T. R. Gottwald

Visual Rating and the Use of Image Analysis for Assessing Different Symptoms of Citrus Canker on Grapefruit Leaves

Citrus canker is caused by the bacterial pathogen Xanthomonas axonopodis pv. citri and infects several citrus species in wet tropical and subtropical citrus growing regions. Accurate, precise, and reproducible disease assessment is needed for monitoring epidemics and disease response in breeding material. The objective of this study was to assess reproducibility of image analysis (IA) for measuring severity of canker symptoms and to compare this to visual assessments made by three visual raters (VR1-3) for various symptom types (lesion numbers, % area necrotic, and % area necrotic+chlorotic), and to assess inter- and intra-VR reproducibility. Digital images of 210 citrus leaves with a range of symptom severity were assessed on two separate occasions. IA was more precise than VRs for all symptom types (inter-assessment correlation coefficients, r, for lesion numbers by IA = 0.99, by VRs = 0.89 to 0.94; for %, r for % area necrotic+chlorotic for IA = 0.97 and for VRs = 0.86 to 0.89; and r for % area necrotic for IA = 0.96 and for VRs = 0.74 to 0.85). Accuracy based on Lins concordance coefficient also followed a similar pattern, with IA being most consistently accurate for all symptom types (bias correction factor, Cb = 0.99 to 1.00) compared to visual raters (Cb = 0.85 to 1.00). Lesion number was the most reproducible symptom assessment (Lins concordance correlation coefficient, ρc, = 0.76 to 0.99), followed by % area necrotic+chlorotic (ρc = 0.85 to 0.97), and finally % area necrotic (ρc = 0.72 to 0.96). Based on the true value provided by IA, precision among VRs was reasonable for number of lesions per leaf (r = 0.88 to 0.94), slightly less precision for % area necrotic+chlorotic (r = 0.87 to 0.92), and poorest precision for % area necrotic (r = 0.77 to 0.83). Loss in accuracy was less, but showed a similar trend with counts of lesion numbers (Cb = 0.93 to 0.99) which was more consistently accurately reproduced by VRs than either % area necrotic (Cb = 0.85 to 0.99) or % area necrotic+chlorotic (Cb = 0.91 to 1.00). Thus, visual raters suffered losses in both precision and accuracy, with loss in precision estimating % area necrotic being the greatest. Indeed, only for % area necrotic was there a significant effect of rater (a two-way random effects model ANOVA returned a P < 0.001 and 0.016 for rater in assessments 1 and 2, respectively). VRs showed a marked preference for clustering of % area severity estimates, especially at severity >20% area (e.g., 25, 30, 35, 40, etc.), yet VRs were prepared to estimate disease of <1% area, and at 1% increments up to 20%. There was a linear relationship between actual disease (IA assessments) and VRs. IA appears to provide a highly reproducible way to assess canker-infected leaves for disease, but symptom characters (symptom heterogeneity, coalescence of lesions) could lead to discrepancies in results, and full automation of the system remains to be tested.

Effect of Simulated Wind-Driven Rain on Duration and Distance of Dispersal of Xanthomonas axonopodis pv. citri from Canker-Infected Citrus Trees

