Concepts in Huanglongbing Epidemiology

'Gottwald, T. R., 2Bergamin-Filho, A. 3Bassanezi, R., 2Amorim, L., 41rey, M, 5Zhao, X, and 6Auberi, B.'USDA, ARS, US Horticultural Research Laboratory, Fort Pierce, Florida, USA; 2ESALQ,

Universidade de São Paulo, Brazil; 3Fundecitrus, Araraquara, Brazil; 4 U Sugar Corporation,Clewiston, Florida, Citrus Research Institute, Chinese Academy ofAgricultural Sciences, Beibei,

Chongquing, China, and 6C1RAD-FLHOR, Montpellier, France ('retired)tgoltwaId('iIushrl. ars. usda.gov

AbstractHuanglongbing (FILB) was discovered in Brazil and Florida in 2004 and 2005 respectively.

Previously, very few quantitative epidemiological studies had been conducted, and thus the increaseand spread of the disease remains incompletely characterized. The perennial nature of the diseasenecessitates a dedication to data collection over multiple years, to obtain spatio-temporal datasufficient to understand increase and spread of the disease. In most areas, removing HLB-diseasedtrees in the attempt to control or contain the disease is one of the main disease mitigation strategies.This makes tracking disease increase and spread very difficult or impossible. This limitation has beencoupled with the inability, until recently, of detection by PCR, requiring a reliance on visualassessment for disease monitoring. Monitoring the occurrence of HLB by visual symptoms alone isproblematic. The lag in time between transmission by psyllid vectors or propagation and the onset ofsymptoms can be quite variable and quantifying the severity of disease symptoms in individual treesis not a true indication of pathogen incidence. Additionally, due to the temporal variation in symptomexpression, trees infected at the same time may express symptoms over one or more years.Considering the above caveats, epidemiological data has been collected from a few locations inReunion Island (RI), southern China (LCF, LART), Taiwan, The Philippines and more recently in SãoPaulo, Brazil and southern Florida, USA. However, to date an assessment of disease progression hasonly been accomplished in China and Reunion Island and resulted in an estimated reduced longevityof HLB-infected sweet orange and mandarin groves. HLB epidemics are multiyear in duration, butrarely progress to an asymptote before removal of the planting occurs. Therefore, both theexponential and the logistic models adequately described disease progress over time. In the LCF plot,an asymptote was reached after 13 yrs. However, for groves that became unproductive, diseaseincidence never reached asymptotic levels prior to removal, but logistic model predictions calculatedthat HLB would reach asymptotic disease after 7 and 13 years.Spatial analysis of the incidence of HLB visual symptoms has been less problematic and wasundertaken to better understand the relationships among infected trees near to and at distance fromeach other and thereby gain some understanding of vector spread of HLB. Combined interpretationsof spatial analyses indicate two spatial mechanisms of vector spread of HLB, within local areas andover longer distances. In most cases vectors apparently spread the disease to either adjacent or nearbytrees. Spatial autocorrelation also identifies a prevalence of reflected clusters or areas of aggregationthat are discontinuous with the main cluster. These are interpreted as indicative of the presence ofsecondary foci that are quite prevalent and are at a distance of about 25-50 m from the main cluster ofdisease and each other. Such a pattern of widely spaced foci perhaps indicates a spatial mechanismassociated with longer distance vector movement. Longer regional scale vector transmission has notbeen investigated.

IntroductionAs with many fastidious arboreal plant pathosystems, the epidemiology of HLB has been

observed and written about, but only relatively few quantitative epidemiological studies have beenconducted. This is due to the perennial nature of the disease requiring a dedication to data collectionover multiple years combined with the inability until recently of detection by ELISA or PCR,requiring a reliance for past disease monitoring on visual assessment. In addition, there has beendifficulty in locating study sites where the disease is allowed to progress unimpeded so epidemics canbe followed without intervention of control activities. This is due to the devastating results ofinfection and the fear of allowing uncontrolled inoculum sources to exist in a region with susceptible

The International Workshop for Prevention of Citrus Greening Disease in Severely Infested Areas I

plantings (1,2).From an epidemiological point of view, monitoring the occurrence of symptoms, can be

