Less than 14% of land in the catchment of the Fleet is non-agricultural. Therefore, it is not surprising that the most significant influence upon changing nutrient loads to surface waters since 1866 has been agriculture. Local trends in land use change, cropping practises, and livestock numbers and inorganic fertiliser use, provide a basis for interpreting changes in the trophic status of the Fleet. These changes, especially the strong growth in nitrogen and phosphorus export to standing waters after 1945, may also be viewed in the context of UK national agricultural history.

The first half of the period studied was one of general economic stagnation 343536, with consequently very little real rise in nutrient inputs from agriculture to the waters of the Fleet. Increases in nitrogen inputs to the lagoon between 1946 and 1986 can then be explained by three local changes which reflect parallel national trends. These are

1. a steady rise in stocking rates of cattle, and recovery of sheep numbers from their minimum 1951, which led to greater amounts of FYM being applied to the land.

2. an increase in land under production, with a decline in the area of rough grazing, and a rise in cereal cropping, leading to intensification of operations.

3. a steep rise in amounts of inorganic nitrogen fertilisers applied to all agricultural land, and in particular to temporary and permanent grass, producing an increase in inputs of inorganic nitrogen.

Areal nitrogen loadings on the Fleet for 1866–2002, are expressed using Vollenweider's first model 37, which explores the relationship between nutrient loading and water body mean depth (Figure 10). Values are shown (in the original model) for bodies which are 'oligotrophic' (in this context, not overloaded with nutrients), and those which are naturally or artificially 'eutrophic' (i.e. overloaded by nutrients, in relation to bodies of that area). Owing to the very high nitrogen loadings of the Fleet, only part of the original model is used here. Loadings are shown for (a) the Fleet as a whole, (b) the West Fleet (the much less tidal part of the lagoon), and (c) Abbotsbury Bay, which receives ca 38% of the total surface drainage (see Site description, above). Rather than gross annual nitrogen loading, Hydrologically Relevant Load (HRL; 38, see Methods, below) is used in order to represent the effective areal loading of the element.

According to this result, the Fleet as a whole has been grossly overloaded with nitrogen throughout the period of study. This is mostly a consequence of its very small mean depth (<0.5 m), along with its very high areal loading (ca 3 – ca 7 g m^-2). Little change occurred until 1945, after which loading increased rapidly (with a slight decline for 1961) until the peak year of 1986, before falling slightly by 2002. In 1946, total annual nitrogen HRL was ca 48 t, whereas by 1986, it had risen to ca 100 t, an increase of 104%. This finding clearly illustrates the local effects of the general late 20th century UK agricultural intensification.

By 2002, as a result of declining numbers of cattle, a reduction in the area of cereals and of temporary grass, and a fall in the rate of inorganic fertiliser application to permanent grassland, loads had returned to those of the early 1970s. It is likely that at least some of this reduction is due to the recent implementation (in the Abbotsbury area) by the Ilchester Estate (which owns much of this part of the catchment), of a Countryside Stewardship scheme, a programme designed to reduce intensity of impact of farming on the environment 39. Non-disclosure of statistics may also be involved, however, with pigs recorded only once between 1988 and 2002. Also, for the period 2000–2002, between 150 and 227 ha were listed as 'unclassified'.

Nitrogen loadings for the West Fleet exhibit similar trends, with little increase between 1866 and 1966. Values for the 1970s reached 10 g m^-2 yr-1, and, during the 1980s and 1990s, a peak of ca 13 g m^-2 (1986). By 2002, they had returned to ca 10 g m^-2 yr-1. Loadings for Abbotsbury Bay are very high indeed, varying between ca 16 and 21 g m^-2 yr-1 for 1866–1966, rising to ca 28 g m^-2 by 1971, and then ca 36 – ca 44 g m^-2 yr-1 for the late 20th century. Values have now (2002) fallen somewhat, close to those of the early 1970s.

