- Top of page
Restoration schemes for enhancing plant species diversity in agricultural grassland have featured in UK and European agri-environment schemes since 1985. In the UK this has been through countryside stewardship and environmentally sensitive area (ESA) schemes, replaced on 3 March 2005 by environmental stewardship. On 1 September 2006 this latest scheme covered 3·1 million ha, with 23 000 live agreements in place, representing first-year payments of more than £105 million. This is about 30% of UK agricultural land eligible for such payments. However, despite the periodic revision of management prescriptions, their effectiveness has recently been questioned (Kleijn & Sutherland 2003; Whitfield 2006). Knop et al. (2006) have demonstrated significantly higher plant species diversity in Swiss meadows under agri-environment schemes, although they acknowledged that the differences may have been present prior to the implementation of the scheme, confirming that rigorous evaluation needs to be incorporated into agri-environment schemes (Kleijn & Sutherland 2003). In the meantime, data on the long-term consequences of different management treatments for mesotrophic grasslands, such as hay-cut dates, fertilizer addition and grazing regimes, plus an understanding of some of the underlying ecological processes, should be used to identify appropriate management treatments for enhancing diversity and to inform policy on the likely time scales. Experiments across a range of grassland types have included tests of single treatments such as grazing regimes (Smith & Rushton 1994). Tests of biological tools have included the use of the hemiparasite Rhinanthus minor to debilitate competitive grasses (Pywell et al. 2004) and the sowing of seed when converting arable land to grassland (Pywell et al. 2002) and when enhancing diversity in improved ryegrass swards (Smith et al. 2000).
Such experiments have identified some of the key management features and ecological processes that control plant species diversity in mesotrophic grassland. They have also shown that it can take many years to increase plant diversity successfully and return grass swards to some semblance of their traditional character. Five-year experiments begun in the 1990s, and subsequently extended, will not fully influence management prescriptions for many years, with yet further delays in putting experimental results into farming practice. Therefore, concern about the effectiveness of grassland agri-environment schemes can be expected, particularly when experiments indicate that the ecological problem is more intractable than first thought and requires investigation of ecological mechanisms through field trial and mesocosm experiments (Bardgett et al. 2006).
We present the phase 2 results from a long-term meadow grassland experiment begun in 1990 at Colt Park, North Yorkshire, England. It was devised to determine the best combination of management practices for the restoration of plant species diversity to agriculturally improved grassland. Specifically, the experiment tested the effects of three grazing, three hay-cut date, two fertilizer and two seed-addition treatments on the plant species composition of a Lolium perenne–Cynosurus cristatus grassland (MG6) (Rodwell 1992; Lolio–Cynosuretum cristati grasslands in European terminology). The first phase (1990–98) demonstrated that all the main treatments produced small but significant changes in plant species diversity (Smith et al. 2000), with a particularly large increase occurring with a combination of autumn cattle and spring sheep grazing, a 21 July hay-cut date and addition of seed of many species. A large decrease in diversity occurred when, in the absence of autumn cattle grazing, hay was cut on 14 June and fertilizer was also applied. Rhinanthus minor spread to most plots after its introduction in the seed treatment, being particularly abundant in 1996 in all treatment combinations that included autumn grazing, no mineral fertilizer and a 21 July hay cut. Populations of > 40 plants m−2 were associated with low hay yields. This first phase of the trial thus demonstrated that the species composition of the meadow was particularly affected by combinations of management treatments.
The addition of seed was important for the restoration of species diversity, as the soil seed bank was very similar in species composition to the above-ground vegetation of agriculturally improved meadows and contained few extra species that could contribute to an increase in sward diversity (Smith et al. 2002). Also, there were many fewer species than in the soil seed bank of a traditionally managed meadow. The hemiparasite Rhinanthus minor (hay rattle) was thought to have been an important driver of the observed vegetation change, as it is known to reduce the dominance of competitive grasses (Joshi, Matthies & Schmid 2000; Pywell et al. 2004) across a range of residual soil fertility and farmyard manure (FYM) applications (Bardgett et al. 2006). This suggests great potential for its use in the restoration of species-rich grassland (Davies et al. 1997; Pywell et al. 2004). In addition, hay rattle has been shown to increase rates of soil nitrogen mineralization (Bardgett et al. 2006) and change other soil microbial properties. At Colt Park, when seed was added to unfertilized plots cut on 21 July, increased species richness had been observed by 2000, primarily through an increase in legumes, stress-tolerant and stress-tolerant-ruderal plant strategists, and was associated with an increase in fungal abundance in the soil (Smith et al. 2003). Our interest in the soil microbial community was based on the possible significance of changes in the abundance of soil fungi and soil fungal: bacterial (F:B) ratios for the restoration of plant species diversity. Reductions in soil F:B ratios have been linked to intensive management and, in particular, the use of mineral fertilizer, and it has been suggested that increases in this ratio are associated with more efficient nutrient cycling, increased abundance of mycorrhizal fungi and enhancement of plant species diversity (Bardgett & McAlister 1999; Donnison et al. 2000; Bardgett et al. 2001; Grayston et al. 2004). These shifts in microbial community structure resulting from intensive management may be reversible, albeit over long time scales, and the introduction of particular plant species into the sward might accelerate this reversal, acting to promote fungal growth in soil (Smith et al. 2003; Bardgett et al. 2006).
