tests. 1426 PHYTOPATHOLOGY The positive relationship with rainfall, but not, or less so, with maximum and minimum temperature, was reinforced by detailed HLB and ACP observations in two smaller scale studies in Florida (M. M. Shimwela, S. E. Halbert, M. L. Keremane, P. Mears, B. H. Singer, W. S. Lee, J. B. Jones, R. R. Ploetz, and A. H. C. van Bruggen, unpublished data). In earlier Florida studies, the abundance of Laspositive ACP populations (Irey et al. 2011; Parkunan et al. 2011) or ACP populations in general (Tsai et al. 2002) were highest during summer and fall, when rainfall and minimum temperatures were relatively high. Elsewhere, ACP populations and Las titers in ACP were also positively correlated with rainfall, rather than temperature or relative humidity (Teck et al. 2011). This may be due to the physiological response of citrus trees to rain (wetting the deep roots) resulting in movement of Las from deep roots and accumulation in the foliage, where the highest Las concentrations were found in summer and fall (Parkunan et al. 2011, Irey et al. 2011). On the contrary, soil wetting by irrigation may be mostly shallow (5 to 15 cm) depending on irrigation frequency and intensity (Kadyampakeni and Morgan 2017), and superficial roots with Las may be exposed to relatively high temperatures resulting in relatively low Las densities at those depths (Doud et al. 2017). However, the difference in Las concentrations in roots exposed to deep versus shallow wetting has not been investigated as far as we know. Despite positive correlations between rainfall and the regional occurrence of HLB, rainfall is currently not included as environmental factor in risk-based survey methods (Gottwald et al. 2014, Parnell et al. 2014). In the future, rainfall could be a useful parameter when assessing risks in areas with highly variable rainfall, such as California (Narouei-Khandan et al. 2016). ACP and HLB were reported first in Southeast Florida (MiamiDade County), and later in the West and North. The initial sightings of ACP and HLB in the state may not be related solely to the sites of initial introduction but to the highly suitable environmental conditions in the area (including high rainfall) for establishment of the disease and its vector. In California, HLB was found first in the Los Angeles area (Kumagai et al. 2013), which has higher precipitation than the in-land valleys. As in Florida, the location of the first HLB sighting may have been related more to the suitable environment for HLB development rather than proximity to the introduction sites (Narouei-Khandan et al. 2016). In the first years of the study period, individual foci could be distinguished in southern and central Florida. These foci may have originated from Las-infected trees or ACP that were moved from the Southeast to various locations in Florida (Gottwald et al. 2006, Halbert et al. 2010, Halbert et al. 2012). Alternatively, Las-infected ACP could have been transported on Murraya exotica, an ornamental plant that was propagated in high numbers in southern Miami-Dade County at the same time when Las was first observed there (Halbert et al. 2012). These plants are susceptible to infection but do not show symptoms of disease and were sold throughout the state in garden centers and for landscaping. The front of the identified foci expanded in the following years until neighboring foci coalesced. The estimated rate of frontal movement was 1.63 m month_1 or 19.5 km year_1. This estimate is very similar to the estimated frontal expansion rate in Brazil (19.3 km year_1) (Gottwald et al. 2006). We believe that evaluating the expansion of individual foci in Florida, rather than the statewide spread of HLB