Abstract
A three-year survey was conducted to characterize the spatial distribution patterns of Pierce’s disease (PD) of grapevines in the Coachella Valley (Riverside County, CA). Spatial structures of PD distributions were characterized with geostatistics and spatial analysis with distance indices (SADIE). The survey confirmed 97 diseased grapevines from seven vineyards in the valley, and the diseased vines were localized within the vineyards. Approximately 82% of diseased grapevines were found near two to six consecutive dead, missing, or replanted grapevines in a row. Six vineyards had low (<0.01%) PD incidence, but one vineyard had 3.8% incidence and the diseased vines were spatially aggregated, forming a significant (p < 0.05) patch. The diseased grapevines within the patch also were aggregated and spatially correlated within 26 m. This is the first report of PD distribution in the Coachella Valley, and these findings provide the foundation for development of PD sampling plans.
Xylella fastidiosa Wells et al., a xylem-limited bacterium, causes Pierce’s disease (PD) of grapevines. It is transmitted by xylem-fluid feeding insects such as sharpshooters (Hemiptera: Cicadellidae) and spittlebugs (Hemiptera: Cercopidae) (Purcell and Frazier 1985). Pierce’s disease has been present in California for more than 100 years (Hopkins and Purcell 2002), and infrequent epidemics have occurred in the north coast of California. These epidemics were associated with populations of the blue-green sharpshooter, Graphocephala atropunctata Signoret (Purcell and Frazier 1985). Recent epidemics in southern California have been linked to the invasion of the glassy-winged sharpshooter, Homalodisca coagulata Say (Blua et al. 1999). The recent outbreak of H. coagulata and PD in the Temecula Valley (Riverside County, CA) has impacted 25% of the vineyard acres, resulting in ~$13 million in damage (Wine Institute 2002).
The Coachella Valley, located in southern California (Riverside County), has ~4,600 ha of vineyards that produce the first domestic table grapes for market in the United States (Winkler et al. 1984). Pierce’s disease has not been a concern to growers until the recent invasion of H. coagulata into southern California. Since H. coagulata first was identified in the Coachella Valley in 1995 (Blua et al. 1999), population densities increased until 2003, when a chemical control program was initiated by the California Department of Food and Agriculture (Park et al. 2006). However, there is still a risk of PD epidemics in the Coachella Valley because both H. coagulata and X. fasitidiosa are present and both can increase in areas where winters are mild (Hopkins and Purcell 2002). Currently, no cures are available for diseased grapevines with PD, and PD management recommendations focus on finding and removing diseased grapevines and controlling vector insects (Varela et al. 2001). Based on these management recommendations, determining the location of diseased grapevines within vineyards is critical. However, with the exception of entire vineyard surveys, no sampling protocols have been developed.
Spatial distribution of infected plants is one attribute that can assist in understanding the epidemiology of plant disease (Weltzien 1972). When the pattern of disease is described, logical hypotheses about factors driving the distribution can be explored. The best method for understanding spatial distribution of disease is to create maps of disease incidence. However, mapping the distribution of disease has not been applied to epidemiological studies until recently because there was a lack of suitable technologies or methods to generate distribution maps of plant diseases in fields. With global positioning systems (GPS), geographic information systems (GIS), geostatistics, and spatial analysis with distance indices (SADIE), it currently is possible to create distribution maps and to statistically analyze spatial distribution of plant disease. In addition to understanding epidemiology, characterization of disease spatial distribution will assist development of sampling protocols to identify where disease exists.
A survey program was initiated in 2002 to characterize the distribution of PD in the Coachella Valley to better understand epidemiology and improve sampling protocols. In this paper, the results from a three-year survey are reported with descriptions of the spatial distribution of PD in the Coachella Valley.
Materials and Methods
Disease survey.
A landscape-scale survey was conducted from 2002 to 2004 in the Coachella Valley. Vineyards were surveyed yearly by visually inspecting grapevines for PD symptoms (chlorosis, necrosis, and progressive marginal discoloration of the leaf) and the presence of X. fastidiosa was confirmed by enzyme-linked immunosorbent assay (ELISA). In 2002, 300 grapevines in each of 25 vineyards were examined. In 2003 and 2004, all vineyards in the Coachella Valley were examined (Figure 1⇓); 80 vineyards and ~571,000 grapevines in 2003; and 82 vineyards and ~616,000 grapevines in 2004. All surveys were conducted in June and July because PD symptoms in vineyards are easier to distinguish from other grape maladies at that time of the year in the Coachella Valley.
