Degree day models to forecast the seasonal phenology of Drosophila suzukii in tart cherry orchards in the Midwest U.S.

Spotted-wing drosophila, Drosophila suzukii (Matsumura) (Diptera: Drosophilidae), is an invasive economic pest of soft-skinned and stone fruit across the globe. Our study establishes both a predictive generalized linear mixed model (GLMM), and a generalized additive mixed model (GAMM) of the dynamic seasonal phenology of D. suzukii based on four years of adult monitoring trap data in Wisconsin tart cherry orchards collected throughout the growing season. The models incorporate year, field site, relative humidity, and degree days (DD); and relate these factors to trap catch. The GLMM estimated a coefficient of 2.21 for DD/1000, meaning for every increment of 1000 DD, trap catch increases by roughly 9 flies. The GAMM generated a curve based on a cubic regression smoothing function of DD which approximates critical DD points of first adult D. suzukii detection at 1276 DD, above average field populations beginning at 2019 DD, and peak activity at 3180 DD. By incorporating four years of comprehensive seasonal phenology data from the same locations, we introduce robust models capable of using DD to forecast changing adult D. suzukii populations in the field leading to the application of more timely and effective management strategies.

Introduction

Spotted-wing drosophila, Drosophila suzukii (Matsumura) (Diptera: Drosophilidae), is an invasive economic pest of soft-skinned and stone fruit in North America, South America, and Europe [1, 2, 3, 4]. Female D. suzukii possess a distinctive, serrated ovipositor that allows them to infest still ripening and ripe fruit [5], unlike other species of drosophilids which can only target overripe, rotting, or damaged fruit. Along with this unique morphological feature, D. suzukii have high rates of reproduction, fast generation times, and quickly adapt to variable climates making them a formidable economic pest [2, 6, 7]. Damage from D. suzukii to susceptible fruit crops in Western U.S. states can result in annual losses up to $511 million [8]. In Minnesota, crop damage resulting from D. suzukii was projected at $2.2 million annually in raspberry alone [9].

Rigorous insecticide application regimens are the most common means of D. suzukii management in agricultural areas of high pest pressure [10, 11]. In the Midwest, frequent insecticide treatments are implemented in D. suzukii affected farms and orchards during the field season with five or more applications under heavy pest pressure [12, 13]. Chemical control is a costly management option not only monetarily, but also environmentally [8, 11, 14, 15]. Growers commonly apply insecticides which specifically target adult D. suzukii and rotate insecticides which incorporate active ingredients that have different modes of action, which helps minimize the number of treatments per season and reduces the risk of developing resistance [12, 13]. A better understanding of this pest’s seasonal phenology and the related underlying mechanisms is an instrumental step in continuously building more efficient and effective integrated pest management strategies against D. suzukii.

Modeling the seasonal population variations of pests from field monitoring data can help determine periods of high pest abundance [16]. Phenology models are particularly effective in describing the population dynamics of pests which maintain relatively few and synchronous generations during the year [17, 18]. Drosophila suzukii, however, is a highly prolific pest which has rapid rates of reproduction and several overlapping generations [7, 19]. Previously developed D. suzukii predictive phenology models navigated this caveat by generating developmental stage-based models [20], or physiological age-structured models approximated by degree-days (DD) [21]. Abiotic factors such as heat units (DD) and relative humidity (RH) are major drivers of D. suzukii development activity [7, 22, 23, 24], and incorporating these effects allows for more descriptive models. Population modeling of D. suzukii can provide a better understanding of this pest’s population trends in a regionally specific sense, having potential management implications such as timing insecticide treatments to coincide with predictions of high pest pressure [20, 21]. Models have previously been produced from data obtained in the Pacific Coast, North Carolina, and Michigan regions [13, 20], but they do not accurately describe D. suzukii population dynamics observed in Wisconsin. No model incorporating DD or RH describing the pest’s seasonal population variation in the Midwest is currently available.

To better understand the complex population dynamics of D. suzukii, we developed two complementing predictive phenology models: a generalized linear mixed model (GLMM), and a generalized additive mixed model (GAMM). Previous research involving elucidating the phenology of insects through population modelling include the use of both linear [25], and additive models [13]. These models are implemented to help observe trends in species phenology, explore the influence abiotic factors have on populations dynamics, and provide general estimates of insect population intensity [13, 16, 20, 25]. In general, GAM models are able to estimate different DD’s corresponding to phenological events such as first detection or peak activity, and GLM models have the capacity to produce trap catch approximations provided a specific DD. Implementing compatible models to emphasize different aspects of D. suzukii’s phenology furthers our understanding of how abiotic factors may driver the pest’s seasonal population dynamics and allows for more precise field population trend predictions that are potentially useful for preemptive management in areas with high levels of D. suzukii overlapping with susceptible crops.

The GLMM and GAMM established from our study is capable of predicting the dynamic seasonal phenology of D. suzukii based on four years of adult monitoring trap data collected in Wisconsin tart cherry orchards during the growing season in relation to DD and RH. By incorporating four years of comprehensive seasonal phenology data from the same locations, we present two models using DD and RH which forecast changing adult D. suzukii populations in tart cherry orchards in Wisconsin.

Materials and methods

Adult seasonal phenology

Field sites

To assess the seasonal phenology of D. suzukii, we sampled four tart cherry orchards from 2015–2018 in northeastern Wisconsin, U.S., located near the urban centers of Maplewood, Sturgeon Bay, Egg Harbor, and Sister Bay (Table 1). Sampling occurred on private land, and permission to conduct research was granted from each land owner of every site. Each year, a total of 13 traps were placed across these locations, and each trap location was re-sampled each year. Three traps were placed in Maplewood, five traps in Sturgeon Bay, two traps in Egg Harbor, and three traps in Sister Bay. All of the orchards grew the ‘Montmorency’ cultivar, while one location grew ‘Montmorency’ and ‘Balaton’. The orchard which grew ‘Balaton’ was retained in the model because the ‘Balaton’ site had an average weekly trap catch throughout the four years (29.95 flies) that fell within the 95% confidence interval of the average weekly D. suzukii trap catch for all the sites throughout the four years (22.26–30.04 flies). Also, both ‘Balaton’ and ‘Montmorency’ cultivars had similar D. suzukii egg and larval infestation levels in lab assays from a previous study [26]. All orchards were conventionally managed for D. suzukii by rotating pyrethroid and organophosphate insecticides, averaging four to five applications throughout the growing season. Adult D. suzukii field monitoring from 2015–2018 typically began in mid-May, and continued until late-August, two weeks post tart cherry harvest.

