Main authors: R.K. Laursen, F. Bondgaard, P. Schipper, K. Verloop, L. Tendler, R. Cassidy, L. Farrow, D. Doody, F. A. Nicholson, J. R. Williams, I. Wright, J. Rowbottom, I. A. Leitão, A. Ferreira, B. Hasler, M. Glavan, A. Jamsek, N. Surdyk, J. van Vliet, P. Leendertse, M. Hoogendoorn and L. Jackson-Blake.
Editor: Jane Brandt
Source document: »R.K. Laursen et al. (2019) Evaluation of Decision Supports Tools. FAIRWAY Project Deliverable 5.2 216 pp

 

A comprehensive review and survey of Decision Support Tools (DSTs) currently in use in the FAIRWAY case studies is described in »Survey and review of existing decision support tools. Of the 36 DSTs  identified as most relevant, 12 were selected for further investigation to see if a tool developed in a particular national context could be used or provide inspiration elsewhere (»Evaluation of decision support tools). Here we describe the tools evaluated for potential use in the Derg Catchment case study.


Contents table
1. Selection of DSTs to evaluate in the Derg Catchment case study 
2. Farmscoper
3. Phytopixal
4. SCIMAP 
[Note: Because of the resolution of the images, it is difficult to see the detail in some of the figures and tables. See the »full report for more legible originals.]

1. Selection of DSTs to evaluate in the Derg Catchment case study

The Derg is a 384 km2 sub-catchment of the Foyle river system located in the north west of the island of Ireland (Figure 18). The catchment is cross-border, with headwaters in the Republic of Ireland (RoI) and drinking water abstraction at the outlet, in Northern Ireland (NI), for treatment and supply of between 16 and 28 ML day-1 to the NW region.

D5.2 fig18
Figure 18

In Northern Ireland (NI) and the Republic of Ireland (RoI) there is considerable concern about MCPA (2-methyl-4-chlorophenoxyacetic acid) contamination of natural waterbodies. The herbicide is widely used in the control of rushes (juncus spp.) and other weeds commonly found in managed grasslands. MCPA usage is of particular concern because rushes grow well in the poorly drained, clay soils common in the Derg catchment and the herbicide is highly mobile in the water phase. Although much progress has been made in updating pesticide usage practice in both NI and the RoI, through the implementation of action plans for sustainable pesticide use (DAFM, 2018), and compulsory training and tighter control on sales, significant concentrations of MCPA are still observed in river water and in drinking water supplies (NIEA, 2017, EPA, 2017). Frequent exceedances of the Drinking Water Directive threshold of 0.1 µg L-1 have been detected during regulatory compliance monitoring by the water company at the abstraction point (Figure 18) while 37% of sub-daily sampling undertaken by the INTERREG VA Source to Tap project (www.sourcetotap.eu) between April and November 2018, also exceeded this threshold.

A better understanding of MCPA export and mitigation can be achieved through the application of quantitative mathematical models of pesticide load or spatial models to identify areas where the risk of pesticide export is highest. The DSTs tested in the Derg case study cover both pesticide load modelling and spatial assessment of source and transport risk:

  1. Farmscoper (http://www.adas.uk/Service/farmscoper) is an advanced export coefficient model which estimates diffuse losses of P, N, sediment and pesticides from single or multiple farms and quantifies the expected impacts and economic costs of mitigating losses to water and air (Gooday et al., 2014). The models can be upscaled to catchment scales by aggregating data for all farms within a specific catchment (Zhang et al., 2012). Parameters of interest, such as soil type, rainfall and farm type have been pre-defined in the model and are based on conditions in England and Wales, the region for which the model was developed.
  2. The Phytopixal protocol (Macary et al., 2014) generates a spatial risk assessment based on a small number of physical catchment characteristics (slope angle, soil propensity to cause overland flow and proximity (as linear distance) to a waterbody) and a measure of pesticide usage (e.g. frequency of pesticide application or mass of pesticide applied to particular land use types). The user is allowed considerable freedom in selection of the data sources used, allowing for customisation of the protocol presented in Macary et al (2014). The protocol was developed in France.
  3. SCIMAP (http://www.scimap.org.uk) is a diffuse pollution risk mapping tool (Lane et al., 2006, Milledge et al., 2012, Reaney et al., 2011) that generates a spatial risk assessment based on hydrological connectivity. Using topographic information the model predicts, for each point in the landscape, the probability of overland flow being generated, and therefore providing a pathway for contaminants export to the river network. Soil erosion potential is the only contaminant considered explicitly, but the operator can use expert knowledge to explore the risk associated with nutrients, sediment, pathogens or pesticides. The user is allowed considerable freedom in selection of the data sources used, allowing for customisation of the protocol. The latest version of this model is hosted on an online platform and covers only Great Britain (England, Scotland and Wales), but an archived version bundled as a toolbox within the open source SAGA GIS software will remain available and can be customised for other regions (available to download from (http://www.scimap.org.uk/2016/02/x64-scimap-for-saga-gis-february-2016/).