somewhat problematic. The typical leaf mottling and chlorosis similar to zinc pattern deficiency areusually followed by retarded growth, but these symptoms are often non distinct and/or restricted toone branch or side of the tree (8,23,27). Eventually, twig dieback and a general decline ensue due tothe pathogens severe effect on the phloem of the host. Fruit symptoms, i.e., small, misshapen andpoorly colored fruit, can also be observed and monitored on diseased trees, especially mandarins andoranges which are considered the most susceptible. Infection by the HLB pathogen is often expressedin sectors on infected trees, indicating incomplete systemic infection or perhaps variable pathogentiter levels throughout infected trees. The lag in time between transmission of the pathogen by psyllidvectors or by propagation and the onset of visual symptoms can be quite variable depending the timeof the year when infection took place, ensuing environmental conditions, tree age, species/cultivarand horticultural health, cultivar, etc. (1,6,11,24,27). Thus visually quantifying the severity orexpression of disease symptoms in individual trees is not a true indication of pathogen content.Additionally, due to the temporal variation in symptom expression, trees infected at the same time(especially trees of several years age) may express the onset of infection with great variability overone or more years. This inherently broad and variable lag period, compromises the accuracy ofspatial and temporal studies to some extent. Because the visual symptoms we see at any point in timeare the expression of infections that have occurred sometime in the past, we are actually visuallyassessing a 'fuzzy history' of infection as it existed at a time period in the past equivalent to the lagperiod. Nevertheless, with these caveats in mind, useful information can still be gleaned from theepidemiological studies conducted to date concerning the spatial processes that give rise to HLBdisease, how it spreads, and how it increases. This information can be used to predict the economicand physical life of a given planting and a means to investigate the influences and efficacy of possiblecontrol interventions.

HLB epidemics can be established by introduction of infected plant materials and bytransmission due to insect vectors. The unintentional introduction of infected plant materialsestablishes the disease in new areas or countries and subsequent unregulated movement can havedisastrous results. Natural transmission appears to be related to high vector populations and theextensiveness of the inoculum reservoir (1,7). Psyllid migrations appear to be highest when hostplants are flushing and psyllid populations sedentary and feeding when foliage is mature (1,4). Thus,natural spread is probably greatest in late spring and perhaps other periods when new flush isavailable and psyllid populations are high (1,4,5). Psyllid vectors are also attracted to yellowwavelengths of light, and thus preferentially to foliage expressing HLB symptoms.HLB Epidemics: The spatial and temporal dynamics of HLB in mature groves has been investigatedin only very few cases, primarily from Reunion Island and China (10,11,15) and is only nowbeginning in Brazil and Florida. In all prior cases, only the Asian HLB pathogen, Ca. L. asiaticus,was known to be present. With the discovery of Ca. L. americanus in São Paulo, Brazil, some dataon disease increase and spread of that organism is beginning to be collected as well (3). Prior studiespresented an opportunity to investigate the disease in citrus groves in two situations, where inoculumwas introduced by infected planting stock, and where clean stock was utilized and the pathogen wasintroduced by the immigration of bacteria-contaminated psyllid vectors. The studies were conductedto establish preliminary rates of disease increase of citrus HLB under endemic conditions in thepresence of vector populations, and develop preliminary temporal models to estimate the expectedlongevity of infected sweet orange and mandarin groves.

To accomplish this, citrus HLB disease was monitored by visual assessment in citrus plotsover a period of several years in four locations (11). The Reunion Island plot (RI) was located on thecoastal leeward side of the island. The plot was established from disease-free 'Valencia' sweetorange planting material in 1970 originating from a French budwood program in Corsica. TheLiuzhou Citrus Farm Plot (LCF) was located near Liuzhou City, Guangxi Province, People'sRepublic of China. This plot was established in October 1953 and consisted of a mix of'Xinhuicheng' and 'Anliucheng' sweet orange, on rangpur lime. The Liuzhou Agricultural ResearchInstitute plot (LARI) was established in Spring 1968 and was also located near Liuzhou City but wasoriginally designed as a rootstock trial, and consisted of 'Xinhuicheng' and 'Anliucheng' sweetorange, 'Tankan' mandarin, and 'Ponkan' tangerine on various rootstocks. The two Liuzhou, China,

2 1 Multilateral Research Network for Food and Agricultural Safety

plantings were probably established from a very low incidence of greening-infected planting materialalthough no greening symptoms were noted at planting.

Minimal insect control programs were practiced in the plots in China, and, therefore, thecontrol programs were probably not sufficient to adequately control or affect the vector populations toan appreciable extent. The same was true for the RI plot where, although an insecticide program waspracticed, psyllids were not a primary target insect. The Asian psyllid, D. citri, and the Africanpsyllid, T. erytreae, were present on Reunion Island during the duration of the experiment. However,D. citri was the only psyllid present in the lower, hotter coastal area where the RI plot was located(1,2), and only D. citri, is presently known to exist in the PRC where LCF and LARI plots werelocated (27). Prevailing wind direction in the two China plots was variable, however, prevailingwinds in the Reunion Island plot was east to west. Disease was also assessed in an additional plot inthe Poutai Citrus Cooperative Grove near Shantou, Guangdong Province, PRC in 1986 and again1988 for spatial analysis of vector spread only (10). The Shantou cooperative plot consisted of ca.20,000 trees of TankanlFuzhou mandarin on land converted to citrus from swamp rice and receivedonly a local standard range of care with primitive hand-powered sprayers for insecticide treatment.The trees were planted on beds running north-south and the planting was established from disease-free planting materials in 1983. Although the data from these plots was published previously, newanalyses are presented here as well which provides greater insight into increase and spread of HLB.

