Scottish Government

Appendix V Method applied to Phase II

1. Introduction

The aim of Phase II of the project was to estimate the extent to which diffuse pollution will be managed by 2015. This required the application of different mitigation measures to common land uses, and an estimation of the likely uptake of each mitigation measure resulting in an overall impact (management outcome) on the pollutant load, as described in other sections of the report.

Here, we set out the methodology adopted to provide quantitative but only indicative estimates of the diffuse nitrate, phosphorus and sediment loss from agricultural land that reflect forecast changes in the structure of the agricultural sector and land management practices. It was done through the utilisation of previous work ( WFD19/77 and the Business as Usual III ( BAUIII) export coefficient models). The BAUIII export coefficient model did not deal with the effectiveness of measures for biochemical oxygen demand ( BOD) or Faecal Indicator Organism ( FIO) fluxes.

2. Previous Work

2.1. WFD19 Database

The first phase of project WFD19 investigated the feasibility of developing a Geographic Information System ( GIS) based screening tool for diffuse pollution at the national scale, involving a review of available modelling methodologies and datasets. Relevant models to address individual pollutant pressures and appropriate datasets were found to have been developed in the past, but application of a screening tool at such a large scale, covering both rural and urban pressures, and considering all pressures, had not been attempted before. Nevertheless, it was concluded that a basic-level screening tool for Scotland and Northern Ireland was practicable and would be a significant contribution to the characterisation of water body catchments under the Water Framework Directive ( WFD).

A national Environment Database was first constructed, containing environmental and agricultural statistical data summarised to 1 km ² suitable for visualisation and querying in a Geographic Information System. The database collates information on specific properties (e.g. land use, agricultural livestock numbers and population counts) controlling pollutant inputs and intrinsic environment properties (e.g. topography, soil physical properties and climate statistics) controlling risk of pollutant mobilisation and delivery. The database covers a land area of 78,770 km ² for Scotland, and includes a modelled monthly water balance and index of landscape connectivity. The final database, including summaries of model results, contains more than 80 tables and 1,000 items of data for each 1 km ² cell.

Modelling methodologies to calculate pollutant pressures and loads delivered to surface water bodies and to the base of the soil profile were developed for a) nutrient nitrate and phosphorus; b) suspended sediment; d) BOD and e) faecal indicator organisms. The methodologies were developed to work with environmental and agricultural data that were available for the whole of Scotland. The models also developed approaches that had previously been applied for policy work in the UK, including elements of the NIRAMS (Nitrogen Risk Assessment Model for Scotland) model of nitrate leaching and the Event Mean Concentration ( EMC) model of pollutants in urban runoff, or were being developed for this purpose, such as the PSYCHIC model (Phosphorus and Sediment Yield Characterisation in Catchments).

The models and Environment Database were linked to calculate pollutant pressures and loads for each 1 km ² cell across each country. Taken together, the models provided estimates of the diffuse pollutant loads derived from a) agricultural land; b) forestry; c) paved urban areas; d) road infrastructure; e) and septic tank diffuse sources. Additionally, estimates of the pollutant load from point sewage treatment discharges were made by use of per capita export coefficients. Summary statistics were calculated and stored in the national Environment Database, giving data on the proportion of the total pollutant load derived from each diffuse source. The calculated pollutant pressures were summed for the catchments of the river, coastal and lake water bodies defined by SEPA and EHS for reporting under the WFD.

Where monitoring data were available, the outputs from the models were validated against observed loads for selected pollutants. Observed loads were collated for 13 catchments in Scotland. The risk of failure was assessed by comparing concentrations with UKTAG recommended standards. In most cases, the selected modelling methodologies led to an over-estimate of observed loadings, as they did not take account retention in the receiving water body. However, it is possible that the limited 'spot sampling' of pollutant concentrations also resulted in an under-estimate of the true loadings.