A consultation was undertaken with representatives from water companies in NI, the Rivers Trust and catchment officers from the INTERREG VA “Source to Tap” project (http://www.sourcetotap.eu) who have daily interactions with farmers and local authorities and NGOs in the catchment.

Both the water companies and Rivers Trust are involved in catchment management across Ireland and are currently implementing a pilot Land Incentive Scheme (LIS) to address MCPA pesticides, colour and turbidity in the Derg case study catchment. Farmers receive financial support to implement a number of mitigation measures, with the expectation that the measures will serve to reduce exceedances of colour and pesticide limits in the raw drinking water supply from the catchment.

At the meeting, an overview of each of the DSTs were presented to the stakeholders and their opinions on the utility of the DST were discussed. The outputs of these discussion are synthesised with the results of testing undertaken in–house by AFBI.

2. Farmscoper

2.1 Assessment and testing

The Farmscoper DST is a series of Microsoft Excel spreadsheets with macro-driven databases that has been designed to allow the generation and customisation of individual farm systems, based on on-farm data or using available census data on livestock, cropping and manure management (exemplar screen shots of the input data sheets are shown in Figure 19 -21). Outputs to water and air are modelled for a range of atmospheric and waterborne contaminants including nutrients, pesticides and sediments (a full list is provided in Table 25). Predictions are based on well-established models which have been used in the UK, including NEAP-N for nitrate (Anthony et al., 1996) and PSYCHIC Davison et al., 2008; Strömqvist et al., 2008) for phosphorus and sediment; MACRO Tool (Jarvis, 1995) and SWAT for pesticides. Contaminant losses are apportioned across source (e.g. dairy, beef, arable products, grass products), pathway (e.g. runoff, preferential flow, leaching) and timescale (short to long term) within the model. Soil types in the model are represented based on soil permeability, and classified based on the requirement for artificial sub-surface drainage (e.g. pipe drains). Three drainage classes are available and used as the basis for generating contaminant export coefficients for farming systems on different soils. Three workbooks in the model (Evaluate, Prioritise and Cost) are used to estimate the environmental impact and cost-effectiveness of one or more mitigation methods, from a library of over 100 options. Model evaluation can be undertaken at farm level or upscaled to catchments, through aggregating individual model output for all farms in a catchment.

D5.2 fig19
Figure 19
D5.2 fig20
Figure 20
D5.2 fig21
Figure 21
D5.2 tab25
Table 25

In this report, the testing of Farmscoper in the Derg case study focusses on pesticides, as it is the primary contaminant causing breaches of the Drinking Water Regulations limits in the catchment. Potential applications for nutrients and sediment, however are also of interest.

Key issues covered in an assessment of suitability/utility included the following points.

i. Differences in pesticide usage between Ireland and England/Wales

Pesticides are represented in Farmscoper as a % of typical plant protection products (PPP) used, based on pesticide surveys and mode of application. Output Table 25 is predicted as a dose unit per litre of whichever pesticides are used as standard on the specified crop type. For England and Wales the dosage of pesticides for a particular land use is based on the Pesticide Usage Surveys for GB (Garthwaite et al., 2005, 2006), which provide the average usage amounts of herbicide, fungicide etc. on different crop types.

