Harvest Weed Seed Control: Seed Production and Retention of Fallopia convolvulus, Sinapis arvensis, Spergula arvensis and Stellaria media at Spring Oat Maturity

If seeds retained on weeds at crop harvest could be collected and removed by the combine harvester, weed infestation could be reduced in the following years. We estimated the proportion of weed seeds that could be removed at oat harvest. The seed production and shedding pattern of Fallopia convolvulus, Sinapis arvensis, Spergula arvensis and Stellaria media, were assessed in two spring oat fields in Denmark during 2018 and 2019. Ten randomly chosen plants of each species were surrounded by a porous net before flowering. The start time of seed shedding was recorded, and the seeds were collected from the nets and counted weekly until oat harvest. Just before harvest, the retained seeds on the weed plants were counted. The ratio between harvestable seeds and shed seeds during the growing season was determined. On average 260, 195, 411 and 316 seeds plant−1 were produced by F. convolvulus, Sinapis arvensis, Spergula arvensis and S. media, respectively, of which in average 44%, 67%, 45% and 56% of the seeds were retained on the plants at harvest. There was a strong, positive correlation between the weed biomass and the total seed production.

Fallopia convolvulus is one of the most troublesome weeds in the world in cereal fields [2]. The climbing habit of the plant allows it to obtain sunlight while growing in stands of grain or other tall crops that may otherwise shade it [3]. The growth of F. convolvulus shoots is positively correlated with the daily temperature curve. However, as days become successively warmer, growth is successively less [4]. A single plant that emerges early in the growing season (April) may produce as many as 30,000 seeds, while individuals emerging two months later (June) may produce 15,000 seeds [2]. Fallopia convolvulus has an indeterminate flowering habit, which can result in flowers, immature seeds and mature seeds present on the same plant [3].
Sinapis arvensis is widely introduced and naturalized in temperate regions around the world. It has a persistent seed bank, a competitive annual growth habit and high fecundity; all characteristics contribute to its weedy nature, and ensure that it will remain a problem. During harvest operations,