Dynamics of dispersal of the bacteria that causes citrus canker (Xanthomonas axonopodis pv. citri) were assessed in simulated wind-driven rain splash. The wind/rain-splash events were simulated using electric blowers to generate turbulent wind (15 to 20 m s-1) and sprayer nozzles to produce water droplets entrained in the wind flow. The splash was blown at an inoculum source of canker-infected trees 1 m downwind. The splash downwind of the source of the infected trees was collected by vertical panel samplers and funnel samplers. The duration over which bacteria were dispersed in spray was assessed in continuous wind at intervals from 0 to 52 h after commencing the simulated rain splash event. In one experiment on 11 February 2003, a total of 1.48 × 106 bacteria were collected by panels 1 m downwind from the inoculum source during the first 10 min of dispersal, but the numbers declined to 3.60 × 105 bacteria after 1 h and ranged between 1.42 × 105 and 1.93 × 104 up to 52 h. In a more detailed study (15 July 2003) of dispersal duration over 4 h, the greatest quantity of bacteria collected by panel samplers were dispersed in the first 5-min period (1.01 × 108 bacteria collected). By 10 min after initiation of dispersal, approximately one-third (3.09 × 107 bacteria collected) of the initial number was being dispersed, and by the end of the first hour, only one-tenth (1.31 × 107 bacteria collected) of the initial quantity was dispersed. Funnel samplers placed at ground level under the trees showed a similar trend. The distance to which bacteria were dispersed in wind-blown splash was also tested under simulated conditions: on 18 September 2003, bacteria were collected by panel samplers at all distances sampled (1, 2, 4, 6, 8, 10, and 12 m) with the greatest number of bacteria deposited at 1 m (4.93 × 106 bacteria collected), while 2.22 × 103 bacteria were deposited over a 10-min period 12 m from the inoculum source. Wind speed declined from 19.5 m s-1 upwind of the trees to 2.8 m s-1 1 m downwind, and by 4 m downwind from the inoculum source, movement was similar to the surrounding air. The data on duration and distance of dispersal were best described by power law regression models compared to exponential models. Citrus canker is readily dispersed in wind-driven rain and is dispersed in large quantities immediately after the stimulus occurs, upon which wind-driven splash can disperse inoculum over a prolonged period and over a substantial distance.

Optimal Strategies for the Eradication of Asiatic Citrus Canker in Heterogeneous Host Landscapes

ABSTRACT The eradication of nonnative plant pathogens is a key challenge in plant disease epidemiology. Asiatic citrus canker is an economically significant disease of citrus caused by the bacterial plant pathogen Xanthomonas citri subsp. citri. The pathogen is a major exotic disease problem in many citrus producing areas of the world including the United States, Brazil, and Australia. Various eradication attempts have been made on the disease but have been associated with significant social and economic costs due to the necessary removal of large numbers of host trees. In this paper, a spatially explicit stochastic simulation model of Asiatic citrus canker is introduced that describes an epidemic of the disease in a heterogeneous host landscape. We show that an optimum eradication strategy can be determined that minimizes the adverse costs associated with eradication. In particular, we show how the optimum strategy and its total cost depend on the topological arrangement of the host landscape. We discuss the implications of the results for invading plant disease epidemics in general and for historical and future eradication attempts on Asiatic citrus canker.

Characteristics of the Perception of Different Severity Measures of Citrus Canker and the Relationships Between the Various Symptom Types

Citrus canker is a disease of citrus and is caused by the bacterial pathogen Xanthomonas citri subsp. citri. Ways of managing the disease are being sought, and accurate, precise, reproducible disease assessment is needed for monitoring epidemics. The objective of this study was to investigate the characteristics of visual assessment of citrus canker symptoms compared with actual disease measured using image analysis (IA). Images of 210 citrus leaves with a range of incidence and severity of citrus canker were assessed by three plant pathologists (VR1-3) and by IA. The number of lesions (L), % area necrotic (%AN), and % area necrotic+chlorotic (%ANC) were assessed. The best relationships were found between %AN and %ANC (r2 = 0.41 to 0.87), and the worst between L and %AN (r2 = 0.27 to 0.66). Bland-Altman plots showed various sources of rater error in assessments, including under- and over-estimation, proportional error, and heterogeneity of variation dependent on actual disease magnitude. There was a tendency to overestimate area diseased, but not lesion counts, and this tendency was pronounced at lower disease severity, with a leaf having more lesions tending to be assessed as having greater area infected compared with a leaf with fewer lesions but equal actual area infected. The rater estimations of disease were less accurate or precise with increasing actual disease severity as indicated by the fit of a normal probability density function-the incidence of extreme values increases with increasing actual disease. For example, for %ANC the kurtosis of the distribution ranged from 17.92 to 1.18, 0.51, and 0.22 in actual disease category ranges of 0 to 10, 11 to 20, 21 to 30, and 31 to 40% area infected, respectively. The log variance of the estimates plotted against log actual disease for all three raters over two assessment occasions gave a linear relationship for L, %AN, and %ANC (r2 = 0.74, 0.65, and 0.74, respectively). Training should improve the accuracy, precision, and reproducibility of raters, and knowledge of the characteristics of disease assessment should help develop and target the training more appropriately and address specific causes and sources of error.

