Derivation and Validation of a Binomial Sequential Decision ... - BioOne

SAMPLING AND BIOSTATISTICS

Derivation and Validation of a Binomial Sequential Decision Plan for Managing Pea Aphids (Hemiptera: Aphididae) as Direct Pests of Dry Pea (Fabales: Fabaceae) in the Pacific Northwest BRADLEY S. STOKES,1 EDWARD J. BECHINSKI,

AND

SANFORD D. EIGENBRODE

Department of Plant, Soil and Entomological Sciences, University of Idaho, 875 Perimeter Drive, M.S. 2339, Moscow, ID 83844-2339

J. Econ. Entomol. 106(6): 2613Ð2620 (2013); DOI: http://dx.doi.org/10.1603/EC13241

ABSTRACT We developed a binomial sequential decision plan that classiÞes the economic status of nonviruliferous pea aphids, Acyrthosiphon pisum (Harris) (Hemiptera: Aphididae), in commercial dry peas, Pisum sativum L. (Fabales: Fabaceae), in the Palouse region of northern Idaho and eastern Washington state. Relationships between mean pea aphid density per plant (x) and the proportion of aphid-infested plants (Pi) were determined by in situ visual counts of 100 plants on each of 27 seasonal dates during 2011 from early vegetative plant growth (stage V105) to late reproductive growth (stage R207) at two Þeld sites near Moscow, ID. The best-Þt Nachman model Pi ⫽ 1 ⫺ exp(⫺0.3616 x0.808) was used to restate the limits of noneconomic and economic infestations from one and three aphids per plant to 30 and 58% aphid-infested plants, respectively. Sequential decision plans were computed using the stop-line formulas of Waters for WaldÕs Sequential Probability Ratio Test. Validation of the sequential decision plan by simulated sampling from the 2011 data as well as from six commercial Þelds sampled during 2012 showed that when observed Þeld densities either exceeded the economic injury level or were less than one third the economic injury level, the plan correctly classiÞed aphid economic status in ⬎99% of the resampling trials. Practical implementation of the plan is discussed. KEY WORDS pea aphid, Acyrthosiphon pisum, dry pea, Pisum sativum, binomial sequential sampling

The Palouse agronomic region of northern Idaho and adjoining eastern Washington annually accounts for 10 Ð30% of total U.S. production of dry peas, Pisum sativum L. (Fabales: Fabaceae), seeded on 40,500 ha and valued at US$20 Ð30 million (U.S. Department of AgricultureÐNational Agricultural Statistics Service [USDAÐNASS] 2013a,b). Pea aphid, Acyrthosiphon pisum (Harris) (Hemiptera: Aphididae), long has been recognized as a key pest throughout the region (Portman and Manis 1954). Infestations annually develop in virtually every commercial dry pea Þeld in the Palouse where they reduce crop yield quality and quantity directly by feeding on phloem sap and indirectly by vectoring two viruses that periodically occur in devastating regional epidemicsÑ bean leaf roll virus (BLRV) and pea enation mosaic virus (PEMV; Clement 2006). Commercial dry pea producers in our region historically have managed pea aphid by using sweepnet sampling and nominal thresholds (sensu Poston et al. 1983, Pedigo 1986) to determine the need for aphid control with foliar insecticides. These subjective thresholds of 30 Ð 40 aphids per sweep (Homan et al. 1992) fail to account for any the dynamic economic 1

Corresponding author, e-mail: [email protected].

factors (such as crop market value, cost of control, and crop yield potential) that together determine the economic injury level (EIL). Accordingly, integrated pest management (IPM) decision tools for pea aphid were ranked among the highest priorities in the Pest Management Strategic Plan for Pulse Crops (chickpeas, lentils, and dry peas) in the United States (Daniels and OÕNeal 2007), a consensus IPM needs-analysis developed by regional growers, pest management advisors, commodity commissions, and others in the cool-season food legume industry. Stokes (2012) responded to that IPM needs-assessment by conducting small-plot Þeld cage studies in north Idaho during 2009 Ð2010 to derive statistical models that described the relationship between yield of U.S. #1 grade dry pea and density of nonviruliferous pea aphids; the resulting EILs ranged from 3 to 19 aphids per plant at the start of crop ßowering, depending on assumptions about cost of control (US dollar per acre), crop market value (US dollar per hundredweight [cwt]), insecticide efÞcacy, and crop yield potential (cwt per acre). However, few if any commercial growers and industry staff in our region make IPM decisions by directly counting aphids on plants. From our personal experience, in situ visual counts tediously can require