By 1998 it was evident that differences in grazing regime and hay-cut date were major drivers of vegetation change in the Colt Park experiment. Consequently, advice for agri-environment and other conservation schemes for the diversification of grassland was that hay should be cut in mid-July and the regrowth (aftermath) grazed with cattle in the autumn. It was also evident that the second phase (1998–2004) needed to include an additional FYM treatment to address a developing management and research issue. FYM is an inevitable waste product of livestock rearing under cover in barns but it has some value as a fertilizer. It is spread on meadows in winter and early spring at various rates. This waste disposal supplies a variable amount of nutrients for plant growth, with an average nutrient composition of 6 kg nitrogen, 3·5 kg phosphate and 8 kg potash t−1 FYM (Ministry of Agriculture, Fisheries & Food 1994; Simpson & Jefferson 1996). The maximum rate of FYM application prescribed for meadows in the Pennine Dales ESA was 12 t ha−1 (Ministry of Agriculture, Fisheries & Food 1992), i.e. a possible 72 kg ha−1 N, 42 kg ha−1 P and 96 kg ha−1 K to meadows. However, only 20% of these nutrients are thought to be available to the crop following spring FYM application, or 0·5, 3·4 and 7·7 times the amounts of N, P and K supplied by the standard ESA maximum acceptable rate of mineral fertilizer, as used at Colt Park.
After the 1998 hay harvest, the trial was continued on the autumn- and spring-grazing treatment plots and the other grazing treatments (two-thirds of the experiment) were abandoned. The 21 July cut-date treatment was then applied to all remaining plots to investigate the effects of restoring traditional hay-cut dates where early and late hay cuts had been applied for 8 years.
A new FYM treatment was applied from 1999 by subdivision of the original plots. These smaller plots were more than twice the size of the 4-m2 quadrats used for sampling, with past treatments producing distinctive changes at this sampling scale. While the wind-blown seed of Rhinanthus minor was expected to disperse between plots (Coulson et al. 2001), other species had demonstrated considerable fidelity to the sown treatment in phase 1. Elsewhere, sharp boundaries between experimental plots under different treatment combinations had demonstrated the dominant effect of treatments as environmental sieves for plant colonization (Shiel & Hopkins 1991). The success of seed sowing in the first phase indicated that more seed of missing species could be sown in the original seeded plots at the start of the second phase.
The split-plot design in phase 2 (Table 1) allowed past 1990–98 treatments to be continued though phase 2, as well as providing for tests of interactions between the various treatment types and allowing an interpretation of the FYM results against a combination of pre-treatments. The following predictions were tested. (1) The magnitude of the effects of past management treatments on plant species composition will change over time, with the additional nutrients in FYM increasing the abundance of competitive species. (2) Any post-1998 changes in vegetation and soil microbial community would be dependent on the sward composition in 1998. (3) Returning to a 21 July hay-cut date after 8 years of 14 June or 1 September cuts would, within 6 years, produce a grassland similar to that cut on 21 July throughout the experiment. (4) The already demonstrated effects of the management treatments on the soil microbial community (expressed as the relative abundance of fungi and bacteria) would have increased by 2004. (5) Correlation between plant species and soil microbial community properties would continue to develop. (6) The sowing of additional species typical of traditional meadows would successfully reintroduce them into the sward in phase 2, even if inclusion in the original 1990–92 seed addition had failed to do so.