For nitrate, phosphorus, and suspended sediment, intermediate empirical statistical models were developed that predicted observed percentile pollutant concentrations from modelled total pollutant load and drainage. These models were then used to predict the likelihood that river water pollutant concentrations were greater than the appropriate standard for unmonitored catchments. The risk of failure due to priority substances and pesticide runoff was assessed by comparison of modelled concentrations in land drainage with standards. The results of these models are included in the Environment Database and allow assessment of the risk of not achieving good ecological status due to diffuse sources only and both point and diffuse sources. These results were combined with an expert assessment of the risk of not achieving good ecological status for acidification, metals and faecal indicator organisms.

The output from the models was used to calculate the relative importance of point and diffuse sources for each of the pollutant pressures, using simple per capita export models to characterise sewage treatment works discharges. Diffuse sources accounted for c. 65% of phosphorus, 85% of nitrate and 85% of sediment losses, but only 30% of faecal coliform inputs to rivers. Roads and urban areas were found to make a significant 10% contribution to the total modelled sediment losses.

Analyses based on modelled nitrate, phosphorus and sediment losses determined that only 30.9% of Scotland waterbodies could be demonstrated to be not at risk of achieving good ecological status with confidence. However, when monitoring results were included this figure rose to 51.4% with confidence for Scotland. It is necessary to emphasise that this analysis is risk averse. A significant land area could not be proven to be either failing or achieving good ecological status, so was included in the area at risk. This area is significantly higher than that identified in both SEPA's and EHS's characterisation report and a priority for future research will be improving confidence in the risk assessment in these areas. Nitrate and sediment were not a major cause of failure according to current water quality criteria. Phosphorus losses resulted in the greatest land area designated at risk, and were primarily associated with agriculture. Other diffuse pollutant sources, including roads and urban areas, were more critical for losses of priority hazardous substances.

2.2. Business as Usual ( BAU) Export Coefficient Model

The BAU export coefficient model provides quantitative but only indicative estimates of the diffuse nitrate, phosphorus and sediment loss from agricultural land that reflect forecast changes in the structure of the agricultural sector (crop areas; stock numbers; and unit yields) and land management practices (timing and quantities of manures and fertilisers applied; implementation of mitigation measures).

Soil and Climate Zones

Diffuse pollutant losses are sensitive to environment conditions. Key factors are rainfall and soil type. Rainfall frequency and intensity determines the volume of soil drainage and likelihood of surface runoff - and hence the solute leaching potential and soil erosion risk. Soil type (characterised by texture) determines the relative importance of surface runoff, preferential and matrix flow pathways, and hence the efficiency of leaching and the likelihood of retention (Haygarth et al., 2005). Mitigation measures that affect a specific mobilisation or delivery process (e.g. riparian buffer zones impact on soil erosion loss by surface runoff) will have geographically variable effectiveness.

Defra project WQ0106 (Anthony, 2006) defined representative soil texture and climate zones for the calculation of diffuse pollutant losses from representative farm systems. The zones reflected the range of rainfall, soil texture and drainage conditions across England and Wales (Table 1). The zones were mapped based on long-term average rainfall and soil drainage calculated by the NEAP-N model (Anthony, 2006) and a simplification of soil texture classes defined by John Hollis using data from the NSRI soils database ( pers. comm.; formerly NSRI).

The same zones were mapped for Scotland using rainfall and modelled soil drainage data from the SNIFFER Screening Tool Database (Anthony et al., 2006), and data on soil wetness class derived from the FAO Soils Map of Europe in order to avoid data licensing issues. The finalised climate and soil zones are illustrated by Figure 1.

Table 1. Definition of climate and soil zones used to interpolate results of field scale models applied to representative farm systems across England and Wales (Anthony, 2006), where CL = clay loam, C = clay, P = peat, S = sand, SL = sandy loam and L = loam.

Figure 1. Soil and climate zones defined for the extrapolation of the results of field scale models, applied to representative farm systems, across the UK (see also Table 1).

