Spatiotemporal Dynamics of Whitefly Bemisia tabaci ...

Sampling and Biostatistics

Journal of Economic Entomology, XX(X), 2018, 1–9 doi: 10.1093/jee/toy110 Research Article

Spatiotemporal Dynamics of Whitefly Bemisia tabaci (Hemiptera: Aleyrodidae) in Commercial Watermelon Crops Carlos H. O. Lima,1 Renato A. Sarmento,1,3 Tarcísio V. S. Galdino,2 Poliana S. Pereira,1 Joedna Silva,1 Danival J. Souza,1 Gil R. dos Santos,1 Thiago L. Costa,2 and Marcelo C. Picanço2 1 Campus Universitário de Gurupi, Universidade Federal do Tocantins (UFT), Gurupi, Tocantins 77402-970, Brazil, 2Departamento de Entomologia, Universidade Federal de Viçosa (UFV), Viçosa, Minas Gerais 36570-900, Brazil, and 3Corresponding author, e-mail: [email protected]

Subject Editor: John Trumble Received 16 January 2018; Editorial decision 2 April 2018

Abstract Spatiotemporal dynamics studies of crop pests enable the determination of the colonization pattern and dispersion of these insects in the landscape. Geostatistics is an efficient tool for these studies: to determine the spatial distribution pattern of the pest in the crops and to make maps that represent this situation. Analysis of these maps across the development of plants can be used as a tool in precision agriculture programs. Watermelon, Citrullus lanatus (Thunb.) Matsum. and Nakai (Cucurbitales: Cucurbitaceae), is the second most consumed fruit in the world, and the whitefly Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae) is one of the most important pests of this crop. Thus, the objective of this work was to determine the spatiotemporal distribution of B. tabaci in commercial watermelon crops using geostatistics. For 2 yr, we monitored adult whitefly densities in eight watermelon crops in a tropical climate region. The location of the samples and other crops in the landscape was georeferenced. Experimental data were submitted to geostatistical analysis. The colonization of B. tabaci had two patterns. In the first, the colonization started at the outermost parts of the crop. In the second, the insects occupied the whole area of the crop since the beginning of cultivation. The maximum distance between sites of watermelon crops in which spatial dependence of B. tabaci densities was observed was 19.69 m. The adult B. tabaci densities in the eight watermelon fields were positively correlated with rainfall and relative humidity, whereas wind speed negatively affected whiteflies population. Key words: Citrullus lanatus, colonization, dispersion, ecology, geostatistics

Understanding the spatiotemporal dynamics of pest insects in crops provides important information that can be incorporated into integrated pest management programs (Vieira et al. 1983, Oliver and Webster 1990, Galdino et al. 2017). Geostatistics is an efficient and increasingly accessible tool for use in spatiotemporal dynamic studies of insects (Galdino et al. 2017). This statistical tool uses a method that characterizes spatial variation by comparing similarities between distant and near points (Matheron 1963, Isaaks and Srivastava 1989). This technique provides results that allow producing maps, zoning the different densities of the pest in the crop. A sequence of these maps throughout the development of the plants may indicate sites that demand greater attention to pest sampling and control. In addition, these studies make it possible to determine the colonization and dispersion patterns of the pest (Rosado et al. 2015, Martins et al. 2017). In the context of precision agriculture, spatiotemporal dynamics studies of crop pests facilitate the reduction of the following:

1) production costs; 2) economic losses; and 3) environmental impacts of the unnecessary use of pest control methods (Pedigo 2002). These advances come as a consequence of these studies, which indicate the places and times more favorable to the pests, and in which parts of the crops it is necessary to control these organisms (Sciarretta and Trematerra 2006, Rosado et al. 2015, Martins et al. 2017). In these spatiotemporal dynamics studies, existing plants in the landscape that can serve as sources of pests for the crops can be identified and located. This determination may be useful to farmers planning the location of crops in the landscape, or even for predicting where the pest attack will start in the crop (Sciarretta and Trematerra 2014, Macfadyen et al. 2015, Martins et al. 2017). Watermelon, Citrullus lanatus (Thunb.) Matsum. and Nakai (Cucurbitales: Cucurbitaceae), is the second most consumed fruit in the world, with an annual production of more than 111 million tons. Watermelon cultivation is carried out in 118 countries, covering the

© The Author(s) 2018. Published by Oxford University Press on behalf of Entomological Society of America. All rights reserved. For permissions, please e-mail: [email protected].

Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy110/4970895 by University of Durham user on 16 April 2018

1

Journal of Economic Entomology, 2018, Vol. XX, No. XX

2 five continents of the world (FAO 2014). Its fruits are mainly consumed with minimal processing, for example, as juices, jellies, sweets, sauces, and salads. They are rich in vitamins (A, B complex, and C), minerals (K, Mg, Ca, P, and Fe), lycopene, and citrulline (an amino acid with medicinal value) (Santos and Zambolim 2011, USDA 2016). The whitefly Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae) is one of the most important pests of watermelon worldwide, causing significant losses to the yield of commercial crops (Santos and Zambolim 2011, Lima et al. 2017). This insect is polyphagous and occurs in areas of temperate, subtropical, and tropical climate (Oliveira et al. 2001, CABI 2015). The whitefly has a wide host range; thus, crops like soybean, cotton, squash, melon, and tomato may serve as alternate host plants to B. tabaci infesting watermelon crops. Bemisia tabaci can transmit plant viruses (e.g., Watermelon chlorotic stunt virus, Squash vein yellowing virus, Cucurbit leaf crumple virus, and Cucurbit yellow stunting disorder virus) and causes physiological disorders in plants via the injection of toxins, reducing the crop yield (Oliveira et al. 2001, Gusmão et al. 2006, Li et al. 2011). The attack of B. tabaci on plants reduces the photosynthetic rate and sugar translocation, causing flower abortion and a decrease in the concentration of soluble solids in fruits (Wintermantel 2004, Gusmão et al. 2005, Stansly and Naranjo 2010). Despite the importance of studying pest spatiotemporal dynamics, as far as we know, there are no published studies on B. tabaci in watermelon crops regarding this subject. Thus, the objective of this study was to determine the spatiotemporal distribution of B. tabaci adults in commercial watermelon crops during the vegetative, flowering, and fruiting stage by using geostatistics.

Materials and Methods Experimental Conditions This work was performed during 2 yr (2014 and 2015) on eight commercial watermelon crops, C. lanatus ‘Manchester’, located in

Formoso do Araguaia (11° 47ʹ48″ S, 49° 31ʹ44″ W, 215 m altitude, tropical climate with dry winter), Tocantins State, Brazil. The characteristics of these fields are shown in Table 1. The cropped land was flat. Cropping areas and their surroundings are shown in Fig. 1. Times of tillage for each crop are shown in Fig. 2. Watermelon (Manchester variety) growing season is 90 d. The development of the crop was divided into three growth stages according to morphological observations: 1) vegetative stage, 2) flowering, and 3) fruiting. The vegetative stage is 27 d after plant emergence. This stage was characterized by the emergence of plants until the beginning of flowering. Thus, plants that did not have flowers, irrespective of the size of their branches or leaf quantities, were considered at this stage. Flowering period is 10 d and was determined by the emission of flowers. At this stage, the plants already have flowers (closed or open), but did not have fruits. We characterized as full bloom when the plants had open flowers. Fruiting was considered as the beginning of fruit formation (45 d after emergence) with a total duration of 19 d and then comes the development of the fruits, which extends from the end of the formation to the harvest, with a total duration of 24 d. At this stage, the plants are still in flowering because watermelon plants do not stop flower production until the end of its cycle. In tropical regions, such as the area where the work was carried out, the first harvest occurs 70 d after plant emergence and the second on the 90th day. The spacing used was 2.80 × 1.45 m, and plants were subjected to recommended management techniques (Santos and Zambolim 2011, Lima et al. 2017). In the landscapes where the eight crops of watermelon were found, other crops and existing natural vegetation were identified and georeferenced. The eight crops (Fig. 1A–H) had different cultivation histories. In the crops A, B, C, and D, the producers used to grow soybean during the harvest season and watermelon during the intercrop period. Crops E and F were previously used as pasture. Crops G and H were cultivated with rice during the harvest period and watermelon during the intercrop period.