Table 1

Orchard and weather station locations.

Orchard	Longitude	Latitude	Weather Station	Longitude	Latitude	Distance
A	-87.5059	44.7623	Nasewaupee	-87.5056	44.7597	0.1 km
A	-87.4541	44.7806	Nasewaupee	-87.5056	44.7597	4.5 km
B	-87.4304	44.757	Nasewaupee	-87.5056	44.7597	6.0 km
C	-87.0992	45.2493	Sister Bay	-87.0662	45.2191	4.3 km
C	-87.0984	45.229	Sister Bay	-87.0662	45.2191	2.8 km
C	-87.0941	45.2064	Sister Bay	-87.0662	45.2191	2.6 km
C	-87.2698	45.0549	Egg Harbor	-87.2598	45.0509	0.9 km
C	-87.24	45.0697	Egg Harbor	-87.2598	45.0509	2.6 km
D	-87.3262	44.8773	Sturgeon Bay	-87.3678	44.8935	3.7 km
D	-87.3217	44.8789	Sturgeon Bay	-87.3678	44.8935	4.0 km
D	-87.3185	44.8796	Sturgeon Bay	-87.3678	44.8935	4.2 km
D	-87.323	44.883	Sturgeon Bay	-87.3678	44.8935	3.7 km
D	-87.3257	44.8802	Sturgeon Bay	-87.3678	44.8935	3.6 km

Longitude and latitude of the 13 monitoring traps from the four orchards. The weather station corresponding to each monitoring trap is also provided with its respective GPS location. ‘Distance’ refers to the straight-line distance between the weather station and monitoring trap.

Monitoring traps

Scentry traps baited with Scentry D. suzukii attractant lures (Scentry Biologicals Inc., Billings, MT, U.S.) were used to monitor adult populations of D. suzukii. At the bottom of each trap, a drowning solution containing 200 mL of water, 0.8 g of boric acid, and 2–3 mL of unscented dish soap (Seventh Generation Inc., Burlington, VT, U.S.) was added to kill and preserve collected specimens. Monitoring traps were placed in the lower canopy fruiting zone of tart cherry trees in the interior of each orchard. Every week, the trap contents were emptied and deposited in 70% ethanol, then returned to the laboratory where adult D. suzukii from each trap were counted. The drowning solution was replaced weekly, and the Scentry lures were substituted every four weeks based on the manufacturer’s recommendations.

Temperature and humidity

Temperature data for each site from 2015–2018 were retrieved from the PRISM Climate Group [27] by entering the GPS coordinates of each site. Daily minimum, mean, and maximum temperatures were retrieved each day from January 1st to December 31st for the years 2015–2018, and degree-days (DD) were calculated using a lower threshold of 7.2° C and upper threshold of 30° C, as D. suzukii development ceases beyond these temperature boundaries [7]. Cumulative weekly DD totals were computed for each site from January 1st of each year until the end of D. suzukii monitoring. Relative humidity (RH) data for each site from 2014–2018 were retrieved from the Michigan State Enviroweather website [28]. Data were downloaded from four weather stations located 0.1–6.0 km from the monitoring sites (Table 1). Relative humidity was averaged for each week during the monitoring periods.

Statistical analysis

Generalized linear mixed model

A generalized linear mixed model (GLMM) was generated from D. suzukii adult trap catch, DD, RH, year, and site data. The model is based on negative binomial regression fit by maximum likelihood using the glmer function in the lme4 package of R version 3.5.3 [29]. Year, site, and the year*site interaction were incorporated into the model as random effects, and DD and RH were added as fixed effects. Degree days were divided by 1000 to fit an appropriate scale with the other effects. The GLMM can be described as follows:

In this representation of the model, (Y) is the estimate of weekly adult D. suzukii trap catch as a function of non-linear time (DD) (x) and RH (x) with year (ε), site (ε), and year*site (ε) added as random effects. Coefficients β and β correspond to the trap catch explained by DD and RH respectively, and δ represents the residual error of the trap catch estimation. The GLMM allowed for random effects to be accounted for prior to estimating the regression coefficients. While separating random from mixed effects creates a much more interpretable model, this approach limits its flexibility as the coefficient estimate will be solely increasing or decreasing [30]. The GLMM is most appropriate for generating an anticipated D. suzukii trap count at any given week during the field season. This model as fit will only predict an increasing or decreasing trap catch throughout the entire seasonal phenology of D. suzukii based on changing DD, but can predict a specific trap catch given a unique DD estimate. The GLMM pairs well with the GAMM which provides DD approximations for critical occurrences in the field such as first adult D. suzukii detection, or peak activity.

Model diagnostics

Scatter plots were produced for trap catch at each site from 2015–2018. These scatter plots mirrored quantile-quantile plots of the same data, meaning the effect of site and year could be assumed to be random and of equal variance [30]. The full GLM model (including RH as a fixed effect with all random effects) and the reduced model (removing RH and the interaction between random effects) were compared to determine which model best fit the data. An Akaike information criterion (AIC) and a likelihood ratio test (LRT) were run to determine the best fit model between the two models. The inclusion of all parameters was justified as the full model had slightly more explanatory power when comparing the two models (full model AIC = 4875.7, reduced model AIC = 4883.7, LRT p = 0.002). A lower AIC value indicates a more parsimonious model.

Generalized additive mixed model

A generalized additive mixed model (GAMM) was also generated from the same D. suzukii adult trap catch, DD, RH, year, and site data. The model is based on negative Poisson regression with a log-link using the predict.gam function in the lme4 and mgcv packages of R version 3.5.3. Degree days, year, site, and the year*site interaction were incorporated into the model as random effects, and RH was added as a fixed effect. The GAMM can be described as follows:

In this model, (Y) is used to describe the seasonal pattern of adult D. suzukii through non-linear time (DD/1000) (x) as a random effect and RH (x) as a fixed effect. Coefficient β corresponds to the trap catch explained by RH and f (d) is included in the model as a penalized cubic regression smoothing function of DD. Year (ε), site (ε), and year*site (ε) were also added as random effects, and δ represents the residual error of the trap catch estimation. Using the GAMM, we have the ability to estimate underlying trends of the D. suzukii population throughout the growing season. Overall, the GAMM has less interpretability than the GLMM [30], but has more flexibility in attributing a specific DD to critical points in the seasonal phenology of D. suzukii including first adult detection, above average trap catch, and peak activity. The GAMM complements the GLMM which can predict D. suzukii trap catches during the field season, provided a specific DD.