0022-0493/13/2613Ð2620$04.00/0 䉷 2013 Entomological Society of America

2614

JOURNAL OF ECONOMIC ENTOMOLOGY

ⱖ15 min for a single plant when pest densities are high and plants are large. Hence, we reasoned that a presence:absence binomial sampling plan in the format of a sequential decision plan (Waters 1955) might present growers and their IPM advisors with a cost-saving alternative sampling method of known precision and accuracy. Binomial sequential sampling plans have been widely recommended for aphids and other pests whose small size and extreme densities can make enumerative plans too tedious for practical IPM decisionmaking (e.g., Wilson et al. 1983, Bechinski and Homan 1991, Boeve and Weiss 1997, Giles et al. 2000, Hodgson et al. 2004, Butler and Trumble 2012). Our objective was to derive and evaluate a presence:absence sequential decision plan that commercial growers and their IPM advisors could use to quickly and accurately classiÞes the economic status of pea aphid infestations in dry pea Þelds in the Palouse production region. We Þrst quantiÞed relationships between aphid incidence and density. We then developed a decision plan using the Sequential Probability Ratio Test (Wald 1945), and formally analyzed plan performance of this plan. Materials and Methods Sampling Plan Development. Development of a binomial sampling program required that the EILs of Stokes (2012) be reexpressed from aphid density per plant (x) to the proportion of aphid-infested plants (Pi). We determined the relationship between aphid density and infested plants by sampling naturally occurring populations of pea aphids in small Þeld plots at the University of Idaho Parker Research Farm 3.2 km (2 miles) east of Moscow, ID, and at the University of Idaho Kambitsch Research Farm 19.3 km (12 miles) south of Moscow, ID. The Parker Research Farm study site was seeded with ÔAragornÕ dry peas on 6 May 2011 at 168 kg/ha (150 lbs per acre) on 15-cm (6-in.) row spacing while Kambitsch Research Farm study site was seeded with Aragorn dry peas on 19 May 2011 at 140 kg/ha (125 lbs per acre) on 25-cm (10-in.) row spacing using conventional tillage practices. Seeds were treated with standard commercial rates of the fungicides mefenoxam and metalaxyl (Maxim [Syngenta Corporation, Wilmington, DE] and Apron [Syngenta Corporation, Wilmington, DE], respectively) and the micronutrient molybdenum (Moly). Crop cultural methods followed standard commercial practices with two exceptions: there were no applications of commercial insecticides, and herbicide applications were limited to spot spraying of weed patches with glyphosate and clopyralid (Round-Up [Scotts Company LLC, Marysville, OH] and Stinger [Dow AgroSciences, Indianapolis, IN], respectively). The sampling universe at Parker was a 1-ha (2.47 acres) square plot within a larger 2-ha (4.94 acres) Þeld, while the sampling universe at Kambitsch was a 46- by 6-m (0.069 acres) rectangular plot. Sampling involved nondestructive visual counts of pea aphids on randomly selected plants; sample-unit size was one randomly selected plant; sample size was 100 U