Disease severity and disease incidence were visually assessed for each tree in RI during 1975,1977, and 1979 (2). To estimate disease severity, each tree was divided into an upper and lowerhemisphere and each hemisphere was subdivided into four equal quadrants resulting in eight sectionsper tree. Sections were scored individually on a 0 to 5 scale that was indicative of the proportion oflimbs expressing disease symptoms and resulted in a severity rating of 0 to 40 for each tree on eachsurvey date (2,11). Disease incidence at the tree level was calculated as the proportion of treesexpressing HLB symptoms. HLB confirmation was periodically done via i) indexing by side graftingsuspect shoots onto seedling mandarin indicator trees in the greenhouse and incubating these graftedtrees for several months, and ii) by examining prepared foliar specimens via transmission electronmicroscopy for bacteria-like organisms in the phloem sieve tubes (2,11). For the LCF, LARI andShantou plots, only disease incidence was assessed by careful visual examination of all trees withinthe plots for HLB symptoms (10, 11,27).Analysis of disease progression. As explained above, HLB epidemics are multiyear induration and it would be a rare instance that a planting would be allowed to progress to ahigh disease incidence before it would become nonproductive and be removed. Therefore,most of the HLB temporal data sets do not include assessments near a disease asymptote (11).Because of the lack of asymptotic data, either the exponential and the logistic modelsadequately described disease progress in these plots over time (Fig. 1). In one ease anasymptote of disease was reached in the LCF plot after 13 yr. Because the groves becameunproductive due to severe HLB infection and the trees were removed, disease incidenceonly reached 0.76 after 6 yr and 0.84 after 9 yr for LARI and RI, respectively. Even so, thelogistic model was used to predict asymptotic levels of disease (0.98 to 0.99) after ca. 6 to 7and 12 to 13 years after planting for LARI and RI plots, respectively (11). Logistic rates ofdisease increase were higher for LCF and LARI when compared to the RI plot (11). Thismay be because vector populations were much greater in the Liuzhou City, China area wherethe disease and vectors have been endemic for many years compared to Reunion where citrusis not as widely planted and the disease and vector were more recent immigrants. These ratesof HLB disease increase are comparable to those for other graft-transmissible disease ofcitrus such as the citrus tristeza/Aphis gossypii pathosystem, and the citrus variegatedchlorosis/sharpshooter pathosystem (12,13,14). For the Shantou plot, a small planting of 24citrus trees associated with a small vegetable garden was located immediately adjacent to afarmhouse on the southern edge of the planting. These trees were all severely infected withgreening and did not originate from disease-free nursery stock as did the rest of the planting.HLB infection in the planting was very rapid and increased from only 16 trees (0.08%) in1986 to 2880 trees (14.4%) in 1988. Unfortunately, HLB progress was not followed beyond

The International Workshop for Prevention of Citrus Greening Disease in Severely Infested Areas 1 3

1988 (10). In São Paulo Brazil, the incidence of HLB-symptomatic trees in four citrusblocks planted in 2002, surrounded by older heavily HLB affected blocks, varied from 0.004to 0.20 in 2004. After one year, the disease incidence in those blocks reached 0.26 to 0.40.At the same farm, another citrus block planted in 2000 showed a disease incidence of 0.24 in2004 and 0.70 in 2005 (Bassanezi unpublished data). In a large commercial planting inSouth Florida, HLB increased from 0.002 to as much as 0.39 disease incidence in 10 months(Irey unpublished data).Spatial analysis of HLB data. Spatial pattern analysis is used to better understand the relationshipsamong infected trees near and at distance to each other and thereby gain some understanding ofvector spread of HLB. This is accomplished by examining the HLB data at various spatial levels(hierarchies) including, i) immediately adjacent trees within and across row, ii) within groups of trees,and iii) among groups of trees over some distance to each other (17,19). Although insufficient timehas elapsed since the discovery of HLB in Brazil and Florida to determine the temporal increase ofthe disease, many maps of the disease have been compiled from visual surveys, especially in Brazilwhere over 155 maps now exist. These have been analyzed at various spatial scales to determine thespatial pattern of the disease as it occurs in Brazil and attempt to draw some conclusions concerningspread (3). The results of this work in the Western hemisphere mirror that previously found inReunion and Asia.