from south to north, provided a better estimate of frontal movement and ensured that frontal rates were not overestimated. The frontal HLB expansion rates within citrus groves are generally much smaller (M. M. Shimwela, S. E. Halbert, M. L. Keremane, P. Mears, B. H. Singer, W. S. Lee, J. B. Jones, R. R. Ploetz, and A. H. C. van Bruggen, unpublished data) than the expansion rates of regional foci suggesting the existence of more than one dispersal mechanism. Differences in dispersal mechanisms would complicate spatiotemporal distribution models at increasing scales (Cunniffe et al. 2015a; Gosme and Lucas 2009; Lee et al. 2015). Finally, our approach to the analysis of large-scale data allowed us to compare HLB disease progress in counties with primarily large (>1,000 ha) groves to counties with progressively smaller groves. The estimated ‘relative disease incidence’ increased at a lower relative rate (0.24 year_1) in counties with primarily large groves than in counties with smaller groves in hotspot areas (0.67 year_1). However, the relative disease progress rate in counties with smaller groves outside of hotspot areas (0.42 year_1) was not significantly different from the other groups of counties due to the large variation among counties. These rates are similar to those calculated from Gompertz models for the incidence of HLB symptomatic plants based on a 100% census in large citrus groves in South Florida (Gottwald et al. 2010), justifying our method of calculating ‘relative disease incidence’. The management practices in large groves include consistent and coordinated insecticide applications and HLB-infected tree removal (Bassanezi et al. 2013b). During the period of our study, tree removal was not universal among smaller scale farms and insecticide applications were not coordinated. Controlled studies in Brazil (Bassanezi et al. 2013a, b) and a simulation study in the United States (Sisterson and Stenger 2013) showed that tree removal would reduce HLB epidemic disease progress only in large groves (larger than about 1,000 ha) at the beginning of the epidemic. Epidemiological models confirm that culling of infected trees might be effective at low infection rates early in an epidemic (Cunniffe et al. 2015b). However, the extensive culling efforts for HLB control in Florida have not been effective (Grafton-Cardwell et al. 2013), possibly due to the mosaic of many smaller scale groves and the favorable conditions for asymptomatic spread of the disease (Lee et al. 2015). In recent years, smaller groves have coordinated insecticide applications (Grafton-Cardwell et al. 2013), but the extent to which ACP control could reduce infection of new flushes and subsequent symptom development is not well known, especially when practically all trees are infected by Las (Shen et al. 2013a). Despite the important findings of these studies, there are a few additional caveats. As mentioned above, the statewide surveys by FDACS-DPI and USDA-APHIS-PPQ were carried out to monitor the progress of HLB and to document incidence for regulatory purposes. Despite about 30,000 entries in this database, the data are incomplete as citrus managers sent samples to be tested for Las to two other labs in South Florida, at the Southwest Florida Research and Education Center of the University of Florida and United States Sugar Corporation/Southern Gardens Citrus, Clewiston, FL. The databases maintained by those labs also contain many thousands of entries and cover a large part of southern Florida (Irey et al. 2011). It is possible that the relative HLB incidence calculated in this study, particularly for Hendry and Collier counties, is an underestimate because samples from these areas may