Vol. 106, no. 6

(plants) each date. Sampling began when plants reached the early vegetative growth stage (stage V105 [5 nodes]) on 7 June 2011 at Parker and on 14 June 2011 at Kambitsch; sampling continued every 3Ð 6 d until the late reproductive growth stage (stage R207 [pod Þll]) on 28 July 2011 (Parker) and on 2 August 2011 (Kambitsch; Knott 1987). This protocol generated 27 observations, 14 dates from the Parker study site and 13 from the Kambitsch study site, respectively. Each sampling date generated an estimate of the mean number of pea aphids per plant and the percentage of aphid-infested plants. The proportion of aphid-infested plants (Pi) was described as a function of mean aphid density per plant (x) by using the empirical model of Nachman (1981) Pi ⫽ 1 ⫺ exp[a ⫻ (x)b] which presumes that the proportion of infested sample units (Pi) increases as an empirical power function of mean density (x). Values for the constants a and b were estimated by linear regression (PROC REG, SAS Institute 2011) with the model ln [⫺ln (1 ⫺ Pi)] ⫽ ln a ⫹ b ln(x), and adequacy was evaluated from the r2 value. Analyses were conducted by pooling together density:frequency data from both Þeld sites (i.e., Parker and Kambitsch) so as to increase sample size and model robustness. We computed Nachman models for four Pi expressions: Pⱖ1 aphid, Pⱖ3 aphids, Pⱖ5 aphids, and Pⱖ10 aphids. These alternative expressions were evaluated because increasing the tally density (T) reduces variance and potential bias in Pi (Nachman 1984, Binns et al. 2000). In particular, there is a threshold density beyond which all plants are infested (i.e., Pi ⫽ 1.0) and the regression model no longer can be used; increasing the tally density extends the range of the model but at the expense of requiring more sampling effort (i.e., time required to assess each plant to determine whether a threshold is exceeded). Southwood and Henderson (2000) pragmatically suggested that restricting Pi values to ⱕ0.8 ensures reliability of estimates. Derivation of binomial sequential decision plans began from the formulas of Waters (1955) for the binomial distribution. Critical pest densities (m1 and m2) were adopted from the EILs of Stokes (2012) by designating the lower limit of an economic infestation (m2) as three aphids per plant and by assigning the upper limit of a noneconomic infestation (m1) as one aphid per plant. The value for m2 was selected because it recommends control action at the smallest EIL derived by Stokes (2012) given conservative assumptions about cost of control (US$10 per acre price plus application expense), crop market value (US$16 per cwt), insecticide efÞcacy (100% control postapplication), and crop yield potential (15 cwt per acre). In contrast, m1 best is considered an arbitrary parameter whose value can be adjusted to optimize sampling plan performance (Pedigo and Zeiss 1996, Binns 2000). We accordingly examined a series of values and ultimately selected m1 ⫽ 1 because the resulting decision plan (in our Extension experience with commercial growers

December 2013

STOKES ET AL.: BINOMIAL SEQUENTIAL PLANS FOR MANAGING PEA APHIDS

2615

Fig. 1. Proportion of dry pea plants infested with at least one pea aphid (Pi ⱖ 1) as a function of mean pea aphids per plant (x) at the University of Idaho Parker Farm and Kambitsch Farm from 7 June 2011 to 2 August 2011. Each data point is the mean of 100 plants. Regressions were computed using data pooled from both Þeld sites (University of Idaho Parker and Kambitsch Farms).

and their IPM advisors) delivered the best compromise between two pragmatic features noted by Binns et al. (2000) as critical to sequential sampling plan performance: maximize representativeness and reliability by specifying a minimum sample size before pest status can be classiÞed while simultaneously minimizing total sampling effort required to classify pest status. Both m1 and m2 were restated as percentage aphid-infested plants. Because commercial pea producers typically consider failure to treat an economic infestation (designated by parameter ␤) more serious

than mistakenly treating a noneconomic population (designated by parameter ␣), error levels ␣ and ␤ were not assigned the same value but instead were speciÞed as 0.01 and 0.001, respectively. Because sequential sampling can continue indeÞnitely if actual pest density falls between m1 and m2, it is necessary to identify a pragmatic maximum sample size for Þeld use of the decision plan. We Þxed maximum sample size as Nmax ⫽ s2/E2x2 (Southwood and Henderson 2000), where s2 is sampling variance, x is mean density, and E is desired sampling precision

Fig. 2. Proportion of dry pea plants infested with at least three pea aphids (Pi ⱖ 3) as a function of mean pea aphids per plant (x) at the University of Idaho Parker Farm and Kambitsch Farm from 7 June 2011 to 2 August 2011. Each data point is the mean of 100 plants. Regressions were computed using data pooled from both Þeld sites (University of Idaho Parker and Kambitsch Farms).