Table 1. Characteristics of the watermelon fields and the mean number of of adult B. tabaci (adults, tomato per leaf) at different phenological stages

Crop ID

Latitude

Longitude

Relative distance

Plot size

Crop size (ha)

A

11° 46ʹ9.99″S

49° 42ʹ42.76″W

–

6075 m2

25.26

B

11° 51ʹ43.43″S

49° 38ʹ50.08″W

15.6 km

6075 m2

12.22

C

11° 51ʹ38.89″S

49° 39ʹ37.61″W

14.0 km

5775 m2

16.38

D

11° 51ʹ45.86″S

49° 39ʹ38.73″W

14.3 km

5775 m2

8.49

E

11° 51ʹ42.97″S

49° 27ʹ33.88″W

38.70 km

5775 m2

10.00

F

11° 51ʹ54.05″S

49° 27ʹ40.19″W

39.3 km

5775m2

15.00

G

11° 49ʹ52.74″S

49° 45ʹ47.72″W

10.8 km

5775 m2

18.00

H

11° 49ʹ30.91″S

49° 45ʹ38.24″W

10.4 km

5775 m2

18.3


Phenological stage

Adult density (means ± SE)

Vegetative Flowering Fruiting Vegetative Flowering Fruiting Vegetative Flowering Fruiting Vegetative Flowering Fruiting Vegetative Flowering Fruiting Vegetative Flowering Fruiting Vegetative Flowering Fruiting Vegetative Flowering Fruiting

0.67 ± 0.06 1.23 ± 0.10 0.57 ± 0.06 1.07 ± 0.07 0.61 ± 0.05 0.45 ± 0.05 3.21 ± 0.21 5.00 ± 0.43 2.52 ± 0.20 13.14 ± 0.73 11.04 ± 0.85 4.66 ± 0.35 53.26 ± 2.55 11.87 ± 0.52 1.15 ± 0.11 32.72 ± 1.37 13.97 ± 0.54 0.67 ± 0.07 0.04 ± 0.01 0.03 ± 0.01 0.06 ± 0.02 0.00 ± 0.00 0.01 ± 0.00 0.08 ± 0.02


3

Fig. 1. Representations of the eight watermelon crops (A–H) and their surrounding landscapes. Formoso do Araguaia, Tocantins State, Brazil.

The following insecticides were applied on watermelon crops: acetamiprid, azadirachtin, buprofezin, imidacloprid, pyriproxyfen, thiacloprid, thiamethoxam, and orange essential oil.

Evaluation of the B. tabaci Densities and Climatic Elements In each of the eight crops, the adult densities of B. tabaci were evaluated when the plants were at the vegetative, flowering, and

fruiting stages. Pest densities were evaluated in 300 plants per crop. The evaluated plants were positioned along a regular grid pattern throughout the crop, in order to obtain systematized sampling points and to avoid directional tendencies (Fig. 3) (Rosado et al. 2015, Galdino et al. 2017, Martins et al. 2017). The evaluated plants were georeferenced, and a branch was assessed from each. On this branch, we evaluated the sixth most apical leaf, using the direct counting technique; the leaves were carefully handled, and


4


Fig. 2. Stages of plants in eight watermelon crops: daily variation of the mean air temperature, mean wind speed, mean relative air humidity, and total rainfall at these locations.

the adults of B. tabaci were counted. We evaluated this particular leaf, using a direct counting method because these are the ideal sample type and technique for evaluating adult densities of B. tabaci in watermelon crops (Lima et al. 2017). Daily data on climatic elements (mean air temperature, mean relative air humidity, mean wind velocity, and total rainfall) were also collected over the 2 yr in which this work was performed, at a meteorological station located at the study site. From these data, curves showing daily variation in climatic elements were made, throughout the duration of crop plantation in 2014 and 2015. The whitefly density data were also submitted to redundancy analysis (RDA) using Canoco 3.1 software (TerBraak and Smilauer 2002). The exploratory variables were the plant stage (vegetative, fruiting, and flowering) and the climatic elements. The significance of RDA ordering was obtained by Monte Carlo permutation and tested with an F test (α = 0.05). The exploratory variables that contributed to the model’s significance were included in the model. The biplot of the ordering was produced by the software Canodraw 3.0 (TerBraak and Smilauer 2002). In this plot, the vector lengths are proportional

to the importance of the variable. Positively correlated variables had vectors in the same direction, whereas negatively correlated variables had vectors in opposite directions. Uncorrelated variables presented vectors with an angle of 90° between them (Rao 1964).