have been sent preferentially to the other labs. Similarly, the databases from the other labs would be incomplete without the data from the surveys used in this study. Ideally, the three databases would be merged to enable a complete analysis of the HLB distribution in Florida over time and in space. Despite these caveats, the results clearly demonstrate the importance of rainfall for HLB risk models and the need for different, scaledependent dispersal kernels in analytical or simulation models (Cunniffe et al. 2015b; Lee et al. 2015). Our Statewide study showed that at least 10 foci were established in the beginning of the epidemic in Florida, probably due to the movement of Murraya nursery plants (Halbert et al. 2012). Thus, network models (Cunniffe et al. 2015a; Jeger et al. 2007) may be more useful than analytical or simulation models when analyzing HLB epidemics at the largest scales. ACKNOWLEDGMENTS We thank the DPI staff K. Richards, J. Carter, M. Couture, and J. Gilbert as well as T. Riley of the USDA for carrying out the statewide surveys; and K. Shin and S. Timilsina for providing independent reviews of the manuscript. LITERATURE CITED Ammar, E., Hall, D. G., and Shatters, R. G. 2016. Ultrastructure of the salivary glands, alimentary canal and bacteria-like organisms in the Asian citrus psyllid, vector of citrus huanglongbing disease bacteria. J. Microsc. Ultrastruct. 5:9-20. Aurambout, J. P., Finlay, K. J., Luck, J., and Beattie, G. A. C. 2009. A concept model to estimate the potential distribution of the Asiatic citrus psyllid Diaphorina citri Kuwayama in Australia under climate change—A means for assessing biosecurity risk. Ecol. Modell. 220:2512-2524. Bassanezi, R. B., Belasque, J., Jr., and Montesino, L. H. 2013a. Frequency of symptomatic trees removal in small citrus blocks on citrus huanglongbing epidemics. Crop Prot. 52:72-77. Bassanezi, R. B., Busato, L. A., Bergamin Filho, A., Amorim, L., and Gottwald, T. R. 2005. Preliminary spatial pattern analysis of huanglongbing in São Paulo, Brazil. Pages 341-355 in: Sixteenth IOCV Conference, Monterrey, Mexico. Bassanezi, R. B., Montesino, L. H., Gimenes-Fernandes, N., Yamamoto, P. T., Gottwald, T. R., Amorim, L., and Filho, A. B. 2013b. Efficacy of area-wide inoculum reduction and vector control on temporal progress of huanglongbing in young sweet orange plantings. Plant Dis. 97:789-796. Bayles, B. R., Thomas, S. M., Simmons, G. S., Grafton-Cardwell, E. E., and Daugherty, M. P. 2017. Spatiotemporal dynamics of the Southern California Asian citrus psyllid (Diaphorina citri) invasion. PLoS One 12:e0173226. Bouwmeester, H., Heuvelink, G. B. M., and Stoorvogel, J. J. 2016. Mapping crop diseases using survey data: The case of bacterial wilt in bananas in the East African highlands. Eur. J. Agron. 74:173-184. Bové, J. 2014. Heat-tolerant Asian HLB meets heat-sensitive African HLB in the Arabian Peninsula! Why? J. Citrus Pathol. 1(1). https://escholarship.org/ uc/item/1665n4x9 Chiyaka, C., Singer, B. H., Halbert, S. E., Morris, J. G., and van Bruggen, A. H. C. 2012. Modeling huanglongbing transmission within a citrus tree. Proc. Nat. Acad. Sci. 109:12213-12218. Cunniffe, N. J., Koskella, B., Metcalf, C. J. E., Parnell, S., Gottwald, T. R., and Gilligan, C. A. 2015a. Thirteen challenges in modelling plant diseases. Epidemics 10:6-10. Cunniffe, N. J., Stutt, R. O. J. H., DeSimone, R. E., Gottwald, T. R., and Gilligan, C. A. 2015b. Optimizing and communicating options for the control of invasive plant disease when there is epidemiological uncertainty. PLOS Comput. Biol. 11:e1004211. da Graça, J. V., Douhan, G. W., Halbert, S. E., Keremane, M. L., Lee, R. F., Vidalakis, G., and Zhao, H. 2016. Huanglongbing: An