Some Consequences of Using the Horsfall-Barratt Scale for Hypothesis Testing

Comparing treatment effects by hypothesis testing is a common practice in plant pathology. Nearest percent estimates (NPEs) of disease severity were compared with Horsfall-Barratt (H-B) scale data to explore whether there was an effect of assessment method on hypothesis testing. A simulation model based on field-collected data using leaves with disease severity of 0 to 60% was used; the relationship between NPEs and actual severity was linear, a hyperbolic function described the relationship between the standard deviation of the rater mean NPE and actual disease, and a lognormal distribution was assumed to describe the frequency of NPEs of specific actual disease severities by raters. Results of the simulation showed standard deviations of mean NPEs were consistently similar to the original rater standard deviation from the field-collected data; however, the standard deviations of the H-B scale data deviated from that of the original rater standard deviation, particularly at 20 to 50% severity, over which H-B scale grade intervals are widest; thus, it is over this range that differences in hypothesis testing are most likely to occur. To explore this, two normally distributed, hypothetical severity populations were compared using a t test with NPEs and H-B midpoint data. NPE data had a higher probability to reject the null hypothesis (H0) when H0 was false but greater sample size increased the probability to reject H0 for both methods, with the H-B scale data requiring up to a 50% greater sample size to attain the same probability to reject the H0 as NPEs when H0 was false. The increase in sample size resolves the increased sample variance caused by inaccurate individual estimates due to H-B scale midpoint scaling. As expected, various population characteristics influenced the probability to reject H0, including the difference between the two severity distribution means, their variability, and the ability of the raters. Inaccurate raters showed a similar probability to reject H0 when H0 was false using either assessment method but average and accurate raters had a greater probability to reject H0 when H0 was false using NPEs compared with H-B scale data. Accurate raters had, on average, better resolving power for estimating disease compared with that offered by the H-B scale and, therefore, the resulting sample variability was more representative of the population when sample size was limiting. Thus, there are various circumstances under which H-B scale data has a greater risk of failing to reject H0 when H0 is false (a type II error) compared with NPEs.

Automated Image Analysis of the Severity of Foliar Citrus Canker Symptoms

Citrus canker (caused by Xanthomonas citri subsp. citri) is a destructive disease, reducing yield and rendering fruit unfit for fresh sale. Accurate assessment of citrus canker severity and other diseases is needed for several purposes, including monitoring epidemics and evaluation of germplasm. We compared measurements of citrus canker severity (percent area infected) from automated image analysis to visual estimates by raters and true values using images from five leaf samples (65, 123, 50, 50, and 200 leaves; disease severity from 0 to 60%). Severity on leaves was measured by automated image analysis by (i) basing threshold values on a presample of leaves, or (ii) replacing healthy leaf color on a leaf-by-leaf basis before automating image analysis. Samples 1 to 4 were assessed by three trained plant pathologists, and sample 5 was assessed by an additional 25 raters. Healthy leaf area color replacement gave the most consistent agreement with the true severity data. Using color replacement, agreement with true values based on Lins concordance correlation coefficient (ρc) was 0.93, 0.79, 0.71, 0.85, and 0.89 for each of the samples, respectively. The range and consistency of agreement was generally less good for automated thresholds based on a presample (ρc = 0.35-0.90) or visual raters (ρc = 0.30-0.94). The constituents of agreement (precision and accuracy) showed similar trends. No one rater or method was best for every leaf sample, but replacing healthy color in each leaf with a standard color before automation of image analysis improved agreement, and was relatively quick (20 s per image). The accuracy and precision of automated image analysis of citrus canker severity can be comparable to unaided, direct visual estimation by many raters.

