• weed population dynamics;
  • seed production;
  • seed burial;
  • seed mortality


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Post-dispersal seed losses in annual arable weed species are poorly quantified, but may be of significance for natural population control, especially if they can be manipulated. We hypothesized that weed seed predation on the soil surface was significant, so we measured rates in the field to estimate annual seed losses due to predation.
  • 2
    Temporal patterns of weed seed losses due to predation (‘demand’) as well as weed seed production (‘supply’) were measured from May to June until harvest in August 1999 and 2000 during 2-week exposure periods in four organic cereal fields in the Netherlands. The proportion of weed seeds lost to predators Mi (number of seeds consumed per number of seeds exposed per 14 days) was measured, using cards containing seeds of Stellaria media, Chenopodium album or Avena fatua. Seed production, Yi (number of seeds per m2 per 14 days), was measured in 2000, using seed traps.
  • 3
    Annual seed loss due to predation, M̄ (number of seeds consumed per number of seeds produced per year), was calculated based on Mi and the exposure period of seeds to predators, starting with seed shed and ending with seed burial. The importance of the length of the exposure period on total seed loss was explored using a model.
  • 4
    The temporal trend in Mi was consistent among farms and years: high in June and early July, lower in the second half of July and negligible in August and after harvest. Total seed production varied considerably among fields, i.e. 800–16 000 seeds per m2 per year. The timing of peak seed production also varied substantially.
  • 5
    Calculated M̄ ranged from 32% to 70%, when assuming continuous exposure of seeds to predators from seed shed till crop harvest. When exposure was limited to 2 or 4 weeks after seed shed, M̄ decreased to 18–57% or 28–67%, respectively. Differences between fields and weed species were mainly due to differences in the timing of seed shed.
  • 6
    Synthesis and applications. Our results suggest that seed predation in organic cereal fields is an important factor shaping the population dynamics of arable weeds. A combination of environmental conditions (hot and dry weather) and agricultural practices (an early crop sowing) can advance weed phenology and postpone seed burial, resulting in higher proportions of weed seed loss to predation in cereals.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Weed management is a major cost and constraint in organic arable farming. For example, Schotveld & Kloen (1996) found that hand weeding on Dutch organic farms in the early 1990s required about 1500 h of labour per farm (approximately 60 ha). In rotations of alternating combine harvested crops and root crops, a major input into weed seed banks is often made in cereal crops (Mertens 1998). Weed management options in cereals are limited, while the direct need for control to prevent damage is relatively low, because cereals can competitively suppress weeds. However, because weed seed production is not completely prevented in cereals, there is an increased risk of high population densities of weeds in subsequent, less competitive crops.

A total of 70–99% of the weed seeds that are produced annually in cereals cannot be retrieved from the seed bank or do not emerge as seedlings (Mitze 1992; Cardina & Norquay 1997). Predation may be responsible for the larger part of these losses. It has been shown in non-agricultural ecosystems that seed predation is an important factor in limiting the population expansion of plant species. (e.g. Reichman 1979). In agroecosystems, high consumption rates of weed seeds by seed-eating animals over short periods have also been recorded (e.g. Table 1). Whether this is high enough to stabilize or at least slow down the build-up of the weed seed bank depends on the annual seed predation rates. Weed population dynamics are strongly affected by seed mortality, and an annual loss of 25–50% seems to be enough to slow down weed population growth substantially (e.g. Firbank & Watkinson 1986; Medd & Ridings 1989). However, annual losses due to predation have as yet not been assessed. If it significantly impacts on seed bank dynamics, the loss of seed predator populations could be detrimental to weed control and should be prevented.

Table 1.  Examples of weed seed predation rates (seeds consumed per seeds exposed per exposure period) in arable fields
Proportion predationCropPeriodReference
0·20–0·90 per 3 weeksOats, grass-cloverJul–AugAndersson (1998)
0·03–0·62 per 7 daysSoybeanOct–NovBrust & House (1988)
0·14–0·99 per 14 daysMaizeJan–DecCardina et al. (1996)
0·49–0·84 per 7 daysMaizeSepMenalled et al. (2000)
0·05–0·35 per 3–4 daysField margins around cerealsSep–OctPovey, Smith & Watt (1993)
0·24–0·90 per 2 weeksCerealsJun–AugTooley et al. (1999)

The annual seed loss due to epigaeic predation is determined by the duration of the exposure period and the rate of predation during that period. Exposure starts with seed shed, although some seed predators will consume ripe unshed seeds from the plant (Kjellsson 1985). The timing of seed shed is species, crop and climate specific and is documented for a number of weed species (e.g. Rauber & Koch 1975; Leguizamón & Roberts 1982). Exposure ends with burial (Thompson 1987; Hulme 1994), or germination of the seeds. Seed burial is usually accomplished by tillage, especially stubble cultivation and ploughing after crop harvest (Cousens & Moss 1990). However, some seed burial will occur during the cropping season due to natural causes. The rate of seed burial seems to depend largely on weather, soil type and seed characteristics (e.g. Peart 1979; Chambers, McMahon & Haefner 1991). Preliminary trials under Dutch conditions indicated that 50% of small-sized seeds of Chenopodium album L. disappeared from the soil surface in about 2 weeks (Seguer Millàs 2002). Burial rate of large-sized species, such as Avena fatua L. and Polygonum convolvulus L., was considerably lower. Germination of newly produced weed seeds prior to crop harvest is limited due to reduced light intensity and quality under a dense canopy (Pons 2000), unfavourable conditions for germination on the soil surface during summer (Grundy & Mead 1998), and primary dormancy in many weed species (e.g. Vleeshouwers 1997). The exposure and vulnerability of weed seeds to seed predation is therefore mainly affected by variation in the timing of seed shed and seed burial. Differences in predation probability among weed species may also result from differences in attractiveness of the seeds to predators, caused for example by physical features, palatability and nutritional value (e.g. Borchert & Jain 1978; Jørgensen & Toft 1997).