Three to five PD symptomatic leaves were taken from each grapevine to test for infection by X. fastidiosa. Leaf samples were collected from 233 and 478 symptomatic grapevines in 2002 and 2003, respectively. This protocol was modified in 2004 by taking at least five basal leaves from different canes on symptomatic grapevines, because there is a higher probability of detecting X. fastidiosa in a diseased grapevine when basal leaves are selected for ELISA testing (Krell et al. 2006). Leaf samples were collected from 187 symptomatic grapevines in 2004. The presence of X. fastidiosa in all samples was confirmed with ELISA, according to the manufacturer’s instructions (Agdia Inc., Elkhart, IN) with the following modifications. Whole petioles were macerated in sample extraction pouches (Agdia) with 1.5 mL general extraction buffer (Agdia). After addition of the peroxidase substrate (o-phenylenediamine dihydrochloride), microtiter plates were incubated for ~30 min and read at 492 nm on a plate reader (model 2550 EIA Reader; Bio-Rad, Richmond, CA). Samples were considered X. fastidiosa-positive if the absorbance value was greater than the mean of the two negative control samples (petioles from a healthy grapevine) plus three standard deviations. The location of each diseased grapevine was determined with a GPS (GeoXT, Trimble Navigations Ltd., Westminster, CO).
When the presence of X. fastidiosa in a vineyard was confirmed with ELISA, the spatial structure of PD distribution was further investigated in two ways. First, an area was sampled of 5 rows by 12 grapevines (60 grapevines) centered around each of the diseased grapevines identified from the survey. From each of the 60 grapevines, five basal leaves from five different canes were collected. These leaves were pooled by grapevine and subjected to ELISA. Missing, dead, or replanted grapevines in each area also were recorded. Second, the spatial distribution of PD in the identified vineyards was investigated with grid-based sampling. A grid pattern of every tenth grapevine in every fifteenth row was applied to the vineyards, and sample locations were georeferenced with a GPS. At each sample location (one grapevine), five basal leaves from five different canes were sampled. The five petioles were pooled and tested for X. fastidiosa with ELISA as described above.
Mapping and spatial analyses.
All survey data were georeferenced with a GPS and ArcGIS 9 (Environmental Systems Research Institute, Redlands, CA). To investigate the relationship between PD distribution and the surrounding environment, all vineyards and citrus groves in the Coachella Valley were mapped with ArcView 3.2 (Environmental Systems Research Institute) based on 1-m resolution aerial digital imagery of the Coachella Valley (Landiscor Aerial Information, Phoenix, AZ). Field information such as grape variety was collected from growers. When PD was identified in a vineyard, a GPS was used for more detailed mapping of the vineyard, and the types of surrounding vegetation were recorded.
Spatial distribution of PD was characterized and mapped with semivariograms and SADIE. A semivariogram is a graph of the spatial dependence and plots one-half of the squared difference of a sample pair (semivariance) against the distance between two points (Davis 1994). Spatial dependence means that two sample values that are close to one another tend to be more similar than two values farther apart (Isaaks and Srivastava 1989). Such spatial dependence can be quantified with a semivariogram function defined as
where γ (h) is one-half of the variance of two sampled values at sample distance, or lag distance, h; Z(xi) is the measured sample value at sample point xi; Z(xi+h) is the sample value at xi+h; and n(h) is the total number of sample pairs for any h. The parameters of the semivariogram include the range, sill, and nugget (Isaaks and Srivastava 1989, Davis 1994). Range is the distance at which the semivariance reaches a maximum and represents the lag distance beyond which samples are spatially independent. The sill is the value of the semivariance at any distance greater than or equal to the range. The nugget is the value of the semivariance when lag distance equals zero. These parameters characterize the spatial structure of dispersion. A semivariogram for each binomial data set (1 for a diseased grapevine and 0 for a healthy grapevine) was generated using GS+ version 7 (Gammadesign, Plainwell, MI). If the number of sample pairs at a certain lag distance was less than 30, then the semivariogram value for that lag distance was excluded from semivariogram modeling (Journel and Huijbregts 1978). The degree of spatial dependence was quantified by calculating the percent variability explained by spatial dependence: (sill-nugget)/sill × 100. When a spatial trend or drift was detected, directional (anisotropic) semivariances for four directions (0°[north-south], 45° [northeast-southwest], 90° [east-west], and 135° [northwest-southeast]) as described in Isaaks and Srivastava (1989) were calculated. The best semivariogram model was selected based on r2 values for semivariogram fit.