Some evidence of aggregation (clustering) among immediately adjacent diseased trees wasdemonstrated by the ordinary runs analysis, which is a unidirectional analysis, in all plots tested, butthis was not particularly strong (3,22). In many cases there was some orientation or direction toaggregation in most plots indicating that within-row aggregation (where trees are planted closertogether) was slightly stronger than across-row. This directional orientation for aggregation was bestdemonstrated in the Shantou plot, where a higher degree of aggregation was indicated(155/199=77.9% of rows tested) in the north-south direction compared with (91/205=44.4% of rowstested) the east-west direction. This greater north-south aggregation was in the same orientation asthe raised planting beds. For the RI and Shantou plots, mapping of isopaths of disease severitydemonstrated higher concentrations of disease initially around the perimeter of the western half of theRI plot and the southern edge of the Shantou plot (Figs. 2 and 3). For the large number of maps ofHLB incidence from Brazil, aggregation among HLB-symptomatic trees was detected by ordinaryruns analysis, and clustering existed in both within- and across-rows directions. However, as with theReunion and China data, the percentage of aggregation within- and across-rows was low (3).

The association of diseased trees 'within groups' was examined by Beta Binominal Analysis(BBD) as an overall assessment of heterogeneity of disease incidence. When fitted to disease dataparsed into quadrats, the superior fit of the beta binomial distribution is indicative of an aggregatedpattern (18,21). Aggregation at the group spatial level was demonstrated for all plots at all locationsand all quadrat (group) sizes for the majority of years. The only exceptions occurred when diseaseincidence was extremely low or high as expected (10,11). In Brazil, the binomial index of dispersionfor various quadrat sizes suggested aggregation of HLB-symptomatic trees for about 40% of the plotstested (3).

Spatial autocorrelation was used to examine the relationship 'among groups' of trees overlonger distances (16,25). The LARI, LCF, Shantou, and Brazil plots were examined at various groupsizes. For the LARI plots, groups of trees were associated with one another up to four groups awaywithin and across rows indicating associations as large as eight trees within and across rows.Discontinuous (reflected) clusters were also seen at oblique angles as far a 14-16 trees apart. TheLCF results were somewhat different. Only immediately adjacent groups were significantlyassociated with one another. However, numerous reflected clusters also existed, some at obliqueangles and as far as 18 tree spaces apart. In São Paulo, Brazil, data from plots ranging in diseaseincidence indicated in 14 of 20 cases, clusters of HLB-infected trees were found to be associated withsecondary clusters whose centers were at distances ranging from 4.2 to 22.1 tree spaces distant. Thusfor all of these locations, there are indications of psyllid vector movement resulting in transmissionboth to nearby trees causing clusters and to trees at considerable distance initiating new foci ofinfection. The Shantou plot was much more extensive in size and allowed HLB data to be examinedon a much larger scale. Significant aggregation was seen at all quadrat sizes examined and core

4 1 Multilateral Research Network for Food and Agricultural Safety

clusters were very large, demonstrating continuous relationships among groups circumscribing areasof local influence of 300 to 1672 tree spaces. Numerous reflected clusters existed as well, from 8 to572 trees in size located 25 to 30 m from the core cluster and demonstrated spatial relationshipsamong many clusters of diseased trees within the planting (Fig. 2). The RI plot was much smaller insize and groups of adjacent trees were associated with each other up to one group away across rowsand two groups within rows but discontinuous groups were also found up to seven groups (14 trees)away. This demonstrates the development of areas of local influence up to 8 tree spaces (-24m) inradius where incomplete aggregations of trees existed. However, discontinuous groups of trees alsowere found to be associated with each other as far as 8 to 16 tree spaces (-24 to 48 m) apart (Fig. 3).Spatial and temporal conclusions and interpretations: The results of spatial analysis of the largeShantou grove were very enlightening (10). A high proportion of aggregation was detected in theplanting by ordinary runs analysis and was much higher 77.9% in the north-south orientation alongrows compared to 44.4% in the east west orientation across rows. This was probably due to thehorticultural conditions within the planting. Because of the mounds on which the grove was plantedand the irrigation ditches between rows running north-south, traffic was almost entirely restricted tothis direction. Psyllids, when disturbed, often move to the next tree or even many trees away in theopposite direction from the disturbance. Infective vectors entering the grove from the south wouldhave been disturbed and moved along the rows primarily to the north. This would account for thepredominance in clustering within north-south oriented rows. The secondary clumps or clusters ofdiseased trees determined by the spatial autocorrelation analysis were also oriented in this directionfrom one another. Vector movement appears to be generally from one tree to the next; however, thereis some indication that longer distances are traversed at times and new foci of disease initiated. Thiscould possibly be stimulated by population crowding, or major disturbances due to horticulturalpractices, combined with insufficient insecticide control. This distance was estimated to be 25-30 m.

