Plant responses to simultaneous stress of waterlogging and shade: amplified or hierarchical effects?

Authors


Author for correspondence: John Lenssen Tel: +31 24 365 33 47 Fax: +31 24 365 24 09 Email: John.Lenssen@sci.kun.nl

Summary

  • Community ecologists often assume a hierarchy of environmental sieves to predict the impact of multiple stresses on species distribution. We tested whether this assumption corresponds to physiological responses using impact of water level and shade in wetland vegetation as a model.
  • Seedlings of four wetland species were grown under full light and simulated canopy shade, both in drained and waterlogged soils. When subject to both stresses simultaneously, waterlogging and shade independently affected growth of the two waterlogging tolerant species. For the intolerant species, however, waterlogging had the largest impact and the additional effect of shade was smaller than the effect of shade in drained soil. Soil flooding decreased specific leaf area but only if plants were in full light. Waterlogging did also not constrain a higher investment in stems of shaded plants.
  • These results demonstrate that light limitation in flooded habitats only plays a role if species can tolerate waterlogging and therefore correspond with the notion that water level determines the potential species pool and that standing crop consequently determines which species can actually persist.

Introduction

Species are often limited by multiple stress factors operating simultaneously, making it difficult to predict their distribution on the basis of physiological responses to single factors (Chapin et al., 1987; Bazzaz & Morse, 1991). A way to overcome this problem may be to assume a hierarchy of environmental sieves. Plant distribution can then be predicted if, besides the response to single environmental factors, it is known which factor overrules the other or, in other words, which sieve has to be passed first before the other becomes relevant. Here we will test the validity of an assumed hierarchy of environmental sieves in wetland vegetation.

A typical feature of wetland vegetation is that both drained and waterlogged soils may support highly productive stands (Wheeler & Shaw, 1991; Grace, 1999). As a consequence, light availability in lower strata of the canopy is limited along the entire water level gradient. Recent field studies suggest that water level and light availability determine species distribution hierarchically and that water level mainly determines which species can potentially occur, whereas light availability determines which species from this potential species pool will actually be present (Grace, 1999; Lenssen et al., 1999; and references therein). In contrast to these field studies, physiological responses to separate effects of waterlogging and shade suggest that both factors interact in such a way that the adversity of shading is amplified in waterlogged soils.

Waterlogging would affect the response to shade because oxygen deficiency in roots, due to water saturation of the soil, induces anaerobic respiration that enhances consumption of carbohydrate reserves (Laan et al., 1990). Flood tolerant species may avoid oxygen deficiency by means of root aerenchyma (Laan et al., 1990) although reduced allocation to shoots in waterlogged soils has been demonstrated for species with high and low root porosities (Visser et al., 2000a). This may be due to oxygen loss to the rhizosphere, which prevents oxygenation of the entire root system (Visser et al., 2000b). In flooded soils, roots of both tolerant and intolerant species may therefore become a strong sink for carbohydrates and thus inhibit investments in above-ground plant parts. In addition, waterlogging may impair leaf area expansion (Talbot et al., 1987; Dale & Causton, 1992). At least for species from open habitats, increased investment of dry weight in stems and leaves and formation of larger, but thinner leaves are typical shade responses aimed at enhancing light interception and reducing respiratory costs (Corré, 1983; Amthor, 1984; Chapin et al., 1987; Rice & Bazzaz, 1989; Schmitt & Wulff, 1993; Van Hinsberg & Van Tienderen, 1997).

Hence, high water levels might prevent adjustments in leaf morphology and biomass allocation. Since these morphological changes are required to optimize growth in the shade, waterlogging may increase the amount of stress if both factors operate simultaneously (sensuGeiger & Servaites, 1991). This would imply that both factors interact in such a way that the total effect on waterlogged, shaded plants is stronger than expected on the basis of both separate effects. Amplification is, however, one out of three possible ways in which waterlogging and shade may affect plant growth when operating simultaneously.

Waterlogging and shade may also affect plant growth independently or interact in such a way that one factor reduces the impact of additional factors. The latter type of interaction is typical for cases where one limitation is so strong that additional limitations hardly depress plant growth any further (Chapin et al., 1987; Wilson, 1988; Urbas & Zobel, 2000). Based on the hierarchy of effects as concluded from field studies (Lenssen et al., 1999; Grace, 1999) independent or reduced effects instead of amplified effects would be expected. Tolerant species would be affected independently by both factors but for species that are intolerant to waterlogging, additional effects of shade would be reduced because waterlogging is the dominant stress factor.