We investigated the potential role of epigaeic seed predation in limiting the expansion of weed populations in organic cereal fields in the Netherlands. We estimate the magnitude of annual seed losses due to predation by determining:

  • 1
    the magnitude and seasonal variability in epigaeic seed predation;
  • 2
    differences in seed predation among weed species differing in seed size and temporal pattern of seed shed; and
  • 3
    the magnitude and seasonal variability in weed seed production.

Because seed burial rate was not included in this study, we used a simple model to explore ranges of seed predation per weed species and per farm under different scenarios of seed burial.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

experimental sites

Trials were conducted in organic cereal fields on four farms in 1999 (farms 1, 2, 3 and 4) and in 2000 (farms 2, 3, 4 and 5). Farm 1, ‘De Korenbloem’, and farm 2, ‘NZ27’, are located on the clay-soils of the polder ‘Zuidelijk Flevoland’, an area of low-lying land reclaimed from the sea and protected by dikes. Farm 3, ‘Dr H.J. Lovinkhoeve’, is located in the ‘Noordoost polder’, also on clay soil. The landscape of both polders is characterized by large fields and small noncrop areas. Farm 4, ‘Deelerwoud’, and farm 5, ‘Klein Boeschoten’, are located in the forested sand hills of the ‘Veluwe’, an area with relatively small fields and large noncrop areas. At each farm, one cereal field was selected. In both years, fields were sown with winter wheat Triticum aestivum L., except at the more drought sensitive farms 4 and 5 in 2000 when triticale Triticosecale was sown, and on farm 3 where spring wheat T. aestivum was sown in both years. Crop development on the fields at farm 3 (spring cereal) was delayed by about three weeks compared with the fields at the other farms. Details on sowing date, harvesting date, row distance, and crop variety are provided in Table 2.

Table 2.  Details concerning the experimental sites
YearFarmLocationSoil typeWheat varietyField size (ha)Row (cm)Date of sowingDate of harvest
  • *

    Triticale Triticosecale instead of wheat Triticum aestivum L.

 1ZeewoldeClayRenan/Tambor 7·5 12·51-12-19982-8-1999
 2ZeewoldeClayTambor14 12·510-11-19983-8-1999
    Baldus  24-3-19993-8-1999
 3MarknesseClayLavett 4·5 3029-4-199924-8-1999
 4HoenderlooSandTambor 5±2024-12-19984-8-1999
 2ZeewoldeClayHereward14 12·54-11-19991-8-2000
3MarknesseClayBaldus 5 3012-4-200028-8-2000
4HoenderlooSandBinova* 5±1523-10-199924-7-2000
5GarderenSandEldorado* 4  8·615-12-19994-8-2000

weed species

Weed species examined were chickweed Stellaria media (L.) Vill., common lambsquarters or fat hen C. album, and wild-oats A. fatua. Seed production of S. media occurs throughout the year (in wheat, ≈ 2600 seeds per plant; Mitze 1992). Seeds are small (diameter 0·8–1·4 mm; Hanf 1982), long-lived (> 5 years, Roberts 1964), and have a thin seed coat. Chenopodium album releases its seeds mainly in autumn (in soybean, ≈ 4000 seeds per plant; Clements et al. 1996). Seeds of C. album are slightly bigger than those of S. media (diameter 0·7–1·5 mm; Hanf 1982), have a hard seed coat, and are also long lived (> 5 years, Roberts 1964). Of the three weed species, A. fatua produces fewest seeds (in cereals up to 120 seeds per plant; Wilson 1981). The seeds are larger (≈ 7 mm long, diameter 1–2 mm) and shorter lived than those of the other two species (< 3 years; Wilson 1981). Seeds of A. fatua are shed in late summer and autumn (Rauber & Koch 1975). The seed coat is thin and offers little or no protection against seed predators, but seeds are somewhat protected from predators by the lemma and palea. Seeds of A. fatua have a hygroscopically active awn, which can facilitate burial by propelling the seed into cracks in the soil surface (Peart 1979).

weed seed predation

Seed predation was assessed by monitoring removal of randomly scattered seeds glued to ‘seed cards’ made of firm, high quality sand paper (4.0 × 9·5 cm, KWB ‘waved’ grain size 60 or 80) sprayed with repositionable glue (3M spray mount, art. no. 6065) modified after Brust & House (1988). The sandpaper did not curl, bend or soak after showers, and seeds did not germinate. The glue ensured that the seeds stayed on the cards under normal weather conditions while seed predators were able to remove the seeds. The remaining glue was covered with fine sand to prevent invertebrates sticking to the glue. Nails were used to secure the seed-cards to the ground. The number of seeds per card was varied to standardize the seed weight to 20–21 g per card: 50 seeds of S. media (0·4 mg per seed), 30 seeds of C. album (0·7 mg per seed) and 10 seeds of A. fatua (2·1 mg per seed).