Scatter plots were produced for trap catch at each site from 2015–2018 which mirrored quantile-quantile plots of the same data allowing us to assume the effect of site and year was random and of equal variance [30]. Similar to the GLMM, the GAM full model (including RH as a fixed effect with all random effects) and the reduced model (removing RH and the interaction between random effects) were compared to determine the best fitting model. A generalized cross-validation (GCV) score was generated for both models using the mgcv package of R version 3.5.3. The GCV score is used to measure model smoothness selection with respect to the smoothing parameters, as well as estimate prediction error [30]. A minimized GCV score indicates a smoother model, and in a sense, a GCV score is comparable to an AIC value in that a lower score equates to a better fitting model. The full GAM model was selected for analysis as overall it had the most explanatory power (full model GSV = 1.63, reduced model GSV = 1.77).

Results

The field populations of D. suzukii generally increased with DD during the field season, then peaked at the accumulation of about 3000 DD when looking at the log-transformed trap catch data from each site from 2015–2018 (Fig 1). Log-normalized average trap catch for each orchard site from the four year study period is presented with best fit curves to better illustrate the trap catch variance between each orchard (S1 Fig). A simple linear regression explained that the log-normalized trap catch had a weak positive correlation with RH (t = 9.23, p < 0.001, R = 0.10, y = 0.12x – 7.16) (Fig 2).

Fig 1

Weekly adult D. suzukii trap catch over degree days.

Relationship between log-normalized adult D. suzukii total trap catch for each trap/week and weekly cumulative degree days over the four year trapping period. A smoothed fit line with 95% confidence bands illustrate the approximate phenology trend of D. suzukii for this dataset.

Fig 2

Weekly adult D. suzukii trap catch over relative humidity.

Relationship between log-normalized total adult D. suzukii trap catch for each trap/week and mean weekly relative humidity at the trap location over the four year trapping period. A weak positive relationship between trap catch and increasing relative humidity is shown by the smoothed fit line with 95% confidence bands.

The GLMM analyzed trap catch from each site over the four year period. The random effect of year explained 24% of the total variance in D. suzukii trap catch, while the effect of site accounted for 5%, and the year * site interaction was responsible for 18% of the trap catch variability of the random effects. The coefficient estimates are 2.21 (Z = 23.95, p < 0.001) for DD/1000 and 0.03 (Z = 0.02, p = 0.08) for RH. This model estimates an increase of e(2.21) or about nine flies per every 1000 DD accumulated and an increase of e(0.03) or about one fly per every increase of one percent RH.

A smoothed curve plot (Fig 3) was generated by the GAMM, incorporating trap catch data from all sites over the four years to analyze seasonal population trends of D. suzukii. The Y-axis is represented by conditional modes (CM), which measure trends at the population level given the effects provided in the model. The plot suggests first adult detection occurs at 1276 DD and periods of above average trap catch began at 2019 DD and continued until 4707 DD. A very distinguishable interval of D. suzukii population increase occurs between 1549 DD and 3180 DD, which roughly corresponds with early July through early September in Wisconsin. This population increase is denoted by the increasing CM from negative (below average) to positive (above average) between 1549 DD and 3180 DD.

Fig 3

GAMM predicted D. suzukii population dynamics over degree days.

GAMM generated smooth curve plot of adult D. suzukii trap catch total for each site in relation to degree day accumulation (DD) over the four year trapping period. Critical DD values are labelled and boxed on the curve. Conditional modes (CM) measure the population level estimations given the effects (positive CM = higher than average, negative CM = lower than average).

The GAMM produced critical DD values (1276, 1549, 2019, and 3180) were plugged into the GLMM with their corresponding RH values and a predicted trap catch was generated (Table 2). Relative humidity values were taken from the sampling event closest in DD to the critical DD values from each site and year and were averaged. The true D. suzukii trap catches were averaged from each site and year from the sampling event closest to each critical DD value (Table 2). One sample t-tests were then performed to determine the validity of the GLMM by comparing the predicted mean trap catches to the true mean trap catches. The GLMM predicted trap catches at 1276 and 3180 DD were statistically similar to the corresponding true trap catches at p < 0.05, but the predicted trap catches at 1549 and 2019 DD were significantly different than the true trap catches at those respective DD at p < 0.05.

Table 2

Comparisons of GLMM predicted adult D. suzukii trap catch with true trap catch.

DD	RH	GLMM trap catch	True trap catch	95% CI	t	df	p
1276	70.7	1.2*	3.1 ± 1.2	0.7–5.6	1.6	51	0.1
1549	71.9	2.5	1.2 ± 0.4	0.4–1.9	-3.7	51	0.001
2019	71.7	6.6	13.7 ± 3.5	6.6–20.7	2.0	51	0.05
3180	75.4	87.1*	95.9 ± 16.2	63.2–128.6	0.5	41	0.6

Drosophila suzukii mean trap catch estimation resulting from the GLMM (generalized linear mixed model) compared to the actual mean trap catch (± SE) from all sites from 2014–2018 at each of the critical DD (degree day) value, and corresponding average RH (relative humidity). Results from the one sample t-test are also included: t-test statistic (t), degrees of freedom (df), p-value (p), and 95% confidence interval (CI) of the true mean trap catch.

* indicates a GLMM predicted D. suzukii trap catch statistically similar to the true trap catch at the corresponding DD and RH (one sample t-test: p < 0.05).