Disease-free planting material was utilized to establish the RI plot. However, greening-infected groves and D. citri were prevalent in the vicinity of the RI during the course of thisexperiment (2,11). It is likely that migrating, infective D. citri populations carried the greeningpathogen from nearby infected plantings to the RI planting early in the life of the planting. However,sufficient symptoms for recognition of the disease did not develop until 5 yr after the plot wasestablished, when visual estimation determined 9.5% of the trees were already infected. Due to thedelay between vector transmission and symptom expression, it is likely that the bacterium had alreadyinfected numerous other trees but the disease was subclinical in those trees at that time. BecauseHLB incidence and severity were greater around the perimeter of the western half of RI, the diseasemay have been introduced by disease-carrying psyllids from infected groves to the west (11).However, the prevailing winds were cast to west. Early in the epidemic, a high degree of aggregationof diseased trees was detected by ordinary runs analysis and spatial autocorrelation in the east-westaxis of the plot, and by surface response map (Fig. 2) indicating disease movement in that orientation.Vector migration would have been against the prevailing wind and, therefore, may have representedactive migration of the vectors rather than passive transport due to wind.

The introduction of diseased planting materials into LARI and LCF plots was likely to havebeen fairly random, although aggregation was detected by ordinary runs, except at very high and verylow disease proportion. Spatial autocorrelations also demonstrated decreasing cluster size over time.

From a temporal point of view, the relative increase of the HLB disease is slow and multiyears in duration, even when vector populations are high and inoculum sources are prevalent. Oneinterpretation is that vector transmission in the field must therefore be inefficient relative to othervectored plant diseases such as vegetable and field crop virus diseases that can reach asymptoticlevels within a single season. Relatively slow epidemics are common among vectored diseases ofcitrus as well as other arboreal pathosystems even though vector populations for some diseases suchas HLB and CTV are often very high.

From a spatial perspective, results of the various spatial analyses conducted to date can beinterpreted in combination to obtain a more comprehensive picture of spatial patterns that existed inHLB-infected plantings. The combined analyses indicate two mechanisms of vector spread of I-ILB,within local areas and over longer distances. Within local areas, aggregations of infected trees occurthat at times can be quite large, i.e., encompassing as many as 1672 trees. This does not mean thatevery tree in these local areas will become infected, but that a high proportion of them will be as

The International Workshop for Prevention of Citrus Greening Disease in Severely Infested Areas I 5

demonstrated by the 'strength of aggregation' calculation associated with spatial autocorrelationresults and define a focus of infection. In this case vectors are apparently spreading the disease toeither adjacent or nearby trees only a few spaces away. Spatial autocorrelation also identifies aprevalence of reflected clusters or areas of aggregation that are discontinuous with the main cluster.These are interpreted as indicative of the presence of secondary foci. These are quite prevalent andare at a distance of about 25-50 m from the main cluster of disease and each other. Such a pattern ofwidely spaced foci perhaps indicates a spatial mechanism associated with longer distance vectormovement. That is, when vectors move, either naturally in search of new feeding opportunities orwhen disturbed as in the Shantou plot, they occasionally do so to other than nearby trees and whenthis occurs they move at least 25 to 50 meters. Longer or regional scale vector transmission has notbeen investigated. It is obvious that this likely happens as well, but is beyond the scope of the studiesconducted to date.Enhancement of analysis of distribution and spread via new detection methods: The spatial andtemporal data presented were collected by visual assessments, and thus a history of infection as itexisted at some point in the past. It is probable that new detection methods will improve ourassessment of disease incidence data, and thus our understanding of disease increase and spreadsignificantly. Serological methods and PCR are to some extent reliant upon pathogen titer in a plantand its distribution relative to where samples are taken (9). In many cases, serological detection isnearly coincident with visual symptom manifestation. The non-uniform distribution of pre- and evenpost-systemic infections of the I-ILB bacteria, means that even if a tree is infected, samples collectedfrom a portion of the tree with low or no titer, will yield a negative assessment. At present PCR isboth complex and time consuming. At this point in time the ability to process the thousands ofsamples necessary to track an epidemic remains manpower and cost prohibitive but significantadvances may make such endeavors possible in the future. However, in a recent experiment, RT-PCR was used to map HLB infections in six 14 X 14 tree plots to date (20). The results are quiteelucidating relative to asymptomatic infections. The authors found nearly the same number ofasymptomatic but RT-PCR positive trees within the plots as visually symptomatic trees (20). To takethis thought process a step further; intuitively, we know that it requires some time after vectortransmission of the bacteria for the bacterial populations to build up locally and become distributedwithin an infected tree. Thus there is an additional latency period between infection and sufficientbacterial population increase and distribution before it is likely to detect asymptomatic HLBinfections via PCR. It also requires some serendipity to inadvertently collect asymptomatic samplesfor PCR assay from a portion of the tree with bacterial infection. Thus, although PCR allows us todetect many asymptomatic infections, we are still only detecting a portion of the more recent butasymptomatic infections in the planting. There is likely some unknown number of infections stillbeyond our ability to detect. At present we do not have the ability to detect an infection soon aftervector transmission. So we know that for any visual assessment of disease incidence there is at leastdouble that number of infections plus an addition population of infected trees that remain subclinicalrelative to PCR assay. This information is quite useful and may become a clue for future diseasemanagement and decision making strategies relative to the productivity of a planting. That isknowledge of the visual disease incidence gives an estimate of the actual disease incidence in theblock and provides a means to determine a threshold of disease incidence beyond which it would bemore economically beneficial to remove an infected block and replant with diseased-free trees than tocontinue to attempt to manage the disease in that planting knowing that there is considerably moredisease than meets the eye.