Usage practice in Northern Ireland and Ireland differs from that in GB as a result of climate, soil types, crops grown and the advice delivered by pesticide companies and advisors. In NI for example, over 94% of land area is grass and in extensively farmed areas rough grassland is regularly treated, primarily with MCPA to kill rushes. Often MCPA is not applied for agronomic reasons but rather to ensure land remains eligible for the Single Farm payment. In the Farmscoper model rough grassland areas do not produce a loading of pesticides so adaptations will be necessary to represent this practice in NI.

Statistical reports on pesticides are available for NI (2016 for arable crops and 2017 for grasslands (https://www.afbini.gov.uk/articles/pesticide-usage-monitoring-reports)) and Ireland (2012 for arable crops and 2013 for grasslands (http://www.pcs.agriculture.gov.ie/sud/pesticidestatistics/)) and can be provided as kg/ha values for each crop type. However, access to the database component of Farmscoper is required (currently restricted) in order to identify what adaptations are necessary to account for “typical” Plant Protection Product (PPP) use on grassland crops (including rough grazing) in Ireland.

A query has recently been submitted to the developer (19/02/2019) and we are awaiting the response (as of 09/05/2019).

ii. Geo-climatic differences between Ireland and England/Wales

Six climate zones (ranging from <500 to >1500 mm yr-1) are defined in Farmscoper based upon the range of long-term average rainfall across England and Wales for 1961-1990. Annual average rainfall in NI over the 1971-2000 period ranged between 700 and 2200 mm, and was highest in the west where the case study catchment is located. The >1500 mm climate zone in Farmscoper is the most applicable option to represent the high rainfall in the case study catchment. However, in the model the annual rainfall value is then distributed into a monthly rainfall pattern which was originally set for England/Wales, and which differs in NI/RoI. This distribution was used when the background PSYCHIC model was run during development (Davison et al., 2008; Strömqvist et al., 2008) and, as such, cannot be adjusted. This has the potential to lead to over/under estimation of drainage flow compared to that observed.

An evaluation was undertaken using data from a monitored catchment in the east of NI, where an input rainfall range of 900-1200 mm generated a modelled drainage flow in Farmscoper (defined as combined runoff + preferential flow + groundwater recharge) of 580 mm yr-1. This drainage flow estimate was verified against local monitoring data on rainfall and evapotranspiration. Rainfall for the 2015 and 2016 hydrologic years (Oct-Sept) at a representative location in the study catchment was 1138 and 1225 mm yr-1, respectively. Evapotranspiration is estimated at ~ 44% for this area which indicates that a runoff of ~662±25 mm yr-1 (14% higher than modelled) is more accurate for the catchment. Future application of the model, particularly if it is to be used by non-specialists, would require modifications to account for these differences through adjusting flow partitioning within the model. A specialist could apply a manual correction to the model output values to allow for the differences in the interim.

In the landscape, soil type and geology are the principal controls on determining the pathways of effective rainfall over the ground surface or at depth within shallow sub-surface or deeper groundwater flow. In Farmscoper these combined characteristics are represented by three defined soil drainage classes, according to the probability of having artificial sub-surface drainage under different crop types. These include:

  • Free draining soils that do not require sub-surface drainage
  • Slowly permeable soils that require sub-surface drainage for arable crops
  • Slowly permeable soils that require sub-surface drainage for both arable crops and grassland.

For each type, contaminant apportionment among pathways is based on pre-calculated values derived for English and Welsh soil types and the 6 climate zones from NEAP-N for nitrate (Anthony et al., 1996) and PSYCHIC (Davison et al., 2008, Strömqvist et al., 2008) for phosphorus and sediment; MACRO Tool (Jarvis, 1995) and SWAT for pesticides. Thus there is no option for the user to modify the model for areas which are outside of the categories for which the model was originally developed.