Discussion

The models presented from this research suggest that seasonal D. suzukii populations in Wisconsin tart cherry orchards follow predictable patterns based on DD and RH. Our models integrate a robust dataset comprising four years of D. suzukii trap catches from the same tart cherry orchards in eastern Wisconsin. The GLMM model is able to forecast adult D. suzukii numbers as weekly trap catches, which is useful when needing quantitative estimates of field populations. The GAMM model generated a smoothed curve plot representing D. suzukii population tendencies throughout the growing season, which is valuable in determining periods of high or low risk for elevated D. suzukii numbers and phases of rapid population increase or decrease. The field population trends described from this work are characteristic of D. suzukii phenology in the Midwest, as studies in Wisconsin and Michigan fruit crops report first adult detection in the early summer, peak activity in the late summer, and decreasing populations in the early fall [13, 23, 26, 31]. It is worth noting that in our study monitoring typically ceased a few weeks after tart cherry harvest, limiting our ability to extrapolate on the population trends beyond early fall and into winter. Our seasonal population trend findings differ from previous D. suzukii phenology work done in California which describe a bimodal population distribution with peak numbers occurring in the early summer and late fall, and a decrease in population during the mid to late summer [32]. It is possible this mid-season period of quiescence in California D. suzukii populations is explained by the flies’ exposure to high temperatures and low humidity; similar conditions are suggested in inducing a vernal-estival dormancy in other Diptera, such as the olive fruit fly (Bactrocera oleae) [33].

The findings of our research are comparable to D. suzukii population modeling studies from Michigan, Washington, Oregon, and North Carolina [13, 20]. In the work done by Wiman et al., 2014, estimates for California and North Carolina populations of D. suzukii followed a bimodal trend of early summer increase, mid-summer decrease, then late summer increase, whereas the Oregon D. suzukii estimates were more analogous to our results of a single peak of high activity late in the field season. However, the data used to predict D. suzukii populations in Oregon were largely different than our Wisconsin data; the highest D. suzukii trap catch was 200 adults per week in Wisconsin compared to 27 in Oregon, and around 4000 DD were accumulated in Wisconsin during the field season while roughly 1900 DD (Tmin = 4° C, no Tmax) were accumulated in Oregon [20]. A predictive generalized additive model (GAM) generated from seven years of D. suzukii trap catch data in Michigan blueberry forecasted a similar population trend to our findings of first D. suzukii detection occurring in the early/mid-summer, populations peaking in the late summer, then numbers decreasing in the late summer/early fall [13]. The GAM developed by Leach et al., 2019 was based off of calendar day, as opposed to our GAMM based off DD, and incorporated parameters including: first D. suzukii capture, spring activity, prior year max activity, the number of days below 0°C and above 10°C in spring and winter, the number of days above 60% RH, and the number of days above 28°C. According to their GAM, first D. suzukii catch was heavily influenced by the number of days below 0°C and above 10°C, and peak capture was strongly influenced by all measured parameters [13]. Our GAMM estimated first detection to occur at 1262 DD (mid to late June) similar to the GAM predictions in Leach et al., 2019. Our GAMM predicted peak activity at 3180 DD (early September) contrasting from that of Leach et al., 2019, which predicted peak activity in late September/early October [13]. The earlier peak activity forecasted by our model is potentially an artifact of tart cherries being an earlier ripening fruit crop than blueberries [13], demonstrating the importance of host crop availability in D. suzukii phenology.

The predictive models presented from our research are supported by regional and crop specific data obtained from Wisconsin tart cherry orchards, and are the first which forecast field D. suzukii population trends in the Midwestern U.S. relative to DD that could be used for preemptive management strategies. Our DD based models allow for a more precise estimation of field D. suzukii populations than calendar day models alone, but incorporating DD may restrict the ability to discern between the different late season generations of this multivoltine pest [7, 20]. One advantage of structuring models with DD is that they do not assume consistent temperatures from year to year on the same date, and can be more easily applied to other locations with dissimilar climates. Since the life cycle of D. suzukii is so heavily dependent on temperature [7, 23, 34], we assumed there would be less variability in trap catch from year to year for a given DD as opposed to calendar day. Relative humidity data from each site each year were also incorporated into the models as RH also plays an important role in D. suzukii’s seasonal activity [22, 23, 24]. Several additional factors such as photoperiod, host crop availability, and presence of natural enemies may impact the phenology of D. suzukii as well [21, 35, 36]. Another factor which may affect D. suzukii field population dynamics that is commonly overlooked is the frequency of insecticide applications. Insecticide applications are typically initiated with first D. suzukii detection in the early summer when fruit is beginning to ripen, and continue until the completion of harvest [10]. Our study, along with previous work done on D. suzukii phenology in the Midwest [13, 23, 26, 31], acknowledge, but did not incorporate pesticide use as an effect when analyzing D. suzukii population dynamics, meaning the influence insecticide application has on trap catch in this region is largely unknown. Future models directed towards approximating the seasonal phenology of D. suzukii should consider all significant biotic and abiotic effects, including factors that remain unresolved such as D. suzukii overwintering habits, impact of native biological control, and host crop characteristics.

The GLMM is slightly more accurate in estimating D. suzukii populations in the early and later portions of the season (Table 2), meaning first detection and peak activity can be more precisely predicted than mid-season numbers in affected Midwest tart cherry orchards. This may be due to the fact that the GLMM will only estimate an increasing or decreasing population trend. As a result, the model was unable to acknowledge the brief decrease in trap catch (1276–1549 DD) after first D. suzukii detection and before rapid population increase.

According to the GAMM, the first detection of D. suzukii is estimated to occur at 1262 DD, meaning control measures should be initiated when that DD accumulation is reached and fruit has reached a susceptible stage in a given year [26]. For reference, 1276 DD fell most closely on June 26th, 24th, 23rd, and 22nd in 2015, 2016, 2017, and 2018 respectively. Field larval infestations of tart cherries in Wisconsin have been documented starting in mid-July, roughly one month following first adult D. suzukii detection [26]. There also exists a specific period in the mid-summer (1549 DD) in which D. suzukii populations begin to increase, then finally peak in the early fall (3180 DD). Extensive management strategies implemented during the transitory 1276–1549 DD period when D. suzukii levels are still low may help to alleviate the impending rapid increase of D. suzukii in the field forecasted by our models. Wiman et al., 2016 also alludes to the importance of targeting the initial D. suzukii adults for control to reduce the opportunity for high populations to build up and fruit damage to increase throughout the season. Though, any management practice directed to a given crop at this time should be imposed only when susceptible fruit is present on the crop, otherwise the attempt may be meaningless.

