Jones JON
209 Life Science Blvd
Dept. of Entomology & Plant Pathology
Auburn University
Auburn, AL 36849
(phone) 3348441956
(fax) 3348441947
lawrekk@auburn.edu
Effects of Pot Material and Soil Volume on Rotylenchulus reniformis and Meloidogyne incognita Population Development
J. R. Jones^{1}, K. S. Lawrence^{2}, E. van Santen^{3}
___________
^{1}National Soil Dynamics Laboratory, USDAARS, Auburn, AL 36832; ^{2}Department of Entomology and Plant Pathology, Auburn University, Auburn, AL 36849; ^{3}Agronomy and Soils Department, Auburn University, Auburn, AL 36849.
Corresponding author: lawrekk@auburn.edu, (phone) 3348441956, (fax) 3348441947
Abstract: The impact of pot material and soil volume on Rotylenchulus reniformis and Meloidogyne incognita race 3 population densities was evaluated in greenhouse tests. In all experiments, the treatments were arranged as a factorial, where (i) pot materials were clay, polypropylene, or polystyrene and (ii) soil volumes per pot were 90, 150, 250, 500, 750, or 1,000 cm^{3}. Based on the reproductive factor (RfF) values, polystyrene pots resulted in produced higher populations of R. reniformis than either the clay or polypropylene. Populations of R. reniformis in clay pots were 41% lower than those in polystyrene but 25% higher than those produced in polypropylene pots. The mean Rf values ranged from 65, 39, toand 29 in the polystyrene, clay and polypropylene pots, respectively, over all soil volumes.
The Rf values for M. incognita consistently increased with increasing soil volumes of 250 cm^{3} and higher regardless of the pot material. The mean Rf values for clay, polypropylene, and polystyrene pots were 57, 44, and 43, respectively. The 150 cm^{3} ConetainerConetainer^{TM}s produced greater numbers (P ≤ 0.05) of R. reniformis vermiform life stages and eggs per gm of root than all 15 of the pot material and soil volume combinations. Opposite reactions were observed between M. incognita J2 and eggs. J2’s declined while eggs per gm of root increased with increasing soil volumes. For each species we combined data across runs (20 replicates total) and resampled them 1000 times simulating experiments with 2 to 13 replicates to arrive at a 90th percentile upper limit for the coefficient of variation (CV). Six replications replicates resulted in a CV < 15% and <10% for R. reniformis and M. incognita respectively, thus this should be adequate to evaluate population increases in the greenhouse.
Key words: Confidence intervals, pot materials, greenhouse screening, Meloidogyne incognita, SAS^{®} PROC GLIMMIX, reniform nematode, rootknot nematode, Rotylenchulus reniformis, soil volumes, standard errors, studentized residuals.
The major plantparasitic nematode pests of cotton (Gossypium hirsutum L.) are Rotylenchulus reniformis (Linford and Oliveira) and Meloidogyne incognita (Kofoid and White) Chitwood. Rotylenchulus reniformis has been found in every cottonproducing state east of New Mexico, and Meloidogyne incognita is found ininfests all cottonproducing states across the cotton belt in the US. (Lawrence and McLean, 2001).
Greenhouse screening procedures involving these nematodes are an important component in nematology research. The need for a standardized greenhouse nematode screening method has become apparent as greater emphasis has been placed on breeding cotton lines for resistance, while procedures for greenhouse evaluations remain unstandardized. Differences in pot type and soil volume may influence the final results and conclusions of experiments. In the recent literature, wide ranges of pot types and soil volumes have been used with numerous crops in greenhouse assays. Bouton et al. (1989) used 70 cm^{3 }polystyrene trays to screen alfalfa cultivars for resistance to M. incognita. Davis et al. (1996) evaluated selected soybean germplasm for resistance to Meloidogyne species, Heterodera glycines, and R. reniformis in 15cmdiam. clay pots. Robinson and Percival (1997) used 500 cm^{3} of soil in 15cmdiam. plastic pots to evaluate resistance of wild cotton accessions to M. incognita and R. reniformis. Fernandez et al. (2001) selected plastic pots containing 700 cm^{3} of soil to evaluate induced soil suppressiveness to M. incognita. CervantesFlores et al. (2002) assessed resistance screening efficiency of sweet potato cultivars to Meloidogyne spp. using 400 cm^{3} square pots and 150 cm^{3} plastic ConetainerConetainerTMs. Diez et al. (2003) used 11.5cmdiam. clay pots filled with 500 cm^{3} of soil to evaluate competition of M. incognita and R. reniformis. Thus, previous research has been conducted with a wide variety of pots and soil volumes. In order to compare and rely on potscreening results, the effects of these different factors should be known.