We report on an outdoor experiment where we measured growth and morphological responses to combined and separate effects of waterlogging and shade on the wetland species, Epilobium hirsutum, Eupatorium cannabinum, Mentha aquatica and Myosotis scorpioides. We hypothesized that waterlogging would inhibit leaf area expansion and dry weight investment in shoots of shaded plants and thus amplify the effects of shade. Our species are all most abundant in open habitats (Güsewell & Edwards, 1999) but inhabit different positions along the water level gradient and were therefore expected to represent a broad range in tolerance to waterlogging. Epilobium and Eupatorium have their highest abundance at drained sites of land-water gradients, whereas Mentha and Myosotis are usually found at waterlogged sites (Fojt & Harding, 1995; Van de Rijt et al., 1996).

Materials and Methods

Plant material

Epilobium hirsutum L., Eupatorium cannabinum L., Mentha aquatica L. and Myosotis scorpioides L. (nomenclature according to Van der Meijden (1996 )) used for this experiment were raised from seeds, collected from single populations along the River Nieuwe Merwede in the Netherlands (51°45′ N, 4°45′ E). Seeds were stored in dry and dark conditions at room temperature until further usage. Thirty-six days before the start of the experiment seeds were germinated on 1 mm glass beads watered by tap water in a chamber with a day/night alternation of 16 h/8 h at 25°C/10°C. On June 8th seedlings with two cotyledons and two true leaves were planted in plastic trays with a homogeneous 1 : 1 mixture of sand and clay and transferred to a glasshouse with a minimum day temperature of 23°C and a minimum night temperature of 15°C (day/night period: 16 h/8 h). Before the start of the experiment plants were removed from trays, soil was carefully rinsed from the roots under gently flowing tap water and f. wt was determined for each individual plant. Ten plants of each species were separated into roots and shoots and d. wts were determined to calculate the d. wt : f. wt ratio.

After f. wt measurement each plant was placed in a PVC cylinder (15 cm diameter, 50 cm height) filled with the same substrate as used in the plastic trays. To avoid nutrient limitation a 7.5-g Osmocote slow –release fertilizer (7.5% NH4-N, 7.5% NO3-N, 7.2% P2O5 and 12% K2O; Osmocote Plus, Grace Sierra International, Heerlen, The Netherlands) was added to each cylinder.

Experimental design

Shade was realized through frames (90 × 90 × 100 cm, length × width × height) that were covered with a moss green filter (Lee no. 122; ProDesign Lighting Pty Ltd, Kingsley, Australia) to reduce the red : far-red (R : FR) ratio. Neutral density net was placed on top of the moss green filter to reduce light intensity to 7% of full daylight. This simulated the conditions under the canopy of tall perennial wetland vegetation: reduced red : far-red ratio is a general effect of any plant canopy (Schmitt & Wulff, 1993) and reduction to 7% of full light matches with light levels measured under canopies of Phragmites (Lenssen et al., 1999). Ambient light levels at the experimental site varied from 1300 µmol/m2 s−1 on sunny days to 995 µmol/m2 s−1 on cloudy days.

Plants receiving full light treatment were placed under frames covered by a colorless plastic film to obtain microclimatic conditions comparable to shaded frames. The top of the cylinders was 10 cm below the bottom of the frames, so that the space in between allowed air circulation while preventing light transmission as was confirmed by light quantity measurements (data not shown).

In order to control for photo-degradation of the moss green filter R : FR light ratio and light intensity were measured under three shade frames and three full light frames at the start of the experiment (July 11), halfway and at the end (November 1) by MACAM Spectro radiometer SR9910-PC (MACAM Photometrics Ltd, Livingston, UK). At the beginning the R : FR ratio was 0.20 ± 0.03 and at the end 0.24 ± 0.01 (both mean ± SE, n = 3).

Water treatments were realized in four adjacent outdoor ponds in the experimental garden of the University of Nijmegen. At the start of the experiment, 1 July, plants were placed in the ponds under the frames and water level in each pond was maintained at 10 cm below the top of the PVC cylinders to guarantee an adequate water supply for roots of the transplanted seedlings. After 1 wk, the water level in two ponds was gradually raised to the top of the cylinders (‘waterlogged treatment’). In the two other ponds the water level was lowered to a level of 20 cm below the top of the cylinder (‘drained treatment’).