There was no evidence of attraction, avoidance or any other behavioural disturbance to any of the seed predators caused by either the sandpaper or the glue. Extensive visual observations during daytime and by infrared video recordings during the hours of darkness, indicated that large seed predators (mainly mice) simply bit the seeds from the seed cards without difficulty (Westerman 2001). Although some of the smaller ground beetle species (Amara spp.) had difficulty freeing the seeds from the glue, this did not appear to inhibit seed consumption on the cards (P.R. Westerman, unpublished data). In addition, a population of large granivorous groundbeetles, Harpalus rufipes DeGeer, was successfully maintained in the laboratory for several weeks by feeding them seeds on seed cards. Furthermore, Brust & House (1988), using a similar method, could not detect any significant effect on predator behaviour.

We assumed that seeds that were removed from the cards by animals were actually consumed or stored in a seed cache and consumed later. However, it is possible that seeds were removed without being eaten or that seeds survived passage through the digestive track of larger predators. We regularly encountered damaged seeds on the cards and piles of husks from wild oats next to cards, characteristic of consumption by ground beetles and small rodents, respectively (e.g. Kjellsson 1985).

Seed cards (R) were placed in each field (40 cards per species for a total of 120 cards per field), such that the 40 cards of each species were divided over four transects (10 cards per transect). Along each transect, 10 cards were spaced 10 m apart with the first card 10 m from the field edge. Transects ran along the rows and thus extended 110 m into each field. The position of the 12 transects in each field was randomised by: (1) dividing the available cross-row space (100–150 m, depending on the dimensions of the field) into 15 adjacent sections (three of which were used in another study); (2) randomising the location of the transect within each section; and (3) randomly assigning species to transects. Seed cards were collected and replaced every 14 days.

In 2000, 10 control cards per weed species (C) were placed in each field to assess seed loss due to causes other than predation (e.g. wind, rain and loss of adhesive power). Three cards (one of each species) were placed at the bottom of a fine meshed sieve (20 cm diameter) with 3-cm high rims. The rims were lined with strong double-sided adhesive tape. The sieves were completely wrapped in a sheet of small-mesh (11 × 11 mm) metal wire netting and supported about 10 cm above the ground with three steel tent pegs to keep out seed predators. The cards in these sieves were replaced every 14 days, simultaneously with the predator-exposed seed cards. The number of seeds remaining on the control cards, Ci, and the field cards, Ri, were counted for each observation period i and weed species and used in the statistical analysis (see below). For the biological interpretation of the data, the counts were converted into estimates of Mi, the proportion seed predation relative to the number of seeds that remained on the control cards (Abbott 1945): Mi = (Ci − Ri)/Ci (number of seeds consumed per number of seeds exposed per 14 days). In the first year of the study, controls with a different design were used. These controls were discarded because not all predators were excluded effectively. The background seed loss in 2000, averaged over farms and observation periods, was used to calculate Mi values for 1999 data.

Because there can be considerable variation between farms in the timing of seed shed of weeds, the first seed cards were usually placed into the fields immediately after the completion of mechanical weed control at the end of May in 1999 and at the end of April in 2000 (Fig. 1). Weed control in organic cereals consists of frequent harrowing in spring until the crop is too high (≈ 50 cm) to let the machinery pass through. At farm 3 in 1999, the first exposure was just before and the second exposure immediately after the final round of harrowing. At farm 4 in 1999, the first seed cards were placed nine weeks after the last weed control. Predation occurring outside the period with natural seed production will illustrate the need for seeds, i.e. predation ‘if seeds had been present’. The last seed cards were collected from each field in July or August, just before harvesting. In 1999, the last exposure period of seed cards at farms 2 and 3 was 11 instead of 14 days, due to early harvests. At farms 2 and 4, an additional 2-week observation was carried out several weeks after harvest in September 1999 to study seed predation after harvest. By that time, the field at farm 2 had been cultivated and sown with yellow mustard Sinapis alba L. The sequence of events in each field is summarized in Fig. 1. The earlier start of the trials in 2000 was due to warmer spring weather.


Figure 1. Sequence of events on each of the farms in 1999 (A) and 2000 (B): time of sowing ([DOWNWARDS ARROW]), harvest ([UPWARDS ARROW]), tillage (▴), and observation periods (—).

Download figure to PowerPoint

weed seed production

Weed seed production was assessed by ‘seed traps’, modified after Forcella, Peterson & Barbour (1996). Each trap consisted of two aluminium trays (23 × 10 × 6 cm; length × width × height) placed lengthways between two crop rows and buried in the soil such that the rim was 0·5–1·0 cm above the soil surface. Total trap surface was 0·046 (m2 per trap). The trays were drained by small holes in the bottom, and lined with fine cloth to collect the seeds. Sheets of fine-mesh (3·5 mm) plastic grid and small-mesh (11 mm) metal wire netting were placed on top to keep out invertebrates and vertebrates, respectively. The most common seed producing weed species were, on farm 2: S. media and Capsella bursa-pastoris (L.) Medicus, on farm 3: S. media, Veronica spp., C. bursa-pastoris, P. convolvulus, Poa spp. and P. persicaria L., on farm 4: Apera spica-venti (L.) P.B., Viola arvensis Murr., C. album, Matricaria spp., and C. bursa-pastoris, on farm 5: A. spica-venti, Vicia cracca L. C. album and Erodium cicutarium (L.) L’Her.