This study focuses on improving greenhouse screening methods in which nematodes are involved. Three common pot materials, clay, polypropylene, and polystyrene, were selected for the experiments. Clay pots are often standard greenhouse growth pots but require washing and sterilization as far as pathogens are concerned. Both processes are time consuming and expensive. Polypropylene pots are common in the greenhouse nursery industry but tend to have a short shelf life, require washing, and can not be heat sterilized. Polystyrene pots are available through the food service industry, are inexpensive and designed for onetime use.
The specific objectives of this study were to: 1) evaluate pot types and soil volumes in the greenhouse to determine their effects and interactions on R. reniformis and M. incognita population development; 2) determine the optimum number of replicates for these nematodes in greenhouse studies; 3) determine if polypropylene ConetainersConetainers™ (Stuewe & Sons, Inc. Corvallis, OR) are a realistic alternative to the standard greenhouse pots; and 4) determine objective rules based on which “true” outliers may be detected.
MATERIALS AND METHODS
Pot material and soil volume: Greenhouse experiments were conducted to evaluate the effects of soil volume and pot type on Rotylenchulus reniformis and Meloidogyne incognita race 3 populations. Separate experiments were conducted for both R. reniformis and M. incognita and each of these experiments was repeated once. In all four experiments, the treatments were arranged in a factorial design where pot materials were (i) clay, polypropylene, or polystyrene, and (ii) soil volumes per pot were 90 cm^{3}, 150 cm^{3},
250 cm^{3}, 500 cm^{3}, 750 cm^{3}, or 1,000 cm^{3}. Polypropylene Ray Leach Conetainers™ (Stuewe & Sons, Inc. Corvallis, OR) at the 150 cm^{3} soil volume was the only “extra” treatment in this augmented factorial treatment design. Each test thus contained 160 experimental units (150 for the complete factorial plus 10 Conetainers™) and was repeated once for a total of 320 experimental units and data values for every parameter measured for each per nematode species. Polypropylene Conetainers (Stuewe & Sons, Inc. Corvallis, OR) at the 150 cm^{3} soil volume was the only “extra” treatment in this augmented factorial treatment design. The experimental design for each run was a randomized complete block with 10 replicates. The experiments were conducted in a greenhouse between March and September, where ambient temperatures ranged from 25°C to 32°C. The soil used in these experiments was classified a loamy sand (72.5% sand, 25% silt, 2.5% clay; pH 6.4) obtained from the Field Crops Unit of the E. V. Smith Research and Extension Center, Shorter, AL. Soil was collected from the top 20 cm of the soil profile and sieved to remove large particles. Soil was sterilized by autoclaving at 121°C and 103 kPa for 2 hrs on two consecutive days. Rotylenchulus reniformis and M. incognita populations were maintained on a susceptible cotton cultivar, Paymaster 1218 BG/RR. Nematode inoculum was extracted from the soil by combined gravity screening (specific gravity = 1.13) and sucrose centrifugal floatation (specific gravity = 1.13) and enumerated with a stereo microscope (Jenkins, 1964). Rotylenchulus reniformis and M. incognita eggs were extracted from the roots by shaking for 4 min in a 0.6% sodium hypochlorite (NaOCl) solution (Hussey and Boerma, 1981).
Each test was planted with the cotton cultivar Paymaster 1218 BG/RR. Three seeds were planted in each individual pot. A border of 500 cm^{3} polystyrene pots containing a single cotton plant each was placed around the test to reduce the potential of environmental influence.
Seven days after planting, pots were thinned to 1 seedling/pot. After thinning, a depression (1cmdiam. and 5cmdeep) was made approximately 1 cm away from the cotton seedling. A suspension of 1,000 R. reniformis juvenile and vermiform adults or M. incognita J2 and eggs was pipeted into each depression, which was then filled with soil to prevent desiccation. Tests were initialed weekly and ran concurrently with supplemental lighting (1000W mental halide lamps with a 15h photoperiod) and temperature range from 21 to 32 C.
Plants were grown in the greenhouse for 60 d following inoculation. Pots were watered manually as needed and fertilized weekly using Peters 201020 watersoluble fertilizer (BWI, Jackson, MS). At harvest, vermiform life stages of the nematodes and eggs were extracted from the soil and plant roots as previously described. Rotylenchulus reniformis and M. incognita eggs were stained to facilitate counting using 20 ml of a 5% red food coloring (# 40) solution and microwaved for approximately 2 min and 35 sec. Cotton plant height, shoot fresh and dry weights, and root fresh and dry weights were also recorded.
