• near-neighbour effects;
  • plant establishment;
  • population dynamics;
  • pre-emergence mortality;
  • weed control


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Weeds play an important role in arable and horticultural habitats, and models are being developed to improve our understanding of their population dynamics. The position of a weed seed in the soil profile influences the probability of a seed germinating, emerging successfully and its relative time of emergence. Identifying a relationship between the shape or weight of a seed and its ability to emerge from depth may allow the development of generic models. The aim of this study was to quantify seed response to burial depth, to improve the wider application of existing seedling emergence models.
  • 2
    A field experiment used weed seeds sown at different depths and densities in a standard substrate. In addition, two laboratory studies used pre-germinated seeds of the same range of species, buried at a range of depths in optimum conditions using the same substrate. These studies explored the effects of seed size, shape and sowing density on seedling emergence and also enabled reserve-dependent pre-emergence mortality to be quantified.
  • 3
    The largest and heaviest of the seeds tested overall, Veronica hederifolia, emerged from the greatest depth (8 cm). In contrast, Tripleurospermum inodorum and Veronica arvensis, the two smallest species, showed a sharp decline when burial exceeded 1 cm. However, the link between seed shape or weight and the ability to emerge from depth suggests a complex relationship. Given optimum conditions, some species (Stellaria media and Chenopodium album) have the physical reserves to emerge from a wider range of burial depths than normally observed in the field, suggesting an ability to exploit opportunities when they occur.
  • 4
    For some species, emergence was reduced at high seed densities (e.g. Veronica arvensis). These responses may be associated with traits that have evolved to counteract sibling competition.
  • 5
    Synthesis and applications. Generic models identifying the maximum depths for seed germination and emergence have a number of practical applications. For example, they can be used to target cultivation to deplete the weed seed bank or to prescribe the optimum depth of mulches to favour certain species. Our model showed that, in general, larger-seeded species emerged from deeper in the soil, but the relationship between seed size and shape and emergence was complex, possibly species specific. Our germination data may also assist our understanding of the relative importance of different causes of seed losses, particularly from different zones of the soil profile, such as the soil surface. Understanding the component processes of seed behaviour and germination is essential to developing sustainable weed management practices in agriculture and horticulture, and the work reported here contributes to a larger programme modelling weed seed bank population dynamics.


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

With the decline in chemical weed control options in horticulture, the development of efficient, sustainable and economical alternatives becomes increasingly important. Optimizing traditional cultivations to modify the weed population is one such option. Carefully chosen cultivations, tailored to the weed species composition in the reservoir of seeds in the soil, can be used either to encourage emergence and hence the premature depletion of this seed bank, or to bury them to depths from which they cannot germinate and successfully emerge (Froud-Williams, Chancellor & Drennan 1983). Species-specific emergence responses of weed seeds to burial depth have already been well-documented (Mohler 1993). More recently, studies using matrix models to simulate the movement of weed seeds following the use of different cultivation implements have linked these emergence responses to burial depth (Grundy, Mead & Burston 1999; Colbach et al. 2000). However, the development of reliable weed control strategies via precision cultivation, stale seed beds and the use of mulches to manage the weed seed bank ultimately depends on an understanding of how weeds respond to burial depth.

The original depth-response modelling study (Grundy, Mead & Bond 1996) raised three major questions. First, the original emergence models showed evidence of a systematic lack-of-fit that may have been due to clustering of seeds. These models were empirically developed using equal numbers of seeds of each of the six weed species sown either in narrow layers or evenly distributed over broad bands within the soil. For some species, a systematic lack-of-fit was observed when the models originally derived from densely sown seeds were used to predict emergence from the same number of seeds in less dense distributions. It is hypothesized in this study that clustering of seeds at high densities may have had an inhibitory effect on the germination of near neighbours, thereby modifying the depth-of-emergence response models.

Secondly, the scope for using seed weight and shape as a parameter for generalizing the response models needs to be investigated to extend their application to a wider range of species. It is generally accepted that larger-seeded species are able to emerge from greater depths than small-seeded species (Chancellor 1964). While this relationship has been recognized, few studies have attempted to incorporate this parameter in predictive weed species emergence models. An exception is Vleeshouwers (1997) where, along with soil penetration resistance, seed weight was included in a model of pre-emergence growth of weed seedlings from different burial depths.

Thirdly, to understand the influence of burial depth on emergence, there is a need to quantify pre-emergence mortality. The previous study was only able to record successful emergence as an indicator of the effect of burial depth (Grundy & Mead 1998). Therefore, no information was available on the proportion of buried seeds that germinated but failed to emerge, so-called ‘fatal germination’. A major cause of seed bank depletion is premature germination at depths from which the seeds cannot emerge successfully (Schafer & Chilcote 1970; Fenner 1985). Quantifying the magnitude of seed bank depletion during this phase would increase the accuracy of the depth-of-emergence models.

The aim of this study was twofold. First, to answer the three questions raised by the earlier study. Second, to improve existing models on seedling emergence and explore some of the potential candidate parameters (seed weight and shape) that may provide a basis on which to develop generic models. The study used a combination of both field and laboratory for the development of robust models.