2616

JOURNAL OF ECONOMIC ENTOMOLOGY

Vol. 106, no. 6

Fig. 3. Proportion of dry pea plants infested with at least Þve pea aphids (Pi ⱖ 5) as a function of mean pea aphids per plant (x) at the University of Idaho Parker Farm and Kambitsch Farm from 7 June 2011 to 2 August 2011. Each data point is the mean of 100 plants. Regressions were computed using data pooled from both Þeld sites (University of Idaho Parker and Kambitsch Farms).

expressed by the SE stated as a proportion of mean density. The value Nmax was computed for 25% desired precision (E ⫽ 0.25) using sample mean and variance from the four site:dates during 2011 and 2012 when observed aphid density was between the two class limits. Nmax ranged from 32 to 95 plants, with a mean of 73 plants required for 25% precision. For simplicity and practicality, the maximum sample size was limited to 50 samples for the sequential decision plan. Sampling Plan Validation. Performance was analyzed in the manner of Naranjo and Hutchinson

(1997) by resampling of 15 data sets: 1) nine select seasonal dates from the 2011 Þeld sampling study used to derive the Nachman Pi :x models, and 2) six independent data sets generated by regionally sampling commercial dry pea Þelds during 2012. The former validation trials with the 2011 data involved deliberate selection of nine of the original 27 data sets: three 2011 dates on which observed aphid density exceeded the lower limit of an economic infestation, three 2011 dates on which observed aphid density was less the upper limit of a noneconomic infestation, and three

Fig. 4. Proportion of dry pea plants infested with at least 10 pea aphids (Pi ⱖ 10) as a function of mean pea aphids per plant (x) at the University of Idaho Parker Farm and Kambitsch Farm from 7 June 2011 to 2 August 2011. Each data point is the mean of 100 plants. Regressions were computed using data pooled from both Þeld sites (University of Idaho Parker and Kambitsch Farms).

December 2013

STOKES ET AL.: BINOMIAL SEQUENTIAL PLANS FOR MANAGING PEA APHIDS

2011 dates on which observed aphid density fell between those limits. The latter validation trials with 2012 independent data involved six commercial dry pea Þelds sampled once each between 23 July and 26 July 2012 along an 80.5-km north-to-south transect from Rosalia to Uniontown, WA. Sampling at each Þeld consisted of visually inspecting 100 haphazardly selected plants and recording mean pea aphid density and percentage plants with pea aphids. This pseudorandom process of plant selection was intended to mimic Þeld scouting as practiced by growers and their IPM advisors. Fields represented the diversity of dry pea crop cultural practices and aphid density conditions typical of the Palouse production region. Pea plant growth stage at the six sites ranged from late vegetative to mid-reproductive and mean plant height ranged from 47 to 117 cm; crop variety was not available. Resampling involved drawing random density:frequency observations without replacement each date and tabulating aphid presence:absence against the stop lines of the sequential decision plan until either pest status was categorized as economic:noneconomic or until a maximum of 50 plants had been selected. We wrote a simple Microsoft Excel routine that used the RAND function to select random observations and that calculated mean sample size and the frequency of correct control decisions after 100 simulated sampling runs from each of the 15 data sets (i.e., six 2012 sites and nine 2011 data sets) for, in total, 1,500 resampling runs. Results Nachman regression analyses delivered highly accountable models for the relationship between absolute aphid density per plant (x) and the proportion of aphid infested plants, Pi (Figs. 1Ð 4). Restriction of models to Pi values ⱕ0.8, as suggested by Southwood and Henderson (2000), limits maximum detectable absolute mean pest density to 6.4 pea aphids per plant, 10.8 pea aphids per plant, 16.1 pea aphids per plant, and 24.2 pea aphids per plant for the tally densities ⱖ1 aphid, ⱖ3 aphids, ⱖ5 aphids, and ⱖ10 aphids per plant, respectively. Given current crop market value (US$16 per cwt) and aphid control costs (US$10 Ð14 per acre), Stokes (2012) computed the EIL as 3Ð5 aphids per plant. Hence, any one of the four Nachman models reliably could be used to restate current EIL values as the proportion of aphid-infested plants, but the tally density of T ⱖ1 aphid per plant is best suited for practical Þeld use because it simply requires scoring plants for aphid presence:absence. In contrast, under extreme low crop market price (ⱕUS$6 per cwt) and high aphid control costs (ⱖUS$16 per acre), tally densities of T ⱖ10 aphids per plant are necessary to reliably convert EILs of 17Ð19 aphids per plant (as computed by Stokes 2012) into the proportion of aphid-infested plants. The Nachman model for Pⱖ 1 aphid in Fig. 1 restates the previously designated critical densities m1 ⫽ 1 aphid per plant and m2 ⫽ 3 aphids per plant as 30%