Analysis of the complete factorial: Generalized linear models (GLM) methodology with the lognormal distribution function was employed to analyze the data (SAS 9.1, Cary, NC). These models extend mixed models methodology to include distributions from the exponential family of distributions. The normal (Gaussian) distribution and the lognormal distribution are members of this family. GLM models consist of three parts: (1) the random component, which deals with the conditional distribution of the dependent variables, given a set of independent predictors; (2) the linear function of the independent variables (the classical model statement); and (3) an invertible link function g(μ_{i}) = η_{i} , which transforms the expectation of the response to the linear predictor (McCullagh and Nelder, 1990). What makes the GLM approach different from the traditional transformation approach to nonnormal data is that the link function and the conditional distribution of the dependent variable are separated. Response variables were vermiform/J2counts per 100 cm^{3} soil volume, eggs per gram root fresh weight, and the reproductive factor (Rf= final population/initial population). All counts were increased by 0.5 to avoid discarding zero counts. The pot materials factors (clay, polypropylene, and polystyrene) and soil volume (90, 150, 250, 500, 750, and 1,000 cm^{3}) and their interaction were considered to be fixed effects, whereas run or repeat experiments and blocks were random effects. Distributional characteristics of the datasets were assessed using the studentized residual graphics panel and fit statistics in SAS PROC GLIMMIX. The lognormal distribution provided the best fit for the current dataset. Once the distributional characteristics were ascertained, the effect of soil volume was modeled directly in PROC GLIMMIX by treating it as a covariate in the analysis. A backwards selection procedure was adopted beginning with the most complex probable model, which included the main effect for pot material and linear and quadratic effects for soil volume plus their interactions (Littell et al., 20060). Nonsignificant terms (P > 0.15) were then excluded and the new model fitted. This process was repeated until all terms were significant at P 0.10.
Removing extreme observations: Because 10 replicates were available for each pot material × soil volume combination in each of the four runs of the experiment, the effect of dropping the replicate with the highest count, lowest count, or both from the statistical analysis was then investigated. The approach taken in this study was to predict Rf for pot material at a given level of the covariate, chosen to be a standard soil volume of 250 cm^{3}.
Outlier detection: Discarding extreme observation without regard to statistical necessity is a wasteful practice that should be avoided. Studentized residuals can be a decision making tool to manage the elimination of extreme observations. Studentized residuals are created by dividing each residual by the overall standard deviation among residuals:
e_{i} is the i^{th} residual, σ^{2} is the estimated variance among residuals, and h_{ii } is the i^{th} diagonal element (ranging between 0 and 1) of the leverage (or “hat”) matrix H. It measures the influence of the i^{th} observation in the matrix and helps to identify influential observations (Littell et al., 2006). Studentized residuals will approximate a normal distribution with mean 0 and variance 1 when residual degrees of freedom for a model get large._{ }An observation was classified as an outlier if its studentized residual exceeded the value ± 3.0.
Number of replicates: For the evaluation of the number of needed replicates, the 20 observations for a given nematode species × life stage x pot material x soil volume combination were treated as coming from a single experiment. The SAS procedure SURVEYSELECT (http://support.sas.com/onlinedoc/913/docMainpage.jsp; verified 30. April, 2007) was then employed to resample the dataset 1000 times using sample sizes from 2 to 13. We then calculated treatment (pot material × soil volume) least squares means and associated standard errors for each replicate of the sampled dataset. From this we calculated the coefficient of variation defined as
CV = 100 * Standard Error / Treatment Mean.
From the 1000 replicate samples, the 90^{th} percentile for the coefficient of variation for sample sizes n = 2 to 13 was calculated.
ConetainersConetainers™ were compared: Lastly, we tried to determine how reliable nematode counts obtained from 150 cm^{3} polyethylene ConetainersConetainers™ were compared to the other pot materials. In order to answer this question, we analyzed the dataset as having 16 treatments (3 pot materials × 5 soil volumes plus Conetainer Conetainers™), ignoring the augmented factorial structure, and calculated differences between the ConetainerConetainer ™ control and the remaining 15 treatments using Dunnett’s test.
RESULTS
Pot material and soil volume: The simplest relationship between the pot material and soil volume was observed for the R. reniformis vermiform life stages (Figure 1, Table 1). The differences in R. reniformis vermiform life stage numbers standardized to 100 cm^{3} over all soil volumes consistently decreased linearly with increasing soil volume in all three pot types. The equivalent performance in R. reniformis vermiform life stage numbers is evidenced by the common slope of the regression lines for all three pot materials (Fig. 1). The counts of vermiform life stages of R. reniformis from polystyrene pots were significantly (P ≤ 0.012) higher than counts for clay and polypropylene pots (Table 2). The difference between clay and polypropylene was significant at P = 0.073. Rotylenchulus reniformis total vermiform life stages numbers averaged over all soil volumes ranged from 2314 for polypropylene, 3242 for clay, and 5359 for polystyrene pots (Table 3).
Rotylenchulus reniformis egg numbers per gram of root for pot materials fit a quadratic model with a linear and quadratic interaction between pot materials and soil volume (Table1, Fig. 1). The polystyrene pots at soil volumes from 250 to 1000 cm^{3 }contained more eggs per gram of root than clay and polypropylene pots. Linear and quadratic regression coefficient estimates for this pot material differed significantly (P ≤ 0.043) from polypropylene but not from clay (Table 2). Rotylenchulus reniformis egg number ranged from 1444, 712, and 573 eggs per gram of root in the polystyrene, clay and polypropylene pots, respectively.