Per pond three shaded and three full light frames were randomly placed in two parallel rows. Under each frame we placed one replicate per species and harvest date. Hence, our experiment had four treatments (i.e. ‘shade drained’, ‘shade waterlogged’, ‘full light drained’ and ‘full light waterlogged’) with six replicates per species and harvest date. There were two ponds per water level and each pond had three frames per light level.

Harvest procedures

Plants were harvested 27, 63 and 92 d after the start of the experiment and harvest date was assigned to each plant at the start of the experiment. Harvested plants were separated into roots, rhizomes, stolons, main stem and branches, leaves and inflorescences. Leaf area was determined using a digital image processor (Imaging Technology Inc., Woburn, MA). Petioles were pooled with branches to allow comparison of stem dry weight between species with petioles (Mentha and Eupatorium) and without petioles (Myosotis and Epilobium). Stolons, only formed by drained Mentha, were included as stem tissue. Rhizomes, in case of Epilobium, and inflorescences (Myosotis and Epilobium) were not further analyzed but weighed for determination of total d. wt. D. wts were determined after drying at 70°C for at least 48 h.

Data analysis

Dry weight production We first tested the effect of water, light, harvest and species on (Loge-transformed) total d. wt by means of ANOVA. In this analysis and the analyses concerning leaf morphology and d. wt partitioning (see below) each frame was treated as a (random) block factor, the experimental pond (used for water level manipulation) was regarded to be confounded in this block factor. Since each frame was also unique for a certain light treatment we treated block as a factor that was nested within the water × light level. Species × water level and species × light interactions were partitioned in six pair-wise, nonorthogonal contrasts to investigate species differences in treatment response. Significance levels were adjusted according to the Dunn-Šidák method, keeping the overall experiment-wise error rate at 5% (Sokal & Rohlf, 1995).

In a second analysis we investigated whether the observed dry weight production in waterlogged, shaded conditions could be predicted by assuming that both waterlogging and shade affect plant growth independently. For each species and harvest date we first determined the response to waterlogging and shade by calculating the fraction of control conditions remaining under waterlogging and shade, respectively:

Waterlogging response (WR) = d. wt waterlogging, full light/d. wtdrained, full light
Shade response (SR) = d. wtdrained, shade/d. wtdrained, full light

As a null model we assumed that both canopy shade and waterlogging would affect plant growth independently. Accordingly, the combined response to canopy shade and waterlogging would be similar to the product of both separate responses:

Total response (TRpredicted) = WR × SR

In accordance with earlier studies (Wilson, 1988; Cahill, 1999) our null model assumed a multiplication instead of a summation of both separate effects. A summation of both separate effects might yield unrealistic values for the predicted total response. For instance, if 60% of control d. wt would remain under waterlogged conditions and 50% under shade conditions the fraction remaining under waterlogged-shaded conditions according to our assumption will be (0.6)(0.5) = 0.30, that is 30%. If both effects would be summed however, the prediction would be that 110% of the control d. wt would remain under waterlogged and shaded conditions. Hence, a null model based on the additivity assumption would result in predictions that are meaningless in biological terms and can therefore be rejected a priori.

Finally, the d. wt that would have been produced if effects of waterlogging and shade had operated independently was calculated by multiplying TRpredicted with the d. wt production in drained, full light at each harvest date and for each species. The predicted d. wt values were compared with the observed d. wt production in ANOVA. Using total d. wt under waterlogged-shaded conditions as dependent variable we tested the effects of species, harvest and calculation method (observed vs predicted). A significant main effect of calculation would indicate that the observed d. wt was significantly different from predictions according to the assumption of independent effects. Higher than predicted d. wt production would indicate that combined effects were weaker than both separate effects, whereas lower than predicted d. wt production would indicate that both effects had amplified each other. A significant species × calculation term would indicate that the presence or type of interaction is species-specific.

Leaf morphology and d. wt partitioning

Differences between species and experimental treatments in allocation to roots, stems and leaves and on leaf area ratio (LAR; leaf surface relative to total d. wt) and specific leaf area (SLA; leaf surface relative to leaf d. wt) were tested with allometrical relationships which allow comparison at a common plant weight (Coleman et al., 1994). The common allometrical relationship between the d. wt of Y (referring to either root-, stem- or leaf dry weight or leaf area) and the total d. wt X (total d. wt or, in case of SLA, leaf d. wt) is Y = aXb, a and b are constants (Gould, 1966). If b is equal to 1, a gives the ratio between Y and X, but if b significantly differs from 1 the ratio of Y and X changes with total (or leaf) d. w.t. We used Loge-transformed values of X and Y to make the allometrical relationships linear, i.e Loge Y = Logea + b Loge X.

