Introduction
Wild or lowbush blueberry (Vaccinium angustifolium Aiton) is an important fruit crop in the state of Maine, USA, and the Canadian provinces of Quebec, New Brunswick, Prince Edward Island, and Nova Scotia (Yarborough Reference Yarborough2023). The crop is Nova Scotia’s largest horticultural crop by acreage and export, with 15,901 ha harvested annually (Anonymous 2021). Lowbush blueberries are grown in a 2-yr production cycle in which fields are pruned to ground level to promote vegetative growth and flower bud development in the first, or nonbearing year, and fruit are harvested in the second, or bearing year. Commercial fields are therefore maintained as perennial no-till monocultures, making weeds an important management challenge (Jensen and Yarborough Reference Jensen and Yarborough2004). The weed flora of lowbush blueberry fields consists primarily of native and non-native woody and herbaceous perennial plants (Jensen and Yarborough Reference Jensen and Yarborough2004; McCully et al. Reference McCully, Sampson and Watson1991), with creeping herbaceous perennials among the most common types of weeds in this cropping system (Lyu et al. Reference Lyu, McLean, McKenzie-Gopsill and White2021).
Narrowleaf goldenrod [Euthamia graminifolia (L.) Nutt.] is a common creeping herbaceous perennial weed in lowbush blueberry fields (Hall Reference Hall1959; McCully et al. Reference McCully, Sampson and Watson1991). The plant reproduces by seed and rhizomes, although the relative contribution of seedlings and ramets to established populations in lowbush blueberry fields is unknown. Euthamia graminifolia occurred in 86% of lowbush blueberry fields surveyed in Nova Scotia between 2017 and 2019 (Lyu et al. Reference Lyu, McLean, McKenzie-Gopsill and White2021), and the plant is now the most common species of goldenrod in lowbush blueberry fields in this province, presumably due to reduced hexazinone efficacy on this weed (White et al. Reference White, Boyd and Van Acker2016). Although direct effects of E. graminifolia on lowbush blueberry are poorly understood, shoots of this weed are taller than lowbush blueberry stems (Farooq Reference Farooq2018) and can reduce lowbush blueberry flower bud formation (Boyd and White Reference Boyd and White2010). Increased cover of other tall weeds such as spreading dogbane (Apocynum androsaemifolium L.) and bracken fern [Pteridium aquilinum (L.) Kuhn] also reduce lowbush blueberry yield (Yarborough and Marra, Reference Yarborough and Marra1997), and growers widely recognize tall weeds such as E. graminifolia as a hindrance to mechanical harvesting. Growers currently rely on postemergence mesotrione applications to suppress E. graminifolia, with application timing based primarily on height of emerged shoots (Boyd and White Reference Boyd and White2010) due to limited knowledge of shoot emergence patterns in lowbush blueberry fields. The height of E. graminifolia shoots also facilitates control by weed wiper applications of nonselective herbicides such as glyphosate. Success with this approach, however, is variable, possibly due to poor application timing and lack of knowledge of when emerged ramets are most susceptible (Farooq Reference Farooq2018).
Ramet emergence patterns and progression of ramet phenological development has been studied for many perennial weeds and used to help improve weed management (Miyazaki et al. Reference Miyazaki, Ito and Urakawa2005; Stoller and Wax Reference Stoller and Wax1973; Webster and Cardina Reference Webster and Cardina1999; Wu et al. Reference Wu, Boyd, Cutler and Olson2013). When emergence and phenology data are related to reliable explanatory variables, such as temperature and/or moisture expressed as growing degree days (GDD) or hydrothermal time, predictive models can be developed to help improve future timing of management practices (Blatt et al. Reference Blatt, De Clerck-Floate and White2022; Donald Reference Donald2000; Ekeleme et al. Reference Ekeleme, Forcella, Archer, Chikoye and Okezie Akobundu2004). Such models have been developed for the lowbush blueberry plant (White et al. Reference White, Boyd and Van Acker2012) and the associated weed species red sorrel (Rumex acetosella L.) (White et al. Reference White, Boyd and Van Acker2015) and A. androsaemifolium (Wu et al. Reference Wu, Boyd, Cutler and Olson2013). These models are currently being used by growers to aid management decisions, but similar models for E. graminifolia are lacking.