SADIE was used to map and test the significance of clustering in the PD distribution within vineyards. A cluster can be represented as a region of relatively large counts close to one another (a patch) or a region of relatively small counts, including zeros, close to one another (a gap) (Perry 1995). To quantify the degree of PD clustering, the overall aggregation index (Ia) described by Perry (1995) was calculated. A value of Ia = 1 suggests a spatially random pattern; Ia > 1 suggests a more aggregated pattern; and Ia < 1 suggests a more regular pattern (Perry et al. 1999). The associated probability (Pa) was calculated from formal randomization tests (Perry et al. 1999) under the null hypothesis that observed counts were arranged randomly among the given sample locations. SADIE statistics were generated with SADIEShell version 1.22 (Rothamsted Experimental Station, Harpenden Herts, UK), and ArcGIS was used to generate maps of PD clustering. If patches were identified, then the distribution pattern of PD was confirmed by taking leaf samples from all grapevines within the patch as described by SADIE.
Results
During the three-year survey, a total of 97 diseased grapevines from seven vineyards with four varieties (Table 1⇓) were confirmed (Figure 1⇑). Five vineyards were located in the eastern side of the valley and two in the western side of the valley. The vineyards with PD were studied further to characterize the within-vineyard spatial structure. All diseased grapevines identified from the survey were highly localized within vineyards that were adjacent to citrus; the mean distance between a citrus orchard and the closest diseased grapevine in the vineyard was 143 m ±86.9 (SD). Spatial distribution of PD showed that ~82% of PD-confirmed grapevines were located near two to six consecutive grapevines with missing, dead, or replanted grapevines in a row. Three example distribution maps are shown in Figure 2⇓.
The PD incidence in all vineyards except vineyard C was <0.01%. Spatial analyses with geostatistics and SADIE were not applicable because of the low disease incidence in those vineyards. Vineyard C had exceptionally high disease incidence (3.8%) compared with other vineyards. This vineyard was relatively flat with 2 m of maximum elevation difference, and it was surrounded by palm trees to the east, citrus and grapes to the north, and arid mountains with desert saltbush scrub to the west and south (Figure 3a⇓). Geostatistical analysis showed that a spatial trend was present, indicating PD incidence systematically increased toward one direction within the vineyard. Because a spatial trend existed, referred to in geo-statistics as anisotropy, directional semivariograms were constructed. Directional semivariograms showed that the nonlinear spatial trend existed only in the 90° direction (Table 2⇓). The fit with a power model indicated a strong aggregation with spatial trend, and the PD incidence increased from west to east in the vineyard (Figure 3a⇓). Nugget models (linear models without slopes) fit the semivariograms for other directions (0°, 45°, and 135°), indicating that the analyses failed to show a departure from randomness. SADIE also showed that the distribution pattern of PD in the vineyard was aggregated (Ia = 2.12, p = 0.013) with a significant gap in the western part and a patch in the eastern edge of the vineyard near palm trees and desert (Figure 3a⇓). The distribution pattern of PD within this patch further was investigated by taking leaf samples, regardless of symptoms, from all grapevines within the patch (12 rows by 16 grapevines, n = 192 grapevines; Figure 3b⇓) and testing with ELISA. Pierce’s disease incidence in the patch was 28.0% and SADIE revealed that diseased grapevines were significantly aggregated (Ia = 1.49, p = 0.027). A spherical model fit the semivariogram, and 54.7% of spatial variability of PD distribution in the patch was explained by spatial dependence. The range of the semivariogram showed that diseased grapevines within 26 m were spatially correlated, suggesting that grapevines within 26 m of an infected grapevine have a high chance of being infected with X. fastidiosa.
Discussion
The vineyards with diseased grapevines were widely distributed across the Coachella Valley, and the main similarity among vineyards was their relatively close proximity to citrus, which is known to be the preferred overwintering and reproductive host for H. coagulata (Blua et al. 1999, 2001). The vineyard with the highest PD incidence (vineyard C) also was bordered by desert vegetation, potentially implicating the native smoketree sharpshooter, Homalodisca liturata Ball (Hemiptera: Cicadellidae), as an important vector. The grape varieties in vineyards with PD were among the most common in the Coachella Valley; Flame Seedless and Perlette are the most abundant varieties and minor varieties include Thompson Seedless, Superior Seedless, Beauty Seedless, Princess, Fantasy, Red Globe, and Black Emerald. Although relative susceptibility of winegrape varieties to PD has been reported (Raju and Goheen 1991, Krivanek and Walker 2005), little is known about the relative susceptibility of table grape varieties to PD.