Vollenweider's second model 40 predicts the trophic status of a water body from mean areal phosphorus loading versus mean depth, divided by water residence time (Figure 11). Values for the last for various parts of the Fleet are estimated from 14 and 15, but may only be approximate. Loadings are defined in terms of critical limits 3840, these being (a) 'permissible' – that loading unlikely to force oligotrophic bodies to change from their present state, and (b) 'dangerous' – that likely to continue to force eutrophic bodies to change 41. Again, rather than total annual load, HRL 38 of phosphorus is used here in order to calculate critical loadings of the element.

Phosphorus loadings for the Fleet as a whole remained within permissible limits throughout the period 1866–1981, only exceeding that value during the interval 1986–1988. As with nitrogen, the last decade (1990–2002) has seen total phosphorus loads decline slightly, approximately to those of the 1970s. Loadings on the West Fleet remained higher than those on the Fleet as a whole, but stayed within permissible limits, until the 1940s. They then increased consistently, reaching a maximum, close to dangerous, during the late 1980s. During the last decade they have fallen, but still remain above permissible limits.

Abbotsbury Bay has consistently received the greatest areal loadings of both nitrogen and phosphorus. Phosphorus loads upon the Bay were already greater than permissible by 1866, and exceeded dangerous during the 1950s, again reaching a maximum during the 1980s, with a decline during subsequent decades. Phosphorus loadings on Abbotsbury Bay would need to be reduced to those of the 1950s in order to fall to dangerous values, and to those of the early C20th even to begin to approach permissible limits.

Phosphorus loads on the Fleet first rose as inorganic fertiliser applications reached a peak during the early 20th century. The most significant increases, however, occurred between 1946 and 1988, when annual HRLs rose from ca 0.5 t yr-1 to ca 0.9 t, an increase of 175%. This rise is explained by two changes in local agricultural practice,

1. an increase in inorganic fertiliser applications to all crops, and grassland, which peaked during 1951, and again in 1971, and

2. a sustained increase numbers of sheep and cattle, and fluctuating pig numbers, which generated peaks of organic phosphorus applied as FYM in 1956, and in 1988.

Together, these changes led to an initial post-1945 increase in phosphorus loading from inorganic fertilisers, and then a later, separate rise in organic inputs from livestock. FYM is the main source of phosphorus to the Fleet in nearly all years of the study, exceeded by the contribution of inorganic fertilisers only during the intervals 1946–1951, and 1961–1971, when application rates rose to maximum values.

After 1988, total phosphorus loads declined, mainly as a result of a reduction in cattle numbers, a contraction in the area under cereals, and a fall in rates of inorganic fertiliser applications to these crops. Again, this change may at least be partly due to implementation of the Countryside Stewardship scheme at the Abbotsbury end of the catchment. Phosphorus loadings on the Fleet as a whole have returned almost within permissible limits, but not quite. As with nitrogen, non-disclosure of pig numbers, and omission of a small proportion of agricultural land from official statistics, may play a part in reducing the modelled total phosphorus load.

The contribution of sewage effluent to total phosphorus load has become increasingly significant since Abbotsbury STW opened in 1966. By 2002, human load on the Fleet had risen by 165% (from ca 0.1 to ca 0.3 t yr-1), contributing approximately 14% of total load. This increase was due mainly to three changes, namely

(a) a marginal rise in local population,

(b) growth in use of phosphorus-based detergents, and

(c) expansion of sewered population, with the commissioning of Langton Herring STW in 1981.

Although human load is small in relation to that from agriculture, it became critical during the years 1986–1988, when permissible limits for areal loading on the Fleet as a whole were exceeded.

As indicated, nitrogen contributions to the Fleet from wildfowl are not particularly important. For phosphorus, however, wildfowl currently contribute a significant proportion of the total budget of the Fleet (ca 35%, or ca 7 × that from sewage effluent). Of this, Abbotsbury Swannery currently produces ca 4.5% of the total phosphorus budget, which is broadly equivalent to the present human input.