Export Coefficient Model

Defra project WQ0106 calculated nitrate, phosphorus and sediment losses from representative model farms using field scale models (Anthony, 2006). The farm types modelled included a specialist cereal, dairy, beef, LFA grazing, lowland grazing indoor pig and poultry farm. Pollutant losses from each model farm, for each combination of soil texture and climate condition, were calculated using a combination of the PSYCHIC, NITCAT, MANNER and N- CYCLE models (Davison et al., in prep; Lord, 1992; Chambers et al., 1999; Scholefield et al., 1991). The loss calculations assumed current farm practice and that no mitigation measures were in place. To represent the impact of timing of manure and fertiliser applications on pollutant losses, simulations were made for applications in each and every month, and weighted in proportion to national statistics on the quantities of manure and fertiliser applied in each month. Pollutant loss was partitioned into losses due to external (fertiliser), internal (soil) and recycled (manure) sources. External sources are the sources of potential pollutant that are notionally imported across the farm gate such as mineral fertilisers, some feedstuffs, and atmospheric deposition. Internal sources describe those that are generated within the soil profile, including the decomposition of organic materials. Recycled sources are generally internal to a farm as part of the production system. This includes farm manures, slurries and dirty water, and also excreta from grazing animals.

To map pollutant losses across the country, the field scale model results were re-expressed as export coefficients, i.e. quantity of pollutant lost per unit land area or per unit of pollutant in total fertiliser applied or total excreta production on the farm.

Table 2 summarises the model derived export coefficients. They express the baseline pollutant losses as a proportion of the total nitrogen and phosphate content of potential pollutants applied to the field area. Note that it is the total pollutant content, including pollutant fixed in organic forms, rather than the readily available (dissolved) pollutant content of the manures. The coefficient for manures is also expressed as a proportion of the total nutrient content of the fresh excreta generated on the farm, and not the content of the manure at time of application. Manure is used to refer to both losses due to excreta and spreading of managed manure collected from buildings. The export coefficients were calculated separately for each soil texture and climate zone. In a significant deviation from export coefficient literature an area based export coefficient was also calculated to represent the loss of nitrogen and phosphorus from soil sources (by soil erosion and leaching of soluble soil components). This soil-derived component represents the background supply of nitrogen and phosphorus that will respond more slowly to changes in the balance of nutrients input and recovered from the farm system, and includes indirect losses associated with fertiliser and manure additions to the soil organic nutrient store. The soil source was also used to represent the sediment losses.

In deriving the export coefficients, the total nitrogen and phosphate in fertiliser and manure on each representative farm system were derived from statistical data on fertiliser use ( BSFP, 2004) and the quantity and quality of excreta produced annually by each livestock type (Smith and Frost, 2000; Smith et al., 2000). Note that to convert a mass of phosphate (P ₂O ₅) to elemental phosphorus (P), multiply by 0.436.

The export coefficients for the poultry and pig farms were calculated only for the handled manure. The coefficients for the associated arable land area receiving the manure would be the same as for the arable farm.

These coefficients are here used to calculate the baseline pollutant losses (assuming no mitigation measures in place) for each of the recognised farm types. A dynamic Excel spreadsheet has been developed that automates this process. This is called the 'Diffuse Calculator' and reports total pollutant losses, using as input the agricultural census data that was disaggregated by farm type. Table 2 shows 'Diffuse Calculator' data for Scotland.

For each 10 by 10km cell, we estimate the proportion of the agricultural land area that is located in each soil and climate zone using the 1km ² zone map , assuming that agricultural land is distributed evenly within each cell.

The agricultural data was summarised on the same 10-by-10 km grid used by the BAU. Ten km is representative of the maximum distance that manures would generally be transported between and within farms, and so does not give false impression of accuracy in the location of livestock derived pollutant emissions. Summarising land use at 10-by-10 km also enables an indicative statistical separation of the total land areas and stock numbers between farm business types that are the basis of the scenario forecasts.

The original BAU work uses indicative farm-types. These farm types are aligned to Defra's (post 2003) Robust Farm Types ( RFT) classification used in the Defra Farm Business Survey. The numbers of holdings by farm-type were obtained from SEERAD on a parish basis (n=891) and summarised to the common 10-by-10 km grid. The resulting data were summarised to the 10-by-10 km grid and expressed simply as relative numbers of farms of each type, rather than the absolute number within a grid cell. It is advised that the relative counts of farm-type were derived from relatively coarse scale spatial data and cannot be considered accurate for an individual grid cell.