Materials and methods

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

depth–density interaction trial

During the autumn of 1996, weed seeds were collected from around experimental fields at Horticulture Research International, Wellesbourne, Warwickshire, UK. For consistency, six species were chosen that had been included previously in the development of the original depth-of-emergence modelling studies (Grundy, Mead & Bond 1996; Grundy & Mead 1998). These species were Polygonum aviculare L., Chenopodium album L., Tripleurospermum inodorum (L.) Schultz-Bip, Veronica persica Poiret, Stellaria media (L.) Vill. and Veronica arvensis L. The species were also chosen for their horticultural significance and because they are representative of the range of seed weights typically found for broad-leaved horticultural weed species (Table 1). Plastic cylinders (length 300 mm, diameter 230 mm) were buried leaving 80 mm visible above the surrounding soil. Nylon mesh (Whaleys (Bradford) Ltd, UK) was placed at the base of each cylinder to prevent soil fauna from entering the cylinders and disrupting the seed positions or predating the seeds. Each cylinder was filled with sieved sterilized sandy silt loam (pH range 6–7, initial moisture content 3–5%, bulk density ranged between 1·0 and 1·5 g l−1, particles retained < 0·002 mm and screened to 8 mm) so that, after a period of natural settling, the soil surface within the cylinder was level with that of the surrounding soil. Ungerminated seeds were sown at each of six different depths including the surface (0, 1, 2, 4, 8, 16 cm). To allow the possible density–depth interaction to be assessed, a range of four densities were used, equivalent to 2407, 9628, 38 510 and 154 040 seeds m−2. This range of densities was comparable with those reported in surveys of weed seed densities (Roberts & Chancellor 1986). The different seed densities were achieved by scattering 100 seeds of a single species evenly within an area indicated by a circular template. The diameters of the templates were 28·75 mm, 57·5 mm, 115 mm and 230 mm, to give the equivalent density m−2.

Table 1.  List of species observed in each experiment (both field and incubator) with estimated mean weights of 100 seeds (with SE in parentheses) and method of pre-germination
SpeciesStudy100 Seed weight§ (mg)Pre-germination protocol
  • *

    Depth–density interaction field trial.

  • †‡

    First and second pre-emergence mortality experiments, respectively, carried out in the laboratory (incubator).

  • §

    Mean 100 seed weights all based on 10 observations each of 100 seeds, except V. persica, based on six observations.

  • All seeds were germinated in sealed Petri dishes using two Whatman Grade 1 filter papers (Whatman International Ltd, Maidstone, UK) with 2·5 ml of distilled water. All germination tests were made in the light. Protocols developed from personal observations.

V. arvensis++ 10·4 (0·08)8 h at 22 °C, 16 h at 12 °C for 3 days, followed by 8 h at 22 °C, 16 h at 7 °C for 3 days
T. inodorum+++ 35·6 (0·46)8 h at 22 °C, 16 h at 7 °C for 7 days
S. media+++ 39·7 (0·26)8 h at 22 °C, 16 h at 7 °C for 7 days
C. album+++ 64·5 (0·33)8 h at 22 °C, 16 h at 7 °C for 7 days
V. persica++ 75·9 (0·89)8 h at 22 °C, 16 h at 7 °C for 8 days
P. aviculare++130·1 (4·59)Pre-chill at 1 °C for 12 days then 8 h at 22 °C, 16 h at 7 °C for 7 days
V. hederifolia++626·1 (10·78)8 h at 22 °C, 16 h at 7 °C for 8 days

Five replicates of each combination of species, depth and density were included in the trial, with each complete replicate occupying 24 plastic cylinders. Each cylinder contained one set of seeds of each species, each species being buried at a different depth. The allocation of species to depths within each replicate followed a 6 × 6 Latin square, with four cylinders in each replicate having the same allocation. The allocation of densities to depths was similarly achieved following an extended 4 × 4 Latin square (with two rows repeated), with six cylinders in each replicate having the same allocation. The final design was achieved by crossing these two designs to obtain each combination of species, depth and density once within each replicate set of 24 cylinders. By ensuring that each species appeared only once in each pot it was simple to determine the depth and density from which seedlings had emerged. When sowing the seeds, care was taken to avoid sowing seeds at high densities directly above each other, reducing the probability of emerging seedlings from lower depths disturbing the upper layer of seeds.

The soil columns in each experiment were left undisturbed for the duration of the recording period. The effects of rain, frost and wind on the weed seed distributions were considered negligible. Weekly weed emergence counts were made over a 2-year period with weed seedlings being removed by cutting at ground level once they had been recorded. The seeds were sown on the 2 December 1996 and the weekly recordings terminated on the 7 October 1998.

statistical methods: the overall effect of seed density

Emergence counts from the depth–density interaction trial were cumulated over each of the 2 years separately and over the whole 2-year experimental period. An initial assessment of treatment effects was obtained using analysis of variance of the cumulative number of emerged seedlings over each period as a percentage of the initial number (100). In all cases, percentages were arcsine transformed prior to analysis to satisfy the assumption of homogeneity of variance. Of particular interest were assessments of the overall differences between species, between densities and the interaction between species and density, but the analysis also allowed the assessment of any three-way interaction between species, density and depth.

statistical methods: developing a depth-response model

For consistency with previous studies, the effect of burial depth on emergence was assessed initially for each combination of species and density separately, using the emergence counts during the first complete experimental year only. In a previous paper (Grundy & Mead 1998) three possible models were proposed for the proportion of emerged seedlings or probability of emergence. All were based on the probit model, originally developed by Finney (1971) for use in insecticide bioassays, assuming a binomial distribution and probit link function and with log-depth as the explanatory variable. The equations describing the three models are:

  • ‘Linear’ probit p= 1 −Φ(a+b× ln(depth)) =Φ(A+B× ln(depth))(eqn 1)
  • ‘Control mortality’ probit p=θ× (1 −Φ(a+b× ln(depth))) =θ× (Φ(A+B× ln(depth))()eqn 2)
  • ‘Quadratic’ probit p= 1 −Φ(a+b× ln(depth) +c× ln(depth)2) =Φ(A+B× ln(depth) +C× ln(depth)2)(eqn 3)

where p is the proportion emerged, θ is the fitted parameter representing the maximum possible proportion of emerged seedlings, Φ is the cumulative normal distribution function, and a, b and c are the fitted parameters for the constant, linear and quadratic terms, respectively. The equations can be re-expressed in the second form using a symmetry argument and setting A, B and C equal to –a, –b and –c, respectively. All models were fitted using the generalized linear model facilities in GenStat for Windows (Lane & Payne 2000); the control mortality model used the probitanalysis procedure (Payne 2000). The best-fitting model for each species was identified, and the lack-of-fit mean deviance for this best-fitting model compared with the estimate of ‘pure error’ mean deviance obtained from the replicate observations on each treatment combination.