The relationship between pot materials and soil volume for R. reniformis Rf values was described with the most complicated model involving all interaction terms (Fig.1, Table 1). Quadratic regressions best illustrated the relationship between soil volume and R. reniformis Rf values (Table 2). Irrespective of pot material, Rfvalues increased until at least 500 cm^{3}. Clay pots differed significantly (P ≤ 0.093) from polypropylene pots for linear and quadratic regression coefficients; none of the other contrasts were significant. The mean Rf values ranged from 65, 39, and 29 in the polystyrene, clay and polypropylene pots, respectively, over all soil volumes.
The relationship between pot materials and soil volume for the M. incognita second stage juveniles (J2) involved a negative linear interaction between pot materials and soil volume (Table 1). Populations of M. incognita J2 when standardized to 100 cm^{3} over all soil volumes decreased as soil volume increased for all pot materials (Fig. 1). Greater numbers of M. incognita J2 were found at the lower soil volume of 90 cm^{3} in the polystyrene pot material as compared to polypropylene and clay. The regression for clay pots did not differ significantly (P ≥ 0.337) from polyethylene pots (Table 2). The intercept for polystyrene pots was greater (P ≤ 0.003) and the slope significantly steeper (P ≤ 0.073) than either of the other two pot materials. The mean J2 count for polystyrene, clay, and polypropylene pots was 2119, 1236, and 1166 J2s per 100cm^{3} soil volume, respectively, over all soil volumes (Table 3). The slopes for the M. incognita J2 numbers versus soil volume in all three pot materials were steeper than those for R. reniformis vermiform life stages (Fig. 1, Table 2).
The relationship between the numbers of eggs per gram of root produced by M. incognita and pot materials and soil volume was similar to that produced by R. reniformis (Table 1). There was significant (P ≤ 0.0001) linear and quadratic interaction between the pot materials and soil volume; thus, the relationship between soil volume and egg populations was best modeled by a separate quadratic regression for each pot material (Table 2, Fig.1). While the overall regression was significant, contrast failed to establish significant differences between pot materials. A look at the graph in Figure 1 shows why this is the case. The mean egg production for polystyrene, clay, and polypropylene pots was 2807, 2392, and 1903 eggs per gram of root, respectively, over all soil volumes (Table 3).
The relationship between the Rf values produced by M. incognita and pot materials and soil volumes was simple (Table 1). There were no significant interactions between the pot material and soil volume. The relationship between soil volume and Rf values was best modeled by a quadratic regression with a common linear and quadratic term for all pot materials; hence, the regression lines in Figure 1 are parallel. Similar to what was seen with the R. reniformis data, Rf values for M. incognita consistently increased with increasing soil volume up to a certain point regardless of the pot material. However, soil volumes greater than 500 cm^{3} did not further increase Rf values. The mean Rf value for clay, polypropylene, and polystyrene pots was 57, 44, and 43, respectively (Table 3).
Removing extreme observations: As might be expected, removal of the replicate with the maximum observation for either vermiform (J2) or egg number decreased the average Rf, whereas removal of the lowest replicate increased it (Fig. 2). However, the change was small, ranging from 97 – 107% compared to using all replicates. Whereas the relationship between removal of observations and magnitude of the mean was predictable and consistent for both species and stages, standard errors decreased in 9/12 cases. If there was an increase it was 3% or less. The maximum decrease in the standard error (9%) for Rf was obtained when the observation with the maximum egg number for M. incognita was removed from the analysis. The effect of removal of extreme observations on the differences between treatment means was even less predictable, but the effects were much larger than for means with relative differences ranging from 82% when both maximum and minimum observations were removed for M. incognita eggs to 134% when the maximum egg count for M. incognita was removed (Fig. 2). The effect of removing extreme observations on standard error of the differences was much smaller, ranging from 96 – 115%. The lack of a clear pattern when extreme values were removed from the dataset reinforces our contention made in the introduction that discarding maximum and/or minimum observations without regard for statistical necessity is a wasteful practice, particularly since there seem to be no predictable benefits.
Outlier detection: The fit of the regression models as judged by Generalized ChiSquare / degree of freedom ratio was quite good (Table 1); a ratio of 1.0 is the target for optimal fit. Hence using studentized residuals in SAS PROC GLIMMIX for outlier detection (> 3 SD) resulted in far fewer observations being discarded from the analysis compared to simply eliminating maximum and/or minimum observations (Table 1). Most outliers were detected for M. incognita (3/300 J2; 1/300 eggs, and 1/300 Rf) than for R. reniformis (1/300 eggs and 1/300 Rf), and all but one were concerned with the minimum observation. This reflects far fewer ‘discarded’ observations than the maximumminimum approach, where either 60 (max. and min. eliminated) or 30 (either max. or min. eliminated) would have been eliminated without regard to the magnitude of the residual. All observation removal had studentized residuals ≤ 3.25, thus the effect of their removal was very small. Again, removal of “extreme” observations using the maximum and/or minimum criterion as compared to the studentized residual criterion ≥ 3.0 is a wasteful practice.