Partitioning of drainage flow (which is defined in Farmscoper as the sum of runoff + preferential flow + groundwater recharge) between runoff and infiltration (preferential flow + groundwater recharge) in the model is an area requiring further evaluation. Farmscoper runoff estimates are lower than would be expected in NI catchments. For monitored field-scale plots on the AFBI CENIT site (Cassidy et al., 2017, Doody et al., 2010, Watson et al., 2007) annual runoff (as overland flow) can account for 40% of the total effective rainfall. This is primarily due to the steep slopes and clay-rich impermeable soils which typify the glacial depositional landscapes that dominate ~43% of the NI landscape. Before Farmscoper can be used for real management scenario analysis in an Irish context, this issue needs to be examined further. For pesticides this may require MACRO and SWAT to be re-run for some scenarios, in collaboration with the model developer.

iii. Data requirements or data deficits

In addition to the representation of geo-climatic factors within the model the availability and accuracy of on-farm data is crucial, and pose the greatest limitation to practical use of the model for pesticides in the Derg case study catchment.

Farm level data required to populate the “CREATE” Farm spreadsheet in the model are not freely available for NI or RoI. All farm data submitted to the departments of agriculture (DAERA in NI and DAFM in RoI) are confidential and cannot be used without consent. Census data for an area or by farm type can be accessed but still requires an agreement with the agriculture departments. For electoral districts with fewer than 10 farms (RoI)/5 farms (NI) this data is not available, affecting use in extensively farmed areas similar to the headwater areas of the Derg case study catchment. These data do not routinely record pesticide usage so it would be less applicable than for nutrients, which are recorded on farm as part of the Nitrates Action Programme and Phosphorus Regulations.

Individual farm data can therefore only be acquired by visits to farmers and on-site survey which would require significant investment in staff time. To correctly represent pesticide use in the Derg catchment it would be necessary to survey farmers and supply them with a table of “typical” pesticide use for their crop types and for them to express how their actual use compares.

iv. Mitigation measure options and costs

Farmscoper’s evaluation of the costs and environmental benefits of each selected mitigation measure (examples of those relating to pesticides are shown in Table 26) uses options provided in English and Welsh agricultural schemes and costs correct at the time of development (Figure 22). The mitigation options for pesticides would be generally applicable to NI but costs will need to be updated to reflect inflation and differences in pricing structures between Britain and NI. The Source to Tap project that is currently being carried out in the Derg catchment is collecting data on the costs of implementing mitigation measures for MCPA. Following the completion of the Source To Tap project in 2021 data may be available to inform the modification of this component of the model. It may also be possible to add additional mitigation options to the list, such as weed wiping or lime applications to inhibit weed growth.

D5.2 tab26
Table 26
D5.2 fig22
Figure 22

2.2 Implementation

At the consultation meeting with stakeholders, an overview of Farmscoper was presented to the stakeholders and their opinions on the utility of the DST were discussed.

  • In practice, is the DST suitable for your case study area? Farmscoper provides an advanced modelling approach to estimate diffuse losses of contaminants from single or multiple farms and farm systems up to catchment scale and quantifies the expected impacts and economic costs of mitigation on those losses to water and air. As such it is a potentially powerful tool to support water managers in prioritising the most effective mitigation options in drinking water catchments. In the consultation with stakeholders all expressed an interest in using the DST and requested that we obtain information on modifications necessary to make it applicable to NI/RoI.
  • If the DST proved useful, please outline the benefits delivered for your target application. In the Derg catchment water companies must treat drinking water at abstraction to remove pesticides (involving an expensive filtration system). As the cost of water treatment increases as contamination increases water companies are considering providing incentives for farmers to improve pesticide practice and use alternatives. In managing a limited budget the water companies need to identify which target groups (e.g. farm types) and mitigation options will deliver the greatest impact on water quality. Farmscoper allows multiple scenarios of mitigation options to be trialed on different farm types and provides a quantitative output as to the costs and environmental impact of those options.
  • How did it compare to what was previously available in your case study area? There are currently no similar DSTs available for this region. There is a gap for such a DST in this region, not just for pesticides but for all diffuse contaminants.
  • Any issues affecting DST utility in your case study area? What steps would be necessary to remove this obstacle to application? Testing identified issues relating to climate, soil and land use practices which will need to be addressed if the model is to be applied in this region.