Summary data on the average crop areas and numbers of livestock for each farm-type were obtained directly from SEERAD. These data were calculated for the project directly from the survey returns for 2004. Average values were not found for every crop or stock class, but there were sufficient to be used as indicators of the likelihood of a general type of crop or animal being present. For example, the reported average number of store pigs was used to index the presence of each of the fattening pig classes.

A spreadsheet is used to re-distribute the baseline 2004 agricultural survey between the farm-types, doing this for each individual 10-by-10 km grid cell. The spreadsheet was set up to allow re-calculation, should improved survey or farm-type data become available.

The survey items within a grid cell were re-distributed between the farm-types in proportion to:

a) the local relative number of farms of each type within the grid cell; and

b) the national indices of average crop areas and livestock numbers on each farm-type.

This simple statistical re-distribution methodology has the advantage of being applicable to the survey datasets available for the whole of the UK. The output is a new survey dataset for each farm-type, listing the crop areas and livestock numbers according to the categories, for each 10-by-10 km grid cell. However, it is emphasised that the split between farm-types is a statistical and imprecise process. The results for any one grid cell are not expected to agree exactly with holding level statistics. The process combines national average data on farm cropping and animal numbers with local estimates of land use and farm-type counts that are imprecise at the scale of the individual cell. The process also assumes that each farm holding of the same type is the same size across the country. No attempt was made to distribute survey data by size of enterprise.

Overall, however, we believe that the process is a useful means for the indicative mapping of land use by general farm-type at the regional and national scales based on data that were readily available.

For each zone and farm type, and for each land use and stock type, we then estimate total fertiliser use and excreta production based on the forecasts developed by SAC and ADAS. This enables our diffuse pollutant loss calculations to be sensitive to the forecast changes in agricultural structure and inputs. Fertiliser inputs and excreta production can be specific to a farm type, e.g. to represent higher inputs on a specialist dairy farm relative to a beef and sheep farm.

The spreadsheet then carries out the appropriate multiplications - using the export coefficients - to calculate the total pollutant loss. This can be reported separately by farm type and soil and climate zone (see Figure 5).

Results of spatially detailed model runs exist for Scotland, England and Wales. They used the NIRAMS, NEAP-N and PSYCHIC models (Anthony et al., 2005; Anthony, 2006; Anthony

Table 2. Summary of model derived export coefficients for nitrate, phosphorus and sediment by climatic and soil zone (Anthony, 2006)

and Lyons, 2006). The Scotland calculations used agricultural census data for 1997, and the England and Wales calculations used data for 2000. These model results provide a useful validation of the simpler export coefficient model and can potentially be scaled by the export coefficient results.

Mitigation Measure Implementation

Defra projects WQ0106 and ES0203 established a short-list of diffuse pollution mitigation measures. For each measure, the percentage reduction in pollutant loss that would occur if it were fully implemented was estimated based on expert opinion and summary of empirical data. The pollutant reduction was expressed relative to baseline losses for representative model farms. The baseline losses were calculated using the same models as for the derived export coefficient model.

The measure effects were estimated only for the model farm systems located in the medium climate zone. To estimate measure effectiveness within other climate zones, the impacts of the mitigation measures were re-expressed as a percentage of the loss due to specific sources that they addressed, namely external (fertilisers), internal (soil) and recycled (manure and excreta). The impact of a measure was limited to no more than 90% of the baseline loss.

To calculate the impact of the mitigation measures, it is first necessary to estimate the percentage of the land area to which they are applied, and the percentage efficiency of implementation. Under Defra project WQ0106 this was done by expert opinion, with the aid of data on scheme option uptake, and was done separately for agricultural land within and outside of legislated scheme areas such as the Nitrate Vulnerable Zones ( NVZs) and the England Catchment Sensitive Farming Delivery Initiative ( ECSFDI) catchments.