A second finding of the study was the distribution of asymptomatic but RT-PCR positivetrees relative to visually symptomatic trees (20). Although the plot sizes were small, by apply anearest-neighbor analysis it was determined that about 90% of the visually negative but RT-PCRpositive trees occurred within approximately 38 m (125 11) of a visually positive tree (20). Thisindicates that there is a considerable amount of secondary spread that although not necessarily nearestneighbor, still occurs predominately within a relatively short distance or "area of influence". Thisdistance also corroborates the findings of spatial autocorrelation that indicate the establishment ofsecondary foci within 24-50 m as described above.

6 1 Multilateral Research Network for Food and Agricultural Safety

Known regional distribution of HLB in the western Hemisphere:In Brazil, both Asian and the newly described American HLB were first discovered in 2004

near the city of Araraquara in the state of São Paulo, near the center of commercial citrus productionin that state which is the largest citrus producing state in Brazil. Subsequent surveys have indicatedthat the highest incidence of disease is also centered in this area providing indirect evidence that thisarea is the most probable point of introduction. An overall assessment of the aspect of the disease inthe area and of trees believed to have the oldest infections, led researchers there to believe that HLBwas likely introduced into the area approximately 10 years ago. Simple measurement of the distancefrom the presumed point of introduction to the advancing edge of the epidemic results in an estimateddistance of spread of approximately 193 km (120 miles) or about 19.3 km (12 miles) per year. Wemust also consider that the disease has undoubted spread beyond the area of visual detection by somedistance as well. However, this does not take into account that the disease may have been transportedinadvertently and established in new locations by contaminated nursery or planting materials prior torealizing the disease existed in the area. This would mean that human assisted long distance transportwas playing a role and thereby would decrease the estimated yearly distance of spread by someunknown amount. Never the less, this is indirect evidence of long distance transport of HLB bycontaminated vectors as well. Up to May 2006, 5,000 samples were collected and sent by growers toFundecitrus. Candidatus Liberibacter americanus is the prevalent species in São Paulo state,comprising 77.2% of all positive PCR samples, against 21.4% for Ca. Liberibacter asiaticus. Theremaining 1.4% samples were positive for both species (26). American HLB is also moredisseminated than Asian HLB in Brazil. 65% of HLB-affected municipalities have only Ca.Liberibacter americanus, 33% have both Ca. Liberibacter americanus and Ca. Liberibacter asiaticus,and 2% only Ca. Liberibacter asiaticus alone (26).

In Florida HLB was first discovered by a Florida Department of Agriculture and ConsumerServices scientist on August 24, 2005 during a routine survey in a pummelo tree in a commercialtropical fruit nursery in Florida City and immediately thereafter in the residential communities ofPinecrest and Coral city. Presumptive PCR positive samples were sent to the USDA, APHISdiagnostic lab in Beltsville and confirmed on September 2, 2005 to be Candidatus Liberibacterasiaticus the causal agent of Asian HLB. The discovery immediately prompted a survey to delimitthe infection. East-west transects were surveyed every five miles northward in an attempt todetermine the northern extent of the disease distribution. A positive detection in a transectimmediately prompted a survey in the next transect five miles to the north. HLB distribution wasquickly confirmed to extend northward to the Fort Pierce residential area, 193 km (120 miles).Concern for the commercial citrus industry prompted simultaneous selected survey of commercialplantings nearest to the known residential distribution along the southeast Florida coast. Suspectssamples from selected commercial plantings also quickly the presence of HLB. This prompted astate-wide spot survey of the entire Florida commercial citrus industry and selected residential areas.At the time of this writing, the known distribution has been determined includes 13 counties in thesouthern half of the Florida peninsula. Preliminary indications are that the disease is most severe inthe southern most residential and commercial areas. This would imply that the initial introduction(s)may have been in this area. However, unlike Brazil, there is no clear indication of a point or even alimited area of introduction, although it is presumed to have been introduced somewhere in thesoutheastern metropolitan residential area. Based on this assumption, measurements from the initialdiscovery point and the most nearest residential infections to the most adjacent commercial plantingare approximately 90 and 55 miles, respectively. This nearest commercial planting is both isolatedand has not acquired nursery materials from outside sources. Thus the assumption is that HLBarrived in this planting via psyllid vector dissemination and transmission. This is further suspectedbecause the distribution of HLB symptomatic trees in this planning is more intense in the southeasternportion of this planting most adjacent to the residential area and separated from it by a citrus-free voidof march, i.e., the Everglades area. Although not conclusive, it is further possible evidence that longdistance movement of HLB is likely by psyllid vectors and in this case could be related to movementof air masses carrying infected vectors during hurricane or tropical storms that have affected Floridarecently over this void to commercial areas.