Euthamia graminifolia can produce dense patches and outcompete surrounding vegetation (Butcko and Jensen Reference Butcko and Jensen2002), and a negative correlation between broadleaf weed cover and blueberry yield has been demonstrated across several studies (Drummond et al. Reference Drummond, Smagula, Yarborough and Annis2012; Marty et al. Reference Marty, Lévesque, Bradley, Lafond and Paré2019; Yarborough and Marra, Reference Yarborough and Marra1997). There is, however, limited knowledge of the general emergence and phenology patterns exhibited by E. graminifolia in lowbush blueberry fields, and little knowledge of the seedbank dynamics or extent and occurrence of seedling recruitment of this weed species in lowbush blueberry fields. The objectives of this research were therefore to (1) develop predictive GDD models for E. graminifolia ramet emergence and phenological development, (2) determine whether E. graminifolia forms seedbanks in lowbush blueberry fields, and (3) establish whether E. graminifolia seedlings emerge in lowbush blueberry fields.
Materials and Methods
Ramet Emergence and Phenological Development
The experiment was established in eight nonbearing-year commercial lowbush blueberry fields with known established E. graminifolia populations (Table 1). Each site was used once to increase spatial replication, and one to three sites were chosen per year from 2017 to 2020 to increase temporal replication. The experiment consisted of monitoring E. graminifolia ramet emergence and phenological development in ten 0.5 m by 1 m quadrats placed in areas containing E. graminifolia within each field. Emerged ramets were counted and marked with colored circular elastics (diameter = 6 mm) twice weekly from early April until no new ramets emerged for 2 consecutive weeks. Specific elastic colors or color combinations were used on each counting date to keep emergence cohorts separate. Emerged ramets at the flower bud and flowering stages were also counted and tagged on each counting date. Ramets were considered to be at the bud or flowering stage when at least one flower bud or open flower was visible, respectively. Flower bud and flowering stage counts also continued until 2 consecutive weeks of zero values occurred. The number of emerged ramets, ramets at the flower bud stage, and ramets at the flowering stage were expressed as percent cumulative data for statistical analysis and presentation.
Table 1. Site locations and dates of experiment establishment for monitoring Euthamia graminifolia ramet emergence and phenological development in lowbush blueberry fields in Nova Scotia, Canada, from 2017 to 2020.a
a A long dash (—) indicates information not available.
b OM, organic matter.
Temperature loggers (Watchdog 1400-series data loggers, Spectrum Technologies; Aurora, IL, USA) were placed at each location to record hourly air temperature and were attached to stakes and placed 0.5 m above the soil surface. Regional air temperature data from the nearest Environment Canada weather station were used to supplement temperature logger data (Table 2) to determine cumulative GDD at each site beginning on April 1 using the formula:
$${\rm{GDD}} = \mathop \sum \nolimits_{i = 1}^n \left( {{T_{{\rm{mean}}}} - {T_{{\rm{base}}}}} \right)$$
Table 2. Date of data logger establishment and duration of temperature data supplementation with air temperature data from the nearest Environment Canada weather station at lowbush blueberry fields used to monitor Euthamia graminifolia emergence and phenology in Nova Scotia, Canada
a Debert weather station located at 45.416667°N, 63.466667°W, elevation 37.5 m; Nappan weather station located at 45.7759556°N, 64.241444°W, elevation 20 m.
b Use of Environment Canada weather data not required at this site.
where T mean is the mean daily air temperature, T base is the lowest air temperature at which we assume ramet emergence or development will not occur, and n is the number of days over which GDD are calculated. In this equation, GDD = 0 if T mean ≤ T base. A base temperature of 0 C was used and was determined using the iterative process described by Izquierdo et al. (Reference Izquierdo, González-Andújar, Bastida, Lezaún and Sánchez del Arco2009) in the nonlinear equations described later. Rainfall data for the 2017, 2018, 2019, and 2020 growing seasons were obtained from the Environment Canada weather station located near Debert, NS, Canada (Table 3).
Table 3. Monthly rainfall (mm) during the period of Euthamia graminifolia emergence and phenology data collection in 2017, 2018, 2019, and 2020a
a Data were obtained from records for the Environment and Climate Change Canada weather station located at Debert, NS, Canada (45.416667°N, 63.466667°W, elevation 37.5 m).