Number plants examined 1 2 3

DO NOT SPRAY RUNNING TALLY: if tally is less than

no. plants with ≥ 1 pea aphid

2617 SPRAY if tally exceeds

__________ __________ __________

4 5 6 7 8

__________

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

__________

__________ __________ __________ __________ __________ __________

1 1 2 2 3 3 3 4 4 5 5 6 6 7 7 7 8 8 9 9 10 10 10 11 11 12 12 13 13 14 14 14 15 15 16 16 17

__________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________ __________

7 7 8 8 9 9 10 10 10 11 11 12 12 13 13 14 14 14 15 15 16 16 17 17 18 18 18 19 19 20 20 21 21 22 22 22 23 23 24 24 25 25 25

designates no decision is possible; continue sampling

Fig. 5. Binomial sequential decision plan for pea aphid management in dry peas at the beginning of plant reproductive growth. “Do Not Spray” and “Spray” differentiate infestations where pea aphid density is no more than one third the EIL (30% plants with ⱖ1 aphids) and at least equal to the EIL (58% plants with ⱖ1 aphids) for ␣ ⫽ 0.01 and ␤ ⫽ 0.001. (Online Þgure in color.)

aphid-infested plants (i.e., Pi ⫽ 0.30) and 58% aphidinfested plants (i.e., Pi ⫽ 0.58), respectively. Substitution of those values into the formulas of Waters (1955) gives the classiÞcation lines for the binomial sequential decision plan as lower classiÞcation limit ⫽ ⫺5.877 ⫹ 0.441x and upper classiÞcation limit ⫽ 3.923 ⫹ 0.441x, where x is the number of plants inspected. Depicted in tabular format for ease of use in

2618 Table 1. plan

JOURNAL OF ECONOMIC ENTOMOLOGY

Vol. 106, no. 6

Results from resampling of select 2011 data and 2012 independent data for validation of the binomial sequential decision

Resampling classiÞcation results (%)a

Observed Þeld sampling data Validation classb

Site

Date

Pi

Pest status

Noneconomic

No decision

Pi ⬍ m1

UI Parker, Moscow, ID UI Kambitsch, Moscow, ID UI Parker, Moscow, ID Chambers, WA Albion, WA McCoy, WA Rosalia, WA Elberton, WA

7 June 2011 14 June 2011 17 June 2011 24 July 2012 26 July 2012 23 July 2012 23 July 2012 23 July 2012 Mean

0.02 0.02 0.21 0.00 0.03 0.03 0.05 0.11 0.06

Noneconomic Noneconomic Noneconomic Noneconomic Noneconomic Noneconomic Noneconomic Noneconomic Noneconomic

0 0 0 0 0 0 0 0 0%

m1 ⬍ Pi ⬍ m2

UI Kambitsch, Moscow, ID UI Parker, Moscow, ID UI Parker, Moscow, ID Uniontown, WA

15 July 2011 20 June 2011 24 June 2011 24 July 2012 Mean

0.32 0.41 0.53 0.34 0.40

Noneconomic Noneconomic Noneconomic Noneconomic Noneconomic

Pi ⬎ m2

UI Kambitsch, Moscow, ID UI Parker, Moscow, ID UI Parker, Moscow, ID

28 July 2011 28 June 2011 20 July 2011 Mean

0.65 0.81 1.00 0.82

Economic Economic Economic Economic

100 100 100 100 100 100 100 100 100% correct (n ⫽ 800) 97 87 61 95 85% correct (n ⫽ 340) 3 0 0 1% (n ⫽ 3)