The International Workshop for Prevention of Citrus Greening Disease in Severely Infested Areas 1 7

References1. Aubert, B.1987.Trioza eiytreae Del Guercio and Diaphorina curl Kuwayama

(Homoptera:Psylloidae), the two vectors of citrus greening disease: Biological aspects andpossible control strategies. Fruits 42:149-162.

2. Aubert, B. 1987. Le greening une maladie infectieuse des agrumes, d'origine bactéreenne,transmise pardes homoptères.Strategie de lutte développée a l'Jle de la Reunion.Circonstances epidemiologigues en Afrique/Asie et modalités d'intervention. IRFA/CIAD-B.P. 180-97455 Saint Pierre Cedex.

3. Bassanezi, R. B., Busato, L. A., Bergamin-Filho, A., Amorim, L. And Gottwald, T. R. 2005.Preliminary spatial analysis of huanglongbing in São Paulo, Brazil. Proc. International Org.Citrus Virol. (In press)

4. Catling, H. D. 1969. The bionomics of the South African psylla, Trioza eiytreae D. G.(Homoptera:Psyllidae) 3. The influence of extremes of weather on survival. J. Entomol. Soc.South Afr. 32(2):273-290.

5. Catling, H. D. 1970. The bionomics of the South African Citrus psylla Trioza e'ytreae_D. G.(Homoptera:Psyllidae). 4. Influence of predators. J. Entomol. Soc. South. Aft. 33(2):341-348.6. Catling, H. D. and Atkinson, P. R. 1974. Spread of greening by Trioza erytreae. (Del Guercio)in Swaziland. Pages 33-39 in: Proc. 6th Conf. Intl. Organ. Citrus Virologists. L. G. Weathersand M. Cohen, eds. IOCV, Div. Agric. Sci., Univ. of Calif., Riverside.

7. Chao, H. Y., Chiang, Y. H., Lee, S. L., Chiu, C. S. and Su, W. F. 1979. A preliminary study onthe relation between the prevalence of the Citrus Yellow Shoot (Huanglungbin) and the Citruspsyllid Diaphorina citri Kuwayama. Acta Phytopathologica Sinica (2), 121-126.

8. Gamier, M., Danel, N., and Bove, J. M. 1984. Etiology of citrus greening disease. Ann.Microbiol. (Inst. Pasteur) 1 35A: 169-179.

9. Gamier, M., Martin-Gros, G., and Bove, J. M. 1987. Monoclonal antibodies against thebacterial-like organism associated with citrus greening disease. Ann. Inst. Pasteur(Microbiology) 138:639-650.

10. Gottwald, T. R., Aubert, B., and Huang, K. L.. Spatial pattern analysis of citrus greening inShantou, China. Proc. 11th Conf. Internat. Org . Citrus Virol. 11:421-427. 1991.

11. Gottwald. T. R., Aubert, B. and X.-Y. Zhao. Preliminary analysis of citrus greening(Huanglungbin) epidemics in the People's Republic of China and French Reunion Island.Phytopathology 79:687-693. 1989.

12. Gottwald, T. R., Cambra, M., Moreno, P. Camarasa, E., and Piquer, J. Spatial and temporalanalyses of citrus tristeza in Eastern Spain. Phytopathology. 86:45-55. 1996.13. Gottwald, T. R., Garnsey, S. M. and Borbón, J. 1998. Increase and patterns of spread of citrustristeza virus infections in Costa Rica and the Dominican Republic in the presence of the browncitrus aphid, Toxoptera citricida. Phytopathology 88:621-636.

14. Gottwald, T. R., Gibson, G., Gamsey, S. M. and Irey, M. 1999. Examination of the effect ofaphid vector population composition on the spatial dynamics of citrus tristeza virus spread viastochastic modeling. Phytopathology 89: 603-608.

15. Gottwald, T. R., Gonzales, C. I., and Mercado, B. G. Analysis of the distribution of citrusgreening in groves in the Philippines. Proc. 11th Conf. Internat. Org . Citrus Virol. 11:414-420.1991.