A more general approach to modelling the emergence response to burial depth requires the identification of a single model form that can be used for all species. The quadratic probit model (equation 3) is sufficiently flexible to describe the range of shapes of response seen, and was therefore used for this purpose. The additional lack-of-fit due to choosing this model over the best-fitting model for each species was assessed.

statistical methods: the effect of density on the depth-response model

The influence of density on the depth response for each species was explored using two separate approaches. The first, simpler, approach assessed whether different models were required for each density by allowing the parameters (A, B, C) of the quadratic probit model (equation 3) to vary with density, using a ‘parallel curves’ analysis. This approach is possible using the generalized linear model facilities in GenStat for Windows (Lane & Payne 2000). At one extreme, all three parameters were constrained to be the same for all four densities (fitting a single curve), while at the other all three parameters were estimated independently for each density (fitting separate curves, as described earlier). Considering the change in deviance between fitting the single and separate curves was then used to assess the need for separate models for each density. Where a single curve appeared to be inadequate, further exploration of which parameters differed between densities was possible within the same parallel curves analysis framework. The single curves fitted here to each species (data pooled across all densities) were then used to assess the relationship between seed size and the emergence response to burial depth.

A second, more biologically realistic, approach incorporated an additional scaling parameter, Di, into the model to allow for the effect of each density, where the subscript i denotes the density:

  • pi=Di×Φ(A+B× ln(depth) +C× ln(depth)2) (eqn 4)

This parameter appears outside the cumulative normal distribution function and acts like the parameter θ in the control mortality probit model (equation 2), scaling the whole response. This modification allows for density to affect the maximum proportion emergence, but neither the depth at which this maximum emergence occurs nor the maximum depth at which emergence occurs is affected. An initial fit for this model was obtained by constraining the non-linear parameters (A, B, C) to the values fitted for the single curve model described above. The fit for this model was then optimized using the generalized non-linear model facilities (fitnonlinear directive) in GenStat for Windows (Lane & Payne 2000), although an iterative two-step approach was required to make any progress in the optimization. In this, the non-linear parameters (A, B, C) were (re-)estimated in the first step whilst keeping the linear parameters (Di) constant, with the linear parameters re-estimated in the second step whilst keeping the non-linear parameters constant. This process was continued until the change in deviance between steps was small.

For both approaches, the relationship between fitted parameter values for the different densities and actual density was assessed using linear regression.

the effect of depth on pre-emergence mortality

A companion study was made of depth of seed burial and the pre-emergence growth phase. The six species used in the field depth–density study were also used in this laboratory study. In addition a large-seeded species, Veronica hederifolia L., was included to provide further information on how responses varied with seed weight. A first experiment included only five species because the development stage of the other two species (V. persica and P. aviculare) was too advanced to allow fair comparison and to avoid the potential risk of damage to the radicle during burial (Table 1). In a second experiment these two species were included; however, this time V. arvensis was too advanced to be included for the same reasons described above (Table 1). Plastic open-ended cylinders (length 9·8 cm, diameter 10·4 cm) were each placed on a tray to prevent soil escaping then filled to the appropriate level with sieved sterile sandy silt loam (as used in the field study). In the first of the two laboratory experiments water was added to achieve a standard starting moisture content of 7·5%, whereas in the second the added water gave a much higher soil moisture of 27·1%. Ten pre-germinated seeds (with visible radicle protrusion) of each of the species were sown in each cylinder at depths of 1, 2, 3, 4, 5, 6, 7 and 8 cm below the surface. An additional treatment included scattering the pre-germinated seeds directly on the upper surface of the loam. Optimum temperature conditions to promote a good germination percentage and to synchronize the timing of germination of each of the seven species were obtained from a previous germination study (Table 1). Seeds were discarded if the radicle protrusion was greater than 2 mm. After sowing the pre-germinated seeds, an appropriate amount of sterile sandy silt loam was added to each cylinder to bring the final level to 1 cm below the upper rim of the cylinder. Cling-film was placed over the upper exposed end of each cylinder to reduce evaporation, although initial cylinder weights were checked daily and any loss in weight was replaced by gentle addition of water to maintain soil moisture. Four replicates of the nine depths of burial treatments were placed in an incubator at 15 °C. Each replicate occupied a separate shelf and the treatments were fully randomized on each shelf to account for any spatial variability in the incubator temperature. Emerged seedlings were counted and removed (by cutting at soil level) on a daily basis for a period of 3 weeks.

statistical methods

The numbers of emerged seedlings of each species from each cylinder were summed across the period of the experiment. These cumulative emergence counts were subjected to probit analysis (using the probitanalysis procedure; Payne 2000) with log-depth as the explanatory variable. Where species were assessed in both experiments (Table 1), curves were fitted for each experiment separately and for the combined data sets. In all cases the lack-of-fit of the fitted curve was compared with the between-replicate variability (pure error).

relationship between seed weight, seed shape and depth parameters

Mean seed weights were calculated from 10 replicate samples each of 100 seeds (Table 1), and mean seed dimensions (measurements on three orthogonal axes, length, width, breadth) for the species in the study were obtained from Holm-Nielsen (1998). An index of seed shape was calculated as the variance of these three scaled dimensions (Thompson, Band & Hodgson 1993). Parametric correlation coefficients were calculated to assess the strength of the relationships between parameters of the fitted emergence and pre-emergence mortality models and these measures of seed size.


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

does seed density affect emergence?

There was a significant effect of seed density on total weed emergence over the duration of the field experiment, when averaged across species and depth. Similar effects were observed when individual years were taken in isolation. For all analyses, increased seed density resulted in a small but significant reduction in emergence across both years of the experiment (P < 0·001; Table 2). This pattern was most noticeable in P. aviculare, while S. media and, in particular, V. arvensis had increased emerg-ence at the lowest seed density only, with no clear pattern detectable between the remaining three higher seed densities. Tripleurospermum inodorum, V. persica and C. album did not show a strong response to density.