The effect of landscape pattern on the optimal eradication zone of an invading epidemic.

A number of high profile eradication attempts on plant pathogens have recently been attempted in response to the increasing number of introductions of economically significant nonnative pathogen species. Eradication programs involve the removal of a large proportion of a host population and can thus lead to significant social and economic costs. In this paper we use a spatially explicit stochastic model to simulate an invading pathogen and show that it is possible to identify an optimal control radius, i.e., one that minimizes the total number of hosts removed during an eradication campaign that is effective in eradicating the pathogen. However, by simulating the epidemic and eradication processes in multiple landscapes, we demonstrate that the optimal radius depends critically on landscape pattern (i.e., the spatial configuration of hosts within the landscape). In particular, we find that the optimal radius, and also the number of host removals associated with it, increases with both the level of aggregation and the density of hosts in the landscape. The result is of practical significance and demonstrates that the location of an invading epidemic should be a key consideration in the design of future eradication strategies.

Wind Speed Effects on the Quantity of Xanthomonas citri subsp. citri Dispersed Downwind from Canopies of Grapefruit Trees Infected with Citrus Canker

The epidemic of citrus canker (Xanthomonas citri subsp. citri) in Florida continues to expand since termination of the eradication program in 2006. Storms are known to be associated with disease spread, but little information exists on the interaction of fundamental physical and biological processes involved in dispersal of this bacterium. To investigate the role of wind speed in dispersal, wind/rain events were simulated using a fan to generate wind up to 19 m·s-1 and spray nozzles to simulate rain. Funnels at ground level and panels at 1.3 m height and distances up to 5 m downwind collected wind-driven splash. Greater wind speeds consistently dispersed more bacteria, measured by concentration (colony forming units [CFU] ml-1) or number sampled (bacteria flux density [BFD] = bacteria cm-2 min-1), from the canopy in the splash. The CFU ml-1 of X. citri subsp. citri collected by panels 1 m downwind at the highest wind speed was up to 41-fold greater than that collected at the lowest wind speed. BFD at the highest wind speed was up to 884-fold higher than that collected at the lowest wind speed. Both panels at distances >1 m and funnels at distances >0 m collected many-fold more X. citri subsp. citri at higher wind speeds compared to no wind (up to 1.4 × 103-fold greater CFU ml-1 and 1.8 × 105-fold the BFD). The resulting relationship between wind speed up to 19 m·s-1 and the mean CFU ml-1 collected by panel collectors downwind was linear and highly significant. Likewise, the mean CFU ml-1 collected from the funnel collectors had a linear relationship with wind speed. The relationship between wind speed and BFD collected by panels was generally similar to that described for CFU ml-1 of X. citri subsp. citri collected. However, BFD collected by funnels was too inconsistent to determine a meaningful relationship with increasing wind speed. The quantity of bacteria collected by panels declined with distance, and the relationship was described by an inverse power model (R2 = 0.94 to 1.00). At higher wind speeds, more bacteria were dispersed to all distances. Windborne inoculum in splash in subtropical wet environments is likely to be epidemiologically significant, as both rain intensity and high wind speed can interact to provide conditions conducive for dispersing large quantities of bacteria from canker-infected citrus trees. Disease and crop management aimed at reducing sources of inoculum and wind speeds in a grove should help minimize disease spread by windborne inoculum.