Key issues covered in an assessment of suitability/utility included:

  • Differences in pesticide usage between Ireland and England/Wales The database behind Farmscoper will need adaptation to the pesticide applications and usage statistics for NI/RoI. The database in Farmscoper is locked so the developer would have to be involved in modifying this, and may require funding to do so. However, the practical element of making these changes would not be difficult and all data required are available in NI/RoI.
  • Geo-climatic differences between Ireland and England/Wales Climate, topography and soil type in the case study is markedly different from England and Wales where the model was developed. Ireland has, in general, higher runoff rates and monthly rainfall and evapotranspiration figures differ from England/Wales. Most of these coefficients are pre-calculated for Farmscoper and would require MACRO and SWAT to be re-run for some scenarios, in collaboration with the model developer.
  • Data requirements and data deficits Farm level data for pesticide use is not routinely collected in farm surveys and general farm data is not freely available in NI or RoI. Negotiations with government departments (who hold such data) would be necessary to allow access for potential users, such as water companies. However, farm census data for various farm types could be used, although the pesticide usage component of such data is limited.
  • Mitigation measure options and costs The mitigation options and costs for the case study area need updating to account for local pricing structures and to allow for the effects of inflation since the model was developed. In addition a number of alternative mitigation measures that are being trialled by the water companies and NGOs could be included in the model options. This is relatively straightforward but would need the developer’s permission and assistance.

Overall, to bring Farmscoper into use in the case study catchment, investment would be necessary to facilitate the developer (ADAS) to make the necessary adaptations. This is something the water companies would consider. We have requested but to date not received feedback from the developer.

Table 29. Advantages and disadvantages of Farmscoper in an Irish context.

Advantages Disadvantages
The capability to evaluate cost-benefits of combinations of mitigation measures is a potentially powerful tool to support water managers in drinking water catchments Greater emphasis on nutrients than pesticides, which is the case study issue. Pesticide usage is simplified and based on UK statistics – different from NI/RoI.
User-friendly Excel-based interface. Outputs as clear graphics and tables. Climatic and soil components of the model are based on England/Wales. RoI/NI are different (e.g. higher runoff rates) so adaptation necessary for further use.
Functionality from single farm to catchment scales. Farm level data availability is limited in NI/RoI due to farm confidentiality.
Strong scientific basis to the model export coefficient approach Mitigation measures and costs need to be updated for NI/RoI

3. Phytopixal

3.1 Assessment and testing

Phytopixal is a GIS protocol proposed by Macary et al. (2014) to evaluate the risk of diffuse pesticide contamination of the rivers located in the Corteaux de Gascogne region of south-west France. The protocol is used to generate a spatial risk assessment for pesticides based on a number of physical characteristics such as inter alia slope, land use and proximity to waterbodies. Slope was derived from a 5 m high resolution digital terrain model (DTM), whilst land use was taken from the CORINE (2012) dataset and the location of waterbodies was taken from the OSNI and OSi mapping databases. Proximity to the waterbody was classified into three risk level (< 30m, 30m - 100m and > 100m) based on the straight-line distance between pixel and waterbodies. These distances were chosen as a first assessment of the impact of distance on the likelihood of fast flow reaching the waterbody. An in-depth analysis of the catchment would require an evaluation of local characteristics. Soil maps were drawn from the Irish National 1:250,000 Soil Map (Irish Soil Information System, 2015) and the General Soil Map of Northern Ireland map (AFBI, 2009), which were merged, and the Standard Percentage Runoff (SPR) (the proportion of rainfall that contributes to the increase in surface runoff) for each soil type in Northern Ireland was adopted. Expert knowledge was used to determine appropriate SPR values for soils in the RoI. All data not presented as a raster dataset was converted to a raster grid by overlaying the non-raster dataset with an empty raster grid that had the same dimensions as that of the DTM. The value to be used in each cell in the blank raster was taken as the most common value in the underlying layer.

A risk profile was developed for each catchment characteristic (Slope, land use, proximity to a waterbody and SPR) through use of expert opinion and the scientific literature. Slope angle was divided into 5 classes based on the Natural Breaks (Jenks) technique (the “Goodness of variance fit”). This technique seeks to group data in such a way that variance in values within a class is minimised, whilst the variance between classes is maximised (Seamon et al., 2013, ESRI, 2016). Land use risk was determined through calculation of the mass of MCPA added to each land use identified in the catchment, based on the national trends reported in the Pesticide Usage surveys for Northern Ireland 2016 (arable) and 2017 (grassland) (Lavery et al., 2017, Lavery et al., 2016).