16. Gottwald, T. R., Richie, S. M., and Campbell, C. L. LCOR2 - Spatial correlation analysissoftware for the personal computer. Plant Disease 76:213-215. 1992.

17. Hughes, G., and Madden, L. V. 1992. Aggregation and incidence of disease. Plant Pathol.41:657-660.

18. Hughes, G. and Madden, L. V. 1993. Using the beta-binomial distribution to describeaggregated patterns of disease incidence. Phytopathology 83: 759-763.

19. Hughes, G., McRoberts, N., Madden, L. V., and Gottwald, T.R. 1997. Relationships betweendisease incidence at two levels in a spatial hierarchy. Phytopathology 87:542-550.20. Irey, M. S., Tim Gast, T., and Gottwald, T. R. 2006. Comparison of visual assessment andpolymerase chain reaction assay testing to estimate the incidence of the Huanglongbingpathogen in commercial Florida citrus. Proc. Florida State Hortic. Soc. (In press).

8 1 Multilateral Research Network for Food and Agricultural Safety

21. Madden, L. V. and Hughes, G. 1994. BBD - Computer software for fitting the beta-binomialdistribution to disease incidence data. Plant Dis. 78:536-540.

22. Madden, L. V., Louie, R., Abt, J. J., and Knoke, J. K. 1982. Evaluation of tests forrandomness of infected plants. Phytopathology 72:195-198.

23. Martinez, A. L. and Wallace, J. M. 1967. Citrus leaf-mottle-yellow disease in the Philippinesand transmission of the causal virus by a psyllid, Diaphorina citri. Plant Dis. Rep. 51:692-695.

24. McClean, A. P. D. and Oberholzer, P. C. J. 1965. Citrus psylla, a vector of greening disease ofsweet orange. S. Afr. J. Agric. Sci. 8:297-298.

25. Modjeska, J. S. and Rawlings, J. 0. 1983. Spatial correlation analysis of uniformity data.Biometrics 39:373-384.

26. Wulff, N. A., D. C. Teixeira, E. C. Martins, A. P. R. Leite, N. R. H. Padiar, A. G. Mariano, A.E. Carmo, D. P. Abrahão, M. C. Sousa, A. J. Ayres & J. M. Bové. 2006. Huanglongbingdiagnosis. Proc. Huanglongbing Greening Int. Workshop: 17.

27. Zhao, Xue Yuan. 1981. Citrus yellow shoot (Huanglungbin) in China: a review. Proc. mt.Soc. Citriculture 466-469.

Figure 1. Analysis of disease progress of citrus huanglongbing in citrus groves in the PeoplesRepublic of China and Reunion Island. LCF = Liuzhou Citrus Farm Plot located near Liuzhou City,Guangxi Province, People's Republic of China. LARI = Luizhou Agricultural Research Institute,Guangxi Province, the People's Republic of China; RI = Reunion Island; EXP = exponential model;Log = logistic model; R2 coefficient of determination; b = slope of regression. Disease percentageswere transformed by In (y) and In (y/( 1 -y)) for exponential and logistic models, respectively.

Tin-A LCFi

• J4TT2

ICU

LO 05355515961 6365

IIyAIP.,1d IL

W14LC

68 69 70 11 7273Year

CR1P.IdItI

.ç .WT4T7ZZ1 '.••L,Wid2 '..Uth1

75 76 77 78 79

4Of

CrCD

m CDLU <

CD

t0

The International Workshop for Prevention of Citrus Greening Disease in Severely Infested Areas 1 9

hL

Figure 2. Response surface representations of Figure 3. Response surface representations ofdisease development and spatial spread of citrus disease development and spatial spread of citrushuanglongbing disease in Reunion Island. The huanglongbing disease in the Shantou Citruscitrus grove was established in 1970.CooperativeplantationnearShantou,Assessments of disease severity were conducted Guangdong Province, PRC in 1986 (Top) andin 1975 (Top), 1977 (Middle), and 1979 again 1988 (Bottom). Note high disease(bottom) panels respectively. Note high disease incidence was predominately along the southernseverity around perimeter of western half of border of the planting in 1986 and apparentlygrove in 1975, early in the epidemic followed by emanated from dooryard plantings immediatelyspread and increase to the east. Final panel is anto the south. Subsequent spread was to theaerial photo of Reunion Island HLB plot. Photo north in the direction of equipment andis in a similar orientation as the response surfacepersonnel travel along north-south orientedmaps in Fig. 2 and was taken about the same raised beds.time as the final disease assessment in 1979.

1975

UP

- aI

kb-- &k

f.2

1977IL

A

1979titJL

1979

-

FM

10 Multilateral Research Network for Food and Agricultural Safety