GDD Model Development
Cumulative ramet emergence, ramets at the flower bud stage, and flowering ramets were plotted as functions of GDD. Fitting of nonlinear equations and parameter estimates for equations was conducted using the Gauss-Newton algorithm in PROC NLIN in SAS (SAS v. 9.3, SAS Institute, Cary, NC, USA). Nonlinear equations were chosen based on goodness-of-fit statistics (described later), visual alignment of model predictions with observed data, and presence of biologically relevant parameters. Percent cumulate ramet emergence (Y) was related to GDD using a four-parameter Weibull equation of the form:
$$Y = a\left\{ {1 - \exp\left[ { - k{{\left( {{\rm GDD} - x0} \right)}^c}} \right]} \right\}$$
where a is theoretical maximum percent cumulative emergence, k is rate of increase in percent cumulative emergence, x0 is the lag phase until the onset of emergence, and c is a shape parameter (Martinson et al. Reference Martinson, Durgan, Forcella, Wiersma, Spokas and Archer2007).
Percent cumulative ramets at the bud stage and flowering stage (Y) were related to GDD using a three-parameter Gompertz equation of the form:
$$Y = a\, \exp\left\{ { - \exp\left[ { - k\left( {{\rm GDD} - x0} \right)} \right]} \right\}$$
where a is theoretical maximum percent cumulative ramets at the bud stage or flowering stage, k is rate of increase in percent cumulative ramets at the0 bud stage or flowering stage, and x0 is the inflection point of the curve on the x axis (Dorado et al. Reference Dorado, Sousa, Calha, González-Andújar and Fernández-Quintanilla2008).
Goodness of fit for the proposed models was determined by calculating the coefficient of determination (R2), adjusted coefficient of determination (R2adj), and root mean-square error (RMSE) as described by White et al. (Reference White, Boyd and Van Acker2012, Reference White, Boyd and Van Acker2015). Models for cumulative ramet emergence, ramets at the flower bud stage, and ramets at the flowering stage were calibrated using data from five randomly selected sites (Debert 2017, Collingwood 2017, Baseline 2019, Westchester 2019, and Webb Mountain 2020). Calibrated models were validated with emergence, ramets at the bud stage, and flowering ramet datasets from the three sites not used in the original model calibration (Portapique 2017, North River 2018, and Mount Pleasant 2020). Ramet emergence, ramets at the flower bud stage, and flowering ramet predictions were calculated with the calibrated models and were plotted against observed data at each site used for model validation. Agreement of model predicted and observed values was based on the R2adj and RMSE.
Seedling Recruitment
The experiment was established in the nonbearing-year commercial lowbush blueberry fields used for ramet emergence at Webb Mountain and Mount Pleasant (Table 1). The experiment was established on April 20, 2020 at each site and consisted of monitoring E. graminifolia seedling emergence in ten 0.5 m by 1 m quadrats placed in areas containing E. graminifolia within each field. Newly emerged seedlings were counted twice weekly starting in early April and continuing until no new seedlings emerged for 2 consecutive weeks. Newly emerged seedlings were marked with circular colored elastics (diameter = 6 mm) on each counting date to prevent double counting.
Seedbank Analysis
The objective of this experiment was to determine the presence or absence of E. graminifolia seedbanks in lowbush blueberry fields. Seedbank presence or absence was determined by collecting 12 soil cores per field from 7 nonbearing-year lowbush blueberry fields located near Londonderry (45.481562°N, 63.577846°W), Camden (45.291023°N, 63.130170°W), Collingwood (Purdy [45.590922°N, 63.811182°W] and Stanley [45.590680°N, 63.820964°W] fields), Westchester (45.640918°N, 63.600899°W), Collingwood (45.573702°N, 63.900311°W), and Mount Pleasant (45.769682°N, 63.841162°W), NS, Canada, in 2019. All fields had known presence of E. graminifolia before sampling. Cores were collected with a 333-cm3 (7.5-cm depth) soil bulk density core sampler, bagged individually in paper bags in the field, and brought back to the lab where they were dried and sieved through a 2-mm soil sieve to remove coarse debris. Each core sample was then combined with 1,000 ml of Pro-Mix Potting Mix (Premier Tech Home and Garden, Brantford, ON, Canada) and placed in germination trays in the greenhouse under a 16-h/8-h (night/day) photoperiod and maintained at approximately 21 C. Trays were watered daily throughout the experiment. Samples were checked weekly, and any emerged seedlings were identified and removed. This continued until no new seedlings emerged for 2 consecutive weeks.