Trap design is known to influence total seed catch and relative catch of different species, and no single design was found to be optimal for all weed species (Johnson & West 1988; Forcella, Peterson & Barbour 1996). With the current trap design, we aimed to combine the advantages of previous designs, i.e. high side walls and buried, ensuring a true reflection of seed production. Furthermore, the buried position of the traps increased the chance of catching seeds of decumbent weed species, such as S. media or Veronica spp. The small grid on top of the trays may have hindered the collection of some of the larger seed, e.g. P. convulvulus or V. cracca, but these heavier seeds usually entered the traps without difficulty. We considered the error in seed catch due to obstruction to be smaller than the potential error due to seed consumption in the absence of the grid.

In 2000, 30 seed traps were placed in each field, such that the total trapping surface was 1·38 m2 per 1–1·5 ha. Traps were placed in six of the 12 transects, with five traps per transect, 1 m to the right of every other seed card. The cloth containing the seeds was replaced every 2 weeks simultaneous with the replacement of the seed cards. Seed trapping started 2 weeks after seed predation measurements and before seed dispersal and was terminated just before harvest (farms 2, 5) or just after harvest (farms 3, 4). Seeds were identified and counted, and converted into estimates of seed numbers per observation period i, Yi, and per weed species (number of seeds produced per m2 per 14 days).

statistical analysis

The number of removed seeds per card was analysed per farm and, when appropriate, per sampling period. A regression model (GLM) was used with a logit link and a binomial variance function allowing for overdispersion (genstat 5; Genstat 5 Committee 1993). The analysis of deviance was performed in two stages: first, the unprocessed seed removal data from the field were compared with data from the controls to affirm the occurrence of significant seed predation. Next, the number of removed seeds per card, except the control cards, was analysed to test for the effect of sampling period and weed species. When these proved significant, the effect of weed species was analysed per sampling period to establish the level of significance. Because the results from the control cards of 1999 were discarded, the first step of the analysis was omitted for the 1999 data. Significance was evaluated in terms of mean deviance ratios, which are in turn evaluated by comparison with F-distributions (α = 0·05) (e.g. genstat 5: McCullagh & Nelder 1989; Genstat 5 Committee 1993). In the case of a significant effect of weed species, a t-test was used to rank the weed species in order of proportion seed loss.

estimates of annual seed loss due to predation

Observed seasonal trends in Mi were combined with observed patterns of seed shed Yi in 2000 to assess total seed losses due to predation over a season, &#x004d;̄. Because the rate of seed burial is unknown, we calculated the annual seed loss assuming three different scenarios with decreasing duration of exposure to seed predators: (1) continuous exposure, from seed shed till harvest; (2) limited exposure period of 4 weeks; and (3) limited exposure period of 2 weeks. It was assumed that seed exposure ends ultimately with stubble cultivation immediately following harvest at the end of August, when most seeds will be buried. Additional seeds may be produced by weeds surviving cultivation or by newly emerged weeds.

The annual proportion seed loss due to predation, &#x004d;̄, was calculated as 1 − S̄, with S̄ the annual proportion of seeds that survive predation:

  • image

With k = n, when assuming continuous exposure, k = min(i + 1, n) when assuming 4-week exposure, and k=i when assuming a 2-week exposure.

Here, inline image is the sum of seed production over one season, while inline image is the surviving number of seeds at the end of the year out of those produced in the i-th period inline image is the overall number of seeds surviving during a season out of those produced.

&#x004d;̄ was calculated for each weed species that produced seeds, &#x004d;̄w. Species-specific seed predation values Mi were used for S. media, C. album, and A. fatua. For the other weed species, no specific predation data were available, and therefore the average value of Mi over the three weed species was used. Species-specific data on seed production Yi were used for all weed species. In addition, &#x004d;̄ was calculated as an average of all weed species on a farm:

  • image

With Yw the total seed production for a given weed species. All calculations were carried out in microsoft excel.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

seasonal pattern in seed loss in 1999

Seasonal patterns of the proportion seed predation, Mi, on the four farms were similar: high in June and early July (0·6–0·95) with lower levels in the second half of July (0·25–0·6). Measurements on farm 3 on 13 and 24 August indicated that predation was zero in August (Fig. 2a). Predation was negligible after harvest (10 September) on farms 2 and 4. The effect of sampling period was significant (P < 0·005) on all farms except farm 1 (farm 2, inline image = 139·1; farm 3, inline image = 141·4; farm 4, inline image = 229·8).


Figure 2. Proportion of weed seeds lost to predators per 2 weeks, Mi (average over three weed species) in cereal fields on four organic farms in 1999 (A) and 2000 (B) (n = 120). Error bars represent the standard error of the mean (SE).

Download figure to PowerPoint

differences in seed loss among weed species in 1999

Ranking of weed species changed over time: sometimes seed predation was highest for S. media (Fig. 3a, 8 July; Fig. 3b,d), sometimes for C. album (Fig. 3a, 23 July; Fig. 3a, 30 July), or A. fatua (Fig. 3a, 25 June; Fig. 3c). This was confirmed by a significant interaction between weed species and sampling period (P < 0·005) on all farms.