The Swannery is therefore currently probably not a major source of nutrient loadings to the lagoon, and historically, its contribution may have been less. One caveat, however, is that introduction of increased phosphorus inputs (both from agriculture, STWs and increased numbers of swans) into the relatively confined waters of Abbotsbury Bay, may be responsible for some of the ecological changes recently observed (e.g. decline of eelgrass 21). The very short water residence time of the Bay (ca 12.5 days) helps counteract such effects. The bird populations of the Fleet as a whole, are clearly an important source of phosphorus, which for an aquatic site of such great nature conservation value, is not particularly surprising. Unlike sewage effluent, they are also, especially the Abbotsbury swans, part of the inherent value of the Fleet 13, and hence its attraction to human visitors.

Differences in loadings between the three sections of the Fleet, can be explained in terms of catchment hydrology, location of point sources, and morphometry of the lagoon. Approximately 1000 ha (38% of the catchment) drains into Abbotsbury Bay, which itself covers only 28 ha (6% of the Lagoon). Hence, in areal terms, the Bay receives by far the greatest annual loadings of any part of the Fleet. Inputs from Abbotsbury STW (the larger of the two discharging into the Fleet) swell this load further, particularly in terms of phosphorus. The narrowness of the catchment from Abbotsbury to the inlet at Smallmouth (Figure 1) produces lower areal loadings from diffuse sources for the West Fleet and Fleet as a whole. Variations in bathymetry along the Fleet, and the weak influence of tidal flushing compared to that at Smallmouth, indicate that loadings upon Abbotsbury Bay are likely to be the most critical, owing to the attenuated influence of the tidal cycle at the western end of the lagoon.

Few previous hindcasting studies have been carried out for the Fleet 4243. Mainstone & Parr 33 used export coefficients in order to model nutrient loads for 1997. Like us, they concluded that agriculture is the main source of nutrients for the Fleet, accounting for >80% of total annual nitrogen, and 56–69% of total annual phosphorus load. Equivalent values from our study (for 2000), are (respectively) 74% and 77%. Although not directly comparable, owing to differences in methods and data sets used, these estimates are broadly similar. Loadings for 2003, calculated on the basis of a detailed land use survey by Hudson 32, are ca 84–98 t nitrogen, and ca 4 t phosphorus.

Historical land use changes in the catchment of the Fleet are similar to those recorded by Johnes et al. 44 in their hindcasting study of ten British river catchments, and also those described by O'Sullivan 6845 in his investigations of Slapton Ley (Devon), Loe Pool (Cornwall) and Bosherston Lakes (Dyfed). Here, intensification after 1945 led to increases in livestock numbers, increased use of inorganic fertilisers, and removal of rough grazing. However, at all sites, there was also evidence of an earlier, late 19th, early 20th century agricultural intensification, and consequent eutrophication, associated with completion of the UK branch line railway network, and connection of previously very rural locations to the national market, particularly in dairy products.

In the data for the Fleet, there is less evidence of this 'railway effect', even though a local branch line did indeed reach Abbotsbury in 1885 46. However, the main line to nearby Weymouth (Figure 1), with several local stations located close to the eastern edge of the catchment of the Fleet, had already been built by 1857, so that, by 1866, the area draining into the Fleet had already been connected to the national market. Any local 'railway effect' may therefore precede the beginning of collection of the data we have analysed here.

Overall, our results suggest that, even before 1866, the Fleet was already grossly overloaded with nitrogen, and that, in Abbotsbury Bay, phosphorus loadings had already exceeded permissible levels. Tithe maps and apportionments in the Dorset Record Office 4243 and records of the Ilchester Estate 28, offer important documentary evidence of earlier historical agricultural practices and land use within the catchment of the Fleet, which could be used to hindcast nutrient loads further.

Vollenweider's first and second models 3740 may also be used in order to evaluate proposed strategies for alleviation of eutrophication 68. Here, three measures designed to mitigate nutrient loads upon the Fleet are modelled (Figures 12, 13; Table 7), namely

(a) implementation of 5 m riparian buffer strips along all watercourses in its catchment, and around the perimeter of the lagoon,

(b) introduction of phosphorus-stripping at both STWs discharging into the Fleet, and

(c) both 31.