Economic 0 0 0 0 0 0 0 0 0%

Avg. sample size 14.7 14.7 25.2 14.0 14.9 14.9 15.4 19.6 16.7

3 6 6 5 5% (n ⫽ 20)

0 7 33 0 10% (n ⫽ 40)

33.2 40.4 38.2 35.6 36.9

3 0 0 1% (n ⫽ 3)

94 100 100 98% correct (n ⫽ 294)

18.5 11.3 8.0 12.6

a

Results of 100 resampling runs for each site:date, where n is number of trials in each classiÞcation category. m1 ⫽ 0.30 and m2 ⫽ 0.58, the upper limit of a noneconomic infestation and the lower limit of an economic infestation, respectively, stated as the proportion of aphid-infested plants (Pi). b

the Þeld (Fig. 5), the binomial sequential decision plan requires inspecting at least 14 consecutive individual plants without detection of pea aphids before one can classify a Þeld as noneconomic and inspection of a minimum of eight consecutive plants positive for pea aphids to classify as economic. This asymmetry in minimum sample size reßects the assignment of differing values for ␣ and ␤ error rates. The binomial sequential decision plan performed best when pest density was far below or above the EIL (Table 1). In particular, resampling from the eight 2011Ð2012 data sets where actual aphid pest status was noneconomic correctly classiÞed sites during each of the 800 simulated sampling runs at a mean sample size of 16.7 plants. In the other extreme, resampling from the three 2011 data sets where actual Þeld density exceeded the EIL correctly classiÞed sites 98% of the time (294 of 300). One of the three 2011 data sets (Kambitsch 28 Jul) entirely accounted for the 2% classiÞcation errors: 1% (3 cases in 100 runs) was misclassiÞed as noneconomic and the other 1% (3 cases in 100 runs) could not be classiÞed within the 50-sample limit. Inadequate Þt of the Nachman model contributed to those decision plan errors; actual aphid density of 5.7 aphids per plant at that site exceeded the absolute EIL of 3.0 aphids per plant, but the observed Þeld frequency of 65% infested plants was substantially less than the 77% predicted value. Table 1 also suggests the binomial sequential decision plan will generate higher-than-expected classiÞcation errors as density approaches the EIL. The four 2011Ð2012 data sets where frequency of infested plants fell between the critical m1 and m2 values correctly

classiÞed sites as noneconomic 85% of the time (340 cases of the 400 resampling runs). The decision plan failed to classify 5% of these noneconomic cases within the 50-sample limit and incorrectly classiÞed them as economic infestations in 10% of the simulated sampling runs. MisclassiÞcation of these noneconomic sites as economic was greatest (33 of 100 resampling runs) at one Þeld (Parker 24 Jun) where the observed frequency of aphid-infested plants was 5 percentage points less than the EIL. Overall, the decision plan conservatively classiÞes infestations by recommending control action when pest density is less than the EIL. Discussion Adoption of the binomial sequential decision plan by dry pea growers and their pest management advisors can potentially maximize decision-making accuracy while minimizing sampling effort and sampling time. The plan should prove especially useful to industry Þeld staff that must balance sampling accuracy and efÞciency as they seasonally scout multiple commercial Þelds over extensive geographic areas. Cullen et al. (2000) reported that professional IPM advisors in California expected Þeld scouting programs to correctly make pest control decisions 85% of the time for an investment of 15- to 30-min sampling effort per Þeld. Validation studies here suggest the binomial sequential decision plan should satisfy similar standards for accuracy in the dry pea industry of the PaciÞc Northwest. One practical matter for implementation is how to make IPM decisions in Þelds that remain between the lower and upper decision limits after 50 plants have been inspected. The frequency of aphid-infested