This research adds a valuable new tool to crop protection against D. suzukii by predicting their field population trends in relation to DD an RH, allowing for the preemptive implementation of integrated pest management strategies, or improved trap deployment. Our models best apply to growers in the Midwest who farm tart cherries in locations with a known presence of the pest, or areas of potential invasion with a comparable climate to the Midwest U.S. Currently, growers in the Midwest are advised to begin management practices at first detection of D. suzukii in their region [37], but in some cases, control measures in some fruit crops may be delayed until fruits reach a susceptible stage of development [26]. We understand that these models are regionally limited, derived from the specific, and varying factors pertaining to Wisconsin tart cherry orchards. However, continued work on population modeling of D. suzukii in different climatic areas with the inclusion of crop susceptibility and fruit field infestations is an integral step towards the effective and efficient management of this pest in all invaded regions.

Weekly adult D. suzukii trap catch separated by site over degree days.

Relationship between log-normalized adult D. suzukii total trap catch for each orchard site/week and weekly cumulative degree days over the four year trapping period. A smoothed fit line shows approximate phenology of D. suzukii for each orchard in this dataset.

(TIFF)

Click here for additional data file.

(XLSX)

Click here for additional data file.

(XLSX)

Click here for additional data file.

(XLSX)

Click here for additional data file.

(XLSX)

Click here for additional data file.

12 Feb 2020

PONE-D-19-35638

Degree day models to forecast the seasonal phenology of Drosophila suzukii (Diptera: Drosophilidae) in Midwest climate

PLOS ONE

Dear Dr. Guédot,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

==============================

Your manuscript has been reviewed by three independent and qualified reviewers. They returned with contrasting recommendations (minor, major revisions and reject). Some comments are apparently serious, e.g., trap position, unconsidered factors (variety, pesticide applications, etc). Therefore, I need to read a full point by point rebuttal letter to the comments below, before taking any further editorial decision.

Specific comments from the academic editor:

- I suggest adding the crop system in the title. Moreover, although Midwest is pretty known I suggest adding US. (Diptera: Drosophilidae) could be deleted.

- the raw data file gives errors in most cells with formulae

==============================

We would appreciate receiving your revised manuscript by Mar 28 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Antonio Biondi, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please ensure that you include a title page within your main document. We do appreciate that you have a title page document uploaded as a separate file, however, as per our author guidelines (http://journals.plos.org/plosone/s/submission-guidelines#loc-title-page) we do require this to be part of the manuscript file itself and not uploaded separately.

Could you therefore please include the title page into the beginning of your manuscript file itself, listing all authors and affiliations.

Additional Editor Comments (if provided):

Your manuscript has been reviewed by three independent and qualified reviewers. They returned with contrasting recommendations (minor, major revisions and reject). Some comments are apparently serious, e.g., trap position, unconsidered factors (variety, pesticide applications, etc). Therefore, I need to read a full point by point rebuttal letter to the comments below, before taking any further editorial decision.

Specific comments from the academic editor:

- I suggest adding the crop system in the title

- the raw data file gives errors in most cells with formulae

- please check for any reviewers' attachment

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Partly

Reviewer #3: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: N/A

Reviewer #2: Yes

Reviewer #3: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The manuscript “Degree day models to forecast the seasonal phenology of Drosophila suzukii (Diptera: Drosophilidae) in Midwest climate” analyzes two predictive models (GLMM and GAMM) on the dynamic seasonal phenology of spotted-wing drosophila (SWD) based on four years of data. Furthermore, the authors of the manuscript have a useful amount of data about the SWD captures through the season to consider the fly abundance during the season. However, the manuscript is a bit confusing, and it is hard, at least for this reviewer, to follow the hypotheses, methodology and results obtained. In this sense, the manuscript would improve if it had a more precise distribution and data analyzed. Otherwise, it is complicated to evaluate the findings. My comments below try to explain the significant part of the shortcuts that I found during the reading.

Introduction:

The introduction needs more information about why the authors used two Generalized different models for the predictions.

Methods:

This section needs more detail about field information (maybe a figure, should help to do some visual idea about the field distribution).

There are 13 different locations with the cherry variety ‘Montmorency’; however, there one location (authors didn’t said which site is) with two different cherry varieties ‘Montmorency’ and ‘Balaton’. For this reason, this location should be excluded from the predictive models. The populations' capture in this field could increase or decrease (attract/repel SWD) in correlation with variety absent in another experimental field.

The abiotic conditions are essential if this study is focused on predictive models that consider temperature and relative humidity as a key factors. However, the authors recorded these data from weather stations “located 1-7 km” from the monitoring sites, which is of low accuracy to make these models reliable.

In general, both models are not well described. Furthermore, the model diagnostics section is confusing.

Results

L 178: “slight positive correlation with RH”; At least for this reviewer, a correlation value of R2= 0.04, is almost no correlation. However, authors in the figure text said: “A positive relationship between trap catch and increasing relative humidity is shown”, which is a bit confusing for the reader.

The mean of SWD trapped predicted in the GLMM (Table 1) does not fit well with the “true trap catch”, which is worrying. At 1549 DD, the GLM model predicts that there are 6.6 D. suzukii adults/trap but in reality, the data showed a twice flies captures (13.7 D. suzukii adults/trap). This data could make the GLMM not a consistent model to follow the SWD dynamics in cherry.

Discussion

Finally, other factors influencing SWD dynamics were mostly not detailed or discussed and final conclusions regarding how predictive models could support crop protection against D. suzukii were premature given the data presented.

Reviewer #2: PONE-D-19-35638

I appreciated the approach of this paper. It is well-written, though I had a few copy edit comments. The main issue is lack of independent, validating data and potential over-reach of the conclusions. The data are limited to a very specific crop system and the results should not be generalized to the landscape level. One of the major oversights is the lack of discussion about the potential impact of the insecticide applications on the number of adult flies trapped, even though the introduction stated that there could be five or more applications. Tart cherries often use rather strong chemistries such as dimethoate that could have a huge influence on the number of flies trapped. Also I would have liked to know if traps were on orchard borders or in interiors, as the spatial distribution of traps is known to influence captures. Finally, it would have been great to have some observations linked to the trap capture data. Most growers are giving up on traps, they are too cumbersome and time consuming. The most important variable is the infestation risk to the crop, not how many are caught in traps so that was perhaps a missed opportunity. It would be great to see if the model could predict first infestation, or total infestation on managed and/or untreated crop. That said, the data and analysis presented are relatively sound and meet the criteria for publication in PLOS. I recommend acceptance after minor changes are implemented. I had two uploaded documents: 1) marked up manuscript, and 2) line number comments corresponding to the marks on 1.