The 'Methods Calculator' spreadsheet takes these matrices (see Figure 5) of measure effect and implementation to calculate a set of multiplier coefficients that describe the pollutant loss relative to the baseline. The coefficients are calculated separately for each pollutant type, by soil, fertiliser and excreta sources, by land use, and for each soil and climate zone. Note that unlike in project WQ0106, there is no optimisation in the selection of mitigation measures. Implementation is determined entirely by expert opinion and such empirical data as is available. For this project multiplier coefficients were determined for 2004 (areas within and outside Nitrate Vulnerable Zones - NVZs) and 2015 (national General Binding Rules ( GBRs), within NVZs, and within High engagement GBR catchments). Figures 2 and 3 show an example of the coefficients derived for mixed farm type and the cereal farm type.

Figure 2. Example of mitigation calculation sheet for cereal farm type, arable land 2015 NVZ scenario, showing measure efficiency and uptake.

Figure 3. Example of mitigation calculation sheet for mixed farm type, arable land 2015 NVZ scenario, showing measure efficiency and uptake.

Once calculated, these loss modifiers are copied across to the 'Diffuse Calculator' spreadsheet (see Figure 5) that then automatically derives a new set of pollutant losses based on this prediction of the effect of implementing mitigation practices.

The effect of a single mitigation measure is expressed as a percentage reduction against the loss due to a specific pollutant source, namely external (fertiliser), internal (soil) and recycled (manure and excreta). The pollutant losses due to each source are calculated using a suite of tier one pollutant models. The reduction R due to the measure is scaled in proportion to the efficiency E of implementation:

where P is the expected reduction under optimal conditions, as determined from the literature review. In this equation and all others, the percentage values are re-expressed as fractions (0 to 1).

If the mitigation measures are applicable to all of the land on a farm, and every farm in the landscape, then the net reduction N due to a suite of measures can be calculated using a multiplicative model as:

where R _i is the reduction due to an individual measure.

In the circumstances that one or more measures are not implemented across the whole farm, we have made an assumption of maximum overlap of measure uptake (the product of applicability and implementation):

where A _1:j is the proportion of the farm area affected by measures 1 to j and N is the net effect of all the measures. This is in effect an area weighted version of equation one.

Figure 4. Schematic of the method of calculating the net effectiveness of multiple measures, assuming maximum overlap of measure uptake.

Figure 4 makes this explicit. Three methods are shown on a graph of measure uptake with values of 80%, 50% and 20%. They are ordered by decreasing uptake. The effectiveness values of the measures are 20%, 50% and 20%, respectively. The net effect is determined from the 30% of the farm area impacted by only the first measure with an effectiveness of 20%; the 30% of the farm area impacted by the first and second measures with a combined effectiveness of 60%; and the 20% of the farm area impacted by all three measures with a combined effectiveness of 68% (Equation 1). The net effect is a weighted sum of these combined effectiveness values, where the weight is the area of overlap. In this case, the combined effectiveness of all three measures (including the area of the farm that is not impacted by any measure) is 37.6% (Equation 2).

The method of calculating the net effectiveness of multiple measures assumes that measures are acting on the same potential pollutant source. Therefore, the gain from additional measures decreases rapidly. This is not a perfect model but is thought to be better than the alternative additive model in which the pollutant source is quickly exhausted and the impact of multiple measures over-estimated. To over-come erroneous competition between measures for effect, the potential pollutant loss is separated by source type.

The source apportionment draws upon the conceptual Cost-Cube model developed under Defra project ES0121, but is incomplete in not explicitly representing the models of pollutant mobilisation and delivery. For example, pollutant loss due to manure application may be due to solubilisation and leaching via sub-surface pathways, or due to particle detachment and surface runoff. A measure such as a riparian buffer zone will be ineffective against the former mobilisation and delivery aspects, but effective against the latter. The relative importance of each mobilisation and delivery aspect will vary with environment circumstances. As a consequence, the effectiveness of the mitigation measures can only be a general estimate.

Furthermore, potential pollutants that are conserved by a mitigation measure that addressed one mobilisation and delivery aspect might be subsequently lost by another. For example, nitrate and phosphate in manure that is not lost by surface runoff due to the riparian buffer zone may be subsequently leached from the soil. This carry-over of potential pollutant losses is not represented by the model. As a consequence, the effectiveness of the mitigation measures may be over-estimated.