December 2013

STOKES ET AL.: BINOMIAL SEQUENTIAL PLANS FOR MANAGING PEA APHIDS

plants in those cases is 34 Ð50%; from Fig. 1, one can compute absolute aphid density as 1.2Ð2.2 pea aphids per plant. Commercial producers can decide within their own perceptions of risk if these densities justify immediate control action or a reassessment with Þeld scouting after a 3- to 5-d interval. We anticipate that minimal Þeld sampling effort will be required to score plants as aphid infested. Maiteki and Lamb (1987) showed that even an enumerative sequential decision plan for pea aphid in dry peas could provide substantial cost savings over Þxed-size sampling plans. The savings that further accrue to presence:absence plants for aphids have popularized them in the Midwest United States as “speed scouting” (Hodgson et al. 2007). Farmer adoption of new IPM technology depends on many factors, including ease of use and compatibility with current pest management practices (Bechinski 2000). The decision plan can be delivered in several formats, including preprinted pads of tally sheets (e.g., Bechinski and Homan 1991) or as a printable online form (e.g., Ragsdale 2012). Delivery in the Þeld by means of mobile Internet devices (as in the manner of SoyPod DSS, McCornack 2011) is particularly attractive because users could customize decision plans as desired for local conditions while managing their scouting data and accessing IPM advice. Acknowledgments We thank Amelia Jurkowska, Erin Coyle, Hongjian Ding, and Ying Wu for their outstanding help in the Þeld and lab with this project. We also thank Damon Husebye for his superior guidance and insight into pea aphid ecology. We thank all the members of the USA Dry Pea & Lentil Council for their support. U.S. Department of AgricultureÐ National Institute of Food and AgricultureÐRisk Avoidance and Mitigation Program (USDAÐNIFAÐRAMP) award 2008-51101-04522 primarily supported funding for this work, with additional funding from (USDA-NIFAREACCH) award 2011-68002-30191.

References Cited Bechinski, E. J. 2000. Designing and delivering in-the-Þeld scouting programs, pp. 683Ð706. In L. P. Pedigo and D. G. Buntin (eds.), Handbook of Sampling Methods for Arthropods in Agriculture. CRC, Boca Raton, FL. Bechinski, E., and H. Homan. 1991. Sequential decision cards: a tool for Russian wheat aphid control. University of Idaho CIS No. 884, ID. Binns, M. R. 2000. Sequential sampling for classifying pest status, pp. 137Ð174. In L. P. Pedigo and D. G. Buntin (eds.), Handbook of Sampling Methods for Arthropods in Agriculture. CRC, Boca Raton, FL. Binns, M. R., J. P. Nyrop, and W. Van Der Werf. 2000. Sampling and monitoring in crop protection. CABI. New York, NY. Boeve, P. J., and M. J. Weiss. 1997. Binomial sequential sampling plans for cereal aphids (Homoptera: Aphididae) in spring wheat. J. Econ. Entomol. 90: 965Ð975. Butler, C. D., and J. T. Trumble. 2012. Spatial dispersion and binomial sequential sampling for the potato psyllid