Number of replicates: Treating the 10 replicates for each of the two runs for a given species as coming from a single experiment in the sampling study would tend to maximize the residual error. We thus adopted the 90^{th} percentile for CVs as our evaluation criterion. The effect of the number of replications replicates on R. reniformis vermiform life stages and M incognita J2’s and Rf values for both species was consistent over all pot materials and pot sizes (Fig. 3). Six replicates in a greenhouse test resulted in a coefficient of variation (CV) < 20 and 10% for R. reniformis Rf values and vermiform life stages (Fig. 3). This number should be adequate to evaluate R. reniformis population increases. The CV for M. incognita was always lower than that observed with R. reniformis. Eight replications replicates in a greenhouse test resulted in a coefficient of variation (CV) < 10% for M. incognita Rf values and J2 life stages (Fig. 3).
How do ConetainersConetainers™ compare? The 150cm^{3} polypropylene ConetainerConetainer ™ produced greater numbers (P ≤ 0.05) of R. reniformis vermiform life stages per 100 cm^{3} and eggs per gm of root as compared to the clay, polypropylene, and polystyrene pot materials (Table 3). Fourteen of the 15 pot materials by soil volume combinations produced lower (P ≤ 0.05) numbers of vermiform life stages per 100 cm^{3} as compared to the ConetainerConetainer.™. ConetainersConetainers™ produced greater (P ≤ 0.05) numbers of eggs per gm of root that all 15 of the pot materials by soil volume combinations. Rf values were greater in the ConetainersConetainers™ (P ≤ 0.05) as compared to 90 and 1000 cm^{3} clay pots, 90, 250, 500, and 750 cm^{3} polyethylene pots, and 90 cm^{3} polystyrene pots. The 150cm^{3} ConetainerConetainer ™ produced higher (P ≤ 0.05) or similar numbers of M. incognita J2 in 14 of the 15 pot material and soil volume comparisons (Table 3). Only the 90 cm^{3} polystyrene pots produced greater (P ≤ 0.05) numbers of M. incognita J2 as compared to the ConetainerConetainer.™. The numbers of M. incognita eggs per gram of root were lower (P ≤ 0.05) in 7 of the than the 15 pot material soil volume comparisons, with none being significantly greater than the ConetainerConetainer.™. ConetainersConetainers™Pots produced lower Rf values for M. incognita as compared to the clay, polypropylene, and polystyrene pot material in 12 of 15 comparisons.
DISCUSSION
A general pattern emerges from this analysis. Rotylenchulus reniformis vermiform life stages, eggs per gram of root and Rf values are generally highest for polystyrene pots, followed by clay and polypropylene pots. ConetainersConetainers™ allowed for optimal reproduction of R. reniformis producing greater Rf values than the all pot materials at similar soil volumes. Thus, ConetainersConetainers™ are ideal for screening large numbers of plants optimizing space limitations. Meloidogyne incognita race 3 second stage juvenile numbers, eggs per gram of root and Rf values were not consistently favored by any pot material. Larger volumes of soil produced greater numbers of J2’s and larger Rf values. Pots allowed for adequate reproduction of M. incognita; however, this nematode did not increase to the levels that were observed with R. reniformis. One can only speculate on the reasons for the observed patterns, but it may have to do with changes in soil temperature and moisture over the course of a day. The insulating property of the pot material increases in the order: polypropylene < clay < polystyrene. Thus, one would expect polystyrene to be more stable over the course of a day than the other materials. Maintaining a constant moisture level in each pot is important in nematode reproduction (Rich et al., 1978). In our observations, clay pots tended to dry out more rapidly than polystyrene pots. Polypropylene pots produced the lowest vermiform and egg population densities compared to clay and polystyrene. On the other hand, variability was less in the polypropylene pots due to lower numbers of vermiform life stages, eggs per gm of root, and Rf values for R. reniformis.
Population densities of R. reniformis vermiform life stages and M. incognita J2 per 100 cm^{3} were observed to decrease as pot size or soil volume increased. The increased space within the root system could potentially explain the reduced population size. Higher population densities occurred in the smaller pots in which root densities were greater; thus, the nematodes were proficient in parasitizing the roots. Consequently, low volume polystyrene and 150cm^{3} ConetainersConetainers™ serve as an efficient, consistent, spacesaving potting medium for R. reniformis and M. incognita greenhouse screening evaluations. The increased population numbers for R. reniformis in ConetainersConetainers™ is of particular importance in resistance screenings, where low counts would indicate resistance or tolerance. Utilizing ConetainersConetainers™ tends to minimize greenhouse space, which is always at a premium.