The effectiveness of mitigation measures were determined from field studies of probably at most three factors. The accuracy of the calculated net effect of multiple measures will therefore decrease rapidly with number due to the measure interactions that have not been investigated. It is not possible to quantify this uncertainty.

The prototype of the 'Methods Calculator' spreadsheet includes an example of mitigation measure implementation on the dairy farm type. The measure effect and implementation data were taken from the baseline forecast for non- NVZ areas developed under Defra project WQ0106. It is calculated that the measures will, for example, result in a 14-18% reduction in nitrate loss from the fertiliser source, a 1-6% reduction in nitrate loss from the soil source and a 25-27% reduction in nitrate loss from the manure source, all on the arable land area associated with the farm.

Prior to measure implementation, the export coefficient model based total nitrate loss from the Dairy farm type was 8.7 kt N. After measure implementation, the loss is reduced to 8.1 k

Figure 5. BAU methodology diagram

3. Methodology

Phase II requires the adjustment pollutant losses from a baseline (2004) to 2015 taking into account:

• Changed land-use and animal numbers
• Uptake of measures and
• Spatial differences in uptake.

The WFD77 project generated loads (and concentrations) for 1 km ² cells and WFD catchments. This is a single number e.g. kg ha ^-1 N lost, without the underpinning source apportionment e.g. the amount lost from grassland or arable land within that area, or the amount lost per unit of fertiliser or manure applied. This immediately makes validation of the WFD77 output against other data (Phase I) and manipulation to account of land use change and uptake of measures (Phase II) complex. Therefore, the approach taken in the methodology used in Phase II will have to be at a moderately coarse scale.

Previous work for Defra ( DPI project) approached this by starting at individual farm levels and scaling up using complex spreadsheet models. Since the request within the specification for this project was not to re-run the WFD77 Screening Tool, then the approach adopted will not be exactly the same but will use some of the experiences in developing the methodology. Therefore the methodology will:

• Focus on pollutant loads
• Provide the most robust estimates for N, P, sediment and FIOs
• Give an output scale of a 10 km ² grid.

Given the limitations of the Screening Tool it is proposed to use the simple export coefficient model from the BAU project (Section 2) to calculate an index of change in the pollution loss from 1997/2000 to 2015 ad from 2004 to 2015. The change index will be responsive only to agricultural land use inputs (fertiliser and manure loadings), and estimated uptake of mitigation measures. The WFD77 database and Screening Tool contents were derived using models that were sensitive to local environment conditions (slope, drainage status, soil type). These environment conditions remain constant even with the introduction of mitigation measures. Therefore, it is appropriate to use the BAU export coefficient model to calculate a change coefficient that is multiplied against the WFD77 database contents to calculate pollutant loads for 2015.

Due to the differences in the BAU model and the WFD77 database (Table 3) a number of subsidiary steps need to be addressed.

Table 3. Comparison between the WFD77 database and the BAU export coefficient model

Currently the Screening Tool uses 1997 agricultural census data for Scotland, whereas the BAU export coefficient model uses 2004 agricultural census data. Table 4 shows the difference between livestock numbers between the two years.

Table 4. Changes in livestock numbers between 1997 and 2004 agricultural census data

The percentage changes between livestock numbers are quite large relative to some of the potential mitigation measures impacts. Therefore, to establish a methodology for scaling the BAU model results need to be adjusted for 1997 agricultural census data.

To adjust the agricultural census data a 10 km ² cell map of Scotland was prepared showing the percentage change for each of the crop/livestock categories recognised by the BAU model. Figure 6 shows an example of the percentage change in total livestock numbers. These indices were then applied to the 2004 census data prepared for the BAU export coefficient model. The BAU model was then run for both 1997 and 2004 to generate an estimate in change in pollutant loading.

Figure 6. Percentage change in total livestock numbers between 1997 and 2004.