overview of a complex pathosystem ravaging the world’s citrus. J. Integr. Plant Biol. 58:373-387. Doud, M. M., Wang, Y., Hoffman, M. T., Latza, C. L., Luo, W., Armstrong, C. M., Gottwald, T. R., Dai, L., Luo, F., and Duan, Y. 2017. Solar thermotherapy reduces the titer of Candidatus Liberibacter asiaticus and enhances canopy growth by altering gene expression profiles in HLB-affected citrus plants. Hortic. Res. 4:17054. Farnsworth, D., Grogan, K. A., van Bruggen, A. H. C., and Moss, C. B. 2014. The potential economic cost and response to greening in Florida citrus. Choices 29:1-6. Flores-Sánchez, J. L., Mora-Aguilera, G., Loeza-Kuk, E., López-Arroyo, E. I., Gutiérrez-Espinosa, M. A., Vélazquez-Monreal, J. J., Domı́nguez-Monge, S., Bassanezi, R. B., Acevedo-Sánchez, G., and Robles-Garcı́a, P. 2017. Diffusion model for describing the regional spread of huanglongbing from first-reported outbreaks and basing an area-wide disease management strategy. Plant Dis. 101:1119-1127. Forbes, G. A., Perez, W., and Andrade-Piedra, J. 2014. Field assessment of resistance in potato to Phytophthora infestans. International Cooperators Guide. International Potato Center (CIP), Lima, Peru. http://cipotato.org/ wp-content/uploads/2014/06/006154.pdf Gosme, M., and Lucas, P. 2009. Cascade: An epidemiological model to simulate disease spread and aggregation across multiple scales in a spatial hierarchy. Phytopathology 99:823-832. Gottwald, T. R. 2010. Current epidemiological understanding of citrus huanglongbing. Annu. Rev. Phytopathol. 48:119-139. Gottwald, T. R., Bergamin-Filho, A., Bassanezi, R., and Amorim, L., Irey, M., Zhao, X., and Auberi, B. 2006. Concepts in Huanglongbing Epidemiology. The International Workshop for Prevention of Citrus Greening Disease in Severely Infested Areas. Gottwald, T. R., da Graça, J. V., and Bassanezi, R. B. 2007. Citrus Huanglongbing: The pathogen and its impact. Plant Health Prog. 8:1. Gottwald, T. R., Irey, M. S., Gast, T., Parnell, S. R., Taylor, E. L., and Hilf, M. E. 2010. Spatio-temporal analysis of an HLB epidemic in Florida and implications for spread. Pages 84-97 in: Proc. 17th Conference IOCV, Adana, Turkey. Gottwald, T. R., Luo, W., and McRoberts, N. 2014. Risk-based residential HLB/ACP survey for California, Texas and Arizona. J. Citrus Pathol. 1: 121-125. http://www.plantmanagementnetwork.org/edcenter/seminars/outreach/Citrus/HLB/presentation.html Vol. 108, No. 12, 2018 1427 Grafton-Cardwell, E. E., Stelinski, L. L., and Stansly, P. A. 2013. Biology and management of Asian citrus psyllid, vector of huanglongbing pathogens. Annu. Rev. Entomol. 58:413-432. Gutierrez, A. P., and Ponti, L. 2013. Prospective analysis of the geographic distribution and relative abundance of Asian citrus psyllid (Hemiptera: Liviidae) and citrus greening disease in North America and the Mediterranean Basin. Fla. Entomol. 96:1375-1391. Halbert, S. E. 2005. The discovery of huanglongbing in Florida. Pages 7-11 in: Proc. 2nd International Citrus Canker and Huanglongbing Res. Workshop, Orlando, FL. Halbert, S. E., Manjunath, K., Ramadugu, C., and Lee, R. F. 2012. Incidence of huanglongbing-associated Candidatus Liberibacter asiaticus’ in Diaphorina citri (Hemiptera: Psyllidae) collected from plants for sale in Florida. Fla. Entomol. 95:617-624. Halbert, S. E., Manjunath, K. L., Ramadugu, C., Brodie, M. W., Webb, S. E., and Lee, R. F. 2010. Trailers transporting oranges to processing plants move Asian citrus psyllids. Fla. Entomol. 93:33-38. Hall, D. G., Richardson, M. L., Ammar, E.-D., and Halbert, S. E. 2012. Asian citrus psyllid, Diaphorina citri, vector of citrus huanglongbing disease. Entomol. Experim. Applic. 