122 N.E. Ellis et al. Agriculture, Ecosystems and Environment 76 1999 121–134
1930 Fuller, 1987, the remaining grasslands being further impoverished of species through drainage,
reseeding, an increased use of agro-chemicals and in- creased stocking densities Hopkins, 1988. Through-
out Europe areas under extensive farming regimes are continuing to diminish Bignal and McCracken,
1996.
During the 1980s, a change in European agricultural policies began to encourage a reduction in agricultural
production through the use of quotas, set-aside arable land and grazing extensification schemes. Whilst some
farm households have left farming others, affected by a consequential decline in farm income, have become
involved in non-agricultural activities to compensate, at least within the UK Cox et al., 1989. Factors deter-
mining the uptake and type of non-agricultural activi- ties are complex which are not always associated with
economic conditions e.g. Bryden, 1994; Edmond and Crabtree, 1994. Off-farm work has been associated
with hobby farmers Gasson, 1988 whereas on-farm diversification has been found to be more likely to
occur along tourist routes at least in Scotland; Ed- mond et al., 1993, where farm households are gen-
erally more qualified Corcoran and Dent, 1994 and the farm plays a strategic role in the income generated
by the household Munton et al., 1989.
Whatever the cause for the uptake of pluriactiv- ity, the rate of uptake in certain areas has increased
markedly. A socio-economic survey of 295 farms in the Grampian region of Scotland showed that the up-
take of pluriactivity had trebled between 1980 and 1991 Ellis, 1994; in Wales, the uptake of off-farm
work was greater after 1986 than in the 5 years be- fore Bateman and Ray, 1994. In 1992, this shift in
farm labour resources was recognised in the provi- sion of aid to farmers for less-intensive farming and
involvement in non-farming activities under the EC Agri-environmental package Regulation No 207892
and is now being further addressed in the latest agri- cultural reform package, Agenda 2000.
Land use systems directly determine the types and mosaic of the land cover across the farm whilst the
level of management within each land use deter- mines the type and number of plant species within
Fig. 1. For example, a purely agricultural man- agement plan may aim to maximise all profitable
land, managing intensively with maximumoptimum agro-chemical inputs and livestock numbers. In the
event of alternative employment pluriactivity, the land management plan may differ according to where
the resources of the household are redirected. In the case of off-farm employment off-farm pluriactiv-
ity, the management plan may be simplified with less emphasis on the use of all land and managed
with fewer resources both time-wise and financially through reduced agro-chemical inputs and fewer live-
stock. The result of a less intensive regime may, therefore, allow the establishment of a greater num-
ber of plant species, whilst in less accessible corners of farmland, semi-natural habitats such as scrub
and wetter areas may increase or appear. Gasson 1988 associated off-farm work with less time and
labour demanding systems such as beef, sheep and cereals rather than the more demanding dairy cows
and non-cereal enterprises. In the case where alterna- tive enterprises are developed on the farm on-farm
pluriactivity, the pluriactivity may diversify the land use e.g. alternative crops, golf courses, horse riding
although some forms of on-farm diversification do not affect the land for example, tourist accommo-
dation and farm shops. With on-farm diversification the intensity of land management may not be notice-
ably reduced because, with the household remaining on the farm, a more efficient use of time may be
made.
The hypotheses above indicate a line of influence from farm household socio-economic characteristics
to land management through to ecological charac- teristics. This study initially ignored the middle of
this line of influence and looked directly for a rela- tionship between the involvement in pluriactivity the
socio-economic aspect and the botanical characteris- tics of grass fields the ecological aspect. Only later
were the components of the socio-economic and land management data used to identify the order of influ-
ence of such on grass sward composition.
2. Data collection
2.1. A socio-economic survey of pluriactivity in Grampian
The extent and types of pluriactivity occurring in the Grampian region of Scotland Fig. 2 were identified
by using a socio-economic survey on 295 farms in
N.E. Ellis et al. Agriculture, Ecosystems and Environment 76 1999 121–134 123
Fig. 1. The theoretical relationship between factors affecting a farm household and their decisions on the use of land, outlining the knock-on effects to landscape and species.
Fig. 2. The location of Grampian region within Britain.
1991. The sample of farms was randomly selected by the Farm Type SOAFD, 1992 and by farm British
Size Unit BSU groups, excluding farms smaller than 4 BSUs and farm estates one BSU corresponds to
2000 ECU of standard gross margin; SOAFD, 1992. Farms were, therefore, selected across a range of land
uses and economic categories.
Each farm household was interviewed to obtain data on the extent of the farm, farm ownership, the com-
position of the household, qualifications, income, the number of labourers employed to work at the farm and
the feasibility of family succession of the farm. Data were also collected on any non-agricultural activities
providing additional income to the household, such as: the household members involved, the time allocated
to farmingnon-farming activities by each household member and the location of the work i.e. whether it
was on or off the farm. Using the information col- lected about the location of the pluriactivity, four farm
groups were recognised
1. farms where no member of the household was involved in non-agricultural activities, termed
Non-pluriactive ;
2. farms where at least one household member was working off the farm but there was no on-farm
non-agricultural activity the Off-farm group; 3. farms where there was at least one non-agricultural
activity such as a Bed and Breakfast, farm shop, caravan sites, livery, etc. but no-one had an
off-farm job - the On-farm group; 4. farms where members were working off-farm and
there were also on-farm non-agricultural activities - the Both group.