Table 1. Experimental design of the autumn- and spring-grazed plots, 1990–2004. Phase 1 anova structure (degrees of freedom): cut date (C), 2; mineral fertilizer (F), 1; seed addition (S), 1; C × F, 2; C × S, 2; F × S, 1; C × F × S, 2; residual, 24. Phase 2 anova structure (degrees of freedom): C, 2; F, 1; S, 1; FYM (M), 1; C × F, 2; C × S, 2; F × S, 1; C × M, 2; F × M, 1; S × M, 1; C × F × S, 2; C × F × M, 2; C × S × M, 2; F × S × M, 1; C × F × S × M, 2; residual, 48
|Phase 1, August 1990–98||Phase 2, August 1998–2004|
|Cut date||Fertilizer||Seed||Cut date||Fertilizer||Seed||FYM|
|F1||S0|| ||F1||S0||M0 M1|
|F1||S0|| ||F1||S0||M0 M1|
|F1||S0|| ||F1||S0||M0 M1|
- Top of page
The field trial, at 300 m altitude, at Colt Park meadows in the Ingleborough National Nature Reserve, UK (national grid reference SD 775782, latitude 54°12′ N, longitude 2°21′ W) was on Lolium perenne–Cynosurus cristatus grassland (MG6) (Rodwell 1992), on a shallow brown-earth soil (pH 5·1) over limestone of moderate–high residual fertility (15 mg P2O5 L−1). Management prior to 1990 consisted of autumn and spring grazing, mineral fertilizer application and a 21 July cutting date for hay.
From 1998, autumn and spring grazing were continued where it had been applied to three 12 × 36-m paddocks (0·043 ha), with livestock having free access to the plots. Grazing regimes were (i) 30 beef cattle (10·9 cattle ha−1) intensively grazed during four 3–5-day periods from mid-July to late August; (ii) 150 cross-bred lambs (7·5 lambs ha−1) throughout August; (iii) 10 beef cattle (0·9 cattle ha−1) throughout October; (iv) 40 young sheep (less than 1 year old) (2 sheep ha−1) from October to mid-March; and (v) 200 ewes (6·9 sheep ha−1) during lambing from mid-April to early-May. This grazing regime was generally followed each year but differed in 2001, when grazing with cattle was not possible because of an epidemic of foot and mouth disease in England.
In phase two, a 21 July hay cut was applied across the trial (Table 1) with a BCS Commander 630ws mower (Tracmaster Ltd, Burgess Hill, UK), with a 1·3-m wide reciprocating cutter bar. The cut hay from a plot was dried on that plot, turning it once to assist drying and seed dispersal, then removed from the trial. This May to mid-July period for the growth of the hay crop was typical of pre-1970 agriculture in the northern England uplands (Smith & Jones 1991; Smith 1997).
The two levels of the fertilizer treatment were continued, with each cut-date plot divided into two 6 × 12-m subplots randomly allocated to no fertilizer or 25 kg ha−1 nitrogen plus 12·5 kg ha−1 phosphorus (P2O5) and potassium (K2O). The fertilizer was a proprietary 20:10:10 NPK brand spread by hand in early May each year. The original fertilizer subplots had been further divided into two subsubplots for the phase 1 seed-addition treatment. This was continued post-1998 with 15·4 kg ha−1 commercial seed (Emorsgate Seeds, Tilney All Saints, Norfolk, UK) of each of Lotus corniculatus, Briza media and Ranunculus bulbosus sown in August 1998 and 0·5 kg ha−1Geranium sylvaticum seed sown in September 1999. In 1999 each seed treatment was further subdivided into two 6 × 3-m subsubsubplots and FYM added to one of each pair (i.e. two FYM treatments per subplot). Treatments of no FYM and 12 t ha−1 FYM were applied in April 1999, then again in November–December from 1999 onwards. The FYM was obtained in March–April each year from a local farm and stored in a covered midden until required the following winter. Three replicates of each treatment combination were split between three blocks.
Vegetation sampling was based on a subdivision of each 18-m2 plot into a central grid of two 4-m2 quadrats, with a surrounding 1–0·5-m wide boundary. All vascular plant species were identified in these quadrats using nomenclature according to Stace (1991) and their percentage contribution to the total vegetation cover estimated by eye during June or early July every 2 years from 1994 to 2004, before the plot was cut for hay. The soil microbial community was analysed using the phospholipid fatty acid analysis (PLFA) technique (Bardgett, Hobbs & Frostegård 1996), with analyses undertaken in 1996 and 2000 being repeated in July 2004 (see Smith et al. 2003 for detailed methods).