Effects of light, water level and species on d. wt allocation to roots, stems, leaves and on LAR and SLA were tested with ANCOVA using (Loge -transformed) values of d. wt of these plant parts or leaf area as dependent variable and (Loge-transformed) values of total d. wt or leaf d. wt as the covariate (Klinkhamer et al., 1990; Coleman et al., 1994). Differences between treatments or species in allocation or leaf morphology would surface as a significant interaction between the treatment or species and total (or leaf) d. wt, and would have been caused by a significant change in the elevation of the allometrical relationship (the intercept, Logea), the allometrical coefficient (slope, b) or both.

Results

Dry weight

All species achieved the highest d. wt production in drained, full light conditions (Fig. 1) Both waterlogging and shading reduced d. wt although the effect of both factors separately was species-specific as indicated by the significant species × water level and species × light interaction (Table 1). Contrast analysis of the species × water level term (results not shown) learned that Epilobium and Myosotis had similar responses to water level, that is a small decrease when waterlogged (Fig. 1). Mentha was more reduced when waterlogged and its response to water level was significantly different from Epilobium. Eupatorium was the least tolerant to flooding and differed significantly in this respect from the other three species. Partitioning the species × light interaction only revealed a significant difference between Eupatorium and the other three species, possibly caused by the relatively low d. wt production of waterlogged Eupatorium in full light (Fig. 1). Hence, the four species clearly differed in tolerance to waterlogging, but with the possible exception of Eupatorium, showed similar shade tolerance.

Figure 1.

Changes in total d. wt (Log e -transformed values) of Epilobium hirsutum , Eupatorium cannabinum , Mentha aquatica and Myosotis scorpioides under drained, full light (open square), waterlogged, full light (open circle), drained, shaded (solid square) and waterlogged, shaded (solid circle) conditions (mean ± SE). Also indicated is total dry weight in waterlogged, shaded conditions predicted on basis of independent effects of waterlogging and shading (solid triangle, dashed line). Asterisks indicate significant differences (tested with least significant difference, P < 0.05) between observed and predicted values for waterlogged, shaded conditions at the given harvest time. If error bars are not visible they fall within the range of the symbol.

Table 1. F -values, significance and error mean squares for effects of water level (W), light treatment (L), Species (S), Harvest (H) and Block (B) on (Log e -transformed values of) total plant dry weight (d. wt). df = degrees of freedom; significance: * P < 0.05; ** P < 0.01; *** P < 0.001
Effectdfd. wt
W1113.4***
L1368.7***
W*L1 13.8**
B(within W*L) (= Error) (MS)200.7
S3 39.4***
S*W3 7.3***
S*L3 15.0***
S*W*L31.5
S*B(within W*L) (= Error) (MS)600.4
H2388.9***
H*W2 9.5***
H*L2 62.0***
H*W*L2 4.8*
H*B (within W*L) (= Error) (MS)400.4
S*H6 11.6***
S*H*W6 7.1***
S*H*L61.8
S*H*W*L61.6
S*H*B(withinW*L) (= residual(MS)800.3

The hypothesis to be tested predicted that response to combined effects of waterlogging and shade would be stronger than the product of both separate effects. As shown by the significant calculation term (Table 2) d. wt production of waterlogged, shaded plants significantly differed from predictions according to independence of both effects. However, the occurrence of interactions between water level and shade was species specific as demonstrated by the significant species × harvest × calculation term (Table 2). In contrast to our hypothesis, response to the multiple stresses of waterlogging and shading was never stronger than the product of both separate responses. For the two most tolerant species, Epilobium and Myosotis, observed d. wt production was equal to predicted values, indicating that waterlogging and shading affected these species independently. For Mentha and Eupatorium responses to combined stress factors were weaker than expected on the basis of both separate effects (Fig. 1).