Materials and Methods
We assessed the seed shattering and seed production of Fallopia convolvulus, Sinapis arvensis, Spergula arvensis and Stellaria media during the growing season of spring oat in two fields with sandy soil at the research station in Taastrup (55 • 38 N, 12 • 17 E), Denmark. The fields were ploughed in the spring and harrowed before sowing. One field was sown 19 April and harvested 2 August 2018 and the other field was sown 2 April and harvested 23 August 2019. The oat cultivar was Dominik and Symphony in 2018 and 2019, respectively, with the sowing rate of 170 and 175 kg ha −1 . No pesticide or fertilizer was applied both years on the study.
Ten plants of each species were selected randomly and surrounded by a trap comprised of a porous net (precision woven open mesh fabrics: SEFAR NITEX 06-475/56, Sefar, Germany; mesh opening: 475 µm-opening area: 56%) before flowering covering an area of approximate 710 cm 2 ( Figure 1). Traps were checked each week to record the start time of seeds shedding. Hereafter, seeds were collected using a portable vacuum cleaner every 6−8 days, depending on weather conditions, and stored in paper bags. The collected seeds were counted until oat harvest. Just before crop harvest, weed plants were cut at the soil surface, and the number of seeds retained on the weed plants was counted. The ratio of harvestable seeds to seeds produced by each weed species was determined.
Agronomy 2020, 10, x FOR PEER REVIEW 3 of 11 weed plants were cut at the soil surface, and the number of seeds retained on the weed plants was counted. The ratio of harvestable seeds to seeds produced by each weed species was determined. Weather data was provided from the research weather station in the area (55°67′ N, 12°30′ E) ( Figure 2). Daily maximum and minimum temperature data were used to calculate the Growing Degree Days (GDD) for the growing seasons: (1) where Tm and b0 represent the mean daily temperature and the base temperature (0 °C), respectively. S1 and S2 are the time of crop sowing and harvesting, respectively. To test whether the total seed production and dry weight of the plants varied significantly between the years, analysis of variance (ANOVA) followed by Fisher's least significant difference (LSD) for means separation was done using R version 3.6.1 [25]. Variance homogeneity was assessed by visual inspection of residual plots. To test whether seed shed differed between the weeks for each species, repeated measurement was used. The analyses were done using the extension packages lme4 [26] and multcomp [27]. Plants were considered as random effect, and the number of shattered seeds over the weeks considered as the response. The significance level was set to 0.05. The relationship between weed seed production and biomass was assessed using linear regression analysis. Weather data was provided from the research weather station in the area (55 • 67 N, 12 • 30 E) ( Figure 2). Daily maximum and minimum temperature data were used to calculate the Growing Degree Days (GDD) for the growing seasons: where Tm and b 0 represent the mean daily temperature and the base temperature (0 • C), respectively. S1 and S2 are the time of crop sowing and harvesting, respectively.
Agronomy 2020, 10, x FOR PEER REVIEW 3 of 11 weed plants were cut at the soil surface, and the number of seeds retained on the weed plants was counted. The ratio of harvestable seeds to seeds produced by each weed species was determined. Weather data was provided from the research weather station in the area (55°67′ N, 12°30′ E) ( Figure 2). Daily maximum and minimum temperature data were used to calculate the Growing Degree Days (GDD) for the growing seasons: (1) where Tm and b0 represent the mean daily temperature and the base temperature (0 °C), respectively. S1 and S2 are the time of crop sowing and harvesting, respectively. To test whether the total seed production and dry weight of the plants varied significantly between the years, analysis of variance (ANOVA) followed by Fisher's least significant difference (LSD) for means separation was done using R version 3.6.1 [25]. Variance homogeneity was assessed by visual inspection of residual plots. To test whether seed shed differed between the weeks for each species, repeated measurement was used. The analyses were done using the extension packages lme4 [26] and multcomp [27]. Plants were considered as random effect, and the number of shattered seeds over the weeks considered as the response. The significance level was set to 0.05. The relationship between weed seed production and biomass was assessed using linear regression analysis. To test whether the total seed production and dry weight of the plants varied significantly between the years, analysis of variance (ANOVA) followed by Fisher's least significant difference (LSD) for means separation was done using R version 3.6.1 [25]. Variance homogeneity was assessed by visual inspection of residual plots. To test whether seed shed differed between the weeks for each species, repeated measurement was used. The analyses were done using the extension packages lme4 [26] and multcomp [27]. Plants were considered as random effect, and the number of shattered seeds over the weeks considered as the response. The significance level was set to 0.05. The relationship between weed seed production and biomass was assessed using linear regression analysis.

Seed Production and Shedding in 2018
Figures 3-5 show the seed shedding pattern of F. convolvulus, Sinapis arvensis and Spergula arvensis, respectively. All three species started seed shedding more than one week before oat harvest in 2018. Fallopia convolvulus started to shed seeds between 12-19 July, and the largest number of shed seeds took place in the week before harvest (26 July-1 August) (Figure 3a). Seed shedding of Sinapis arvensis started between 20-27 July. The largest number of seeds was shed between 27 July-2 August, one week before harvest (Figure 4a). Spergula arvensis started seed shedding between 26 June-3 July, and the greatest number of shed seeds took place between 9-16 July (Figure 5a). On average 200, 109 and 697 seeds plant −1 were produced by F. convolvulus, Sinapis arvensis and Spergula arvensis, respectively, of which 57.5%, 27.7% and 39.0% of the seeds were shed before harvest. The difference between the weekly numbers of shed seeds for these three species was statistically significant (p = 0.001).

Seed Production and Shedding in 2018
Figures 3-5 show the seed shedding pattern of F. convolvulus, Sinapis arvensis and Spergula arvensis, respectively. All three species started seed shedding more than one week before oat harvest in 2018. Fallopia convolvulus started to shed seeds between 12-19 July, and the largest number of shed seeds took place in the week before harvest (26 July-1 August) (Figure 3a). Seed shedding of Sinapis arvensis started between 20-27 July. The largest number of seeds was shed between 27 July-2 August, one week before harvest (Figure 4a). Spergula arvensis started seed shedding between 26 June-3 July, and the greatest number of shed seeds took place between 9-16 July (Figure 5a). On average 200, 109 and 697 seeds plant −1 were produced by F. convolvulus, Sinapis arvensis and Spergula arvensis, respectively, of which 57.5%, 27.7% and 39.0% of the seeds were shed before harvest. The difference between the weekly numbers of shed seeds for these three species was statistically significant (p = 0.001).