Table 2.  Total percentage emergence over 2 years for six species at four seed densities. Percentages shown are back-transformed means, with arcsine-transformed means in parentheses, and responses have been averaged across sowing depth
SpeciesSeed density m−2
2407962838 510154 040Mean
  1. SED are based on 572 d.f. and are on the arcsine-transformed scale. SED for comparing species means = 0·781. SED for comparing seed density means = 0·638. SED for comparing means for combinations of species and seed density = 1·563.

V. arvensis12·4(20·62)8·5(16·91)9·4(17·83)7·8(16·19)9·4(17·89)
T. inodorum 8·3(16·73)6·4(14·61)7·2(15·57)6·5(14·78)7·1(15·42)
S. media 5·5(13·56)3·4(10·67)3·7(11·03)3·3(10·39)3·9(11·41)
C. album 7·4(15·75)6·8(15·10)5·6(13·73)6·7(14·97)6·6(14·89)
V. persica 6·8(15·16)6·7(14·99)5·8(13·91)5·9(14·08)6·3(14·53)
P. aviculare 6·6(14·93)5·4(13·41)4·7(12·53)4·2(11·86)5·2(13·18)
Mean 7·7(16·12)6·1(14·28)5·9(14·10)5·6(13·71)  

can a single model form be used to describe the response for all species?

As in the previous study (Grundy & Mead 1998), different models were found to be most appropriate for the responses of the different species in the field experiment. For T. inodorum the linear probit model (equation 1) gave the best fit, while for V. arvensis and V. persica the control mortality model (equation 2) was the most appropriate. For all three of these species maximal emergence was observed at the soil surface (Fig. 1). The remaining three species (S. media, C. album and P. aviculare) showed evidence of inhibition at the soil surface and were best described by the quadratic probit model (equation 3). The flexibility afforded by the quadratic model described the responses for the other species without significantly increasing the lack-of-fit (data not shown). The percentage emergence from the top 2 cm from both the best fitting and the quadratic probit models (assuming an even distribution of seeds) showed little difference. The proportion of emergence predicted from the quadratic model was always greater (with the exception of T. inodorum at the highest density) but the difference in proportions never exceeded 0·02.


Figure 1. Quadratic probit curves for proportion emergence [fitted against loge(depth)] plotted against depth, for each combination of species and seed density separately. Species: (a) V. arvensis, (b) T. inodorum, (c) S. media, (d) C. album, (e) V. persica and (f) P. aviculare. Seed densities: solid circles, solid line, 2407; open circles, dotted line, 9628; solid squares, dashed line, 38 510; open squares, dashed-dotted line, 154 040 seeds m−2.

Download figure to PowerPoint

is the response to burial depth affected by density?

The mean predicted emergence from the top 2 cm in the field experiment showed differences between the proportions obtained from the model allowing parameters to vary with density (Table 3, separate curves), compared with the proportion obtained from the single model fitted to all densities (Table 3, single curve; Fig. 2). In particular, for V. arvensis, the predicted proportions ranged from 0·40 to 0·27 for the low and high densities, respectively, with the combined model giving a proportion of 0·32. For other species there was somewhat less variability, for example C. album (predicted proportions ranged from 0·31 to 0·28). The comparison of these models, based on the reduction in residual deviance, showed no benefit from using the more complex, separate curves model for any of the species (Table 3).

Table 3.  Comparison of (a) single quadratic probit model with more complex models allowing parameters to vary with seed density, including (b) a comparison of mean predicted proportion emergence from an even distribution of seeds from the top 2-cm soil layer for single and separate curves
 d.f.V. arvensisT. inodorumS. mediaC. albumV. persicaP. aviculare
(a) Summary of model fits
Between-replicate mean deviance96 7·98 5·232·073·232·52 3·92
Lack-of-fit mean deviances
 Single curve2117·22 6·6711·317·803·0713·29
 Separate curves1224·36 9·8312·5911·704·4921·21
 Scaling model (A, B, C fixed)1717·89 7·6812·359·253·3615·79
 Scaling model (optimized)1711·70 3·528·335·023·3311·42
Mean deviance for improvement from single curve
 Separate curves 9 7·70 2·459·592·601·18 2·71
 Scaling model (A, B, C fixed) 414·40 2·386·881·621·83 2·63
 Scaling model (optimized) 440·7220·0623·9519·611·9621·20
Significance of approximate F-test for improvement against between-replicate mean deviance
 Separate curves 9 0·473 0·892< 0·0010·6110·893 0·715
 Scaling model (A, B, C fixed) 4 0·134 0·7680·0140·7350·575 0·614
 Scaling model (optimized) 4 0·001 0·006< 0·001< 0·0010·543 0·001
(b) Predicted proportion emergence from top 2-cm soil layer
Individual density models (separate curves)
 2407 seeds m−2  0·403 0·2100·2390·3040·194 0·247
 9628 seeds m−2  0·288 0·1890·2010·3090·203 0·247
 38 510 seeds m−2  0·315 0·1900·2670·2770·183 0·231
 154 040 seeds m−2  0·272 0·1560·1660·2880·177 0·198
Single curve for all densities  0·318 0·1850·2150·2940·189 0·231

Figure 2. Single quadratic probit curves for proportion emergence [fitted against loge(depth)] plotted against depth for each species separately, for all four densities together. Species: (a) V. arvensis, (b) T. inodorum, (c) S. media, (d) C. album, (e) V. persica and (f) P. aviculare. Seed densities: solid circles, 2407; open circles 9628; solid squares, 38 510; open squares, 154 040 seeds m−2.