The first measure focuses on diffuse loads of both nitrogen and phosphorus; those which are mainly delivered to surface waters in particulate form during winter, by increased runoff. The second is aimed at lowering phosphorus loads from point sources (i.e. STWs). These are delivered mainly in soluble form, and at constant rates throughout the year (with a possible increase in this case for the tourist season), and therefore tend to reach peak concentrations in inflowing rivers during the growth season. Both kinds of source are important, however, for estimating loadings to standing waters 3347, where nutrients delivered in particulate form are stored as sediments, and eventually become available for growth via resuspension and remobilisation. Values for buffer strip efficiency were taken from 48 for nitrogen, and 49 for phosphorus. The retention rate of phosphorus stripping at STWs is that used by Johnes 50.

Modelling of results for introduction of 5 m buffer strips for nitrogen (Figure 12) shows that it would be possible to reduce current (2002) non-point loadings upon the Fleet (ca 79 t) approximately to those of 1951 (ca 50 t), a fall of 37%. This is a considerable reduction, but would still leave the lagoon grossly overloaded with the element, emphasising the magnitude of increases in nitrogen loadings since 1945. Reductions on the West Fleet, and on Abbotsbury Bay, are of the same order (36–38%). Unlike his second model 40, Vollenweider's first 37 does not take residence time into account, but even when we include this variable in our calculations, we find that diffuse nitrogen inputs to each part of the lagoon continue to exceed dangerous loadings.

The effect of introduction of 5 m buffer strips on diffuse inputs of phosphorus (Figure 13) is greater (43–49% decrease), with loadings on the Fleet as a whole, and on the West Fleet, approaching those of 1866, leading to reduction of inputs to the West Fleet below permissible loading. In Abbotsbury Bay, the effect is attenuated, owing to the residual influence of current sewage inputs, which would prevent reduction of loadings below those of 1916, a value still beyond the dangerous limit for this part of the Fleet.

Installation of phosphorus stripping at Abbotsbury and Langton Herring STWs (Figure 13) would lead to a 27% reduction in point source inputs to the Lagoon (Table 7), to ca 1.3 t yr-1, a loading equivalent to that of 1951. As loadings of this element to the Fleet as a whole already lie within permissible limits, such reduction would be beneficial, but not critical. For the West Fleet, the equivalent reduction is 20% (to ca 1 t yr-1), which would still leave loadings on this part of the lagoon just beyond permissible limits. For Abbotsbury Bay, as calculated here, the reduction achieved is also 27%, but marginal effects are much less, with inputs to this part of the lagoon remaining beyond dangerous on the basis of phosphorus stripping alone. Phosphorus stripping may still be an important option to consider for installation at Abbotsbury STW, however, in that the works may have been operating 'above consent' for some years 20.

Joint implementation of both phosphorus stripping and establishment of 5 m buffer strips (Figure 13) would achieve further reductions in phosphorus loadings (66–77%) on all parts of the lagoon, and would serve to reduce inputs to the West Fleet to values further within permissible limits. However, in the case of Abbotsbury Bay, loadings would still lie beyond permissible limits. Such measures might still be worth considering, as the extent of any further reductions in phosphorus loading required, in order to bring this part of the lagoon within permissible limits (e.g. a voluntary ban on use of phosphate detergents), would be quite modest. Even greater reductions could be achieved, for example, by diversion of all sewage inputs from the Abbotsbury end of the Fleet, and further 'extensification' of agriculture.

Data on agricultural land use and livestock numbers for the catchment of the Fleet for the period 1866–1988 were compiled using annual Parish Summaries of the June Agricultural Census (England and Wales) located at the Public Record Office, Kew, London. Nine civil parishes, records from which were used to develop data on land use history, lie partly or mainly within the catchment (Table 8). Of these, Abbotsbury, Langton Herring and Fleet make up more than 70% of its total area.