In our study, numbers of M. incognita were similar in the 150cm^{3} ConetainerConetainer ™ and 250cm^{3} clay, polypropylene and polystyrene pots. However, the numbers of R. reniformis vermiform life stages and eggs were greater (P ≤ 0.001) in the 150cm^{3} ConetainerConetainer.™. The root systems of sweet potatoes grown in ConetainersConetainers™ were thicker, longer, and less dense that those formed in 400cm^{3} square polypropylene pots (Flores et al., 2002). The cotton root elongates rapidly (Smith and Cothren, 1999), placing the apical meristem tissues at the base of the ConetainerConetainer,™, 22.9 cm from the point of inoculation within 3 to 5 days after planting. Rotylenchulus reniformis will enter the root and initiate syncytia formation along the entire root surface (Jones and Dropkin, 1975; Diez, et al., 2003), while M. incognita selects the apical meristem root tissues for giant cell induction (Dropkin and Nelson, 1960; Diez et al., 2003). These differences in colonization and feeding preferences could explain the differences in population increases between the two nematodes in the 150cm^{3} ConetainerConetainer.™. The root architecture was limited by the narrow cone shape of the 2.54 × 22.9cm ConetainerConetainer,™, which leads to long thin roots with root tips at the base of the ConetainerConetainer.™. Rotylenchulus reniformis nonselectivity in root tissues colonization and the limited migration distance imposed by the area of the ConetainerConetainer ™ contributed to the greater Rf values and vermiform life stage population densities produced within the ConetainerConetainer ™ as compared to 8 and 13 of the 15 pot material × soil volume combinations, respectively. Thus, the selective root colonization of M. incognita and the greater migration distance to the apical meristem tissues imposed by the ConetainerConetainer ™ probably restricted population densities of M. incognita. The ConetainerConetainer ™ did not produced higher M. incognita Rf values as compared to the 15 pot material × soil volume combinations.
We have shown that a criterion based on studentized residuals is preferable to simply discarding maximum and/or minimum observations for a given treatment. Although the practice of discarding outlier observations is rarely acknowledged in the literature, and thus hard to document, it nevertheless occurs. The advantage of this approach is twofold: (1) a reduction in the number of ‘wasted’ observations and (2) limits that can be set by the researchers, e.g., 2 standard deviation (SD) vs. 3 SD or 4 SD vs. 3 SD, based on professional experience with the particular screening system (e.g., host species, nematode species). As an aside, no observation would have been discarded in our experiment had we set the limit at 3.25 rather than 3.0.
In agronomic research, the number of replicates needed is of utmost importance in greenhouse and field tests. In both types of experimentation, space is usually limited. Increased numbers of experimental replicates and larger pots are generally thought to result in lower coefficients of variation (CV) in the greenhouse; however, larger pots take up more space and result in fewer replicates or treatments. In our study, we found that the lower volume pots (< 500 cm^{3}) were preferred over the higher volumes (> 500 cm^{3}) for increased populations of R. reniformis but not M. incognita in the 60 day test duration. From a population dynamics viewpoint, the significance of the reduction in the CV can easily be appreciated, as these differences tend to increase with increased populations. Six replications replicates in a greenhouse test resulted in a CV < 15% (Fig. 4C). This number should be adequate to evaluate R. reniformis population increases. The CV for M. incognita was always much lower than that of R. reniformis. Smaller CVs optimize the ability of a study to detect differences between treatments in experiments. Conversely, since smaller pots are feasible and six replicates are adequate, the researcher could add another factor or level to an experiment, thereby increasing the scope of inference.
Finally, as shown by Weaver et al. (2007) generalized linear models enable the researcher to choose a conditional distribution from the exponential family that is appropriate for the data at hand rather than making an a priori assumption of using the lognormal or log transformation when a negative binomial distribution would make more sense because the data were overdispersed.
LITERATURE CITED
Bouton, J. H., Hussey, R. S., and Smith, S. R. 1989. Greenhouse screening of alfalfa cultivars for resistance to southern rootknot nematode. Crop Science 29:823825.
CervantesFlores, J. C., Yencho, G. C., and Davis, E. L. 2002. Efficient evaluation of resistance to three rootknot nematode species in selected sweet potato cultivars. Hortscience 37:390392.
Davis, E. L., Koenning, S. R., Burton, J. W., and Barker, K. R. 1996. Greenhouse evaluation of selected soybean germplasm for resistance to North Carolina populations of Heterodera glycines, Rotylenchulus reniformis, and Meloidogyne species. Journal of Nematology 28:590598.
Diez, A., Lawrence, G. W., and Lawrence, K. S. 2003. Competition of Meloidogyne incognita and Rotylenchulus reniformis on cotton following separate and concomitant inoculations. Journal of Nematology 35:422429.