Download figure to PowerPoint

Adopting the more biologically realistic model (equation 4) initially showed that the inclusion of a simple scaling parameter also gave no improvement relative to the single curve [Table 3, scaling model (A, B, C fixed)]. However, with the exception of V. persica, by minimizing the residual deviation using the optimization process described, a significant improvement was achieved over the single curve [Table 3, scaling model (optimized)]. Although there was no strong relationship between the fitted values of the scaling parameter, Di, and seed density, there was a tendency for larger values of the parameter to occur at low densities. While the weakness of this relationship does not justify or encourage the use of this more complex model to predict proportion emergence, it can provide some quantification of the variability caused by changes in seed density in the field.

effect of depth on pre-emergence mortality

For all species an increase in pre-emergence mortality with increasing depth was observed in the two laboratory experiments. However, the fitted probit curves for each illustrate that the rate of increase varied between species (Fig. 3 and Table 4). For example T. inodorum showed a rapid increase in mortality, with no seedlings able to emerge from a depth greater than 1 cm. In contrast, V. hederifolia showed a very gradual increase, with more than 50% of pre-germinated seeds still emerging from a depth of 8 cm.


Figure 3. Probit curves for proportion pre-emergence mortality [fitted against loge(depth)] plotted against depth for each species. Species: (a) V. arvensis, (b) T. inodorum, (c) S. media, (d) C. album, (e) V. persica, (f) P. aviculare and (g) V. hederifolia. Experiments: solid circles, first pre-emergence mortality experiment; open circles, second pre-emergence mortality experiment.

Download figure to PowerPoint

Table 4.  Summary of emergence response using pre-germinated seeds
SpeciesBetween-replicate mean deviances (27 d.f.)Lack-of-fit (7 d.f.)Intercept (SE)Slope (SE)
  • *

    These species are common to both laboratory experiments, therefore combined data are presented. For these species the between-replicate mean deviance and lack-of-fit data are based on 54 d.f. and 16 d.f., respectively.

V. arvensis0·827 0·571 0·02(0·193)1·87(0·304)
T. inodorum*0·347 1·191 0·00(0·782)5·75(0·550)
S. media*2·610 5·714−2·25(0·342)1·72(0·224)
C. album*1·29610·772−3·61(0·531)2·22(0·318)
V. persica0·835 2·647−0·62(0·205)1·37(0·173)
P. aviculare0·843 1·407−1·86(0·243)2·07(0·195)
V. hederifolia*1·641 5·241−1·98(0·300)0·70(0·182)

Where species were included in both laboratory experiments there were slight differences in the fitted slopes of the mortality responses. These could be attributed to differences in the soil moisture content for experiment 1 (7·5%) and experiment 2 (27·1%). The bulk density of the soil in experiment 1 (1·25 g cm−3) was slightly higher than in experiment 2 (1·02 g cm−3). Generally, mortality was lower in the first experiment than in the second. In addition, the rate of emergence varied between species (data not shown) with, for example, C. album being consistently faster to emerge from all depths.

relationship between seed weight, seed shape and depth parameters

In the field and laboratory experiments, seed weight ranged from 10·38 mg 100 seeds−1 to 626·08 mg 100 seeds−1 for V. arvensis and V. hederifolia, respectively. Species such as C. album and V. hederifolia have more spherical seeds, indicated by the relatively small seed shape index, while, for example, T. inodorum has a more cylindrical shape, indicated by the relatively high seed shape index (Table 5). Correlations of depth parameters with seed weight were generally poor, with the notable exception of those with the probit slope for the laboratory experiments and the depth below which pre-emergence mortality exceeded 50% (Table 5). However, the removal of substantially the heaviest species, V. hederifolia, from the calculation significantly reduced this latter correlation to 0·24. In contrast, correlations with the seed shape index were generally higher for all parameters, with strong correlations with the fitted quadratic probit parameters and predicted optimal depth of emergence. There was a lack-of-fit for some species (V. arvensis and P. aviculare; Fig. 2) due to the use of the same quadratic probit model to describe the effect of burial depth on emergence across all species. This lack-of-fit in turn influenced the correlations with seed weight and shape (Table 5).

Table 5.  Summary of average seed weight and shape and their correlation with fitted parameters from both the field and laboratory experiments
V. arvensisT. inodorumS. mediaC. albumV. persicaP. aviculareV. hederifoliaWith seed weightWith seed shape index*
Hundred seed weight (mg)10·3835·5639·6764·4675·92130·1626·1  
 Length (mm)1·01·81·11·31·72·5  2·5  
 Width (mm)0·70·71·11·31·31·4  2·1  
 Depth (mm)0·30·70·60·80·81·4  1·5  
 Variance (Thompson, Band & Hodgson 1993)0·1230·1250·0690·0490·0700·065  0·041−0·576 
Fitted parameters (field experiment): single quadratic probit
Calculated properties (field experiment): single quadratic probit
 Optimal depth (mm)2·20·05·15·42·17·10·640−0·783
 Depth below which emergence < 5%40·121·229·837·634·144·90·505−0·410
Fitted parameters (laboratory experiments)
 Probit slope1·875·751·722·221·372·61  0·700·986−0·645
Calculated properties (laboratory experiments)         
 Depth below which pre-emergence mortality > 50% (mm)8·68·835·749·814·523·3168·60·960−0·654


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

modelling the burial depth response

The quadratic model appeared to provide the best fit overall for emergence response to burial depth in the field experiment. This agrees with previous work where different seed-lots were used (Grundy & Mead 1998). While seed-lot differences may modify the depth response, the general shape of the quadratic curve remained valid. There was less agreement with previous work in the soil surface layers. Possible explanations are that the top 1 cm of the soil profile is vulnerable to rapid changes in moisture and temperature (Fenner 1985) and is also more exposed to predation pressure (Crawley 2000). It is also possible that seeds near the surface were exposed to low temperatures shortly after sowing that may have contributed to the variability of the observed response at shallow depths. These variables are difficult to predict and incorporate into simple empirical models. Clearly, the surface layer, which represents the interface between the seed bank and the environment, represents a crucial and variable zone in population dynamics. Indeed the inconsistency between experiments in the emergence from the surface layer may justify the need for future studies that model this part of the response independently.