• Highlight, page 9

L28 prominent - common?

• Highlight, page 9

L31 with five or more applications under heavy pressure

• Highlight, page 9

L35 risk of developing resistance

• Highlight, page 10

L46 heat units

• Highlight, page 10

L50 to coincide with predictions of

• Strike Out, page 10

L53 the

• Highlight, page 10

L54 any reasons why phenology might be different in the midwestern US?

• Highlight, page 11

L69 would be nice to report GPS points for sites, latitude info could prove useful to others

• Highlight, page 11

L88 What resolution and time scale were the PRISM data?

• Highlight, page 11

L90 degree-days (DD) - what method used for DD?

• Highlight, page 13

L33 but wouldn't this be more in the SWD foraging environement rather than ambient? Were the orchards irrigated?

• Strike Out, page 14

L154 we?

• Highlight, page 16

L183 explanation for the confidence interval?

• Highlight, page 16

L186 explanation for CI missing

• Highlight, page 17

L206 smoothing protocol? Loess?

• Highlight, page 18

L244 but presumably if there was a later crop or host, the populations would likely increase well beyong the monitored period (Oct), or at least you don;t know from this data set, these data are specific to the tart cherry crop and should not be expanded to generalize the whole population

• Highlight, page 21

L297 Careful, no data on crop infestation are presented. What if damage preceded trap captures? Unhappy growers.

• Highlight, page 21

L309 Would be important to link the model to 1) crop susceptibility (color), 2) infestation/damage.

Reviewer #3: Degree day models to forecast the seasonal phenology of Drosophila suzukii

(Diptera: Drosophilidae) in Midwest climate (Kamiyama et al.)

The paper seeks to develop a general model for SWD in the Midwest by

fitting GLMM and GAMM statistical models to four years of field trap catch

data using degree days, RH, year, and site as independent variables. The

paper is well written and the statistical analyses well done.

The approach of putting field data on a dd scale is old (e.g., R.D. Hughes

and N. Gilbert @1968), and normalizing the data for each location and year

helps to put the phenology on a common scale based on the temperatures

experienced at the different sites and years (L 274-276). The first question

is what is the starting time for accumulating dd, and further how were the

dd calculated (the term non-linear time gives direction but not precise

information). Statistical models tend to be time and place specific

explaining why similar models from other locations do no work for the

Midwest (lines 252-265). In fact, the proposed model(s) may not work in

other areas of the Midwest. The parameter values should be given for the

models.

Mechanistic models have been developed that use biodemographic functions to

estimate the effects of physical factors (temperature, RH, photoperiod,

etc.) at short time steps on the dynamics of the target poikilotherm. In an

age -stage context, these models use abiotic and biotic factors as drivers

of the dynamics (doi:10.1016/j.ecolmodel.2016.05.014 and

doi:10.1007/s10530-016-1255-6). No model should be used to forecast

densities, hence the comment L 268-271 is inappropriate – at best one can

attempt to predict phenology “ for preemptive – IPM”. The authors recognize

this limitation L-308-309.

Some specific points:

L178- (t = 5.50, p < 0.001, are high relative to R2 = 0.04 – please check.

L201-203 This result is the critical result, and merits deeper explanation

of the conditional mode.

The work seeks to develop IPM relevant information, but fail to provide

accessible rules. Figures 1 and 2 were not particularly useful, showing a

very wide scatter of data interpreted statistically in the text.

The data reported in the supplemental materials is for 2018.

*** Note from Editorial Office: please note that the responses to the following questions are 'partly' but this is not an option in the reviewer form and 'No' has therefore been selected:

2. Has the statistical analysis been performed appropriately and rigorously? Partly

3. Have the authors made all data underlying the findings in their manuscript fully available? Partly -2018 ***

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: PONE-D-19-35638_markup.pdf

Click here for additional data file.

Submitted filename: PONE-D-19-35638_notes.pdf

Click here for additional data file.

20 Mar 2020

Authors Response to Review Comments

PONE-D-19-35638

Degree day models to forecast the seasonal phenology of Drosophila suzukii (Diptera: Drosophilidae) in Midwest climate

PLOS ONE

Dear Dr Biondi,

We are pleased to submit for publication the revised version of “Degree day models to forecast the seasonal phenology of Drosophila suzukii (Diptera: Drosophilidae) in Midwest climate”. We appreciate all the thoughtful and constructive criticisms from the reviewers. We have addressed each of their concerns as detailed below as much as possible. The reviewers’ comments were italicized with our response directly below. We believe that the revised version can meet the criteria for publication in PLOS ONE.

Specific comments from the academic editor:

- I suggest adding the crop system in the title. Moreover, although Midwest is pretty known I suggest adding US. (Diptera: Drosophilidae) could be deleted.

We agreed with the suggestions and added the crop system and country to the title. “(Diptera: Drosophilidae)” was deleted.

- the raw data file gives errors in most cells with formulae

This was fixed.

Could you therefore please include the title page into the beginning of your manuscript file itself, listing all authors and affiliations?

A title page was included within the main text file.

Please check for any reviewers' attachment

Attachments were reviewed.

Response to Reviewer 1’s comments

The manuscript “Degree day models to forecast the seasonal phenology of Drosophila suzukii (Diptera: Drosophilidae) in Midwest climate” analyzes two predictive models (GLMM and GAMM) on the dynamic seasonal phenology of spotted-wing drosophila (SWD) based on four years of data. Furthermore, the authors of the manuscript have a useful amount of data about the SWD captures through the season to consider the fly abundance during the season. However, the manuscript is a bit confusing, and it is hard, at least for this reviewer, to follow the hypotheses, methodology and results obtained. In this sense, the manuscript would improve if it had a more precise distribution and data analyzed. Otherwise, it is complicated to evaluate the findings. My comments below try to explain the significant part of the shortcuts that I found during the reading.