146:207-223. Hussain, M., Lin, Y., and Wang, L. 2016. Effect of temperature on longevity of Diaphorina citri (Hemiptera: Liviidae) studied by microcalorimeter. J. Therm. Anal. Calorim. 127:1245-1252. Irey, M. S., Gast, T., Cote, J., Gadea, P., Santiago, O., Briefman, L., and Graham, J. H. 2011. Seasonal variability in HLB testing data in plant and psyllid samples in Florida. Second International Research Conference Huanglongbing (IRCHLB), Orlando, FL. http://www.plantmanagementnetwork. org/proceedings/irchlb/2011/ Jagoueix, S., Bové, J. M., and Garnier, M. 1996. PCR detection of the two ‘Candidatus’ Liberobacter species associated with greening disease of citrus. Mol. Cell. Probes 10:43-50. Jeger, M. J., Pautasso, M., Holdenrieder, O., and Shae, M. W. 2007. Modelling disease spread and control in networks: Implications for plant sciences. New Phytol. 174:279-297. Johnston, K., VerHoef, J. M., Krivoruchko, K., and Lucas, N. 2001. Using ArcGIS Geostatistical Analyst. ESRI, New York. Kadyampakeni, D. M., and Morgan, K. T. 2017. Irrigation scheduling and soil moisture dynamics influence water uptake by Huanglongbing affected trees. Sci. Hortic. (Amsterdam) 224:272-279. Kumagai, L. B., LeVesque, C. S., Blomquist, C. L., Madishetty, K., Guo, Y., Woods, P. W., Rooney-Latham, S., Rascoe, J., Gallindo, T., Schnabel, D., and Polek, M. 2013. First report of Candidatus Liberibacter asiaticus associated with citrus Huanglongbing in California. Plant Dis. 97:283. Leal, R. M., Barbosa, J. C., Gregolin Costa, M., Belasque, J., Jr., Yamamoto, P. T., and Dragone, J. 2010. Spatial distribution of huanglongbing (greening) on citrus using geostatistics.Rev. Brasil. Fruticult. 32:808-818. Lee, J. A., Halbert, S. E., Dawson, W. O., Robertson, C. J., Keesling, J. E., and Singer, B. H. 2015. Asymptomatic spread of huanglongbing and implications for disease control. Proc. Natl. Acad. Sci. 112:7605-7610. Li, W., Hartung, J. S., and Levy, L. 2007. Evaluation of DNA amplification methods for improved detection of ‘Candidatus Liberibacter species’ associated with citrus huanglongbing. Plant Dis. 91:51-58. Li, W., Levy, L., and Hartung, J. S. 2006. Quantitative real-time PCR for detection and identification of Candidatus Liberibacter species associated with citrus huanglongbing. J. Microbiol. Methods 66:104-115. López-Collado, J., López-Arroyo, J. I., Robles-Garcı́a, P. L., and Márquez-Santos, M. 2013. Geographic distribution of habitat, development, and population growth rates of the Asian citrus psyllid, Diaphorina citri, in Mexico. J. Insect Sci. 13:114. Luo, W., Pietravalle, S., Parnell, S., van den Bosch, F., Gottwald, T. R., Irey, M. S., and Parker, S. R. 2012. An improved regulatory sampling method for mapping and representing plant disease from a limited number of samples. Epidemics 4:68-77. Mundt, C. C., Sackett, K. E., Wallace, L. D., Cowger, C., and Dudley, J. P. 2009. Aerial dispersal and multiple-scale spread of epidemic disease. EcoHealth 6:546-552. Narouei-Khandan, H. A., Halbert, S. E., Worner, S. P., and van Bruggen, A. H. C. 2016. Global climate suitability of citrus huanglongbing and its vector, the Asian citrus psyllid, using two correlative species distribution 1428 PHYTOPATHOLOGY modeling approaches, with emphasis on the USA. Eur. J. Plant Pathol. 144: 655-670. Nelson, T. A., and Boots, B. 2008. Detecting spatial hotspots in landscape ecology. Ecography 31:556-566. Ormsby, T., Napoleon, E., Burke, R., Feaster, L., Groessl, C., and Bowden, L. 2010. Getting to Know ArcGIS Desktop: Updated for ArcGIS 10. 