The catchment model employed is that of Jørgensen 52, as adapted by 5150 and 44. Estimates of external nutrient loads upon a waterbody are calculated by summing outputs from all potential diffuse and point sources of nitrogen and phosphorus within and over the catchment, as in

(a) natural (i.e. diffuse/non-point) losses

n

LP = Σ·A × EP (where EP = Ep_f + Ep_ag + EP_u)

i = 1

where LP = loss of phosphorus (kg ha-1 yr-1)

A = area of watershed (ha)

EP = export coefficient for phosphorus, where:

_f = forest, _ag = agricultural land, _u = urban areas

(b) direct input from precipitation

IP = CP × A₀

where IP = input of phosphorus

CP = P concentration in precipitation (mg l^-1)

A₀ = area of water body (ha)

(c) load from sewage

LP = (D_caP × H × 365 × M × B × R_S)/10⁶

where LP = phosphorus loss (t a^-')

D ca = phosphorus discharge person^-1 (g day^-')

H = sewered human population

M = loss during mechanical treatment (10–15%)

B = removal during biological treatment (10–15%)

C = removal during chemical treatment (80–90%)

Rs = retention coefficient of filter bed (1–88%)

Here, the original model has been modified in three ways. First, estimated contributions of animal wastes, and of organic and inorganic fertilisers to nutrient loads exported from agricultural land, are quantified separately. Second, agricultural land is not treated as an homogenous unit. Instead, agricultural land use types are differentiated, reflecting variations in predicted rates of nutrient application rates, and hence of nutrient export. These modifications were initiated by 51, and further developed by 50 and 44. A third modification is that we have introduced coefficients for inputs from wildfowl (including waterfowl), in order to take into account the importance of these populations for the Fleet.

Amounts of inorganic fertiliser applied to agricultural land use types within the catchment of the Fleet were calculated using data from four principal sources. These are: -

1. National average crop applications for the period 1976–2002 from the British Survey of Fertiliser Practice (BSFP; 57). Between 1942 and 1991, these were compiled by Rothamsted Experimental Station, Harpenden, Hertfordshire, and from 1992 until 1998, by the University of Edinburgh Data Library. In 1999 they became the responsibility of the Rural Business Unit of the University of Cambridge.

2. A report on fertiliser practice in England and Wales presented to the "Closed" Conference of Advisory Soil Scientists Fertiliser and Lime Committee by Church & Webber 58, containing trends for major land use types for the period 1957–69.

3. National trends in consumption of total inorganic nitrogen and phosphorus fertilisers for the period 1913–1956, published in Fertiliser Statistics of the Fertiliser Manufacturers Association [FMA; 59].

4. National trends in artificial fertiliser use for the period 1851–1956, as derived by Brassley 60.

The BSFP uses a stratified sampling procedure in order to produce regional and national statistics for fertiliser application rates. Owing to lack of regional data for West Dorset, application rates of nitrogen and phosphorus fertilisers used here are based on national figures. Mean application rates for all cereals, all arable (including bare fallow and market gardening), temporary grass and permanent grass, were derived from BSFP national statistics for individual crop types and land uses for the years 1976–2002. Church & Webber 58 did not differentiate between cereals and arable, however, so their summary statistics refer to all tillage. For the period before 1976, it was not possible to derive separate cereal and arable application rates. National trends in inorganic fertiliser use reported by 59 and 60, were used in order to extrapolate application rates for the years 1866–1951.

Where values fall below 1 kg ha-1 yr-1, applications were considered to be nominal, and figures for previous years not calculated. All application rates are given as total nitrogen (kg N ha-1 yr-1), and total phosphorus (kg P ha-1 yr-1). Total inorganic phosphorus applied was calculated assuming that 1 kg of inorganic fertiliser phosphorus pentoxide (P₂O₅), contains 0.4327 kg of phosphorus. Appropriate export coefficients for losses of inorganic fertiliser from agricultural land (Table 9) were selected from 4514944 and 54.

Annual amounts of nitrogen and phosphorus produced by livestock in the catchment of the Fleet were calculated using figures derived by 51 from a comprehensive review of literature sources (Table 10). Goats – not included in their study – were assumed equivalent to sheep 37. Unlike inorganic fertilisers, animal waste is not applied in total to the land. Instead, some is voided directly whilst animals are grazing, whereas that dropped in stock houses is collected, stored, and applied as farmyard manure (FYM) or slurry. Coefficients derived by 50 are therefore calibrated in order to account for losses of volatile forms of nitrogen (mainly ammonia) during storage, and uptake by crops.