Although the generic quadratic model provided an adequate description for all species studied, other forms of model may provide better fits to individual species. For example, alternative models proposed in this paper provided a better fit for species that had maximal emergence at the surface. When developing models it is important to identify their end purpose as different models have different strengths. While species-specific models may better describe biology and ecology at the species level, generic models may provide a better basis for broad practical application in weed management.


An influence of seed density on germination response was observed in the field experiment, but this did not affect the fitted curves markedly. The use of an optimized scaling parameter to account for changes in density improved the fits while retaining the overall shape of the curve, but the degree of improvement did not justify the added complexity of the model. While the density effect on germination and emergence was not large enough to warrant inclusion within the final model, it was significant for some species. There can be several disadvantages to the sibling competition that may occur when large numbers of seeds from the same species are scattered within a relatively small area. For example, the progeny would be likely to have identical requirements for resources and, at maturity, inbreeding could be encouraged (Willson & Traveset 2000). Avoidance tactics may have evolved, such as small seed size to aid spatial dispersal, dormancy to spread germination of siblings over time (Venable 1989) and the direct inhibition of germination of seeds of the same species. Manipulating the dormancy status of neighbouring seeds could be one such inhibition mechanism. Such chemical inhibition has been termed autotoxicity (Hilhorst & Karssen 2000); however, there is little evidence and little knowledge of how it may operate. One of the few examples of this phenomenon is the weed Parthenium hysterophorus, in which the seeds contain organic inhibitors in sufficient quantities to inhibit germination when seeds are present in high densities (Picman & Picman 1984). In the present study, V. arvensis is an example of a small compact adult plant producing up to 17 000 seeds plant−1 under optimal conditions, with many remaining close to the parent plant (Grime, Hodgson & Hunt 1988). This could lead to situations where high densities of seeds are scattered at the foot of the parent plant. A mechanism, such as autotoxicity, that reduced successful germination and emergence at unfavourably high seed densities may give a significant advantage to species where there is intraspecific competition for limited resources.

As intraspecific density effects were observed in the study, it is possible that there were also interspecific effects of density between different species in the same pot. However, the randomization of combinations of species, density and depth within the experiment should have ensured that no pair of treatments occurred together regularly to produce a consistent effect.

reserve-dependent pre-emergence mortality

For most of the species in the two laboratory-based studies, emergence declined rapidly below the top 5 cm. An apparent lack of sufficient pre-emergence reserves accounted for a sharp depletion of pre-germinated seeds below the species-specific optima. This was particularly acute for the smallest seeded species. For example, with T. inodorum and V. arvensis, despite germination having taken place, seedling mortality increased rapidly to more than 50% at a burial depth of less than 1 cm. Pre-emergence mortality in the field experiment may have been caused by a number of factors such as predation or insufficient soil moisture. Hence, no attempt was made in the field study to discriminate between the relative importance of these different factors. In contrast, the two laboratory experiments allowed reserve-dependent mortality to be isolated from other factors. The use of pre-germinated seeds in the two laboratory experiments made it possible to isolate the pre-emergence phase from the germination phase, as all sown seeds were known to have germinated successfully. The subsequent burial of these pre-germinated seeds in the sterile loam (uncompacted with small particles and no large clods to hinder growth), in predator-free, optimum conditions, made it possible to isolate pre-emergence mortality caused specifically by the exhaustion of seed reserves.

In these laboratory experiments, were the seeds being forced to emerge from depths at which germination would not naturally occur? For some of the light-requiring species used in the study (e.g. T. inodorum) it could be argued that germination would not even have been initiated from depths such as 8 cm. Indeed, to do so, given the limited seed reserves, would be a waste of reproductive capacity (Chancellor 1964). At the same time it would be reasonable to assume that, in an agricultural situation, cultivation during seed bed preparation may provide sufficient stimulus prior to burial to initiate the germination process in some seeds. This could lead to germination at depths from which emergence is unlikely to be achieved. Similarly, the depth-sensing mechanism, thought to prevent germination from depths that they cannot physically emerge, has been shown to be less than 100% effective. Although there is little quantitative evidence in the literature for this, 40% of Achyranthese aspersa seeds buried at or below 8 cm were lost due to fatal germination (Fenner 1995).

It is useful to compare the emergence depths achieved from the field experiment with those using the pre-germinated seed in the two laboratory experiments. Even taking into account different soil conditions (the field experiment was open to the natural environment while the laboratory conditions were controlled) and potential population variation between the studies, striking differences were observed. In the laboratory experiments, S. media and C. album were able to emerge from much greater depths than observed in the field. This disparity suggested that, in the field experiment, as well as having greater exposure to mortality factors such as predators, pathogens and the drying and wetting of the seed bed, there may have been mechanisms that prevented germination occurring below a given depth. It also suggests that, under optimal conditions (as experienced in the laboratory experiments), some seeds had a capacity to exploit germination opportunities over a wide depth range. Chenopodium album was observed in this study to be the fastest species to emerge across the range of burial depths (data not shown). A previous study by Vleeshouwers (1997) also noted a high relative emergence rate for this species. The emergence rate of a species could provide a useful quantitative measure of its relative weediness, as early establishment encourages competition with a crop.

For P. aviculare and V. persica, the maximum depths of emergence for both the field and laboratory experiments were in agreement. However, the seeds of V. arvensis and T. inodorum failed to emerge from relatively shallow depths of burial, even in the absence of predators and using pre-germinated seed. Even if germination had been initiated, these seeds did not have the reserves to emerge from any great depth.

seed size and shape

Seed size represents a degree of investment made by the maternal plant. Large-seeded species are generally able to establish over a wider range of conditions (e.g. greater burial depths, as seen for V. hederifolia in this study). However, smaller-seeded species have a greater tendency towards dormancy, enabling them to make the most of their limited reserves and only germinate when they encounter conditions providing the best chance of successful establishment. In the two laboratory experiments, the large-seeded V. hederifolia also tended to have a very slow growth, even from shallow depths, compared with the other species (data not shown). This supports the theory that smaller-seeded species generally have a more rapid relative growth rate to exploit their establishment opportunities (Leishman et al. 2000). This can also be a weakness, as such plants tend to produce less dense tissue with a greater susceptibility to physical damage. This may contribute to the large mortality rates observed in the laboratory experiments for the smallest seeded species (V. arvensis and T. inodorum), even at the shallowest burial depths.