Figure 3. Proportion weed seeds lost to predators per 2 weeks and per weed species &#x004d;̄w, in 1999 [asterisks above the Figure indicate significant differences on that date (GLM, t-test; *P < 0·05; **P < 0·01; ***P < 0·005)]: farm 1 (A), farm 2 (B), farm 3 (C) and farm 4 (D) (n = 40).

Download figure to PowerPoint

background seed loss in 2000

The proportion seed loss from cards due to wind, rain and loss of adhesive power of the glue varied from 0·018 to 0·373 (Fig. 4). Averaged over farms, sampling periods, and weed species, background seed loss in 2000 was 0·23 ± 0·009 (mean ± SE). Background seed loss was significantly lower for A. fatua (0·14 ± 0·023) than for S. media (0·26 ± 0·013) and C. album (0·20 ± 0·015). The number of recovered seeds on control cards was significantly higher than on seed cards in the field, on all farms and all sampling periods (inline image > 5·0; P < 0·05).


Figure 4. Proportion background seed loss, per 2 weeks, in cereal fields on four organic farms in 2000 (n = 30). Error bars represent the standard error of the mean (SE).

Download figure to PowerPoint

seasonal pattern in seed loss in 2000

As in 1999, the general trend was very similar for the four farms (Fig. 2b). Mi was low in May (0·25) on farm 2. In June, Mi was high on all farms (0·63–0·93), and dropped to 0·24–0·57 in early July, remaining at that level for most of July. Results from farms 3 and 5 suggest that Mi declined further thereafter. The effect of sampling period was significant on all farms (P < 0·005).

differences in seed loss among weed species in 2000

As in 1999, ranking of weed species changed over time. However, seeds of A. fatua suffered the highest losses on almost all farms and periods (Fig. 5a–d). Seeds of C. album usually suffered least from predation. The interaction between weed species and sampling period was significant on all farms.


Figure 5. Proportion of weed seeds lost to predators per 2 weeks and per weed species &#x004d;̄w, in 2000 [asterisks above the figure indicate significant differences on that date (GLM, t-test; *P < 0·05; **P < 0·01; ***P < 0·005)]: farm 2 (A), farm 3 (B), farm 4 (C) and farm 5 (D) (n = 40).

Download figure to PowerPoint

weed seed production in 2000

Seed shed was variable in time and among farms (Fig. 6a–d). When expressed as numbers of seeds per m2, estimates of total seed production were 1171 on farm 2, 800 on farm 3, 1109 on farm 4, and 16 263 on farm 5. The low seed production on farm 3 may be related to the shorter growing period in spring cereals compared with winter cereals. Peak seed production occurred on 30 June on farm 2, 21 July on farm 4 and farm 5, and 11 August on farm 3 (Fig. 6a,b). Combine harvesting did not greatly increase seed shed on farms 3 and 4 on 28 August and 24 July, respectively (Fig. 6b,c).


Figure 6. Weed seed production (seeds per m2 per 14 days) on farm 2 (A), farm 3 (B), farm 4 (C) and farm 5 (D) in 2000 (n = 30). Error bars represent the standard error of the mean (SE).

Download figure to PowerPoint

estimates of annual seed loss due to predation

When comparing &#x004d;̄w for continuous exposure to seed predators, some weed species within a particular field suffered much greater seed losses than others (Table 3). For species other than S. media, C. album, and A. fatua, estimates of &#x004d;̄w are based on the timing of seed shed associated with the average estimate of predation at that time for the three weed species tested. On farm 3, Poa spp. and C. bursa-pastoris could loose 56% of their seeds to predators, whereas only about 18–20% of the seeds of Veronica spp., P. convolvulus and P. persicaria would be consumed. The calculated differences are related to the timing of seed shed. Seeds of Poa spp. and C. bursa-pastoris were released relatively early in the season (Fig. 6b) and were therefore exposed to high levels of seed predation in early summer, with exposure lasting throughout the growing season. Furthermore, the estimated &#x004d;̄w of a particular weed species differed between farms. On farm 2, the annual proportion seed predation of S. media was 70%, whereas on farm 3 it was only 30% and on farm 5 it was 61%. Similarly, &#x004d;̄w for P. persicaria was 51% on farm 2, 18% on farm 3 and 38% on farm 4. Again, this was caused by a difference in the timing of seed shed between farms. Estimates of &#x004d;̄ were sensitive to the seed burial scenario used (Table 3). The estimates were always highest for continuously exposed seeds, intermediate for the 4-week exposure and lowest for the 2-week exposure.

Table 3.  Estimates of annual seed loss due to predation for the most abundant weed species in cereals on four farms in 2000, M̄w, and averaged over all weed species, M̄, assuming that seeds are exposed to seed predators continuously, for 4 weeks or for 2 weeks
FarmWeed speciesw (seed consumed per seed produced per year)
Exposure period
Continuous4 weeks2 weeks
2Capsella bursa-pastoris0·760·700·55
Chenopodium album0·640·570·40
Polygonum persicaria0·510·480·45
Stellaria media0·700·550·35
All species, M̄0·700·560·36
3Capsella bursa-pastoris0·570·470·30
Poa spp.0·560·470·31
Polygonum convolvulus0·180·180·12
Polygonum persicaria0·180·170·13
Stellaria media0·300·250·16
Veronica spp.0·200·200·12
All species, M̄0·320·280·18
4Apera spica-venti0·540·540·38
Capsella bursa-pastoris0·640·580·42
Chenopodium album0·590·560·44
Galeopsis tetrahit0·590·590·38
Matricaria spp.0·470·470·38
Polygonum persicaria0·380·380·38
Viola arvensis0·690·660·49
All species, M̄0·550·540·39
5Apera spica-venti0·670·670·57
Capsella bursa-pastoris0·820·810·72
Chenopodium album0·860·850·72
Erodium cicutarium0·540·540·48
Polygonum convolvulus0·480·480·44
Stellaria media0·610·590·49
Vicia cracca0·410·410·38
All species, M̄0·670·670·57