Redundancy analysis (RDA), available within canoco (version 4.5; Ter Braak & Smilauer 2002), was used to assess the proportion of the variability in plant species composition attributable to each treatment at each 2-year sample period from 1994 to 2004 (Ter Braak & Smilauer 2002). This tested prediction 1, which was also tested by analysis of the principal response curves (Ter Braak & Smilauer 2002) through phases 1 and 2 within plots that had consistently been grazed in the autumn and spring and cut for hay on 21 July. Treatments were compared over time against the species composition of the pre-1990 management regime, i.e. plots that had just received mineral fertilizer, were cut for hay on 21 July but had not been treated with FYM or had seed added.
A repeated-measures analysis of variance (SAS Statistical Software) was used to investigate changes over time (2000–2004) and the effects of the interaction of time with the applied treatments on the species richness, Ellenberg fertility and similarity of the vegetation to a target plant community (prediction 3). Epsilon values were calculated according to the method of Greenhouse & Geisser (1959) to adjust the degrees of freedom based on the divergence of the covariance matrix from homogeneity.
The target plant community was Anthoxanthum odoratum–Geranium sylvaticum grassland (MG3b) (Rodwell 1992) and the similarity was assessed with the TABLEFIT coefficient (Hill 1996). Upper and lower target values were defined from a random set of pseudoquadrats for subcommunities of MG3, MG4, MG5, MG6 and MG7 vegetation using methods described by Smith et al. (2003). These lower and upper limits were, respectively, 1 SD below and above the mean value for a plant community in this data set. Ellenberg fertility was assessed for each quadrat from the sum of the cover values for each species, weighted by Ellenberg's fertility index, recalculated for British conditions by Hill et al. (2000).
anova was also used to assess treatment differences in 2004 for the PLFA variables, including derived variables such as the ratio of fungal-to-bacterial fatty acids (F:B). The F:B ratio was used as an indicator of changes in the relative abundance of fungi relative to bacteria (Bardgett, Hobbs & Frostegård 1996). Fungal PLFA was estimated as the abundance of the fatty acid 18:2ω6 and bacterial PLFA as the sum of the abundance of the fatty acids i15:0, a15:0, 15:0, i16:0, 17:0, i17:0, 17:0cyclo, 18:1ω7 and 19:0cyclo (Frostegård, Bååth & Tunlid 1993). This anova tested prediction 4. As described earlier, a repeated-measures analysis (using SAS) was used to investigate changes over time (1996–2004) and the interaction of time with the applied treatments on F:B ratios (prediction 4) on plots that had been autumn and spring grazed and cut on 21 July without the addition of FYM throughout the 1996–2004 period. This tested the effects of fertilizer and seed addition on F:B ratios in 1996, 2000 and 2004. Significant interactions between treatments and time would indicate that the response of the vegetation to post-1998 treatments was dependent upon the composition of the vegetation in 1998 (prediction 2). As with the other variables, a test of normality (the Anderson–Darling test) was applied to the residuals of each anova. If this deviated significantly from normality, a suitable transformation was applied to the original data. Interactions between treatments were tested as part of the anova. Prediction 2 was also tested by RDA of the 1998–2004 vegetation, with the characteristics of the 1998 vegetation, defined as the sample quadrat scores on the first four axes of a principal components analysis of the 1998 vegetation data, used as covariables to remove the effects of the phase 1 management treatments. The blocks were covariables within which the random permutations were made for the Monte Carlo tests in canoco (Ter Braak & Smilauer 2002).
Prediction 5 was tested by RDA of the 2004 plant species cover and soil PLFA data, the latter being used as ‘environmental’ data with management treatments used as supplementary nominal variables, positioned within the ordination after construction. Presentation of the resultant species–PLFA–treatment triplots was simplified by removing the PLFA variables that made little contribution to the variation in plant species. Prediction 6 was tested by inspection of the biplots to see if the traditional meadow species sown in 1998–99 had appeared in the sward by 2004.