Identifying a single measure of seed size for use as a parameter is difficult, although seed mass (i.e. seed weight excluding any dispersal mechanisms), as used in the present study, has been suggested as the best measure of resource availability (Westonby 1998). Assuming that V. hederifolia is typical of large-seeded species, the results from this study indicate that seed mass could contribute to a prediction of depletion depth (based on depth with > 50% mortality). However, seed mass was not the only measure in this study that was correlated with ability to emerge from depth. The correlation between optimum emergence depth and the seed shape index (Thompson, Band & Hodgson 1993) suggested that the relationships between seed mass, seed shape and burial depth response are complex. Seed shape is related to surface area that provides contact with the external environment, whereas seed mass gives a measure of resource availability. Each is likely to influence the response to burial depth in a different way.

future work

In this study we presented several approaches towards developing a practical generic model of emergence behaviour for a wide range of weed species. The generally good fit of the quadratic probit model for the emergence response to burial depth indicated the possibility of using a single form of model for all species. However, the lack-of-fit shown by some species suggests that a more complex form of model may be necessary to achieve this. There is quantitative evidence in this study that the shape of the depth-of-burial response is related to seed weight or shape and hence these parameters could be used to predict the behaviour of species not included in the present study. Ultimately, further investigation over a wider range of seed weights and shapes is required than was possible within this experimental programme. It is also important to combine the results of this study with the effect that different soil structures may have on modifying the depth–response models. For example, Cussans et al. (1996) demonstrated that finer soil aggregate size reduced total emergence and slowed the time to 50% emergence, while larger seeds appeared to be relatively insensitive to aggregate size in their emergence response to burial depth.

Seedling establishment is clearly a crucial phase in plant population dynamics. Freckleton & Watkinson (2002), in their analysis of whether weed population dynamics are chaotic, describe processes in the life cycle of plant communities where there is evidence of density dependence providing stability. The proportion of seedlings emerging from a seed bank is one such example: emergence generally declines with increased density, probably as a result of reduced light levels (Fenner 1995). In the present study emerged seedlings were recorded and removed regularly, so the density effect was not thought to have been caused by shading. Hence, we suggest that the direct inhibition of germination of seeds of the same species at high sowing densities may be another factor influencing population stability. An effect of seed density was observed in subsequent emergence. However, it was not found to influence the depth of burial model sufficiently to warrant its inclusion.

practical application

For some species (e.g. T. inodorum) the pre-emergence phase accounts for a substantial loss of germinated seeds and strongly influences its emergence response and ecology. Therefore, understanding the factors that influence establishment may help to develop more efficient ways of depleting the arable weed seed bank at this vulnerable stage. Models could be used to identify the maximum depths from which seeds can germinate and physically emerge, providing opportunities for targeting cultivation or prescribing the optimum depth of mulches. Alternatively, the information could be used to help to maximize the establishment success of beneficial species. For example, surface-sown seeds in the field experiment showed poor emergence while they were clearly able to establish better in the laboratory, where drying out of the surface and predation were prevented. For many weed species, seeds left on the soil surface are in a vulnerable zone with a reduced chance of establishment. They may pose a greater problem to land managers when incorporated into the seed bank, where they may persist. The seeds of several species also provide an essential food source for birds and insects (Marshall et al. 2001), hence soil management (type, timing and intensity of cultivation) needs to be carefully balanced against these potential benefits. Combined with studies of seed predation, the approaches used in this study could be used to understand better the relative importance of different sources of seed loss from these surface layers.

The present study forms part of a larger programme concerned with modelling weed population dynamics, including emergence and germination, weed growth and competition, seed production and the incorporation of seeds into the soil. Linking the findings of the present study to weed-seed germination models based on hydrothermal time (Roman et al. 1999; Grundy et al. 2000) may further contribute to our understanding of the temporal spread and magnitude of flushes of weed emergence relative to climatic data. The modelling of other important stages that contribute towards the dynamics of weed populations and how they may be incorporated are reported elsewhere (Grundy, Mead & Burston 1999; Park et al. 2001). Crucially, it is the integration of these component processes (of which the response to seed burial depth described here is just one module) that should provide a powerful framework for simulating weed population dynamics. While the variability associated with weed population dynamics (especially when component models need to be combined) makes prediction of actual numbers on a local scale an unreasonable expectation, models will aid our understanding of how broad-scale changes in agricultural practice impact on weed species (Freckleton & Watkinson 2002). Simulating the probable outcome of intervention at different life-cycle stages would be valuable for evaluating the sustainability of weed control scenarios. Finally, it is important that we do not disregard the importance of the population dynamics of agricultural weeds as ‘a significant part of European biodiversity is associated with this habitat’ (Robinson & Sutherland 2002). Changes in agriculture and horticultural practice since the 1940s have reduced the biodiversity of these areas, mediated, at least in part, by effects on weeds.