The estimated annual seed loss due to predation per farm, averaged over all weed species, was high on farm 2 and 5 (70% and 67%, respectively) and relatively low on farm 3 (32%; Table 3), when assuming no seed burial. Limiting the exposure period of all weed seeds on a farm to 4 or 2 weeks after seed shed reduced the overall seed loss per farm to 28–67% or 18–57%, respectively. For farms with predominantly late maturing weeds (e.g. farm 5) the reduction of the exposure period had little effect on the estimated annual seed losses.

&#x004d;̄ was lowest on farm 3, which also had the lowest total seed production (800 seeds per m2), and highest on farms 2 and 5 which had moderate (1171 seeds per m2) and high seed production (16 263 seeds per m2), respectively. However, &#x004d;̄ could not cancel out the effect of the high seed production; seed return on farm 5 (5415 seeds per m2), was 10 times higher than on farm 3.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Estimates of &#x004d;̄, annual weed seed losses due to predation in cereals, ranged from 32% to 70% per farm, when assuming continuous exposure to seed predators (Table 3). When the exposure period of seeds was limited to 4 or 2 weeks after seed shed, &#x004d;̄ still reached levels of 28–67% and 18–57% per farm, respectively. These figures indicate that epigaeic seed predation is responsible for a substantial part of the unaccounted seeds in cereals (Mitze 1992; Cardina & Norquay 1997). In fact, seed predation seems to be a more important loss factor than mortality in the seed bank due to physiological ageing and microbial activity (e.g. Roberts & Feast 1972). It is clear from this study that seed predation contributes substantially to the containment of weed population growth on at least three of the four farms participating in this study. From a population dynamics point of view, seed predation appears to be as effective as mechanical methods of weed control (e.g. Wilson, Wright & Butler 1993).

As the trend in seed demand is very consistent over farms and years, differences in the calculated estimates of &#x004d;̄ between farms are mainly caused by variation in the temporal patterns of exposure, which in turn depends on the timing of seed shed and seed burial. Environmental conditions and agricultural practices that advance weed phenology, cause early seed shed and postpone seed burial, should result in higher proportions of seed loss in cereals. For example, estimates of &#x004d;̄ were appreciably higher for crops sown in winter than for the crop sown in spring (Table 3). However, because total seed production was much higher in the winter-sown crops, seed return was ultimately higher in winter than in spring cereals. Exceptionally warm and dry weather conditions, causing early ripening and seed shed, may have been responsible for the high seed losses reported by Mitze (1992). Furthermore, the composition of the weed vegetation influences the timing of seed shed, because usually only one or a few species are responsible for the majority of the seeds produced on a farm (Jones & Naylor 1992). For example, in this study seeds of A. spica-venti, the dominant weed on farm 5, were shed late, while seeds of S. media, the dominant weed in farm 2, were shed relatively early. Total weed seed production was comparable to that usually found in cereals (e.g. Jones & Naylor 1992; Moorcroft et al. 2002).

In this study, observed differences in seed preference between weed species were small, although larger differences have been reported in the literature (e.g. Borchert & Jain 1978). Calculated differences in &#x004d;̄w can therefore largely be ascribed to differences in the timing of seed shed. The early maturing C. bursa-pastoris was predicted to have suffered the highest seed losses due to predation on all farms. Species that started to disseminate seeds early but continued to do so during most of the growing season, such as S. media, V. arvensis, and in this study also C. album, suffered substantial losses as well. Although seed burial rate was assumed to be identical for all weed species, preliminary data indicate that small-sized seeds are more quickly incorporated into the soil and thus exposed to predators for a shorter period of time than large-sized seeds (Seguer Millàs 2002). In this study, all weed species with large-sized seeds, such as P. convolvulus and V. cracca, were late maturing weed species, shedding seeds in late summer and autumn, when predation pressure is lowest. The production of large seeds late in the season may be an adaptation for predation avoidance (Brown & Venable 1991). Differential seed predation in cereals may influence the composition of the weed flora in the long term. However, in the Netherlands, cereals are grown in rotation with other crops such as potatoes and sugar beet, which may exert a different selection pressure and counteract any shift in weed composition induced in cereals.

Invertebrates, mainly granivorous ground beetles, were the dominant seed predator on farm 4 in 1999, while vertebrates, presumably wood mice Apodemus sylvaticus (L.), were the main seed predator on the remaining fields (Westerman 2001; Westerman et al. 2003). We do not know why seed predation decreased during the cropping season. We anticipated an increase in the numbers of seed predators, as a result of reproduction in spring and summer (Zhang et al. 1997; Ouin et al. 2000), and thus a gradual increase in seed demand, as observed in maize fields in Ohio by Cardina et al. (1996). Other studies on seed predation in arable fields are too fragmentary in time to detect seasonal patterns (Table 1). Seed demand decreased when seed availability increased and it is possible that seed predators simply became satiated (e.g. Cardina et al. 1996).