Seed Production and Shedding in 2018
Figures 3-5 show the seed shedding pattern of F. convolvulus, Sinapis arvensis and Spergula arvensis, respectively. All three species started seed shedding more than one week before oat harvest in 2018. Fallopia convolvulus started to shed seeds between 12-19 July, and the largest number of shed seeds took place in the week before harvest (26 July-1 August) (Figure 3a). Seed shedding of Sinapis arvensis started between 20-27 July. The largest number of seeds was shed between 27 July-2 August, one week before harvest (Figure 4a). Spergula arvensis started seed shedding between 26 June-3 July, and the greatest number of shed seeds took place between 9-16 July (Figure 5a). On average 200, 109 and 697 seeds plant −1 were produced by F. convolvulus, Sinapis arvensis and Spergula arvensis, respectively, of which 57.5%, 27.7% and 39.0% of the seeds were shed before harvest. The difference between the weekly numbers of shed seeds for these three species was statistically significant (p = 0.001).

Sinapis arvensis 2018
Sinapis arvensis 2019 Seeds of S. media started to shatter one week before harvest (26 July-2 August) (Figure 6a). Stellaria media produced on average 52 seeds plant −1 , of which 16.2% were shattered before harvest.

Seed Production and Shedding in 2019
All four species started seed shedding more than one week before oat harvest in 2019. Fallopia convolvulus started to shed seeds between 12-19 July, and the largest number of shed seeds took place the week before harvest (16-23 August) (Figure 3b). Seed shedding of Sinapis arvensis started between 2-9 August. The greatest number of seeds was shed between 16−23 August, also one week before harvest (Figure 4b). Spergula arvensis started seed shedding between 24 June-1 July, and the largest number of shed seeds took place between 31 July-6 August (Figure 5b). Seed shedding of S. media started between 5-12 July. The largest number of seeds was shed between 19-26 July (Figure 6b Seeds of S. media started to shatter one week before harvest (26 July-2 August) ( Figure 6a). Stellaria media produced on average 52 seeds plant −1 , of which 16.2% were shattered before harvest. Seeds of S. media started to shatter one week before harvest (26 July-2 August) ( Figure 6a). Stellaria media produced on average 52 seeds plant −1 , of which 16.2% were shattered before harvest.

Seed Production and Shedding in 2019
All four species started seed shedding more than one week before oat harvest in 2019. Fallopia convolvulus started to shed seeds between 12-19 July, and the largest number of shed seeds took place the week before harvest (16-23 August) (Figure 3b). Seed shedding of Sinapis arvensis started between 2-9 August. The greatest number of seeds was shed between 16−23 August, also one week before harvest (Figure 4b). Spergula arvensis started seed shedding between 24 June-1 July, and the largest number of shed seeds took place between 31 July-6 August (Figure 5b). Seed shedding of S. media started between 5-12 July. The largest number of seeds was shed between 19-26 July (Figure 6b

Seed Production and Shedding in 2019
All four species started seed shedding more than one week before oat harvest in 2019. Fallopia convolvulus started to shed seeds between 12-19 July, and the largest number of shed seeds took place the week before harvest (16-23 August) (Figure 3b). Seed shedding of Sinapis arvensis started between 2-9 August. The greatest number of seeds was shed between 16−23 August, also one week before harvest (Figure 4b). Spergula arvensis started seed shedding between 24 June-1 July, and the largest number of shed seeds took place between 31 July-6 August (Figure 5b). Seed shedding of S. media started between 5-12 July. The largest number of seeds was shed between 19-26 July (Figure 6b). On average 321, 282, 125 and 580 seeds plant −1 were produced by F. convolvulus, Sinapis arvensis, Spergula arvensis and S. media, respectively, of which 53.2%, 38.1%, 69.5% and 70.7% of the seeds were shed before harvest. The difference between the weekly numbers of shed seeds for all four species was statistically significant (p = 0.001).