Results and Discussion
Ramet Emergence
Euthamia graminifolia ramets emerged between 25 and 71 GDD (early to mid-April) and emergence continued until 1,047 to 1,665 GDD at all sites (Figure 1A). Emergence to approximately 90%, however, was rapid, and this threshold was reached by 312 to 705 GDD across sites (Figure 1A). Shoot emergence beyond 90% consisted primarily of sporadic shoot emergence over several weeks until cessation of emergence in late summer, particularly at Webb Mountain (Figure 1A). The proposed Weibull model fit the field data well and accurately predicted emergence in the field as a function of GDD (Figure 1A; Table 4). Model prediction for the onset of emergence was 72 GDD and 10%, 50%, and 90% emergence was predicted to occur at 86, 179, and 458 GDD, respectively. Model predicted emergence also agreed closely with observed emergence at the sites used for model validation (Figure 2A–C), suggesting the model can accurately predict E. graminifolia emergence to aid weed management decisions. For example, Donald (Reference Donald2000) suggested his GDD model for Canada thistle [Cirsium arvense (L.) Scop.] could be used to estimate 80% shoot emergence and therefore guide the use of mechanical and chemical weed control for this weed species. Similarly, our model could be used to revise postemergence mesotrione application timing for E. graminifolia management based on shoot emergence rather than shoot size (Boyd and White Reference Boyd and White2010). When considered in the context of crop–weed competition, E. graminifolia ramets emerge approximately 172 GDD earlier than lowbush blueberry ramets (White et al. Reference White, Boyd and Van Acker2012). Weeds that emerge before crop plants cause larger yield reductions than weeds emerging later in the season (Swanton et al. Reference Swanton, Mahoney, Chandler and Gulden2008), suggesting E. graminifolia may be competitive with lowbush blueberry. Emergence before lowbush blueberry, however, may provide an opportunity for early-season E. graminifolia suppression with nonselective herbicides, and this could be considered in future research.
Figure 1. Percent cumulative Euthamia graminifolia ramet emergence (A), ramets at the flower bud stage (B), and ramets at the flowering stage (C) in relation to growing degree days (GDD) calculated from air temperature (T base = 0 C) at sites used for model calibration in Nova Scotia, Canada. Symbols are the mean of 10 observations. Lines are fitted regression equations. A Weibull equation of the form Y = a{1 − exp[−k(GDD − x0)c]} was fit to percent cumulative ramet emergence. A Gompertz equation of the form Y = a exp{−exp[−k(GDD − x0)]} was fit to percent cumulative ramets at the bud stage and percent cumulative flowering ramets. Parameter estimates for each equation are provided in Table 4. RMSE, root
Table 4. Parameter estimates for the proposed Weibull and Gompertz equations describing the relationship between growing degree days (GDD) calculated from air temperature (T base = 0 C) and percent cumulative Euthamia graminifolia ramet emergence, ramets at the flower bud stage, and ramets at the flowering stage in nonbearing-year wild blueberry fields at Debert 2017, Collingwood 2017, Baseline 2019, Westchester 2019, and Webb Mountain 2020 in Nova Scotia, Canada
a The Weibull equation was of the form Y = a{1 − exp[−k(GDD−x0)c]}and the Gompertz equation was of the form Y = a exp{−exp[−k(GDD − x0)]}.
b Weibull parameters, a = theoretical maximum percent cumulative emergence, k = rate of increase in percent cumulative emergence, x0 = the lag phase until the onset of emergence, and c = a shape parameter (Martinson et al. Reference Martinson, Durgan, Forcella, Wiersma, Spokas and Archer2007). Gompertz model parameters, a = theoretical maximum percent cumulative ramets at the flower bud stage or flowering stage, k = rate of increase in percent cumulative ramets at the flower bud stage or flowering stage, and x0 = the inflection point of the curve on the x axis (Dorado et al. Reference Dorado, Sousa, Calha, González-Andújar and Fernández-Quintanilla2008). Values represent the parameter estimate ± 1 SE. A long dash (—) indicates a parameter is not a component of the equation.