Density-dependent responses to local differences in seed supply may be involved in seed predation (e.g. Reichman 1979). Further studies are required to elucidate which density dependent mechanisms may be involved, and whether they might have biased our estimates of annual seed loss. Our experimental set-up may have been sensitive to density-dependent effects, as we created areas of high and low food availability, via (1) seeds on cards, e.g. 10, 30 or 50 seeds per card; (2) the uneven distribution of transects over the field; and (3) the different sizes of the observation areas, viz. 1 or 1·5 ha. On average, we introduced 0·24 seeds per m2 or 0·00017 g m−2 every 2 weeks. These numbers are negligible compared with the natural seed production per 2 weeks, which was at least 100 times higher than our maximum. In addition, the distribution of naturally produced seeds was highly variable spatially (data not shown). We therefore conclude that the error introduced by our experimental set-up is insignificant.

The protection and encouragement of naturally occurring seed predators must be worthwhile, although at this stage we do not know how to achieve this goal. Despite big differences between fields, in for instance the level of seed production, weed species composition, type of cereal, sowing date and density, field size, soil type, row spacing or type of field edge vegetation, none of these factors resulted in a noticeable change in trend or level of seed predation. Landis & Marino (1999) predicted that seed predation would be higher in landscapes with a higher abundance of noncrop habitats such as the Veluwe area included in this study, because these should support a more abundant and diverse fauna of seed predators. However, their prediction was not confirmed in this study. Apparently, the seed predator populations are robust and relatively insensitive to changes in management or environments. They may therefore already be at their maximum. If they are not, we need to know which resources are essential for their survival and reproduction. Only then can habitat management be directed to maximize weed control by seed predators.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This study was financially supported by the graduate school for Production Ecology and Resource Conservation (PE & RC) of Wageningen University. We wish to thank the farmers, P. van Andel, D. Monsma, A. Siepel, V.G.F. Repelaer and R. Joppe for allowing us to use their cereal fields, and for their patience and hospitality. Furthermore, we want to thank Ms Ans Hofman and our colleagues of the University farm ‘Unifarm’ for assistance and Dr Paul Marino for critically reviewing an early version of the manuscript.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Abbott, W.S. (1945) A method of computing the effectiveness of an insecticide. Journal of Economic Entomology, 18, 265267.
  • Andersson, L. (1998) Post-dispersal seed removal in some agricultural weeds. Aspects of Applied Biology, 51, 159164.
  • Borchert, M.I. & Jain, S.K. (1978) The effect of rodent seed predation on four species of California annual grasses. Oecologia, 33, 101113.
  • Brown, J.S. & Venable, D.L. (1991) Life history evolution of seed-bank annuals in response to seed predation. Evolutionary Ecology, 5, 1229.
  • Brust, G.E. & House, G.J. (1988) Weed seed destruction by arthropods and rodents in low-input soybean agroecosystems. American Journal of Alternative Agriculture, 3, 1925.
  • Cardina, J. & Norquay, H.M. (1997) Seed production and seedbank dynamics in subthreshold velvetleaf (Abutilon theophrasti) populations. Weed Science, 45, 8590.
  • Cardina, J., Norquay, H.M., Stinner, B.J. & McCartney, D.A. (1996) Postdispersal predation of velvetleaf (Abutilon theophrasti) seeds. Weed Science, 44, 534539.
  • Chambers, J.C., McMahon, J.A. & Haefner, J.H. (1991) Seed entrapment in alpine ecosystems: effects of soil particle size and diaspore morphology. Ecology, 72, 16681677.
  • Clements, D.R., Benoit, D.L., Murphy, S.D. & Swanton, C.J. (1996) Tillage effects on weed seed return and seedbank composition. Weed Science, 44, 314322.
  • Cousens, R. & Moss, R. (1990) A model of the effects of cultivation on the vertical distribution of weed seeds within the soil. Weed Research, 30, 6170.
  • Firbank, L.G. & Watkinson, A.R. (1986) Modelling the population dynamics of an arable weed and its effects upon crop yield. Journal of Applied Ecology, 23, 147159.
  • Forcella, F., Peterson, D.H. & Barbour, J.C. (1996) Timing and measurement of weed seed shed in corn (Zea mays). Weed Technology, 10, 535543.
  • Genstat, 5 Committee (1993) Genstat Release 3 Reference Manual. Clarendon Press, Oxford, United Kingdom.
  • Grundy, A.C. & Mead, A. (1998) Modelling the effects of seed depth on weed seedling emergence. Aspects of Applied Biology, 51, 7582.
  • Hanf, M. (1982) Ackerunkräuter Europas Mit Ihren Keimlingen und Samen. BASF Aktiengesellschaft, Ludwigshafen.
  • Hulme, P.E. (1994) Post-dispersal seed predation in grassland: its magnitude and sources of variation. Journal of Ecology, 82, 645652.