Comparison of Assessment of Citrus Canker Foliar Symptoms by Experienced and Inexperienced Raters

Citrus canker (Xanthomonas citri subsp. citri) is destructive in many citrus production regions in tropical and subtropical parts of the world. Assessment of canker symptoms is required for diverse reasons, including monitoring epidemics, evaluating the efficacy of control strategies, and disease response in breeding material. The objectives were to compare the ability of experienced and inexperienced raters at assessing citrus canker, to identify factors that affect the quality of the assessment, to determine common sources of error, and to discern how error is related to actual disease magnitude. Two-hundred digital leaf images (0 to 37% area infected) were assessed once by 28 raters, five of whom were experienced plant pathologists (PPs), and 23 who had no experience in disease severity assessment (NPPs). True disease (lesion number [LN], % necrotic area [%N], and % chlorotic+necrotic area [%CN]) was measured using image analysis on a leaf-by-leaf basis, and each parameter was estimated by the 28 raters. LN was neither severely over- nor underestimated, while %N was greatly overestimated, with a lesser tendency to overestimate %CN over the true severity range of these two symptom types. A linear relationship existed between estimate of the disease and true disease for all measures of severity. Data were heteroscedastic and error was not constant with increasing true disease. Agreement between rater estimates and true disease was measured with Lins concordance correlation coefficient (ρc). LN showed greatest agreement (ρc = 0.88 to 0.99), followed by %CN (ρc = 0.80 to 0.95) and %N (ρc = 0.19 to 0.84). Greater lesion number resulted in overestimation of area infected for both %N and %CN. Overestimation was particularly noticeable at low disease severities. There was a linear relationship between log variance and log true disease for LN (r2 = 0.71), %N (r2 = 0.85), and %CN (r2 = 0.88), and raters tended to estimate disease above 10% to the nearest 5 or 10%. GLM analysis showed differences between PP and NPP groups in assessing disease. For LN, precision of assessment for both groups was similar (r2 > 0.92 and 0.94, respectively), but for estimates of %N and %CN, the PPs were more precise (%N and %CN, r2 = 0.61 and 0.73, respectively) compared to NPPs (%N and %CN, r2 = 0.45 and 0.58, respectively). Absolute error for mean LN was low. The absolute error of %N and %CN showed overestimation to approximately 8% area infected. Above 8%, absolute error increased, but comprised both over- and underestimation. For %N and %CN, relative error was almost exclusively positive and dramatic at severity <8% (up to approximately 600%), but at severity >10% it was relatively small. Error in rater estimates of canker severity is ubiquitous. Understanding these sources of error will aid in the development of both appropriate training and relevant rating aids.

A generic risk‐based surveying method for invading plant pathogens

Invasive plant pathogens are increasing with international trade and travel, with damaging environmental and economic consequences. Recent examples include tree diseases such as sudden oak death in the Western United States and ash dieback in Europe. To control an invading pathogen it is crucial that newly infected sites are quickly detected so that measures can be implemented to control the epidemic. However, since sampling resources are often limited, not all locations can be inspected and locations must be prioritized for surveying. Existing approaches to achieve this are often species specific and rely on detailed data collection and parameterization, which is difficult, especially when new arrivals are unanticipated. Consequently regulatory sampling responses are often ad hoc and developed without due consideration of epidemiology, leading to the suboptimal deployment of expensive sampling resources. We introduce a flexible risk-based sampling method that is pathogen generic and enables available information to be utilized to develop epidemiologically informed sampling programs for virtually any biologically relevant plant pathogen. By targeting risk we aim to inform sampling schemes that identify high-impact locations that can be subsequently treated in order to reduce inoculum in the landscape. This damage limitation is often the initial management objective following the first discovery of a new invader. Risk at each location is determined by the product of the basic reproductive number (R0), as a measure of local epidemic size, and the probability of infection. We illustrate how the risk estimates can be used to prioritize a survey by weighting a random sample so that the highest-risk locations have the highest probability of selection. We demonstrate and test the method using a high-quality spatially and temporally resolved data set on Huanglongbing disease (HLB) in Florida, USA. We show that even when available epidemiological information is relatively minimal, the method has strong predictive value and can result in highly effective targeted surveying plans.