- Top of page
In phase 1, differences in hay-cut dates dominated the plant species composition of the vegetation of the autumn- and spring-grazed plots (Fig. 1) but this effect was greatly reduced within 2 years of imposing a common 21 July hay-cut date. The FYM treatment accounted for most of the variability in species composition in phase 2, particularly in 2002, when all applied management treatments together accounted for about 35% of the variation. The impact of FYM was also seen in the principal response curves (Fig. 2) for plots cut on 21 July from 1990 to 2004. These demonstrated that, in the absence of mineral fertilizer, the vegetation through phases 1 and 2 included Ranunculus acris, Rhinanthus minor and Anthoxanthum odoratum, plus others that had been sown into the sward with the seed-addition treatment. FYM application in phase 2, particularly in combination with mineral fertilizer, was associated with increases in Poa trivialis and Lolium perenne. This species composition was very different from that of the baseline treatment, particularly in 2002.
Figure 1. Variance in plant species composition attributable to management treatments within autumn- and spring-grazed plots, 1994–2004. Open diamonds, hay-cut date; infilled diamonds, FYM; open squares, fertilizer; open triangles, seed addition; infilled circles, treatment interactions
Download figure to PowerPoint
Figure 2. Principle response curves for the 1994–2004 change in species composition, relative to that in the control treatment (mineral fertilizer addition), within autumn- and spring-grazed treatments with a 21 July (1990–2004) hay-cut date treatment. F, fertilizer applied; S, seed added; M, FYM applied; O, no treatments. Species codes (emboldened codes indicate those species sown into the sward as part of the seed-addition treatment): Am, Achillea millefolium; Ap, Alopecurus pratensis; Ao, Anthoxanthum odoratum; As, Anthriscus sylvestris; Bh, Bromus hordeaceus; Cm, Conopodium majus; Cc, Cynosurus cristatus; Dg, Dactylis glomerata; Fr, Festuca rubra; Hs, Heracleum sphondyllium; Hl, Holcus lanatus; Lh, Leontodon hispidus; Lc, Lotus corniculatus; Lp, Lolium perenne; Pl, Plantago lanceolata; Pt, Poa trivialis; Pv, Prunella vulgaris; Ra, Ranunculus acris; Rr, Ranunculus repens; Rm, Rhinanthus minor; Tp, Trifolium pratense; Tr, Trifolium repens; Vc, Veronica chamaedrys; Vs, Veronica serpyllifolia.
Download figure to PowerPoint
The repeated-measures analysis of all cut date, fertilizer, seed and FYM treatment combinations in phase 2 (2000–04) demonstrated that greater species richness was separately associated with the addition of seed and the absence of mineral fertilizer (Table 2) and probably persisted from phase 1 (Fig. 3). The similarity of vegetation to the target MG3b plant community increased throughout phase 2 and was greatest where plots had been cut on 14 June in phase 1, when fertilizer was not applied and when seed had been added (Table 2 and Fig. 4). Interactions between treatments were demonstrated over time (Table 3), with the sward consistently increasing its similarity to the target vegetation when FYM was applied in the absence of mineral fertilizer and vice versa. These increases stopped in 2002 when both FYM and fertilizer were absent or applied together.
Table 2. Phase 2 mean vegetation characteristics associated with main treatments, 2000–04; significant differences assessed with repeated-measures anova
|Treatment|| || || ||F||d.f.||P|
|Past cut date||14 June||21 July||1 September|| || || |
|MG3b similarity (%)||59·06||57·81||57·00|| 9·51||2, 4||< 0·05|
|Fertilizer addition||No fertilizer||Fertilizer|| || || || |
|Species richness (spp. 4 m−2)||21·4||19·3|| || 12·97||1, 6||< 0·05|
|Ellenberg fertility|| 4·57|| 4·67|| || 84·65||1, 6||< 0·001|
|MG3b similarity (%)||60·19||55·73|| || 15·42||1, 6||< 0·01|
|Seed addition||No seed||Seed|| || || || |
|Species richness (spp. 4 m−2)||18·7||22·0|| ||147·72||1, 12||< 0·001|
|MG3b similarity (%)||54·76||61·17|| || 51·10||1, 12||< 0·001|
|Year||2000||2002||2004|| || || |
|Ellenberg fertility|| 4·74|| 4·65|| 4·48|| 50·58||2, 86·5||< 0·001|
|MG3b similarity (%)||53·56||59·16||61·16|| 63·24||2, 93·7||< 0·001|
Figure 3. 1994–2004 change in species richness where 2000–04 treatment differences were significant (Table 2). Vertical bars are standard errors; horizontal line at 26·2 represents the target species richness for restoration management; open circles, no fertilizer applied; infilled circles, fertilizer applied; open triangles, no seed added; infilled triangles, seed added.