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

The authors thank Mrs T. Overs for technical assistance with the experimental work. This study was funded by the Department of Environment, Food and Rural Affairs.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Chancellor, R.J. (1964) The depth of weed seed germination in the field. Proceedings of the 7th British Weed Control Conference, Brighton, pp. 607613. British Weed Control Council, London, UK.
  • Colbach, N., Roger-Estrade, J., Chauvel, B. & Caneill, J. (2000) Modelling vertical and lateral seed bank movements during mouldboard ploughing. European Journal of Agronomy, 13, 111124.
  • Crawley, M.J. (2000) Seed predators and plant population dynamics. Seeds: The Ecology of Regeneration in Plant Communities (ed. M.Fenner), 2nd edn, pp. 167182. CABI Publishing, Wallingford, UK.
  • Cussans, G.W., Raudonius, S., Brain, P. & Cumberworth, S. (1996) Effects of depth of seed burial and soil aggregate size on seedling emergence of Alopecuus myosuroides, Galium aparine, Stellaria media and wheat. Weed Research, 36, 133141.
  • Fenner, M. (1985) Seed Ecology. Chapman & Hall, London, UK.
  • Fenner, M. (1995) The ecology of seedbanks. Seed Development and Germination (eds J.Kigel & G.Galili), pp. 507528. Marcel Dekker, New York, NY.
  • Finney, D.J. (1971) Probit Analysis, 3rd edn. Cambridge University Press, Cambridge, UK.
  • Freckleton, R.P. & Watkinson, A.R. (2002) Are weed population dynamics chaotic? Journal of Applied Ecology, 39, 699707.
  • Froud-Williams, R.J., Chancellor, R.J. & Drennan, D.S.H. (1983) Influence of cultivation regime on buried weed seeds in arable cropping systems. Journal of Applied Ecology, 20, 199208.
  • Grime, J.P., Hodgson, J.G. & Hunt, R. (1988) Comparative Plant Ecology: A Functional Approach to Common British Species. Unwin-Hyman Ltd, London, UK.
  • Grundy, A.C. & Mead, A. (1998) Modelling the effect of seed depth on weed seedling emergence. Aspects of Applied Biology, 51, 7582.
  • Grundy, A.C., Mead, A. & Bond, W. (1996) Modelling the effect of weed-seed distribution in the soil profile on seedling emergence. Weed Research, 36, 375384.
  • Grundy, A.C., Mead, A. & Burston, S. (1999) Modelling the effect of cultivation on seed movement with application to the prediction of weed seedling emergence. Journal of Applied Ecology, 36, 663678.
  • Grundy, A.C., Phelps, K., Reader, R.J. & Burston, S. (2000) Modelling the germination of Stellaria media using the concept of hydrothermal time. New Phytologist, 148, 433444.
  • Hilhorst, H.W.M. & Karssen, C.M. (2000) Effect of chemical environment on seed germination. Seeds: The Ecology of Regeneration in Plant Communities(ed. M.Fenner), 2nd edn, pp. 293310. CABI Publishing, Wallingford, UK.
  • Holm-Nielsen, C. (1998) Frø fra det dyrkede land. Ministeriet for Fødevarere, Landbrug og Frikeri, Danmarks Jordbrugs Forskining, Flakkebjerg, Denmark.
  • Lane, P. W. & Payne, R.W. (2000) Regression analysis. The Guide to Genstat, Part 2: Statistics (ed. R.W.Payne), pp. 79234. VSN international Ltd, Oxford, UK.
  • Leishman, M.R., Wright, I.J., Moles, A.T. & Westonby, M. (2000) The evolutionary ecology of seed size. Seeds: The Ecology of Regeneration in Plant Communities(ed. M.Fenner), 2nd edn, pp. 3158. CABI Publishing, Wallingford, UK.
  • Marshall, E.J.P., Brown, V., Boatman, N., Lutman, P. & Squire, G. (2001) The Impact of Herbicides on Weed Abundance and Biodiversity. A report for the UK Pesticide Safety Directorate. Report No. PN0940. IACR Long Ashton Research Station, Long Ashton, UK.
  • Mohler, C.L. (1993) A model of the effects of tillage on emergence of weed seedlings. Ecological Applications, 3, 5373.
  • Park, S.E., Benjamin, L.R., Aikman, D.P. & Watkinson, A.R. (2001) Predicting the growth interactions between plants in mixed species stands using a simple mechanistic model. Annals of Botany, 87, 523536.
  • Payne, R.W. (2000) probitanalysis. Genstat Release 4.2 Reference Manual. Part 3. Procedure Library PL12 (eds R.W.Payne & G.M.Arnold), pp. 260264. VSN International Ltd, Oxford, UK.
  • Picman, J. & Picman, A.K. (1984) Autotoxicity in Parthenium hysterophorus and its possible role in control of germination. Biochemical Systematics and Ecology, 12, 287292.
  • Roberts, H.A. & Chancellor, R.J. (1986) Seed banks of some arable soils in the English Midlands. Weed Research, 26, 251257.
  • Robinson, R.A. & Sutherland, W.J. (2002) Post-war changes in arable farming and biodiversity in Great Britain. Journal of Applied Ecology, 39, 157176.
  • Roman, E.R., Thomas, A.G., Murphy, S.D. & Swanton, C.J. (1999) Modelling germination and seedling elongation of common lambsquarters (Chenopodium album) Weed Science, 47, 149155.
  • Schafer, D.E. & Chilcote, D.O. (1970) Factors influencing persistence and depletion in buried seed populations. II. The effects of soil temperature and moisture. Crop Science, 10, 342345.
  • Thompson, K., Band, S.R. & Hodgson, J.G. (1993) Seed size and shape predict persistence in soil. Functional Ecology, 7, 236241.
  • Venable, L.D. (1989) Modelling the evolutionary ecology of seed banks. Ecology of Soil Seed Banks (eds M. A.Leck, V.T.Parker & R. L.Simpson), pp. 6797. Academic Press, London.
  • Vleeshouwers, L.M. (1997) Modelling the effect of temperature, soil penetration resistance, burial depth and seed weight on pre-emergence growth of weeds. Annals of Botany, 79, 553563.
  • Westonby, M. (1998) A leaf-height-seed (LHS) plant ecology strategy scheme. Plant and Soil, 199, 213227.
  • Willson, F.M. & Traveset, A. (2000) The ecology of seed dispersal. Seeds: The Ecology of Regeneration in Plant Communities (ed. M.Fenner) 2nd edn, pp. 85110. CABI Publishing, Wallingford, UK.