Charge to the Panel: October 24-26, 2006 FIFRA Scientific Advisory Panel

EPA has prepared a memorandum, dated September 26, 2006, that contains
the Agency’s technical assessment of the likelihood of pink bollworm
resistance under the conditions of Arizona’s pink bollworm eradication
program.  Any reference to figures, tables, specific page numbers, and
appendices (bold type) mentioned below are from this memorandum. 

Estimations of Pink Bollworm Populations Using Pheromone Trapping and
Spatial Analysis.

Geospatial maps of the Bt and non-Bt cotton fields were provided by the
Arizona Cotton Research and Protection Council (Appendix 4). There were
4,626 total fields (approx. 156,000 acres) in the eradication zone. 
There were 334 non-Bt fields (6.92%) and 4,292 Bt fields (93.08%) (Table
2).  Each field is numbered.  These fields are the target areas for the
sterile moth releases, pheromone, and insecticide treatments. There were
a total of 4,541 pheromone traps placed in all fields with 3,541
pheromone traps placed in Bt fields and 1,000 pheromone traps place in
non-Bt fields.  The protocol for the sterile moth releases is found on
p. 11 of EPA’s technical assessment.  The actual sterile moth release
rates through August 25, 2006 are found in Table 4.  The number of traps
per field ranged from 0 to 14.  The scheme for using the trapping and
map data is shown in Figure 5.   

To present the trapping data as a predicted surface of PBW numbers, the
Kriging method was used to calculate a predicted value for areas between
the known values of each field.  Kriging is a regression technique used
in geostatistics to estimate the optimal interpolation of these points
across the spatial domain.  This method handles spatial autocorrelation
and is not sensitive to uneven sampling in specific areas, such as the
distribution of cotton fields in the eradication program.  Ordinary
Kriging using a spherical model was applied to trap counts for each week
(see Volume 2, Table 1 of the submission, MRID# 469048-02 for the data)
to develop a predictive surface model encompassing the cotton fields. 
Kriging constructs a weighted moving average that estimates the value of
a spatially distributed variable from adjacent values while considering
the interdependence.  Kriging results in a smoothing effect in which
high original values are underestimated and low original values are
overestimated.  It is a best linear unbiased estimator because it
minimizes the variance of the estimation errors.    

The Kriging maps of native and sterile PBW populations in Arizona’s
eradication program from June 25 through July 22, 2006 are found in
Figures 6A-H.  This analysis indicates that the sterile PBW adult
populations are more abundant, consistent and more widely distributed
than the native population.  The native populations are limited to 1-5
moths per trap with 3-5 areas as “hot spots” (PBW captures > 25)
during this four week sampling period.  The sterile PBW populations are
more abundant with captures > 50 in many areas.  Early results from the
eradication program indicate that the sterile releases have been quite
successful in reducing native PBW populations.

The Panel is asked to comment on the accuracy and precision of the
estimates of native and sterile pink bollworm population levels in the
eradication zone in Arizona in 2006 using the described pheromone
trapping and boll sampling methods, and spatial analysis (Kriging
method).  Factors that may affect the native and sterile pink bollworm
populations estimates include, but are not limited to:

Number and location of Bt and non-Bt fields 

Size of fields (e.g., 15 ha)

Use of pheromone traps, only males sampled, to estimate overall
population size (and therefore population suppression estimates)

Number and placement of traps within a field

Number of traps in Bt vs. non-Bt fields

Use of the centroid of the field to estimate trap location within a
field vs. exact location of the field using GPS coordinates for the
spatial analysis (Kriging method), e.g., to address “hot spots”
within a field or set of small fields 

Sterile release rates (Bt vs. non-Bt)

Frequency of sterile releases

Predictions about the ratios of sterile: native moths

Estimations of overwintering larvae per field

Boll sampling data (not available for these analyses, too early in the
season)

Please identify major sources of uncertainty in the estimates of PBW
population levels and comment on whether the estimates tend to overstate
or understate actual levels of native and sterile PBW.

Given the discussion in a) above, what suggestions does the Panel have
to strengthen the accuracy and precision of the native and sterile pink
bollworm population estimates?  

Simulation Modeling.