Figure 2. Observed and model predicted cumulative Euthamia graminifolia ramet emergence (A–C), ramets at the flower bud stage (D–F), and ramets at the flowering stage (G–I) in relation to growing degree days (GDD) calculated from air temperature (T base = 0 C) at sites used for model validation in Nova Scotia, Canada. Symbols are the mean of 10 observations. Lines are calibrated model predictions. The calibrated model for predicting percent cumulative ramet emergence was a Weibull equation of the form Y = a{1 − exp[−k(GDD − x0)c]}. The calibrated model for predicting percent cumulative ramets at the flower bud stage and flowering stages was a Gompertz equation of the form Y = a exp{−exp[−k(GDD − x0)]}. RMSE, root mean-square error.
Ramets at Flower Bud Stage
Emerged ramets were first observed at the flower bud stage between 710 and 871 GDD (June 21 to 28) and approximately 90% of emerged ramets were at the flower bud stage between 1,303 and 1,956 GDD (July 17 to August 24) across sites (Figure 1B). The proposed Gompertz model fit the field data well and accurately predicted ramet development to the flower bud stage as a function of GDD (Figure 1B; Table 4). Model prediction for the onset of the flower bud stage was 644 GDD and 10%, 50%, and 90% of emerged ramets at the flower bud stage were predicted to occur at 801, 1,074, and 1,522 GDD, respectively. Except for North River (Figure 2E), model predictions for emerged shoots at the flower bud stage agreed closely with the observed shoots at the flower bud stage at each site (Figure 2D and 2F). Model deviation from the observed values at North River is not fully understood, but we did observe what appeared to be insect feeding damage to the apical meristem region of emerged stems at this site that seemed to disrupt the onset of the flower bud and flowering stages. This site was not actively managed by the landowner at the time of data collection, and hence there were no herbicide, fungicide, or insecticide applications made to the field, which may explain the injury and variation observed. Validation results from the other two sites, however, suggest the model can be used to predict ramets at the bud stage and thus can be used to aid weed management. For example, spot applications of glyphosate and flazasulfuron to E. graminifolia at the flower bud stage reduce ramet density in the year after application (Farooq Reference Farooq2018; White Reference White2021), and the model can be used to help growers plan weed wiper or spot applications of these herbicides for E. graminifolia control.
Flowering Ramets
Emerged ramets began flowering between 1,418 and 1,626 GDD (July 30 to August 7), and approximately 90% of emerged ramets were flowering between 1,992 and 2,225 GDD (August 27 to September 12) (Figure 1C). The proposed Gompertz model fit the field data well and accurately predicted ramet development to the flowering stage as a function of GDD (Figure 1C; Table 4). Model prediction for the onset of flowering was 1,369 GDD, and 10%, 50%, and 90% of emerged ramets at the flower stage were predicted to occur at 1,521, 1,777, and 2,113 GDD, respectively. Validation of the model gave similar results to the flower bud model, in that model predictions deviated from observed values at North River (Figure 2H) but aligned well with observed flowering at the other two sites (Figure 2G and 2I). Much like ramet development to the flower bud stage, it is likely that flowering was also disrupted by insect damage observed on the apical meristem of emerged ramets at the North River site. Validation results from the other two sites, however, suggest the model can accurately predict flowering and could thus be used to aid weed management. For example, glyphosate applications to E. graminifolia at the flowering stage reduce shoot density in the year after application (Farooq Reference Farooq2018), and the model can be used to help growers plan weed wiper or spot applications of glyphosate or other symplastic herbicides for late-season E. graminifolia control.
From a beneficial perspective, late-season E. graminifolia flowering is asynchronous with the early-season flowering of lowbush blueberry (White et al. Reference White, Boyd and Van Acker2012), suggesting that E. graminifolia is not competitive for pollinators of lowbush blueberry. On the contrary, we observed significant bee activity on open E. graminifolia flowers in late summer, and goldenrods are frequently visited by a range of managed and native pollinators (Dibble et al. Reference Dibble, Drummond and Stack2020; McCallum and McLean Reference McCallum and McLean2017; Stubbs et al. Reference Stubbs, Jacobson, Osgood and Drummond1992). As such, late-flowering species such as E. graminifolia may be beneficial for pollinators by providing late summer food sources. Growers may therefore consider sparing patches of E. graminifolia to improve late-summer forage opportunities for pollinators.