Table 2. F -values, significance and error mean squares for effects of Calculation (observed vs. predicted), Species, and Harvest on (Log e -transformed values of) total plant dry weight (d. wt) of plants subject to combined effects of waterlogged and shade. df = degrees of freedom; significance: * P < 0.05; ** P < 0.01; *** P < 0.001
Effectdfd. wt waterlogging-shade
Calculation122.81***
Species310.95***
Harvest216.90***
Calculation × Species31.72
Calculation × Harvest2 5.30**
Species × Harvest2 5.33***
Calculation × Species × Harvest6 2.27*
Error (MS)860.68

Leaf morphology

The interaction between light, species and total d. wt had a marginally significant effect on leaf area (Table 3). This suggests that light treatment affected LAR, that is the relationship of leaf area vs. total d. wt, and that this relationship was species-specific. All species increased LAR when shaded, but this increase seemed smallest for Myosotis (data not shown).

Table 3. F -values, significance and error mean squares for effects of water level (W), light treatment (L), Species (S) and Block (B) on (Log e -transformed values of) leaf area and dry weight of leaves, roots and stems. Total plant dry weight (D. WT, Log e -transformed) was used as a covariate. Leaf area was tested with total plant d. wt as a covariate (‘Leaf area-T’) and with leaf dry weight as a covariate (‘Leaf area-L’) to test for differences in leaf area ratio and specific leaf area, respectively. df = degrees of freedom; significance: † P < 0.1 (i.e. marginally significant); * P < 0.05; ** P < 0.01; *** P < 0.001
EffectdfLeaf area-TLeaf area-LLeaf weightStem weightRoot weight
W1 16.05***0.82 20.34***6.081.08
L1 30.96*** 11.36**2.57 30.56*** 29.88***
W*L12.21 7.33*1.700.021.15
B (W*L) (= MS)180.090.100.040.130.49
S3 2.641.322.06 6.49*** 2.8*
W*S31.080.650.610.871.02
L*S3 2.56 2.61 5.52**0.931.89
W*L*S30.281.300.330.130.24
B*S (W*L) (= MS)470.140.090.040.060.17
D. WT1618.88***598.06***2584.93***2394.54***254.39***
W*D. WT10.11 5.04* 5.41*0.42 6.95*
L*D. WT10.020.360.97 4.00* 12.00**
W*L*D. WT11.39 4.85*0.310.012.19
B*D. WT (W*L) (= MS)200.120.100.030.070.23
S*D. WT31.69 2.53 4.52** 10.93***0.52
W*S*D. WT31.081.010.330.211.28
L*S*D. WT3 2.30 3.75* 2.390.220.75
W*L*S*D. WT30.151.210.191.000.86
B*S*D. WT (W*L)510.100.070.030.05 0.18(= MS)

Specific leaf area, given by significance of leaf d. wt on leaf area, was significantly affected by the interaction between water level and light treatment, indicating that the effect of shade on SLA depended on water level (Table 3). In addition, changes in SLA due to light treatment were also species specific, as indicated by the significant light × species × leaf d. wt interaction. For each combination of species, water level and light treatment we inspected the slopes of the linear relationship between loge-values of leaf area and leaf d. wt and found that they were never significantly different from 1. SLA can therefore be considered as independent of total leaf d. wt and is here presented as the ratio of leaf area and leaf d. wt in order to further elucidate the interactions (Fig. 2).

Figure 2.

Specific leaf area (SLA; mean ± SE; all harvests pooled) of drained (DR) and waterlogged (WL) Epilobium hirsutum , Eupatorium cannabinum , Mentha aquatica and Myosotis scorpioides grown in full light (open symbols) or shade (closed symbols). Asterisks indicate a significant difference (tested with least significant difference) in SLA between shade and full light plants for the indicated species and water level.

Waterlogging decreased SLA of three of the species, but only under full light conditions. When shaded, waterlogging no longer affected SLA. Although the interactions between waterlogging and shading were not species specific (Table 3), effects of waterlogging on SLA were not found for Epilobium (Fig. 2). The significant interaction between species, light and leaf d. wt (Table 3) may have been caused by the relatively small increase in SLA when shaded.

Dry weight partitioning

The amount of d. wt of leaves, stems and roots was always significantly affected by the amount of total d. wt (Table 3) and linear regressions with loge-transformed values of d. wts fitted the data very well (Figs 3, 4 and 5). Relationships of d. wt of leaves, stems and roots with total d. wt were not significantly affected by an interaction between water and light (Table 3). This demonstrates that waterlogging did not interfere with the response to shade and that shade did not affect allocation responses to high water levels. Instead, allocation responses to both factors were independent of each other.