  • Johnson, C.K. & West, N.E. (1988) Laboratory comparisons of five seed-trap designs for dry, windy environments. Canadian Journal of Botany, 66, 346348.
  • Jones, N.E. & Naylor, R.E.L. (1992) Significance of seed rain from set-aside. BCPC Monograph, 50, 9196.
  • Jørgensen, H.B. & Toft, S. (1997) Food preference, diet dependent fecundity and larval development in Harpalus rufipes (Coleoptera: Carabidae). Pedobiologia, 41, 307315.
  • Kjellsson, G. (1985) Seed fate in a population of Carex pilulifera L. II. Seed predation and its consequences for dispersal and seed bank. Oecologia, 67, 424429.
  • Landis, D.A. & Marino, P.C. (1999) Landscape structure and extra-field processes: impact on management of pests and beneficials. Handbook of Pest Manangement (ed. J.Ruberson), pp. 79104. Marcel Dekker, New York.
  • Leguizamón, E.S. & Roberts, H.A. (1982) Seed production by an arable weed community. Weed Research, 22, 3539.
  • McCullagh, P. & Nelder, J.A. (1989) Generalized Linear Models. Chapman & Hall, London.
  • Medd, R.W. & Ridings, H.I. (1989) Relevance of seed kill for the control of annual grass weeds in crops. Proceedings of the VII International Symposium on the Biology Control of Weeds, 6–11 March 1988, Rome, Italy, pp. 645650.
  • Menalled, F.D., Marino, P.C., Renner, K.A. & Landis, D.A. (2000) Post-dispersal weed seed predation in Michigan crop fields as a function of agricultural landscape structure. Agriculture, Ecosystems and Environment, 77, 193202.
  • Mertens, S.K. (1998) Weed communities on six ecological farms in Flevoland. MSc Thesis. Wageningen University, the Netherlands.
  • Mitze, H. (1992) Bilanzierung wichtiger populationsdynamischer Parameter augewählter Unkrautarten im Winterweizen. MSc Thesis. University of Göttingen, Germany.
  • Moorcroft, D., Whittingham, M.J., Bradbury, R.B. & Wilson, J.D. (2002) The selection of stubble fields by wintering granivorous birds reflects vegetation cover and food abundance. Journal of Applied Ecology, 39, 535547.
  • Ouin, A., Paillat, G., Butet, A. & Burel, F. (2000) Spatial dynamics of wood mouse (Apodemus sylvaticaus) in an agricultural landscape under intensive use in the Mont Saint Michel Bay (France). Agriculture, Ecosystems and Environment, 78, 159165.
  • Peart, M.H. (1979) Experiments on the biological significance of the morphology of seed-dispersal units in grasses. Journal of Ecology, 67, 843863.
  • Pons, T.L. (2000) Seed responses to light. Seeds: the Ecology of Regeneration in Plant Communities (ed. M.Fenner), 2nd edn, pp. 237260. CAB International, Wallingford, UK.
  • Povey, F.D., Smith, H. & Watt, T.A. (1993) Predation of annual grass weed seeds in arable field margins. Annals of Applied Biology, 122, 323328.
  • Rauber, R. & Koch, W. (1975) Zur Populationsdynamik des Flughafers (Avena fatua L.) unter dem Aspekt der langfristigen Befallsprognose. Proceedings of the European Weed Research Society Symposium ‘Status, Biology and Contrological of Grassweeds in Europe’, 1, 113123.
  • Reichman, O.J. (1979) Desert granivore foraging and its impact on seed densities and distributions. Ecology, 60, 10851092.
  • Roberts, H.A. (1964) Emergence and longevity in cultivated soil of seeds of some annual weeds. Weed Research, 4, 296307.
  • Roberts, H.A. & Feast, P.M. (1972) Fate of seeds of some annual weeds in different depths of cultivated and undisturbed soil. Weed Research, 12, 316324.
  • Schotveld, E. & Kloen, H. (1996) Onkruidbeheersing in een multifunctionele vruchtwisseling. Rapport 74. Instituut voor Agrobiologisch en Bodemvruchtbaarheidsonderzoek (DLO). (In Dutch).
  • Seguer Millàs, J. (2002) Influence of weather conditions and seed features on the burial rate of weed seeds on the soil surface. MSc Thesis. Wageningen University, the Netherlands.
  • Thompson, K. (1987) Seeds and seed banks. New Phytologist, 105, 2334.
  • Tooley, J.A., Froud-Williams, R.J., Boatsman, N.D. & Holland, J.M. (1999) Weed seed predation in arable field margins by carabid beetles (Carabidae: Coleoptera). Aspects of Applied Biology, 54, 211216.
  • Vleeshouwers, L.M. (1997) Modelling weed emergence patterns. PhD Thesis. Wageningen University, the Netherlands.
  • Westerman, P.R. (2001) Weed seed predation by the woodmouse, Apodemus sylvaticus.
  • Westerman, P.R., Hofman, A., Vet, L.E.M. & Van Der Werf, W. (2003) Relative importance of vertebrates and invertebrates in epigaeic weed seed predation in organic cereal fields. Agriculture, Ecosystems and Environment, 95, 417425.
  • Wilson, B.J. (1981) A review of the population dynamics of Avena fatua L. in cereals with special reference to work at the Weed Research Organization. Grass Weeds in Cereals in the United Kingdom Conference, 514.
  • Wilson, B.J., Wright, K.J. & Butler, R.C. (1993) The effect of different frequencies of harrowing in the autumn or spring on winter wheat, and the control of Stellaria media (L.) vill., Galium aparine L. & Brassica napus L. Weed Research, 33, 501506.
  • Zhang, J., Drummond, A., Liebman, M. & Hartke, A. (1997) Phenology and dispersal of Harpalus rufipes DeGeer (Coleoptera: Carabidae) in agroecosystems in Maine. Journal of Agricultural Entomology, 14, 171186.