Pierce’s disease incidence and the vector population are currently low in the Coachella Valley, which might be an ideal time to remove all diseased grapevines to eradicate or minimize PD spread in the future. Diseased grapevines found from this study were removed to prevent further spread in the Coachella Valley because no effective cures for X. fastidiosa-infected grapevines are currently available (Blua et al. 1999). In addition to removing diseased grapevines, a chemical control program for H. coagulata was initiated by the California Department of Food and Agriculture in February 2003 targeting H. coagulata in citrus. After the control program began, a 99.7% reduction in mean trap catches of H. coagulata was reported (Park et al. 2006).
Xylella fastidiosa can be transmitted by more than 20 species of xylem fluid-feeding sharpshooters (Hemiptera: Cicadellidae) and spittlebugs (Hemiptera: Cercopidae) (Purcell and Frazier 1985). Historically, PD in California was linked to the blue-green sharpshooter (Graphocephala atropunctata Signoret) along the north coast and the green sharpshooter (Draeculocephala minerva Ball) and the red-headed sharpshooter (Xyphon fulgida Nottingham) in central California vineyards (Purcell and Frazier 1985, Blua et al. 1999). Observational studies (Hewitt et al. 1949, Purcell 1979) have shown that PD occurred on the edge of vineyards adjacent to riparian areas, pastures, and weedy hayfields where the vector insects overwintered and reproduced. Tubajika et al. (2004) reported spatial aggregation of grapevines with PD symptoms in the lower San Joaquin Valley where they thought H. coagulata was the vector. In the Coachella Valley, H. coagulata and H. liturata are considered the major vector insects of X. fastidiosa. Compared to other vector insects, H. coagulata and H. liturata are extremely mobile (Blua et al. 1999) with the capacity to disperse over large distances (Blackmer et al. 2004). They also have a wide host range that includes many common ornamental and crop plants (Powers 1973, Hoddle et al. 2003). Thus, the within-vineyard distribution of X. fastidiosa transmitted by these insects might vary depending on surrounding environment and disease incidence.
Current PD sampling relies solely on census of vineyards and no alternative PD sampling programs are available for growers and researchers. Results from the present study have two implications for sampling and monitoring PD. First, conventional sampling assumes independent samples, that is, samples are not spatially correlated. This study showed that diseased grapevines were aggregated when disease incidence was high (vineyard C). In these cases, biased sampling in the aggregated and nonaggregated areas should be avoided. With spatial sampling and semivariogram modeling, diseased grapevines in the patch were spatially correlated within 26 m (Table 2⇑), indicating that there is a high chance of finding other diseased vines within 26 m of a diseased grapevine. Conversely, two sample points >26 m apart are predicted to be spatially independent. This result is consistent with Perring et al. (2004), who found spatial autocorrelation within about 23 m from vineyards in the lower San Joaquin Valley. These spatial dependences suggest the minimum distance of 23 to 26 m to take independent samples for PD or the size of a grid for detailed spatial sampling. Second, the results of this study provide the foundation for developing sampling strategies for landscape-scale PD monitoring.
This study showed that diseased grapevines were spatially associated with two or more dead, replanted, or missing adjacent grapevines. These “signatures” could be used to locate priority areas for sampling grapevines to test for X. fastidiosa infection. The identification of these signatures using remote sensing technology is under evaluation. If successful, this will greatly increase the efficiency of PD-monitoring programs over large geographic regions.
Conclusion
Determining the location of grapevines with PD in vineyards is a major objective of growers and researchers. However, field census is the only sampling protocol currently available. This study suggests that spatially oriented sampling protocols to locate diseased vines could be developed by using signatures of PD and by describing spatial structures of PD incidence in vineyards.
Footnotes
Acknowledgments: This research was funded by the University of California Pierce’s Disease Program, California Department of Food and Agriculture, and Desert Grape Administrative Committee in the Coachella Valley.
- Received July 2005.
- Revision received October 2005.
- Copyright © 2006 by the American Society for Enology and Viticulture