2nd ed. ESRI Press, Redlands, CA. Parkunan, V., Ebert, T., Wang, N., Rogers, M., and Dewdney, M. 2011. Seasonal variation of Candidatus Liberibacter asiaticus in citrus branches and in vector, Diaphorina citri, in central Florida sweet orange groves. Phytopathology 101(suppl.):S138. Parnell, S., Gottwald, T. R., Irey, M. S., Luo, W., and van den Bosch, F. 2011. A stochastic optimization method to estimate the spatial distribution of a pathogen from a sample. Phytopathology 101:1184-1190. Parnell, S., Gottwald, T. R., Riley, T., and van den Bosch, F. 2014. A generic risk-based surveying method for invading plant pathogens. Ecol. Appl. 24: 779-790. Parry, M., Gibson, G. J., Parnell, S., Gottwald, T. R., Irey, M. S., Gast, T. C., and Gilligan, C. A. 2014. Bayesian inference for an emerging arboreal epidemic in the presence of control. Proc. Nation. Acad. Sci. 111: 6258-6262. Razi, M. F., Keremane, M. L., Ramadugu, C., Roose, M., Khan, I. A., and Lee, R. F. 2014. Detection of citrus Huanglongbing-associated ‘Candidatus Liberibacter asiaticus’ in citrus and Diaphorina citri in Pakistan, seasonal variability, and implications for disease management. Phytopathology 104: 257-268. Shen, W., Cevallos-Cevallos, J. M., Nunes da Rocha, U., Stansly, P. A., Roberts, P. D., and van Bruggen, A. H. C. 2013a. Relation between plant nutrition, hormones, insecticide applications, bacterial endophytes, and Candidatus Liberibacter Ct values in citrus trees infected with huanglongbing. Eur. J. Plant Pathol. 137:727-742. Shen, W., Halbert, S. E., Dickstein, E., Manjunath, K. L., Shimwela, M. M., and van Bruggen, A. H. C. 2013b. Occurrence and in-grove distribution of citrus huanglongbing in north central Florida. J. Plant Pathol. 95:361-371. Shimwela, M. M., Narouei Khandan, H. A., Halbert, S. E., Keremane, M. L., Minsavage, G. V., Timilsina, S., Massawe, D. P., Jones, J. B., and van Bruggen, A. H. C. 2016. First occurrence of Diaphorina citri in East Africa, characterization of the Ca. Liberibacter species causing huanglongbing (HLB) in Tanzania, and potential further spread of D. citri and HLB in Africa and Europe. Eur. J. Plant Pathol. 146:349-368. Sisterson, M. S., and Stenger, D. C. 2013. Rogueing with replacement in perennial crops: Conditions for successful disease management. Phytopathology 103:117-128. Smith, M. J., Goodchild, M. F., and Longley, P. A. 2007. Geospatial Analysis: Comprehensive Guide to Principles, Techniques and Software Tools. Matador, Leicester, UK. Taylor, R., Mordecal, E., Gilligan, C. A., Rohr, J., and Johnson, L. 2016. Mathematical models are a powerful method to understand and control the spread of Huanglongbing. PeerJ Preprints 4:e2059v1. Teck, S. L. C., Fatimah, A., Beattie, A., Heng, R. K. J., and King, W. S. 2011. Seasonal population dynamics of the Asian citrus psyllid, Diaphorina citri Kuwayama in Sarawak. Am. J. Agric. Biol. Sci. 6:527-535. Tiwari, S., Mann, R. S., Rogers, M. E., and Stelinski, L. L. 2011. Insecticide resistance in field populations of Asian citrus psyllid in Florida. Pest Manag. Sci. 67:1258-1268. Torres-Pacheco, I., López-Arroyo, J., Aguirre-Gómez, J., Guevara-González, R., Yänez-López, R., Hernández-Zul, M., and Quijano-Carranza, J. 2013. Potential distribution in Mexico of Diaphorina citri (Hemiptera: Psyllidae). Vector Huanglongbing Pathogen. Fla. Entomol. 96:36-47. Tsai, J. H., Wang, J. J., and Liu, Y. H. 2002. Seasonal abundance of the Asian citrus psyllid, Diaphorina citri (Homoptera: Psyllidae) in southern Florida. Fla. Entomol. 85:446-451. Udell, B. D., Monzo, C., Paris, T. M., Allan, S. A., and Stansly, P. A. 2017. Influence of limiting and regulating factors on populations of Asian citrus psyllid and the risk of insect and disease outbreaks. Ann. Appl. Biol. 171: 70-88. Xie, F., Gao, S., Yu, D., and Zhang, F. 2016. A citrus huanglongbing model with varying pulse rogueing and general incidence. J. Adv. Appl. Math. 1: 222-233.