We appreciate the review and comments and will address the review based on the specific comments below.

Introduction:

The introduction needs more information about why the authors used two Generalized different models for the predictions.

More detailed reasoning behind implementing the two different models provided L74 – 87.

Methods:

This section needs more detail about field information (maybe a figure, should help to do some visual idea about the field distribution).

There are 13 different locations with the cherry variety ‘Montmorency’; however, there one location (authors didn’t said which site is) with two different cherry varieties ‘Montmorency’ and ‘Balaton’. For this reason, this location should be excluded from the predictive models. The populations' capture in this field could increase or decrease (attract/repel SWD) in correlation with variety absent in another experimental field.

Table added (Table 1) for field distribution and trap set up clarification. Justification added for retaining the ‘Balaton’ site in the models L104 – 109. The weather station locations and distance to the specific traps were added to the table for clarity.

The abiotic conditions are essential if this study is focused on predictive models that consider temperature and relative humidity as a key factors. However, the authors recorded these data from weather stations “located 1-7 km” from the monitoring sites, which is of low accuracy to make these models reliable.

We realize that the distance from the weather stations to the traps may be a concern, however, the selected weather stations were the closest relative humidity measuring devices to the trapping sites as there were no individual relative humidity tracking devices placed in the field near each trap. The sites are large orchards with weather stations on site or near the sites and using data from the weather stations for our models was our most accurate option. The furthest distance between weather station and trap was 6.0 km, all the other weather stations were between 0.1 and 4.5 km from their respective traps. The temperature data for each site was retrieved from the PRISM Climate Group by entering the GPS coordinates of each site. We added Table 1 to display each site and weather station.

In general, both models are not well described. Furthermore, the model diagnostics section is confusing.

We added some more detail to the model descriptions and diagnostics to help with this comment. The other two reviewers did bring this up so we feel that the descriptions are appropriate as edited.

Results

L 178: “slight positive correlation with RH”; At least for this reviewer, a correlation value of R2= 0.04, is almost no correlation. However, authors in the figure text said: “A positive relationship between trap catch and increasing relative humidity is shown”, which is a bit confusing for the reader.

The mean of SWD trapped predicted in the GLMM (Table 1) does not fit well with the “true trap catch”, which is worrying. At 1549 DD, the GLM model predicts that there are 6.6 D. suzukii adults/trap but in reality, the data showed a twice flies captures (13.7 D. suzukii adults/trap). This data could make the GLMM not a consistent model to follow the SWD dynamics in cherry.

We thank you for pointing this out. The linear regression was performed on ‘trap catch’ and ‘RH’ as opposed to ‘log-normalized trap catch’ and ‘RH’. The results of the linear regression have been updated and the new R2 (R2=0.1) value has been provided L227-228.

At 1549 DD, the GLMM predicts 2.5 D. suzukii/trap. This is likely because our data show a slight decrease in trap catch from 1276 DD to 1549 DD (3.1 to 1.2 flies). However, the GLMM predicts a strictly increasing trap catch throughout the season so the GLMM was unable to predict the slight drop in flies. This is a potential oversight in the model that we added to the Discussion (L360 – 363). At 2019 DD, the GLMM predicts 6.6 D. suzukii/trap, which is on the cusp of the 95% CI of the true trap catch (6.6 – 20.7 flies). Figure 3 was also updated so the critical points in the figure now reflect those reported in the text.

Discussion

Finally, other factors influencing SWD dynamics were mostly not detailed or discussed and final conclusions regarding how predictive models could support crop protection against D. suzukii were premature given the data presented.

More details about alternative factors influencing D. suzukii dynamics added to the Discussion (L344 – 357), and final conclusions were tempered back.

Response to Reviewer #2’s comments

I appreciated the approach of this paper. It is well-written, though I had a few copy edit comments. The main issue is lack of independent, validating data and potential over-reach of the conclusions. The data are limited to a very specific crop system and the results should not be generalized to the landscape level. One of the major oversights is the lack of discussion about the potential impact of the insecticide applications on the number of adult flies trapped, even though the introduction stated that there could be five or more applications. Tart cherries often use rather strong chemistries such as dimethoate that could have a huge influence on the number of flies trapped. Also I would have liked to know if traps were on orchard borders or in interiors, as the spatial distribution of traps is known to influence captures. Finally, it would have been great to have some observations linked to the trap capture data. Most growers are giving up on traps, they are too cumbersome and time consuming. The most important variable is the infestation risk to the crop, not how many are caught in traps so that was perhaps a missed opportunity. It would be great to see if the model could predict first infestation, or total infestation on managed and/or untreated crop. That said, the data and analysis presented are relatively sound and meet the criteria for publication in PLOS. I recommend acceptance after minor changes are implemented. I had two uploaded documents: 1) marked up manuscript, and 2) line number comments corresponding to the marks on 1.

Thank you for the comments. The conclusions made in the Discussion have been clarified to apply to specifically tart cherry crop systems. Potential impact of insecticides was acknowledged (L351 – 354). Added ‘host crop characteristics’ to list of unresolved effects that future models should consider to account for tart cherry chemistry. Trap location was added to Methods. We agree, it would have been informative to measure fruit infestation along with trap catch throughout the season. However, it is difficult to collect tart cherries when they are in a susceptible stage of development. In our past work with tart cherries and D. suzukii field larval infestations, we only managed two collections which yielded larval infestations (collecting fruit every other week). Growers want to minimize the amount of time ripe fruits are exposed in the field so they are harvested quickly reducing the time to sample. In the future, multiple trees should be left un-harvested for fruit sampling throughout the season for every year of the phenology study to gain a better understanding of field larval infestations.

• Highlight, page 9

L28 prominent - common?

Changed.

• Highlight, page 9

L31 with five or more applications under heavy pressure

Changed.

• Highlight, page 9

L35 risk of developing resistance

Fixed.

• Highlight, page 10

L46 heat units

Changed.

• Highlight, page 10

L50 to coincide with predictions of

Added.

• Strike Out, page 10

L53 the

Removed.

• Highlight, page 10

L54 any reasons why phenology might be different in the midwestern US?

Presumably a result of climate, more detail provided in the Discussion.

• Highlight, page 11

L69 would be nice to report GPS points for sites, latitude info could prove useful to others

GPS location included in new Table 1.