Dropkin, V. H., and Nelson, P. E. 1960. The histopathology of rootknot nematode infestations in soybean. Phytopathology 50:442447.
Fernandez, C., RodriguezKabana, R., Warrior, P., and Kloepper, J. W. 2001. Induced soil suppressiveness to a rootknot nematode species by a nematicide. Biological Control 22:103114.
Hussey, R. S., and Boerma, H. R. 1981. A greenhouse screening procedure for rootknot nematode resistance in soybeans. Crop Science 21:794796.
Jenkins, W. R. 1964. A rapid centrifugalflotation technique for separating nematodes from soil. Plant Disease Reporter 48:692.
Jones, M. G. K., and Dropkin, V. H. 1975. Cellular alterations induced in soybean roots by three endoparasitic nematodes. Physiological Plant Pathology 5:119124.
Lawrence, G. W., and McLean, K. S. 2001. The reniform nematode. Pp. 4244 in T. L. Kirkpatrick and C. S. Rothrock, eds. Compendium of cotton diseases. 2^{nd} ed. St. Paul, MN: APS Press.
Littell, R. C., Milliken, G. A., Stroup, W. W., Wolffinger, R. D., and Schabenberger, O. 2006. SAS system for mixed models, 2^{nd} ed. Cary, NC: SAS Institute, Inc.
McCullagh, P, and Nelder, J. 1990. Generalized Linear Models. London: Chapman and Hall.
Oosterhuis, D. M., and Jernstedt, J. 1999. Morphology and anatomy of the cotton plant. Pp. 175206. In C. W. Smith and J. T. Cothren, eds. Cotton: Origin, history, technology, and production. New York: John Wiley and Sons, Inc.
Rich, J. R., McKinney, H. E., and Thomason, I. J. 1979. Technique for screening cowpea germplasm for resistance to rootknot nematodes. Journal of Nematology 11:201.
Robinson, A. F., and Percival, A. E. 1997. Resistance to Meloidogyne incognita Race 3 and Rotylenchulus reniformis in wild accessions of Gossypium hirsutum and G. barbadense from Mexico. Supplement to Journal of Nematology 29:746755.
Smith, C. W., and Cothren, J. T. 1999. Cotton: Origin, history, technology, and production. New York: John Wiley and Sons, Inc.
Weaver, D. B., Lawrence, K. S., and van Santen, E. 2007. Reniform nematode resistance in upland cotton germplasm. Crop Science 47:1924.
Table 1. Significance of fixed effects using the analysis of covariance approach to estimate regression parameters with the SAS procedure GLIMMIX with a lognormal distribution function. For each response variable we first fitted a full interaction model up to a 2^{nd} order polynomial. Nonsignificant terms (P > 0.10) were removed and the most parsimonious model fitted. In a final step, outliers with a studentized residual > 3.0 were removed from the dataset before refitting the final model.
 R. reniforminsReniform 
 M. incognitaRootknot  
Source of variation  Vermiform  Eggs  Rf 
 J2  Eggs  Rf 
Pot Material (PM)  < 0.0001  < 0.0001  < 0.0001 
 < 0.0001  < 0.0001  < 0.0001 
Soil volume linear (SVL)  0.0002 




 < 0.0001 
PM x SVL 
 0.0007  < 0.0001 
 < 0.0001  < 0.0001 

Soil volume quadratic (SVQ) 





 < 0.0001 
PM x SVQ 
 0.0123  < 0.0001 
 0.0032  < 0.0001 

Gener. ChiSquare / DF  all observations  1.75  1.44  0.94 
 0.61  0.83  0.65 
Gener. ChiSquare / DF  w/o outliers  1.75(0)^{†}  1.40(1)  0.89(1) 
 0.55(3)  0.83(1)  0.62(1) 
^{†} Number of outliers removed is given in parenthesis.
Table 2. Estimates of regression coefficients, associated standard errors, and contrasts among pot materials for the graphs in Figure 1.