2.2. The botanical survey Seventy-one farms were selected out of the 295
farms visited with a socio-economic questionnaire in order to obtain details on the land use and the botanical
characteristics of the four farm groups. Seventy-one farms was the maximum number that could be botan-
ically surveyed by two people over two summer sea- sons 1991 and 1992.
The initial aim was to stratify this sample by an equal number of farms from each of the farm groups.
124 N.E. Ellis et al. Agriculture, Ecosystems and Environment 76 1999 121–134
Table 1 The number of farms used in the ecological survey in Grampian Region, Scotland, stratified by the location of the non-agricultural
activityies of the household and by Land Class Land
Farm group Class
a
Non-pluriactive Off-farm
On-farm Both
TOTAL 25
Lowlands with variable use, mainly arable 6
6 4
4 20
26 Fertile lowlands with intensive agriculture
6 4
4 2
16 27
Fertile margins with mixed agriculture 7
4 3
3 17
28 Fertile margins with heterogeneous land use
6 5
2 5
18 TOTAL
25 19
13 14
71
a
The Land Classes were defined by geology, relief, climate and human artefacts. The landscape descriptors were later obtained through field survey see Bunce et al., 1996 for further details.
However, a delay in the start of the socio-economic survey meant that the botanical survey started before
all data on the 295 farms had been collected which resulted in an unequal number of farms in each group
although random sampling within each group was maintained. A second stratification of the sample was
also used to reduce any potential bias of pluriactivity to the physical characteristics of the land. This was
achieved by selecting the farms in each group equally across four Institute of Terrestrial Ecology ITE Land
Classes Table 1. These Land Classes have been de- fined by a computerised database which has classified
each 1 km square within Britain into one of 32 Land Classes using data on climate, topography, geology
and human artefacts such as the length of main and minor roads in the square Bunce et al., 1996. The
four selected Land Classes account for 60 of the total Grampian area the lowland area the rest of the
region being mountainous and therefore not under pri- mary agricultural use. Lowland Grampian covers in-
tensive arable areas towards the east and north coasts, and marginal upland areas where heather-dominated
land mainly found in the west of the region has been reclaimed by reseeding and used for
livestock.
The boundaries of the 71 farms were copied onto 1 : 10,000 scale maps before the botanical survey visit
using the sketches made on 1 : 25,000 scale maps by the farm household during the socio-economic survey.
On arrival at the farm, a short interview 10–20 min using the maps was undertaken with the farmer or
manager to obtain data on the use and management of each grass field: a whether the field had been per-
manently grass or part of an arable rotation over the previous 10 years and b how often the grass had
been reseeded in that time, c the main uses of the field over the previous 4 months, d how much inor-
ganic N had been applied over the last 12 months, e whether organic manures farmyard manure, slurry or
silage effluent or f herbicides been applied in the last twelve months and, g if the field has been stocked
during the last 12 months, whether it had been stocked to its potential according to the farmer’s perception.
To minimise the interview time, farmers were asked to indicate which management category each field fell
into rather than obtaining specific data; for example, it was asked whether the field had received 0, 1–149,
150–249 or 250 kg N ha
−1
in the last twelve months rather than the exact amount of inorganic N. The field
management categories were taken from work pre- viously done on British grasslands Hopkins et al.,
1985.
Botanical data were collected from grass fields using 10 quadrats at each farm. Quadrats were ran-
domly positioned on the farm map before the visit. At the farm, quadrats were located as close to the
pre-determined positions as possible but always at least 3 m from the edge of the field to minimise the
influence of the field boundaries. About 5 were found to be positioned outside grassland but were
relocated to the nearest grass field along a cardinal point from the initial location on the map i.e. either
due north, east, south or west, at least 3 m from its boundary. Full species lists were made within each
2 m × 2 m quadrat and an estimate made of each species percent cover. Nomenclature of vascular plants
follows Stace 1991 and of bryophytes Watson, 1981.
N.E. Ellis et al. Agriculture, Ecosystems and Environment 76 1999 121–134 125
2.3. Data management and analyses Two-way analyses of variance ANOVA using the
farm as the unit were used to determine differences between the farm groups. Post hoc testing between
each pluriactive group and the non-pluriactive group used Dunnett’s test in SAS SAS Institute, 1985. Data
expressed as percentages were analysed after angular transformation.
Associations between pluriactivity and socio- economic and land management characteristics were
made using principal components analysis PCA in the GENSTAT package using the correlations
method option Payne et al., 1987 Table 2. The socio-economic and land management data were
thereafter analysed with the species data separately socio-economic with species data, land manage-
ment with species data because of the otherwise high ratio of variables to farm sites. These asso-
ciations were made using Redundancy Analysis RDA within the CANOCO program Ter Braak,
1988, taking all defaults. Two variables, the pres- ence of on-farm pluriactivity and the percentage time
given to agricultural activities, were omitted due to collinearity. Statistical tests of significance on the
RDA axes and ordinated variables were carried out by Monte Carlo tests using 99 random permuta-
tions. With the relatively high number of variables for the number of farms, significance levels between
p
0.05 and p 0.1 were considered as weakly sig- nificant, particularly as the number of farms was also
low.
3. Results