The Agency required that simulation modeling be used to compare the
impact of pink bollworm population suppression vs. resistance risk over
the four-year period of the eradication program.   The simulation model
used was a revised version of the spatially-explicit, stochastic model
discussed in Sisterson et al. (2004).  The simulations examined
population suppression (number of pink bollworm per ha) and risk of
resistance to Bt cotton (rate of increase of resistance allele
frequency).  This model assumes that resistance controlled by a single,
recessive gene.  This model is based on PBW resistance to Cry1Ac. 
Modifications to the model include the release of sterile moths. A
variety of scenarios were simulated using the best estimates of the
parameter values as well as more optimistic and more pessimistic
scenarios.

Preliminary modeling, even  using more “worst-case/pessimistic”
parameter assumptions, predict that the four-year eradication program in
Arizona will suppress pink bollworm without creating a problem with
Cry1Ac resistance to Bt cotton.  In 11 of 12 sets of assumptions
examined, the simulated eradication program eliminated the PBW from the
4096 fields modeled in two years or less without causing a resistance
problem.   In the one exception, PBW was not removed from the region
when the model simulation assumed no release of sterile moths in Bt
fields, 90% Bt cotton, and r = 0.01 in all five replications.  In this
case, the population density declined by 98% (460 final overwintering
larvae per field/29,000 starting overwintering larvae per field) and the
resistance allele frequency increased from 0.01 to 0.02 after four
years, but was still far lower than the 0.50 value typically used as a
criterion for a resistance problem.  The model assumes that population
suppression will occur if the mean PBW density in the region is equal to
or less than 0.1 overwintering larvae per 15 ha (=0.0067 larvae per ha).


The Panel is asked to comment on the certainty of the preliminary
outcomes of the modeling simulations using worst-case assumptions, in
many cases, that pink bollworm populations will be suppressed and there
will be no resistance to Cry1Ac during the four years of the eradication
program simulations.

Dr. Bruce Tabashnik (University of Arizona) plans to conduct additional
simulations using field data collected in 2006 as model inputs in place
of certain assumptions used in the 2005 simulations. 

b)  The Panel is asked to comment on whether there is any reason to
expect these additional simulations of pink bollworm resistance over the
four-year period of the eradication program will change the predicted
outcome to indicate a greater risk of the development of resistance than
seen in the 12 other simulations in which no resistance was seen.

			

PBW resistance to the Cry2Ab2 toxin was not considered in either the
simulation modeling or DNA screening analyses.  Additional consideration
of PBW resistance to the Cry2Ab2 toxin would only be important if the
selection pressure dramatically increases in the next three years, i.e.,
much more Bollgard II planted in the eradication zone.  If some or all
of Arizona’s Bt cotton had two toxins, Cry1Ac + Cry2Ab, evolution of
resistance would be much less likely than it is with only Cry1Ac. 
Modeling resistance to cotton that produces only Cry1Ac is the more
pessimistic scenario.  The modeling predictions (using only Cry1Ac
resistance), therefore, are conservative, i.e., they tend to
overestimate resistance risk.  Based on simulation models examining the
likelihood of insect resistance to pyramided toxins in Bt crops (e.g.,
Roush, 1998; Zhao et al., 2005), even if Bollgard II acreage
substantially increases, the likelihood of PBW resistance to both the
Cry1Ac and Cry2Ab2 toxins would remain low during the four-year PBW
eradication program in Arizona.  

c)  The Agency asks the Panel to comment on the likelihood of Cry2Ab2
resistance given that percentage of Bollgard II planted in Arizona has
been <5% of the total Bt cotton acreage, and simulation modeling
predicts that the likelihood of insect resistance to pyramided toxins in
Bt cotton would be extremely low.

Likelihood of Pink Bollworm Resistance in Future Years of the
Eradication Program in Arizona.

The Panel is asked to comment on the scientific validity of whether the
preliminary field data, spatial analysis, and simulation modeling are
adequate to provide reasonable certainty that the likelihood of pink
bollworm resistance to the Cry1Ac and Cry2Ab2 toxins will be extremely
low during the four years of Arizona’s eradication program.

 PAGE   

 PAGE   1 