Download figure to PowerPoint
Figure 4. 1994–2004 change in similarity to target MG3b vegetation where 2000–04 treatment differences were significant (Table 2). Vertical bars are standard errors; horizontal line at 65 represents the target TableFIT score for restoration management; open circles, no fertilizer applied; infilled circles, fertilizer applied; open triangles, no seed added; infilled triangles, seed added; squares, year.
Download figure to PowerPoint
Table 3. Phase 2 interaction effects of fertilizer treatments, with FYM over time on MG3b similarity scores; F2,93·7 = 4·13, P= 0·0198
| ||No fertilizer||Fertilizer added|
|Year||No FYM||FYM added||No FYM||FYM added|
These changes were similar, although in the opposite direction, to those of the Ellenberg fertility scores. Ellenberg fertility was lowest when fertilizer was not applied and in the later years of phase 2 (Table 2 and Fig. 5). There were a number of interactions such that Ellenberg fertility scores were high when plots had been cut on 14 June 1990–97 and low where hay had been cut on 21 July from 1990, in the absence of fertilizer (Table 4). FYM application increased Ellenberg fertility scores where fertilizer had been added in the absence of seed (Table 5). This joint effect was particularly high in 2002 (Table 6).
Figure 5. 1994–2004 change in Ellenberg fertility score where 2000–04 treatment differences were significant (Table 2): vertical bars are standard errors; horizontal line at 4·79 is the lower threshold for MG6a grassland, the line at 4·67 is the upper threshold for MG3b grassland as targets for restoration management; open circles, no fertilizer applied; infilled circles, fertilizer applied; squares, year.
Download figure to PowerPoint
Table 4. Phase 2 interaction effects of fertilizer and past cut date treatments on weighted Ellenberg fertility scores; F2,6 = 12·41, P < 0·01
| ||No fertilizer||Fertilizer added|
|Cut 14 June 1990–97, 21 July 1998–2004||4·687||4·714|
|Cut 21 July 1990–2004||4·502||4·662|
|Cut 1 Sept 1990–97, 21 July 1998–2004||4·525||4·645|
Table 5. Phase 2 interaction effects of fertilizer, seed addition and FYM treatments on weighted Ellenberg fertility scores; F1,24 = 8·04, P < 0·01
| ||No fertilizer||Fertilizer added|
|No seed||Seed added||No seed||Seed added|
Table 6. Phase 2 interaction effects of fertilizer, seed-addition and FYM treatments over time on weighted Ellenberg fertility scores; F4,86·5= 6·95, P= 0·0023
| ||No fertilizer||Fertilizer applied|
|No seed||Seed applied||No seed||Seed applied|
|Year||No FYM||FYM||No FYM||FYM||No FYM||FYM||No FYM||FYM|
F:B ratios and the amount of fungal PLFA in 2004 were greater in the absence of FYM and mineral fertilizer (Table 7), although there were no significant interactions between these treatments in that year. However, F:B ratios generally increased from 1996 in plots consistently grazed in the autumn and spring and cut on 21 July without FYM addition (Table 8), and there was an interaction between treatments such that F:B ratios were particularly high when seed had been added in the absence of fertilizer and low in the presence of fertilizer (Table 8). There was a significant association (F = 32·67, P= 0·002) between vegetation and PLFA measures along the first canonical axis of the RDA species–PLFA–treatment triplot (Fig. 6). This axis linked the absence of FYM with an increase in fungal PLFA and higher F:B ratios, and with plant species typical of traditional MG3 grassland, e.g. Rhinanthus minor, Anthoxanthum odoratum and Ranunculus acris. High F:B ratios were also associated with the presence of five species sown into the sward as part of the seed-addition treatment. Lower F:B ratios were associated with species typical of improved fertilized grasslands, e.g. Poa trivialis, Alopecurus pratensis and Dactylis glomerata. The seed-addition treatment in 1990–92 had been supplemented in 1998 with four additional species, two of which, Briza media and Geranium sylvaticum, occurred in only one and two sample quadrats, respectively, in 2004. However, Ranunculus bulbosus and Lotus corniculatus were more frequent in 2004, with frequencies of 6·9% and 14·6%, respectively. Lotus corniculatus was found where FYM had not been used and with other species typical of traditional grassland management (Fig. 6).
