162 C.N.R. Critchley, J.A. Fowbert Agriculture, Ecosystems and Environment 79 2000 159–174
establishment method, region, site age and distance from field boundary were fixed factors, and distance
was the repeated measure. Field margin sites 11 sites with shorter transects were omitted. Species
richness data were transformed to log
10
x+1 before analysis.
Relationships between vegetation and environmen- tal factors were analysed for 36 naturally regener-
ated sites for which management information could be obtained from farmers, using Canonical Correspon-
dence Analysis CCA ter Braak, 1987. Within-site species percentage frequencies were calculated, using
the 30 quadrats from set-aside blocks and 25 from field margin set-aside, respectively. All species and amal-
gams were included except bryophytes and unidenti- fied seedlings. Environmental data were region, soil
class, previous crop in the year before set-aside estab- lishment, management of set-aside vegetation in the
previous 7 months and age of each site in years. Soil data were derived from the Soil Survey and Land Re-
search Council dominant soils map of Great Britain at the 1 km scale, and re-classified as described by
Firbank et al. 1998. Management activities grazing, cutting and herbicide application were those carried
out during the 7-month period preceding the field sur- veys September 1995–May 1996. No sites had been
subjected to cultivation.
Stability of the analysis under re-ordering of data was checked as recommended by Oksanen and
Minchin 1997, and found to be satisfactory. En- vironmental variables were added individually by
forward selection, and significance testing done for each variable in turn, and for the overall trace, us-
ing 999 Monte Carlo permutations ter Braak, 1988, 1990. Only environmental variables that showed a
significant relationship with species were added to the model.
3. Results
Vegetation was established by natural regeneration at 45 sites and by sown cover at 52 sites Table 1. The
ages of the set-aside ranged from 1 to 9 years with all intermediate ages represented but not necessarily
for each establishment type. Eleven sites were field margins seven in the mixed region, four in the arable
region.
Table 1 Numbers of sites surveyed by age, region and establishment
method Age years
Mixed Arable
Total Total
Total Total
mixed arable
G
a
N
b
G N
G N
1 1
1 2
1 3
2 2
2 5
2 1
5 3
7 1
3 9
9 2
7 18
9 11
16 4
8 6
7 3
14 10
15 9
5 5
4 9
5 4
6 1
1 1
7 2
1 2
3 3
5 4
4 8
4 2
3 3
6 6
7 5
9 1
1 2
2 2
1 3
Total 30
19 22
26 49
48 52
45
a
G=sown cover.
b
N=natural regeneration.
3.1. Species composition and cover components A total of 229 species or amalgams were recorded
Appendix A. Of the nine species occurring in over 50 of sites, six were grasses; Poa trivialis was found
at 90 and L. multiflorumperenne at 88 of sites. The other most frequently found monocotyledons
were Elytrigia repens, P. annua, Agrostis stolonifera and Holcus lanatus. Cirsium arvense was the most
frequently recorded dicotyledon, found at 65 of sites, followed by Ranunculus repens and Trifolium
repens. However, the majority of dicotyledons were found at very few sites; 140 species were recorded
from less than 10 of sites and 91 species from less than 3. Only one nationally rare Stace, 1991
and 15 declining Rich and Woodruff, 1996 species were recorded Appendix A. Crop volunteers were
recorded but only at a minority of sites; the most fre- quent was Triticum sp. in 10 of sites. Bare ground,
litter and seedlings were very widespread, found at 95, 92 and 92 of sites, respectively.
Mean percentage plant cover was 94.0 S.D. 1.64, with strawlitter 4.9 S.D. 1.45 and bare
ground 1.2 S.D. 0.38. Monocotyledons mean 81.4; S.D. 2.02 and perennials mean 81.5; S.D.
2.40 accounted for the majority of the total cover. Mean species richness 24.6; S.D. 8.40 and vegeta-
tion height 19.1 cm±12.60 S.D. showed relatively high variation among sites.
C.N.R. Critchley, J.A. Fowbert Agriculture, Ecosystems and Environment 79 2000 159–174 163
3.2. Effects of establishment method, region and site age
No differences were detected in the multivariate tests for the main cover components between regions
Rao R
3,87
=1.75, n.s., establishment method Rao R
3,87
=0.35, n.s. or site age Rao R
3,87
=1.16, n.s., nor were there significant univariate F tests for
any of the individual dependent variables. There were significant differences in the multivariate tests for
cover of species functional types between regions, es- tablishment methods and site age Table 2, but only
for site age in relative frequency of species functional types Table 3. None of the interaction terms were
significant.
Differences in individual dependent variables uni- variate F tests for species functional types were
detected for all main factors Tables 2 and 3. Sites in the mixed agriculture region had more cover of
dicotyledons, greater cover and relative frequency of basal species, and lower cover of arable species than
those in the arable region. Sites established by natural regeneration had greater cover and relative frequency
of annuals and leafy species, lower cover of peren- nials, and greater relative frequency of non-arable
species than did those with sown cover. Older sites had more cover of leafy species and monocotyledons
and greater relative frequency of perennials, leafy species, and non-arable species than younger sites.
Younger sites also had greater relative frequency of arable species. In the arable region, young sites
had greater relative frequency of annuals than old sites, but the converse was true in the mixed region
region×age F
1,89
=4.16, p0.05. Perennials in the arable region also had higher relative frequency in
older sites, but in the mixed region there was little difference between the two age classes region×age
F
1,89
=4.13, p0.05 Fig. 1. The regions, establishment methods and age classes
were strongly differentiated by the individual species analysed Table 4. As expected, the agricultural
species L. perenne, T. repens were much more fre- quent in sown cover. The three species found more
frequently in the mixed region than the arable region A. stolonifera, R. repens, T. repens were perennials
which typically occur in relatively moist, mesotrophic grasslands Grime et al., 1988. E. repens, more char-
acteristic of fertile, disturbed habitats, was more fre- quent in the arable region. The only annual analysed
P. annua was more frequent in younger sites, with four of the other perennial species more frequent in
older sites.
Species richness declined progressively with in- creasing distance from the field boundary Rao
R
5,74
=4.45, p0.005 and was significantly higher in older sites than younger sites F
1,78
=17.37, p0.0001 Fig. 2.
3.3. Environmental factors Within the sub-sample subjected to CCA, there was
approximately equal representation of sites from the two regions Table 5. Only 14 sites had been subject to
some form of management during the specified period. The most common soil class was loamy brown soils
15 sites, other soil classes being represented by only a few sites each. Winter wheat was the commonest
crop preceding set-aside 16 sites, with four or fewer sites for each of the other crops.
Although age of site, cutting and groundwater gley soils had a significant relationship with the vegetation,
most variation in the vegetation was not accounted for Table 6. Many perennial species typical of grass-
lands e.g. Alopecurus geniculatus, Carex hirta, Fes- tuca pratensis, Lotus pedunculatus were associated
with greater age of sites Fig. 3. However, a small number of annuals e.g. Tripleurospermum inodorum
were also associated with older sites. Although both cutting and groundwater gleys were significant factors
in the analysis, no ecologically important patterns as- sociated with either were discernible from biplots of
the various combinations of axes 1–4. For example, species associated with groundwater gleys included
some normally associated with dry, sandy soils e.g. Spergula arvensis, Erodium cicutarium.
4. Discussion