Seedling Recruitment and Seedbank Characteristics
Mean cumulative E. graminifolia seedling emergence at Mount Pleasant and Webb Mountain was 2.4 ± 0.8 and 4 ± 1 seedlings m−2, respectively. Euthamia graminifolia seedlings therefore emerge in lowbush blueberry fields, but density is low. Our seedling emergence data are, however, limited to two sites in 2020 when there was generally less rainfall in April, May, and June relative to these same months in 2017 to 2019 (Table 3). Rainfall in April, May, and June of 2020 was also 57%, 9%, and 66% below the 1991 to 2020 Canadian Climate Normals for rainfall in the region of Nova Scotia containing our study sites. As such, we can conclude that E. graminifolia seedlings emerge in lowbush blueberry fields, but additional research is required to better estimate seedling density.
Euthamia graminifolia seedling density from soil core samples ranged from 38 ± 25 to 10,940 ± 1,456 seedlings m−2 (Table 5). These results suggest that E. graminifolia forms seedbanks in lowbush blueberry fields, despite the low levels of seedling emergence observed. Leck and Leck (Reference Leck and Leck1998) found an increase in both E. graminifolia percent cover and seed density in the seedbank in a 10-yr study of an abandoned hayfield. Similarly, conversion of an abandoned hayfield to agriculture reduced aboveground occurrence of E. graminifolia, but seeds were present in the seedbank after 5 yr of crop production (Hill et al. Reference Hill, Patriquin and Vander Kloet1989). In contrast, E. graminifolia was a common component of the aboveground vegetation but lacking from the seedbank in a tallgrass prairie ecosystem (Zylka et al. Reference Zylka, Whelan and Molano-Flores2016). Euthamia graminifolia is therefore capable of forming seedbanks but appears to do so on some sites and not on others. Given that lowbush blueberry fields appear to be sites where E. graminifolia seedbanks can form, management of this weed species should likely include strategies that target both newly emerging seedlings and established plants. Seedling emergence should also be determined at additional sites and paired with local seedbank data to improve our understanding of the potential importance of seedbanks and seedling recruitment for E. graminifolia establishment in lowbush blueberry fields.
Table 5. Total Euthamia graminifolia seedling emergence from soil cores collected from nonbearing-year lowbush blueberry fields in Nova Scotia, Canada, in 2020a
a A total of 12 soil cores were collected from each site using a 333-cm3 soil bulk density core sampler. Cores were sieved to remove coarse debris, combined with 1000 ml of Pro-Mix potting mix, placed in greenhouse trays, watered daily, and inspected for seedling emergence weekly.
b Values represent the mean ± SE of 12 replications.
Euthamia graminifolia ramet emergence and development to the flower bud and flowering stages were accurately predicted by GDD. Predictive models have been developed to facilitate use of GDD by growers to improve the timing of management decisions for this weed species. Growers can now base postemergence mesotrione application timing on percentage ramet emergence rather than ramet height, and this should be evaluated in future research. It is also anticipated that the timing of weed wiper and spot applications of symplastic herbicides such as glyphosate and flazasulfuron can now be based on model predictions for bud stage and flowering of emerged ramets to maximize efficacy of these herbicides on E. graminifolia. It is also acknowledged that E. graminifolia flowering is asynchronous with that of lowbush blueberry and that E. graminifolia flowers may provide an important late-season flower source for wild and managed pollinators. Seedling recruitment of E. graminifolia in lowbush blueberry fields was minimal despite the presence of an E. graminifolia seedbank at several sites. Seedling emergence of this weed species should therefore be studied in greater detail in lowbush blueberry fields, and management efforts should focus on control of both seedlings and established plants.
Acknowledgments
Field research sites were provided by Bragg Lumber Company, Tom Rudolph, Joe Slack, Chesley Walsh, Purdy Resources, and Steven Parks. The authors would also like to acknowledge field research assistance from Hugh Lyu, Jianan Lin, and Vanessa Deveau.
Funding statement
This research was supported by a Collaborative Research and Development Grant (CRDPJ 500615-2016) from the Natural Sciences and Engineering Research Council of Canada in collaboration with the Wild Blueberry Producers Association of Nova Scotia.
Competing interests
The authors declare no conflicts of interest.
Open access