2619

(Hemiptera: Triozidae) on potato. Pest Manage. Sci. 68: 865Ð 869. Clement, S. L. 2006. Pea aphid outbreaks and virus epidemics on peas in the US paciÞc northwest: histories, mysteries, and challenges. Plant Health Prog. (doi:10.1094/ PHP-2006-1018-01-RV). Cullen, E. M., F. G. Zalom, M. L. Flint, and E. E. Zilbert. 2000. Quantifying trade-offs between sampling time and precision in commercial IPM sampling programs. Agric. Syst. 66: 99 Ð113. Daniels, C. H., and S. O. Coates. 2007. Pest Management Strategic Plan for Pulse Crops (Chickpeas, Lentils, and Dry Peas) in the United States. United States Department of Agriculture, Washington, DC. Giles, K. L., T. A. Royer, N. C. Elliott, and S. D. Kindler. 2000. Development and validation of a binomial sequential sampling plan for the greenbug (Homoptera: Aphididae) infesting winter wheat in the southern plains. J. Econ. Entomol. 93: 1522Ð1530. Hodgson, E. W., E. C. Burkness, W. D. Hutchinson, and D. W. Ragsdale. 2004. Enumerative and binomial sequential sampling plans for soybean aphid (Homoptera: Aphididae) in soybean. J. Econ. Entomol. 97: 2127Ð2136. Hodgson, E. W., B. P. McCornack, K. A. Koch, D. W. Ragsdale, K. D. Johnson, M. E. O’Neal, E. M. Cullen, H. J. Kraiss, C. D. DiFonzo, and L. M. Behnken. 2007. Field validation of speed scouting for soybean aphid. Plant Management Network. J. Crop Management. (doi: 10.1094/CM-2007-0511-01-RS). Homan, H. W., R. L. Stoltz, and D. J. Schotzko. 1992. Aphids on peas and lentils and their control. CIS No. 748. University of Idaho, Moscow, ID. Knott, C. M. 1987. A key for stages of development of the pea (Pisum sativum). Ann. Appl. Biol. 111: 233Ð245. Maiteki, G. A., and R. J. Lamb. 1987. Sequential decision plan for control of pea aphid, Acyrthosiphon pisum (Homoptera: Aphididae), on Þeld peas in Manitoba. J. Econ. Entomol. 80: 605Ð 607. McCornack, B. 2011. SoyPod DSS information site up and running. (http://mccornack.info/node/130). Nachman, G. 1981. A mathematical model of the functional relationship between density and spatial distribution of a population. J. Anim. Ecol. 50: 453Ð 460. Nachman, G. 1984. Estimates of mean population density and spatial distribution of Tetranychus urticae (Acari: Tetranychidae) and Phytoseiulus persimilis (Acarina: Phytoseiidae) based on the proportion of empty sampling units. J. Appl. Ecol. 21: 903Ð913. Naranjo, S. E., and W. D. Hutchinson. 1997. Validation of arthropod sampling plans using a resampling approach: software and analysis. Am. Entomol. 43: 48 Ð57. Pedigo, L. P. 1986. General models of economic thresholds, pp. 41Ð57. In L. G. Higley and L. P. Pedigo (eds.), Economic Thresholds for Integrated Pest Management. University of Nebraska Press. Lincoln, NE. Pedigo, L. P., and M. R. Zeiss. 1996. Analyses in insect ecology and management. Iowa State University Press. Ames, IA. Portman, R. W., and H. C. Manis. 1954. Idaho Recommendations for Insect Control. University of Idaho, Moscow, ID. Poston, F. L., L. P. Pedigo, and S. M. Welch. 1983. Economic injury levels: reality and practicality. Bull. Entomol. Soc. Am. 29: 49 Ð53. Ragsdale, D. 2012. Soybean aphid speed scouting. (http:// www.soybeans.umn.edu/crop/insects/aphid/aphid_ sampling.htm).

2620

JOURNAL OF ECONOMIC ENTOMOLOGY

Southwood, T.R.E., and P. A. Henderson. 2000. Ecological methods. Blackwell Sciences, Inc., Malden, MA. Stokes, B. S. 2012. Bioeconomics of pea aphid, Acyrthosiphon pisum Harris, on commercial dry peas, Pisum sativum ÔAragornÕ. M.S. thesis, University of Idaho, Moscow, ID. (USDA–NASS) U.S. Department of Agriculture–National Agricultural Statistics Service. 2013a. Crop Production 2012 Summary. USDA, Beltsville, MD. (USDA–NASS) U.S. Department of Agriculture–National Agricultural Statistics Service. 2013b. Crop Values 2012 Summary. USDA, Beltsville, MD.

Vol. 106, no. 6

Wald, A. 1945. Sequential tests of statistical hypotheses. Ann. Math. Stat. 18: 117Ð188. Waters, W. E. 1955. Sequential sampling in forest insect surveys. For. Sci. 1: 68 Ð79. Wilson, L. T., C. Pickel, R. C. Mount, and F. G. Zalom. 1983. Presence-absence sequential sampling for cabbage aphid and green peach aphid (Homoptera: Aphididae) on Brussels sprouts. J. Econ. Entomol. 76: 476 Ð 479. Received 20 May 2013; accepted 3 September 2013.