The weightings used for each parameter are shown in Table 27 and their distribution across the catchment are shown in Figure 23.

D5.2 tab27
Table 27
D5.2 fig23
Figure 23

Figure 23 (A) shows that, whilst there are steeper slopes distributed across the catchment, there is a higher concentration of these slopes in the headwaters of the catchment. Figure 23 (B) indicates that higher risk soil categories are distributed across the catchment, but particularly in the west and in the headwaters of the catchment. Figure 23 (C) demonstrates the extensive network of streams and waterbodies across the catchment to which buffer zones of < 30m, 30m – 100m and >100m have been applied. These equate to areas of high risk for contamination of water body, moderate risk and low risk respectively. Figure 23 (D) illustrates the distribution of land use across the catchment. This data suggests that the highest risk activities are located in the centre and the east of the catchment, in the lowland areas of the catchment. There are also some moderate risk land uses at the very top of the catchment.

The risk assessment calculation was carried out according to Eq. 1 for each cell within the raster grid and the results are shown in Figure 24.

D5.2 eq01

Where S is Slope risk, B is Buffer zone risk, So is Soil association risk, L is Land use risk

D5.2 fig24
Figure 24
D5.2 fig25
Figure 25
D5.2 tab28
Table 28

The areas of highest risk (categories 4 and 5) are predominantly located in the east of the catchment, particularly close to the larger watercourses. Although urban areas were given a relatively high risk weighting in land use because of the low levels of domestic user training associated with garden centre bought herbicides, the area around Castlederg (the largest urban area in the catchment) still returned a moderate risk value indicating that agricultural use is of greater significance. The importance of land use in this protocol can also be seen in the west of the catchment where there are areas of intermediate to high risk (3 – 5) identified.

The presentation of the data may also be altered to better suit the target audience, as is illustrated in Figure 25 (A) and (B). Figure 25 (A) shows the Derg catchment divided into 10 sub-catchments according to the location of a spatial grab sampling regime carried out in the catchment in 2018. The risk of pesticide contamination in each sub-catchment is calculated according to Eq. 2 and the resulting cumulative weighting values may then be ranked. In this case the results suggest that the most-downstream sub-catchments are the most risky with respect to MCPA contamination of water and that the central sub-catchments are at moderate risk.

D5.2 eq02

where r is the risk value for each cell within the sub-catchment A is the area of the sub-catchment

Figure 25 (B) shows the impact of re-sampling the original data to a cell of 250 000 m2, using Eq. 3.

D5.2 eq03

where r is the risk value of the 5m cells within the new 500m raster a is the area of the 500m raster cell Figure 25 (B) retains a slightly more spatially explicit presentation of the data in that the importance of land use remains visible in the central area of the catchment, but the fine detail is smoothed out in favour of a broad-brush presentation of the data.

A higher resolution map of land use in the NI section of the catchment was developed, based on a visual analysis of 2015/16 aerial Imagery (Figure 25 (C)) that allowed the assignment of land use to individual fields. The categories of land use and their associated risk class are shown in Table 28. The risk analysis in Eq. 1 was then repeated in order to determine if the increased resolution of land use data would significantly alter the final output (Figure 25 (D)).

Whilst the overall pattern of risk is unchanged by the adoption of higher resolution data, it is clear that this approach can identify smaller areas of land that pose a threat and so facilitate the more focussed selection of targets for the deployment of mitigation strategies.

3.2 Implementation

At the consultation meeting with stakeholders, overviews of SCIMAP and Phytopixal was presented to the stakeholders and their opinions on the utility of the DST were discussed. As SCIMAP and Phytopixal are similar in approach their implementation is considered jointly.