Figure 3.

Relationship between leaf d. wt and total plant d. wt (both Log e -values) for Epilobium hirsutum , Eupatorium cannabinum , Mentha aquatica and Myosotis scorpioides grown in drained, full light (dashed line, open square), waterlogged, full light (solid line, open circle), drained, shaded (dotted line, solid square) and waterlogged, shaded (dashed/dotted line, solid circle) conditions.

Figure 4.

Relationship between stem dry weight and total plant dry weight (both log e -values) for Epilobium hirsutum , Eupatorium cannabinum , Mentha aquatica and Myosotis scorpioides grown in drained, full light (dashed line, open squares), waterlogged, full light (solid line, open circles), drained, shaded (dotted line, solid squares) and waterlogged, shaded (dashed/dotted line, solid circles) conditions.

Figure 5.

Relationship between root dry weight and total plant dry weight (both Log e -values) for Epilobium hirsutum , Eupatorium cannabinum , Mentha aquatica and Myosotis scorpioides grown in drained, full light (dashed line, open squares), waterlogged, full light (solid line, open circles), drained, shaded (dotted line, solid squares) and waterlogged, shaded (dashed/dotted line, solid circles) conditions.

Shading alone significantly changed the relationship between stem and total d. wt and root and total d. wt (Table 3). At a common plant weight, shaded plants seemed to invest more d. wt in stems (Fig. 4). When pooled across species and water level the intercept (logea), slope (b) and R2 as found for plants in full light was: logea = −1.26(−1.33, −1.19); b = 1.14(1.11, 1.17); R2 = 0.98 and for plants in shade: logea = −0.85(−0.93, –0.77); b = 1.16(1.11, 1.22); R2 = 0.93 (for all regression coefficients values in parentheses refer to upper and lower limits of 95% confidence interval).

Increased investment in stems due to shading was most likely at the expense of root d. wt since allocation to roots was strongly reduced by shading (Fig. 5). When pooled across water levels and species we found for plants in full light: logea = −1.2(−1.39, −1.19); b = 0.89(0.84, 0.93); R2 = 0.93 and for shaded plants: logea = −2.24(−2.37, −2.10); b = 0.68 (0.59, 0.77); R2 = 0.64. There was also a significant main effect of water level on allocation to roots (Table 2). With data from different species and light levels pooled, regression coefficients (loge (root d. wt) vs. loge (total d. wt)) for drained plants were: logea = −1.71(−1.84, −1.58); b = 0.96(0.89, 1.02); R2 = 0.88 and for waterlogged plants logea = −1.58(−1.69, −1.47); b = 0.94(0.88, 1.00); R2 = 0.89. Hence, differences between drained and waterlogged plants were small but allocation to roots tended to be higher in waterlogged plants.

Waterlogging also induced a significant shift in allocation to leaves (Table 3) resulting into less leaf d. wt per g total d. wt for waterlogged plants (Fig. 3). For drained plants coefficients of the allometrical relationship were: logea = −0.90(−0.86, −0.94); b = 0.89(0.87, 0.91); R2 = 0.98, for waterlogged plants: logea = −1.1(−1.6, – 0.96); b = 0.89(0.86, 0.92); R2 = 0.97 (both water levels pooled across species and light levels). High water level may have induced a subtle allocation shift from leaves to roots.

Discussion

The main objective of this study was to investigate whether physiological responses to waterlogging and shade match with field studies that indicated a dominance of water level over standing crop (determining light availability) as factors regulating species distribution in wetland vegetation (Lenssen et al., 1999; Grace, 1999). We considered amplified effects of waterlogging and shade on plant growth to be in contrast with this assumed hierarchy since this would imply that a species’ shade tolerance depends on water level and that a species’ flooding tolerance would depend on the prevailing light conditions. Amplified effects were, however, anticipated because waterlogging and shade may induce conflicting morphological responses and thus cause an imbalance between source and sink functioning when multiple stresses operate simultaneously (Geiger & Servaites, 1991).

Earlier studies have shown that waterlogging inhibits the possibility of increasing both LAR and SLA (Talbot et al., 1987; Dale & Causton, 1992) and may thus prevent a more efficient light interception under shaded conditions (Rice & Bazzaz, 1989). We saw no effect of water level on the relationship between leaf area and total d. wt (i.e. LAR). However, when expressed as SLA, that is on a leaf weight basis, leaf area was decreased by soil flooding, but surprisingly this only happened in full light conditions. When shaded, plants always increased specific leaf area and effect of water level was no longer significant (Fig. 2). The impact of canopy shade on specific leaf area, therefore overruled the potentially deleterious effects of waterlogging on leaf morphology. Hence, our study provided no indication that soil flooding would affect the ability for light interception in shaded habitats.