Table 7. Treatment effects on soil microbial characteristics 2004
| ||No FYM||FYM||F1,48||P|
|Fungal PLFA||48·28||39·97|| 5·23||< 0·05|
|F:B ratio|| 0·06978|| 0·05907|| 4·94||< 0·05|
| ||No fertilizer||Fertilizer||F1,48||P|
|Fungal PLFA||48·22||39·98||11·01||< 0·05|
|F:B ratio|| 0·07163|| 0·05721|| 7·18||< 0·05|
Table 8. Mean F:B ratios at 4-year intervals (1996–2004) in plots consistently grazed in autumn and spring and cut on 21 July without FYM addition
|Time effect (F2,19·9 = 61·418, P < 0·001)|
| ||No seed||Seed added|| |
|Fertilizer–seed interaction (F1,24 = 11·108, P < 0·05)|
|No fertilizer||0·0474||0·0539|| |
|Fertilizer added||0·0435||0·0369|| |
Figure 6. Redundancy analysis of 2004 plant species composition (solid arrows with small heads), with PLFA (environmental variables represented by dotted arrows with large heads) and management treatments. The latter supplementary variables are represented by symbols: diamonds, FYM; triangles, past hay-cut dates; squares, fertilizer; circles, seed. Only the most important species and PLFA variables are included and PLFA names follow Frostegård, Bååth & Tunlid (1993). Species codes (emboldened codes indicate those species sown into the sward as part of the seed-addition treatment): Alopprat, Alopecurus pratensis; Anthodor, Anthoxanthum odoratum; Bromhord, Bromus hordeaceus; Cardprat, Cardamine pratensis; Dactglom, Dactylis glomerata; Festrubr, Festuca rubra; Holclana, Holcus lanatus; Lotucorn, Lotus corniculatus; Lolipere, Lolium perenne; Planlanc, Plantago lanceolata; Poatriv, Poa trivialis; Ranuacri, Ranunculus acris; Ranurepe, Ranunculus repens; Rhinmino, Rhinanthus minor; Trifprat, Trifolium pratense; Trifrepe, Trifolium repens.
Download figure to PowerPoint
- Top of page
These results show that the effects of phase 1 management treatments on plant species composition did not continue into phase 2 but changed over time and were predominantly a consequence of FYM application, particularly in combination with mineral fertilizer (prediction 1). The particularly large increases in Poa trivialis and Lotus perenne in this combination of treatments in 2002 may have been a response to the foot and mouth disease outbreak in England. The field trial site was within the area where livestock movement was restricted, so cattle were not available for autumn grazing in 2001. Interactions between treatments had a major influence on vegetation in phase 2, just as in phase 1 (Smith et al. 2000), despite accounting for a relatively small proportion of the variation in species composition.
There was the possibility that any post-1998 vegetation change related to FYM application might be dependent upon the species composition of the sward at the end of phase 1 (prediction 2). The anova demonstrated that FYM did not have a significant effect on the vegetation as a phase 2 main treatment but interacted with other treatments to change species composition. The inclusion of time in the fertilizer–FYM treatment interaction might support prediction 2. However, the demonstrated effect of time was not a consequence of linear change through phase 2 but a peak in 2002. This suggests an alternative explanation and we favour the idea that the 2002 change in species composition of the sward was a temporary consequence of the lack of cattle grazing in 2001 because of the foot and mouth epidemic. If this was the case, then phase 2 vegetation change linked to FYM applications was independent of the sward composition in 1998. This may be because of the relatively small, albeit significant, differences in the vegetation in 1998.
The major effects of hay-cut date on vegetation in phase 1 were replaced by a major FYM effect, the effect of cut date on plant species composition progressively declining in phase 2 (Fig. 1). However, there were still some significant but small past cut-date effects on the similarity of the sward to the target MG3b vegetation (Table 2) and a phase 2 cut date–fertilizer interaction for Ellenberg fertility scores (Table 4). Therefore, while the imposition of a common hay-cut date across the trial in phase 2 greatly reduced the vegetation differences created in phase 1, it did not completely remove them after 6 years (prediction 3). The 21 July cut date applied to all plots in phase 2 will have facilitated the spread of Rhinanthus minor to those plots previously cut on 14 June. This will probably be a major factor in vegetation development and facilitate the post-1998 convergence in species composition through the known hemiparasitic effects of Rhinanthus minor on other species (Bardgett et al. 2006) and the possible facilitation of colonization by other species, as demonstrated elsewhere by Pywell et al. (2004).