Analysis of the Spatial Distribution of B. tabaci in Watermelon Crops This analysis was performed using the GS+ Software (Gamma Design Software, Plainwell, Michigan, USA—evaluation version) (Robertson 2008). The semivariograms, which are the geostatistical tool used to evaluate the spatial dependence between two points, were determined (Matheron 1963). The semivariograms were calculated from the data originally sampled, and they allow the detection of differences (using semivariance) between pairs of points sampled, in relation to their distances. This procedure was used to fine-tune the theoretical semivariogram (Warrick and Myers 1987, Rijal et al. 2014). When semivariance increases, this means that there is a spatial dependence relationship between the densities of the studied insect and



5

Fig. 3. Grid of the sampling points at which adult B. tabaci were sampled. Each dot denotes a watermelon plant.

the points sampled. After some distance, the semivariogram curve plateaus; the exact point where the semivariance becomes constant is indicated by the ‘sill’ value (C) (Rosado et al. 2015, Galdino et al. 2017). The equivalent distance to C is represented by the ‘range’ value (A); the range represents the maximum distance between two locations where there is spatial dependence between the data. The height on the y-axis at which the semivariogram originates is called the nugget effect (C0) (Matheron 1963, Vieira et al. 1983, Isaaks and Srivastava 1989, Liebhold et al. 1993). The nugget effect and the sill value were calculated for each of the adjusted model (spherical, exponential, and Gaussian). To check whether the spatial dependence was the same (isotropy) or not (anisotropy) for all directions, semivariograms were constructed for the directions of 0°, 45°, 90°, and 135°. Subsequently, we tried to fit models that best described the spatial dependence of the data (Matheron 1963, Isaaks and Srivastava 1989). The adjusted models were spherical, exponential, and Gaussian (Vieira et al. 1983); all three models were adjusted to the populations of B. tabaci in each crop at the different phenological stages of the crop. To prepare spatial distribution maps of the insect in the crops, kriging was used. For that, a linear regression was performed between the observed values and those estimated by the best semivariogram model. The kriging technique was chosen to give better predictions and smaller variations than other interpolation techniques (Vieira et al. 1983, Oliver and Webster 1990, Galdino et al. 2017). After the map preparation, we performed cross-validation; in this process, β0 (intercept) and β1 (slope) were obtained to adjust the linear regression models (Vieira et al. 1983, Liebhold et al. 1993). For the selection of the best models, the following criteria were used: lower values of intercepts (β0) and residual sum of squares (RSS), and higher coefficients of determination (R2) and slopes (β1) (Liebhold et al. 1993, Galdino et al. 2017). The spatial dependence rate (SDR) for each of the spatial distribution models was determined using the formula: SDR = C0/(C0 + C), where C0 represents the nugget effect and C the sill value. The spatial dependence of each semivariogram was

classified as strong when SDR ≤ 0.25, moderate when 0.25 < SDR ≤ 0.75, and weak when SDR > 0.75 (Cambardella et al. 1994, Sciarretta and Trematerra 2006). Spatial distribution maps of adult B. tabaci were prepared for each crop, and for each stage of the plants (vegetative, flowering, and fruiting). In these maps, the predominant direction of the winds is shown. Additionally, maps were made representing the landscape around each watermelon crop. In the landscape maps, the watermelon crops, other crops, and natural vegetation in the area were represented.

Results Of the 72 models that were processed, 24 were selected. These models were selected because they had the lowest intercept values (β0) and RSS, and highest coefficients of determination (R2) and slopes (β1). Of the selected models, 12 were exponential, 8 Gaussian, 3 spherical, and 1 pure nugget effect (Table 2). All selected models were isotropic (i.e., the spatial autocorrelation was the same in all directions). Differences in the variogram parameters (nugget, sill, and range) were observed in the eight fields. A nugget effect was evidenced by all the variograms. The ranges of the models varied from 1.80 to 19.69 m. The maximum ranges were 10.80, 19.69, and 9.00 for the vegetative, flowering, and fruiting stages, respectively (Table 2). The SDR of the models ranged from 0.0003 to 0.500. It indicates no significant effect of the nugget effect on the interpolation. These proportions showed that the spatial component accounted for 50.00–99.99% of the total spatial variance. Of the selected models, 87.50% presented a strong SDR (