Plant Dry Weight
The average plant dry weight at crop harvest was only different between the years for Spergula arvensis (1.03 g in 2018; 0.28 g in 2019; p ≤ 0.002). The average plant dry weight at crop harvest for F. convolvulus, Sinapis arvensis and S. media was 1.6, 0.8 and 4.8 g, respectively.
For all species, there was a positive correlation between weed plant dry weight and total seed production ( Figure 7). Agronomy 2020, 10, x FOR PEER REVIEW 7 of 11 before harvest. The difference between the weekly numbers of shed seeds for all four species was statistically significant (p = 0.001). Seed production was significantly different between the years for Spergula arvensis (p = 0.0056), Sinapis arvensis (p = 0.03) and S. media (p = 0.0016), but not for F. convolvulus (p = 0.25).

Plant Dry Weight
The average plant dry weight at crop harvest was only different between the years for Spergula arvensis (1.03 g in 2018; 0.28 g in 2019; p ≤ 0.002). The average plant dry weight at crop harvest for F. convolvulus, Sinapis arvensis and S. media was 1.6, 0.8 and 4.8 g, respectively.
For all species, there was a positive correlation between weed plant dry weight and total seed production ( Figure 7).

Discussion
A high fraction of weed seed retained on the weed plants at harvest increases the potential for HWSC methods. We determined both the amounts of shed seeds and the retained seeds on the weeds to find the potentially harvestable ratio. The weed species showed different patterns in seed production and shedding, which also varied between the years. Different weather conditions characterized the two growing seasons. In 2018, the summer was unusually dry, warm and sunny, with many days having temperatures greater than 30 °C. It was the warmest summer since 1874 [28]. The oat plants became drought-stressed, resulting in less growth and a thinner oat stand, creating more space and light for the weeds. In the dry and warm weather in 2018, weeds and crop plants matured earlier than in 2019, resulting in a three weeks earlier harvest. In the dry season, the dry