The Screening Tool does not consider the impact of any mitigation measures such as buffer strips or minimum tillage. The models used in the Screening Tool were primarily driven by inputs, such as the BSFP fertiliser rates. Although there was some calibration and validation against observed loads, it might be argued that the models resulted in an overestimate of losses. It is known from the BAU and previous work that the impact of measure implementation on pollutant losses is small (<10%). The largest effect in the previous work was the change in animal numbers following CAP reform. This is demonstrated in Table 5 where the difference in pollutant loads between 2004 BAU model results run with and without any mitigation measures are only between 1-8%. Therefore the BAU model can be used to estimate the impact of baseline mitigation measures implementation.

Table 5. Average BAU pollution load results with and without the implementation of mitigation measures for 2004.

To run the BAU export coefficient model for 2015, estimated livestock numbers were needed. These were obtained from the FAPRI-adjusted BAU analysis and used to scale the BAU livestock numbers. The model was then re-run using the new livestock numbers and mitigation measure uptake and efficiency numbers to give a scalar to apply against the Screening Tool results.

In summary the methodology was:

1. Use ArcMap to spatially sum the 1 ha WFD77 pollutant values up to the 10 by 10 km ² grid square

2. Update the BAU model to take into account any changes in export coefficient values arising from the model evaluation exercise

3. Run the BAU model to produce 2004 pollutant loads (without measures)

4. Scale the WFD77 pollutant loads (no measures) to take into account the changes in agricultural census data between 1997 and 2004.

5. Use the ratio between the WFD77 pollutant loads (no measures) and the 2004 BAU pollutant loads (with measures) to scale the WFD77 pollutant loads to take into account baseline measure uptake for 2004.

6. Points 3 and 4 will give the project a set of baseline pollutant losses including current mitigation measure uptake.

7. List measures for 2015 under NVZ, GBR and HEGBR scenarios

8. Estimate measure effectiveness, percent implementation and efficiency of implementation for NVZ, GBR and HEGBR measures

9. Update the BAU 'Methods Calculator' to include the new mitigation measures (for NVZ, GBR and HEGBR), measure effectiveness, area implementation and efficiency of implementation values arising from this work

10. Use the 'Methods Calculator' to calculate scenario multiplier coefficients describing pollutant losses relative to baseline

11. Incorporate scenario multiplier coefficients into the 'Diffuse Calculator'

12. Incorporate 2015 changes in livestock numbers from the BAU and FAPRI-adjusted BAU analysis by scaling the BAU agricultural census data.

13. Re-distribute BAU and FAPRI-adjusted BAU agricultural census data by farm type and 10 by 10km ² grid square

14. Input this data in to the census files used in the 'Diffuse Calculator' and re-run the BAU export coefficient model (see Section 2). This will give new pollutant losses for 2015 for NVZ, GBR and HEGBR for both BAU and FAPRI adjusted BAU data

15. Scale the WFD77 pollutant losses with the new 2015 BAU values for NVZ, GBR and HEGBR

16. Compare 2015 results to baseline data to give a percentage change against the baseline for AAG region and farm type.

2.1. Methodology Evaluation

In order to use the proposed methodology it needs to be demonstrated that the BAU export coefficient model and the Screening Tool model results are similar for the baseline and have similar assumptions about the inputs. Figure 7 indicates that there is a good match between the two models in regards to total arable land.

Figure 7. Comparison between BAU and the Screening Tool total agricultural areas.

Figure 8 shows a comparison between the BAU and Screening Tool results for both N and P. It can be seen that although the N results are comparable, there is a significant difference between the P results. To try and improve the comparison two main parameters have been investigated; annual soil drainage and the export coefficient values used within the BAU model.

Figure 8. Comparison between WFD77 and BAU results for N and P.

Annual Soil Drainage

The BAU export coefficient model assumes fixed quantities of soil drainage for each of the soil-climate zones, whereas there is a large variation in rainfall and drainage within each of the zones according to the Screening Tool database and models. Also, the maximum drainage on the hills is considerably greater than assumed for the 'wet' climate zone (1,500 vs 500mm) (Figure 9).