We detected significant differences in the composition of soil microbial communities between management treatments in 2004. In particular, the ratio of F:B fatty acids was greatest in the absence of FYM and mineral fertilizer, demonstrating a continued development of the soil microbial community towards fungal dominance. This finding is consistent with comparative studies of grassland types, which show that traditional management is associated with fungal-dominated microbial communities with high F:B PLFA ratios, whereas intensive management is associated with bacterial dominance of the microbial community and low F:B ratios (Bardgett & McAlister 1999; Donnison et al. 2000; Bardgett et al. 2001; Grayston et al. 2004). Furthermore, the values of F:B ratios detected in these treatments in 2004 were of similar magnitude to those of traditional meadows (Bardgett & McAlister 1999); this indicates that it is possible to convert bacterial-dominated soil microbial communities of intensively managed grassland systems to fungal-dominated communities, more typical of traditional systems, over 14 years. The increased F:B ratios associated with seed addition in the absence of fertilizer (Table 8) suggests that these below-ground changes might be driven by the above-ground changes in the plant species composition of the sward (prediction 5) (Wardle et al. 2004). While this suggestion of a cause–effect link between vegetation and soil microbial community change is not proved by the data presented here, it remains a possible explanation for the demonstrated associations between plant species and soil fatty acids (Fig. 6) and the F:B ratios found under different treatment combinations (Tables 7 and 8). Also, previous studies have shown strong associations between different plant species of grassland and their microbial communities (Bardgett et al. 1999; Innes, Hobbs & Bardgett 2004), suggesting that changes in the dominance of plant species in the community have the potential to alter the microbial community structure of soil. The establishment of Ranunculus bulbosus and Lotus corniculatus in phase 2, after failure to establish during phase 1 (prediction 6), might be a precursor to further species colonization and increase in sward diversity, after sowing of additional species in 2006 during the next phase of this meadow trial from 2004. However, other factors may have been influential here, such as non-viable seed in phase 1 and the spread of Rhinanthus minor in phase 2.
These results have management implications in terms of the (i) expected time scale for vegetation change, (ii) appropriate treatment combinations, (iii) use of FYM, (iv) vegetation response to change of hay-cut date, (v) role of seed addition and (vi) link with soil microbial communities. (i) Overall, our findings indicate that biodiversity goals for upland meadows need to plan beyond the typical 5–10-year management agreement period of agri-environment schemes. There appears to be a limit to what is rapidly achievable. It has taken 14 years for the apparent fertility (from Ellenberg scores) to decline and the vegetation to become similar to that of the target MG3b community in the most effective treatment combinations, but the species diversity was still well below target. Twenty-year management agreements might need to be the minimum expectation for policy planning.
(ii) Various types of management are applied to meadow grasslands and it is important to apply a combination of a mid-July hay-cut date, autumn grazing with cattle, spring grazing with sheep and no mineral fertilizer. This combination provides a niche for the hemiparasite Rhinanthus minor, with major consequences for the relative abundance of other members of the plant community.
(iii) FYM at the rate applied here (12 t ha−1 every year in winter or early spring) is not appropriate. It is particularly deleterious with mineral fertilizer, even when this is at ESA application rates, when it increases the abundance of species associated with high soil fertility.
(iv) A return to a mid-July cut date after 8 years of early (mid-June) and late (early September) hay cuts can rapidly, within 6 years, change the vegetation to a more traditional composition, particularly when there is a clear and close seed supply of Rhinanthus minor and possibly other species to occupy new niches.
(v) The addition of seed of missing species is essential for an increase in sward diversity. Niches created by appropriate sward management will remain unoccupied by additional species unless they can disperse into the sward naturally or artificially from sown seed or strewn hay. The continued phased sowing of seed is probably a sound long-term strategy, although the Colt Park experiment does not yet provide direct evidence for this.
(vi) We have demonstrated that soil F:B ratios increase over time in the absence of fertilizer and in appropriately grazed and cut plots without FYM. Future increases in species diversity might require further changes in soil microbial communities, currently linked to the growth of Lotus corniculatus, Trifolium pratense, Trifolium repens, Anthoxanthum odoratum, Ranunculus acris and Rhinanthus minor at Colt Park. We have no data on the mechanisms involved. We have yet to prove it, but grassland restoration might need to be viewed as a long-term succession from species-poor to species-rich grassland that requires facilitation of the fungal component in soil microbial communities. Early successional facilitator plant species may then provide niches for successful colonization by mid- and late successional species to recreate traditional species-rich swards in the long term (> 20 years).