Discussion
A high fraction of weed seed retained on the weed plants at harvest increases the potential for HWSC methods. We determined both the amounts of shed seeds and the retained seeds on the weeds to find the potentially harvestable ratio. The weed species showed different patterns in seed production and shedding, which also varied between the years. Different weather conditions characterized the two growing seasons. In 2018, the summer was unusually dry, warm and sunny, with many days having temperatures greater than 30 • C. It was the warmest summer since 1874 [28]. The oat plants became drought-stressed, resulting in less growth and a thinner oat stand, creating more space and light for the weeds. In the dry and warm weather in 2018, weeds and crop plants matured earlier than in 2019, resulting in a three weeks earlier harvest. In the dry season, the dry weight of F. convolvulus, Sinapis arvensis and S. media decreased by 68.3%, 40.8% and 11.3%, and the total seed production decreased by 90.9%, 61.3% and 37.5%, respectively.
In field trials in the United Kingdom, Wright et al. [29] evaluated the influence of two different soil moisture regimes on the competitive ability of Sinapis arvensis in spring wheat. Under dry conditions, the competitiveness of Sinapis arvensis measured as plant dry weight and seed production was significantly reduced. The dry weight and total seed production of Spergula arvensis increased by 72.9% and 81.9% in 2018 compared to the rainy season in 2019, as it was a rather weak competitor to oat [10].
On average, 195 seeds were produced by each plant of Sinapis arvensis in the two years. Forcella et al. [30] estimated the total viable seed production of Sinapis arvensis to be 2475 seeds m −2 in corn fields in two years. Seeds were completely dispersed before corn harvest in the warmest year, whereas in the cold year, one-third of seeds were retained on the plant and dispersed via combines during harvest [30]. Matured pods of Sinapis arvensis usually remain intact until the crop is harvested. During harvesting operations, seeds fall in the vicinity of the parent plants, or are most likely gathered with the crop seeds and afterwards spread with the chaff within the field during the harvesting operation [5,7]. We recorded that seeds started to shatter at 1583 GDD in the first year and 1816 GDD in the second year. Before harvest, 27.7% and 38.1% of the seeds were shed, while Burton et al. [31] reported that seed shatter of Sinapis arvensis began at 1110 GDD in spring wheat in 2015 in Saskatchewan, Canada. Only 10.6% of the total seed production of Sinapis arvensis shattered before harvest.
During the growing seasons, F. convolvulus climbs upwards, twining around the crop plants and in the case of cereals, it can cause lodging and make combine harvesting difficult [2]. Seeds started to shatter almost at the same period both years (1406 and 1418 GDD in 2018 and 2019, respectively) and almost with the same amount of seed shatter. Before harvest, 57.5% and 53.2% of seed shed. Burton et al. [31] reported from Saskatchewan, Canada, that seed shatter of F. convolvulus began at 1120 GDD in spring wheat in 2014 and 1060 GDD in 2015. They have observed a considerable variation in the seed shatter of F. convolvulus between the two years (31% of seed shattered before harvest in 2014 and 4.7% in 2015). They found that the high percentage of seed shattering in 2014 was caused by the dry conditions with periods of wind gusts close to the harvest time. However, we only observe a small variation in the seed shattering patterns. In both years, about 50% of the seed shed happened before harvest, but the total seed production was significantly reduced in the dry season. Dosland and Arnold [32] found that the supply of soil moisture was an essential factor for the competition between F. convolvulus and cereals. In a year with low precipitation, the weed germinated earlier and developed leaf area and dry weight, rapidly contributing to the early depletion of water in the field. The growth of the crop proceeded slowly, and there was an early loss of leaf area at heading time, resulting in a reduced yield [32].
There is limited information on seed retention and the possibility of harvesting Spergula arvensis and S. media seeds by a combine harvester. On average, 411 and 316 seeds were produced by Spergula arvensis and S. media in the two years, respectively, of which 45% and 56% was retained on the plants at harvest. In the dry season (2018), S. media started to shed seeds at 1559 GDD, while in the rainy year (2019), shattering started at 1300 GDD with 16.2% and 70.7% of seeds shed before harvest, respectively. Spergula arvensis started to shed seeds at 1097 and 1136 GDD in 2018 and 2019, respectively, with 39.0% and 69.5% of seeds shed before harvest. Tidemann et al. [33] reported that seed retention decreased as GDD increased. Seed retention over time varies by species, site, year and treatment. Many factors may contribute to the variation in seed retention of a plant species such as soil condition, drought, thunderstorm, wind, rainfall and competition between plants, and variation between biotypes [33].
The efficiency of HWSC relies on the proportion of the weed seed production that is retained at crop maturity [34]. Seed retention higher than 80% at crop maturity happens for many agronomically important weed species creating a unique opportunity to target these weed seeds and prevent them from becoming a part of the weed seed bank in the soil [34]. Delays in crop harvest can result in fewer weed seeds being captured because of a higher rate of seed shatter [35,36].
Fifteen cm reflects the practical harvest height for many growers. This height does not ensure that a large fraction of retained seeds on the weeds can be harvested for all the weed species. Sinapis arvensis is a tall plant (30-60 cm) with erect branching stems [14], making it possible to harvest the retained seeds by a combine harvester. This is also possible for Spergula arvensis (15-40 cm), which also has ascending or erect stems [14]. The stem of F. convolvulus (height: up to 2 m) is prostrate, but climbs the stems of other plants [14], which also make it possible to harvest retained seeds. However, S. media, which may become 20-60 cm tall, has decumbent to erect stems [14], which may make it difficult to collect a large proportion of the retained seeds at crop harvest. It is also likely that some seeds spread and fall to the soil surface during the harvesting process.
We observed a strong positive correlation between the weed biomass and the total seed production. The larger the plant, the more seeds were produced [37]. Schwartz et al. [37] also reported a strong correlation between the weed biomass and total seed production of Amaranthus tuberculatus and A. palmeri in soybean fields. A strong correlation between biomass production and seed production has been documented for many weed species [38][39][40][41]. Regardless of the species, the majority of smaller plants had low seed production, indicating that these plants were late-emerging cohorts [42].
We have now shown that a large proportion of seeds produced during the growing season of common weed species potentially can be collected and removed or destroyed [43][44][45] by a combine harvester at crop harvest. The next step will be to test how large a fraction of this potential a combine harvest actually collects, as it depends on several factors such as harvest height and the number of seeds dropping to the soil surface under the harvesting process.
Author Contributions: C.A. was responsible for funding acquisition and the design of the experiment. Z.B. conducted the practical work, data processing and wrote the first draft of the manuscript. Both authors reviewed, edited and accepted the final manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This work was a part of the project: 105 SWEEDHART-Separation of weeds during harvesting and hygienisation to enhance crop productivity in the long term. The activity was conducted under the "Joint European research projects in the field of Sustainable and Resilient Agriculture" under ERA-NET Cofund FACCE SURPLUS 2015. We thank Innovation Fund Denmark for financial support.