Figure 9. Comparison between annual soil drainage.

As phosphorus losses calculated by the PSYCHIC model are approximately linearly correlated with the quantity of drainage by the surface and sub-surface (tile drain) pathways, the export coefficient model results were therefore scaled in proportion to the ratio of the drainage totals calculated by the two modelling approaches for each of the 10 by 10 km cells. In general, this significantly increased the export coefficient model predictions in the upland areas (Figure 10).

Figure 10. Phosphorus loading comparison between BAU and WFD19 before (A) and after ratio scaling (B).

As for phosphorus, differences in the total drainage calculated by the two modelling approaches may explain differences in the modelled total nitrate losses. However, unlike phosphorus, nitrate is source limited and the total quantity of drainage in high rainfall areas in Scotland is usually sufficient to elute the soil profile. The additional drainage in the upland area, relative to the export coefficient model, would have only a small impact on the quantity of leached nitrate. To represent this, the quantities of total subsurface drainage (via drains and to groundwater) calculated by the two models were input to the SLIMMER leaching function that calculates the proportion of nitrate leached as a function of drainage volume. The ratios of the proportions leached were then used to scale the output from the export coefficient model (Figure 11).

Figure 11. Nitrate loading comparison between BAU and WFD19 before (A) and after ratio scaling (B.)

Export Coefficients

To further improve the BAU fit with WFD77 results it was also found necessary to increase the P and sediment export coefficients for rough grazing by 50%. This was to reflect the increased landscape connectivity of rough grazing area in Scotland that would have been properly represented in the Screening Tool model. However, in order to investigate the sensitivity of total P loads to over-estimations in P losses from upland areas (see section 2.5.1 in the main report), we also ran the model separately with the BAU P and sediment export coefficients unchanged (i.e. lower than predicted by WFD77).

For all other land uses and fertiliser and manure practices it was found necessary to decrease the export coefficients by 40%. This can be justified on the basis of less than comprehensive tile drainage.

These combined changes produce a good fit to the Screening Tool model results (Figure 12 and 13).

Figure 12. Comparison between WFD77 and revised BAU results for phosphorus.

Figure 13. Comparison between WFD77 and revised BAU results Nitrate-N.

Published Export Coefficient Data

Validation for the export coefficients used the original BAU project validation carried out against published data. Further validation of the coefficients was also carried out as part of this work. . Table 6 shows the new export coefficient values for this project.

Table 6. New nitrate and phosphorus export coefficients by pollutant source, for each soil and climate zone. Note that losses due to manures are expressed per kilogramme of total nitrogen and phosphate in animal excreta generated on a farm.

Within Phase I of the project the WFD77 export coefficients were compared to a number of published data sets. The most comprehensive data set is that published by Johnes (1996). It can be seen from Table 7 that the new coefficients are found to be broadly comparable to those calculated by Johnes (1996) from catchment studies, with total excreta nitrogen being within 3% and total fertiliser nitrogen within 8% of published values. Total excreta phosphorus is within 2% and total fertiliser phosphorus within 1% of published values.

Table 7. Export coefficients for nitrogen and phosphorus derived by Johnes (1996) for use in the AERC National Export Coefficient Model. The coefficients express pollutant loss per unit of elemental nitrogen and phosphorus in animal manures and fertiliser applied to land. The colours indicate geoclimatic regions: intensive arable (brown); mixed arable and dairying regions on permeable substrate (blue); lowland dairying regions (green); mixed arable and dairying regions on impermeable substrate (pink); extensive livestock and upland regions (yellow); and urban and non-agricultural regions (red). The brown region broadly compares to our 'dry' climate region; the green and blue to our 'medium' climate region; and the pink to our 'wet' climate region.

4. FIO Methodology

The methodology for national projections of losses of pathogens to water is less well developed than for nutrients. Indicative effects at the national-level were based on excretion data and number of livestock for baseline and 2015 figures. This will identify the size of the potential pollutant source and changes in the size of the source. In terms of the mitigation measures, the loss modifiers calculated for manure was used to multiply baseline pathogen loss for the 2015 scenarios.