• Highlight, page 11

L88 What resolution and time scale were the PRISM data?

Added.

• Highlight, page 11

L90 degree-days (DD) - what method used for DD?

Degree-day thresholds listed (L135-136).

• Highlight, page 13

L33 but wouldn't this be more in the SWD foraging environement rather than ambient? Were the orchards irrigated?

L133 removed. Yes, orchards were irrigated.

• Strike Out, page 14

L154 we?

Corrected.

• Highlight, page 16

L183 explanation for the confidence interval?

Added.

• Highlight, page 16

L186 explanation for CI missing

Added.

• Highlight, page 17

L206 smoothing protocol? Loess?

Smoothing protocol (penalized cubic regression) noted in the GAMM description in the Methods.

• Highlight, page 18

L244 but presumably if there was a later crop or host, the populations would likely increase well beyong the monitored period (Oct), or at least you don;t know from this data set, these data are specific to the tart cherry crop and should not be expanded to generalize the whole population

This is true. Included statement explaining we cannot estimate population trends expanding beyond our monitoring times.

• Highlight, page 21

L297 Careful, no data on crop infestation are presented. What if damage preceded trap captures? Unhappy growers.

Reference included backing statement (L367 – 369).

• Highlight, page 21

L309 Would be important to link the model to 1) crop susceptibility (color), 2) infestation/damage.

Included these measurables L388.

Response to Reviewer #3’s comments

The paper seeks to develop a general model for SWD in the Midwest by fitting GLMM and GAMM statistical models to four years of field trap catch data using degree days, RH, year, and site as independent variables. The paper is well written and the statistical analyses well done.

The approach of putting field data on a dd scale is old (e.g., R.D. Hughes and N. Gilbert @1968), and normalizing the data for each location and year helps to put the phenology on a common scale based on the temperatures experienced at the different sites and years (L 274-276). The first question is what is the starting time for accumulating dd, and further how were the dd calculated (the term non-linear time gives direction but not precise information). Statistical models tend to be time and place specific

explaining why similar models from other locations do no work for the Midwest (lines 252-265). In fact, the proposed model(s) may not work in other areas of the Midwest. The parameter values should be given for the models.

Mechanistic models have been developed that use biodemographic functions to estimate the effects of physical factors (temperature, RH, photoperiod, etc.) at short time steps on the dynamics of the target poikilotherm. In an age -stage context, these models use abiotic and biotic factors as drivers

of the dynamics (doi:10.1016/j.ecolmodel.2016.05.014 and doi:10.1007/s10530-016-1255-6). No model should be used to forecast densities, hence the comment L 268-271 is inappropriate – at best one can

attempt to predict phenology “ for preemptive – IPM”. The authors recognize this limitation L-308-309.

We appreciate your comments. Degree-day accumulation began on January 1st for each year (L134), and information on the calculation of DD are provided in the temperature and humidity section in the Methods (L133 – 138). We understand no model based on regionally specific data can accurately predict phenology of D. suzukii in a different region, and this is especially true for an insect such as D. suzukii that has such a different phenology in different regions, hence the need for regional models. We believe these models can be applied the Midwest and to regions growing tart cherries experiencing similar climates to the Midwest to get a general sense of D. suzukii population trends throughout the field season. Parameters for the GAM referenced provided (L321 – 323). We agree with your comment on L268 – 271, the sentence was changed to be more appropriate (L334 – 337).

Some specific points:

L178- (t = 5.50, p < 0.001, are high relative to R2 = 0.04 – please check.

Thank you, corrected.

L201-203 This result is the critical result, and merits deeper explanation of the conditional mode.

Indeed, explanation provided L249 – 251 and L254 – 256.

The work seeks to develop IPM relevant information, but fail to provide accessible rules. Figures 1 and 2 were not particularly useful, showing a very wide scatter of data interpreted statistically in the text.

We prefer to keep Figures 1 and 2 because we feel that providing a visualization of the complete four year trap catch dataset is helpful for readers to go along with the trends explained statistically in the text.

The data reported in the supplemental materials is for 2018.

Thank you, data now includes all years.

*** Note from Editorial Office: please note that the responses to the following questions are 'partly' but this is not an option in the reviewer form and 'No' has therefore been selected:

2. Has the statistical analysis been performed appropriately and rigorously? Partly

3. Have the authors made all data underlying the findings in their manuscript fully available? Partly -2018 ***

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

9 Apr 2020

Degree day models to forecast the seasonal phenology of Drosophila suzukii in tart cherry orchards in the Midwest U.S.

PONE-D-19-35638R1

Dear Dr. Guédot,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Antonio Biondi, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Dear authors,

Your manuscript has been reviewed by one of the referees that had reviewed the original version. Such reviewer found the manuscript not worthy of publication. However, after reading the new comments, the (very carefully)revised manuscript and the rebuttal letter, I totally disagree with such recommendation and, as a consequence, I am promoting the acceptance of this version for publication in PONE.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: N/A

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors have fixed many of the comments that the reviewers had suggested to them. And the article has improved a lot, compared to the original version. However, at least for this reviewer, there are two key points that are important in interpreting this research, that is not fixed properly.

The authors have included an orchard with different cherry varieties in the study. Although they should be explained better the differences between these varieties (phenology, organoleptic, etc..). However, authors have justified by demonstrating that there is no difference between varieties of SWD catches during the study. At least for this reviewer, this could be justify 100% with statistics to show that there are no differences between cherry varieties on D. suzukii catch.

Finally, the second key point, is the most important of the article, because this manuscript is focused on the degree days models. At least for this reviewer, the weather stations are far from the fields where the traps were placed, and maybe this could be a reason to obtain strong confident models. However, this is a methodology mistake, that can't be fixed.

Other comments:

I also agree with the other reviewer about the insecticide treatments, which are necessary to consider to estimate the level of D. suzukii in relation to the treatments during the experiments.

Finally, no data on crop infestation are presented, which is crucial to know the damage on fruit in correlation with the DD models that authors did it.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

15 Apr 2020

PONE-D-19-35638R1

Degree day models to forecast the seasonal phenology of Drosophila suzukii in tart cherry orchards in the Midwest U.S.

Dear Dr. Guédot:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Antonio Biondi

Academic Editor

PLOS ONE