A second reason for hypothesizing amplified effects of waterlogging and shade followed from the notion that waterlogged roots may act as a strong carbohydrate sink (Laan et al., 1990) and thus inhibit investments in above-ground plant parts. Decreased allocation to roots in response to shade has been demonstrated in many studies (Corré, 1983; Dale & Causton, 1992; Schmitt & Wulff, 1993; Van Hinsberg & Van Tienderen, 1997), but literature is less conclusive about the impact of waterlogging on allocation to roots. Reported responses vary from enhanced allocation to roots (Visser et al., 2000a), no impact (Rubio & Lavado, 1999) or even decreased investment to roots due to waterlogging (Dale & Causton, 1992). None of these studies corrected for differences in plant size, although this may affect the outcome of studies on biomass allocation under influence of environmental factors (Coleman et al., 1994; Gunn et al., 1999). Our results were not confounded by size differences and suggest that waterlogging and shade might induce conflicting interests within plants because waterlogging induced an increased allocation to roots, whereas shade decreased the allocation to roots. However our allometrical analyses also suggest that allocation shifts induced by shade are much stronger than those induced by waterlogging, and that both allocation shifts are regulated independently. Waterlogging mainly affected the relative allocation between leaves and roots, whereas shade operated on the allocation between stems and roots.

Hence, waterlogging did not affect morphological responses to canopy shade, which may explain why independent effects on d. wt production were found in the majority of cases. However, reduced effects under simultaneous stress were also found. This type of interaction usually indicates that one factor limits plant growth so strongly that a second limitation hardly has any additional impact (Chapin et al., 1987; Wilson, 1988). In particular for Eupatorium, waterlogging must have exerted the strongest limitation on plant growth because observed d. wt under combined stress was closest to d. wts of waterlogged, full light plants. In addition, subaddditive effects were not found for Epilobium and Myosotis but were continuously found for the least tolerant species Eupatorium (Fig. 1).

To our knowledge, the study by Dale & Causton (1992) is the only previous experiment investigating interactions between waterlogging and shade. They also found significant interactions between waterlogging and shade, but did not distinguish between different types of interactions. However, their conclusion that waterlogging had a stronger impact on plant growth than shading also points towards a hierarchy of limiting factors. A further important difference with the present study is that the three Veronica species tested by Dale & Causton (1992) were equally susceptible to waterlogging, but instead differed in shade tolerance. By including a wider range of flood tolerances we have demonstrated that presence or absence of interactions may be determined by the degree to which a species is limited by waterlogging.

This study has demonstrated that, where both factors operate simultaneously, either flooding or shade is the most important factor determining d. wt and, possibly, related fitness variables such as reproduction and survival. By explicitly addressing the type of interaction we were also able to identify a hierarchy of both factors. Water level may be regarded as a stronger environmental sieve than light availability because, in waterlogged soils, light limitation mainly affected the tolerant species (Epilobium and Myosotis in particular), whereas flooding had such a large impact on the intolerant Eupatorium that additional light limitation did not seem to matter. This hierarchy is only based on physiological responses but is in agreement with results from field studies in wetland vegetation that demonstrated that vegetation removal favored more species at drier sites than at flooded sites of land water gradients (Keddy, 1989; Lenssen et al., 1999). Both these field studies and the physiological responses reported here provide a firm experimental basis for community models (e.g. Grace, 1999) that regard flooding as a factor determining the potential species richness and standing crop as the factor determining which species from this pool can actually persist.

Acknowledgements

We acknowledge M. Fransen, D. Waasdorp, B. A. M. Peters, G. van Weerden and colleagues for their practical assistance and P. H. Van Tienderen and H. P. Koelewijn for their statistical advice. C. W. P. M. Blom, J. F. Stuefer and two anonymous reviewers critically improved earlier versions of the manuscript. This study was financed by the Department for Road and Hydraulic Engineering, Department Zuid-Holland of the Directorate-General for Public Works and Water Management and Water- and Civil Board De Brielse Dijkring. Publication 3095 N100-KNAW Netherlands Institute of Ecology.

Ancillary