 Regression coefficient estimates  Contrast Pvalues  
Term  Clay (C)  Polypropylene (PP)  Polystyrene (PS)  SE  C vs. PP  C vs. PS  PP vs. PS 
R. reniformis  vermiform 





 
Intercept  8.48  8.14  8.95  0.180  0.073  0.012  <0.0001 
Linear  0.00089^{†}  0.000232 


 
R. reniformis  Eggs 





 
Intercept  6.14  5.79  5.96  0.346  0.454  0.703  0.710 
Linear  0.002  0.001  0.006  0.002  0.424  0.155  0.028 
Quadratic  0.0000023  0.0000001  0.0000038  0.000001  0.209  0.432  0.043 
R. reniformis  Rf 





 
Intercept  1.90  1.87  2.35  0.278  0.931  0.228  0.201 
Linear  0.007  0.005  0.006  0.001  0.093  0.519  0.317 
Quadratic  0.000006  0.000003  0.000004  0.000001  0.038  0.315  0.280 
M. incognita  J2 





 
Intercept  7.8859  7.9423  8.5562  0.1741  0.784  0.002  0.003 
Linear  0.00182  0.00214  0.00275  0.000243  0.337  0.007  0.073 
M. incognita  Eggs 





 
Intercept  6.68  7.11  7.06  0.306  0.253  0.323  0.906 
Linear  0.00389  0.00288  0.00405  0.00125  0.556  0.925  0.502 
Quadratic  0.000003  0.000003  0.000003  0.000001  0.610  0.751  0.252 
M. incognita  Rf 





 
Intercept  2.73  2.50  2.54  0.165  0.047  0.105  0.726 
Linear  0.004298^{†}  0.000624 


 
Quadratic  0.000003^{†}  0.000001 



^{†} A common regression line was fit for all pot materials.
Table 3. Comparing counts for ConetainersConetainers™ to other pot materials and soil volumes based on Dunnet’s test. The number for R. reniformis vermiforms and M. incognita J2s are given in count per 100 mL soil. Egg counts are given as number of eggs per g of root fresh weight. The reproductive factor is unitless .
Pot  Soil  R. reniformis  M. incognita  
material  volume  Vermi.  Prob > t  Eggs  Prob > t  Rf  Prob > t  J2  Prob > t  Eggs  Prob > t  Rf  Prob > t  
ConetainerConetainerTM  150 
 12817 
 12284 
 72 
 1859 
 3533 
 19 
 
Clay  90 
 2839  < 0.001  528  < 0.001  9  < 0.001  2400  0.326  1034  < 0.001  16  0.499  
Clay  250 
 5751  0.045  807  < 0.001  45  0.122  1670  0.684  2008  0.053  40  0.006  
Clay  500 
 3056  < 0.001  978  < 0.001  55  0.367  1056  0.030  2863  0.463  73  < 0.001  
Clay  750 
 2505  < 0.001  667  < 0.001  48  0.180  584  < 0.001  3131  0.673  73  < 0.001  
Clay  1000 
 2057  < 0.001  578  < 0.001  35  0.019  471  < 0.001  2926  0.511  81  < 0.001  
Polyethylene  90 
 3290  < 0.001  464  < 0.001  10  < 0.001  2338  0.378  1384  0.001  14  0.218  
Polyethylene  250 
 2482  < 0.001  421  < 0.001  18  < 0.001  1636  0.623  2730  0.369  41  0.005  
Polyethylene  500 
 2625  < 0.001  517  < 0.001  32  0.008  957  0.011  1931  0.036  45  0.001  
Polyethylene  750 
 1384  < 0.001  773  < 0.001  37  0.024  570  < 0.001  2319  0.143  65  < 0.001  
Polyethylene  1000 
 1786  < 0.001  689  < 0.001  47  0.148  330  < 0.001  1149  < 0.001  53  < 0.001  
Polystyrene  90 
 6195  0.068  590  < 0.001  14  < 0.001  6535  < 0.001  1702  0.015  21  0.784  
Polystyrene  250 
 8865  0.355  1208  < 0.001  54  0.337  2031  0.733  2652  0.317  40  0.006  
Polystyrene  500 
 5911  0.053  2985  < 0.001  88  0.491  1030  0.024  3005  0.572  40  0.005  
Polystyrene  750 
 3287  < 0.001  2090  < 0.001  77  0.824  530  < 0.001  4662  0.334  61  < 0.001  
Polystyrene  1000 
 2536  < 0.001  2348  < 0.001  92  0.412  467  < 0.001  2014  0.054  54  < 0.001 
Fig.1. Regression of Rotylenchulus reniformis or Meloidogyne incognita nematode count (vermiform or J2, respectively), egg count, and reproductive factor on soil volume for PM 1218 inoculated at 7 d with 1000 vermiform (J2) life stages and eggs per pot. Regression coefficients are given in Table 2.
Pot material means
Standard error of the mean
0
Pot material differences
Standard error of the difference
0
Fig. 2. Effect of removal of replicate
with the highest count, lowest count or highest and lowest count
(vermiform, J2, eggs) on the relative magnitude of the reproductive
factor (Rf) compared to 10 replicates within a run of the
experiment.
Number of replicates
Fig. 3.
Relationship between number of replicates and coefficient of
variation (CV = 100 * STDERR/MEAN) for R. reniformis
(vermiform, Rf) and M. incognita (J2, Rf) based on sampling
all 20 replicates from two complete experimental runs for each
species 1000 times. Data for each sampling x sample size combination
were analyzed by Proc Mixed and the CV calculated. Data points given
correspond to the 90^{th} percentile of all CVs from 1000
sampling events. The horizontal lines represent CV benchmarks of 20
and 10%.