  • In practice, is the DST suitable for your case study area? SCIMP and Phytopixal are both GIS-based and so their applicability to an area is determined by the availability and quality of data the user inputs. Whilst this is a point that we return to later, in principle both Phytopixal and SCIMAP are suitable for implementation in NI/RoI.
  • If the DST proved useful, please outline the benefits delivered for your target application. In both cases SCIMAP and Phytopixal allow for the visualization of risk across the area of interest, based on the spatial data provided. This means that it is possible to identify areas of the catchment that pose the greatest threat to water quality and thus targeted mitigation measures (both physical and educational). This approach also generates visual imagery which may be more readily accessible to a wider audience than tabulated data. SCIMAP and Phytopixal also both offer approaches to modelling the catchment that require only basic GIS skills and catchment information. Phytopixal allows the user more latitude in selecting the input data and thus to optimise the output towards their contaminant of interest.
  • How did it compare to what was previously available in your case study area? As GIS is a well-established technology Phytopixal and SCIMAP have not provided significant technological advance over previous approaches as the same GIS tools are used as would previously have been adopted by the analyst. The advantage of these approaches is, however that SCIMAP and Phytopixal both provide a documented approach using standardized inputs and thus should ensure that results between studies are more easily compared.

Key issues affecting DST utility in your case study area and the steps necessary to remove the obstacle to application included:

  • Applicability to the Irish landscape Spatial risk mapping for nutrients and sediment risk in overland flow has been modelled and tested extensively in an Irish context (e.g. Thomas et al., 2016a, Thomas et al., 2016b, Thompson et al., 2013) so both approaches would be easily adapted to the case study area. The models are both focussed on surface water/erosion potential and so neither is suitable for use in areas where groundwater contamination is a serious concern. However, these models may be helpful in all areas to identify regions of the landscape where surface water pools and so where the water may be potentially passing to groundwater. In areas where surface water movement is important, both models may be used to differentiate areas by risk through the use of simple catchment parameters that are likely to be available at the start of a project. In the case of NI/RoI, the biggest challenges to model quality currently is the quality of input data. Satellite derived elevation models tend to be less accurate and at a lower resolution than LiDAR datasets, but are cheaper to generate. No national LiDAR datasets are available in NI or RoI, unlike other EU States. The author of SCIMAP have confirmed that, whilst the online version of the software is the focus of their development efforts, this will only cover Great Britain (England, Scotland and Wales). The previous version of SCIMAP that was bundled with SAGAGIS remains available, but will no longer be supported. Going forward this may impact on the applicability of this approach to the NI/ROI area.
  • Data quality During this exercise the modellers noted that the quality of the DTM was not the same across all parts of the Derg (Figure 28). In the upland areas utilised for forestry, it was noted that the topography was presented as being very flat, whilst similar, but unforested parts of the catchment were more rugged. This is an artefact of DTM preparation that will adversely impact on the quality of slope angle calculations. The fact that the Derg is a cross-border catchment also posed problems during the analysis as the Northern Ireland and Republic of Ireland datasets do not always record the same parameters and obtaining the data can incur considerable costs.

Interest around Phytopixal and SCIMAP was focussed on the potential of these approaches to simulate risk in a catchment using already available data, and thus before projects undertook extensive fieldwork campaigns to gather very high resolution data. Stakeholders did raise concerns, however, about the vulnerability of both approaches to low quality data currently available and the time and cost that would be associated with developing higher quality data. Of particular concern was land use as the CORINE dataset does not represent Irish land use patterns well (Cawkwell et al., 2017).

  • Will you be able to implement the DST in practise? Or elements of it? Which elements? Overall this investigation has shown that both Phytopixal and SCIMAP are suitable for implementation in NI/RoI and that the results are useful in developing an increased understanding of the threat of diffuse pesticide pollution. However, currently their utility is significantly limited, by the quality of the available data.

It has also been shown that both models are appropriate for use at a variety of scales as the output is spatially explicit and can be managed to suit the audience. It was noted that if the methodology used to divide the calculated risk value into risk categories was altered to percentiles (e.g. the top 20% of risk values assigned to risk category 5, the next 20% of risk values to 4, and so on) and if the same risk profiles were used, then comparison of results between sites would be possible. However, these approaches use subjective assessment of risk, rather than observed data and so will always be advisory in nature.

Whilst SCIMAP makes explicit mention of pesticides, the user does have the opportunity to award different risk weightings to individual land uses if required which allows for expert knowledge to be used to increase the impact of high risk land uses. This is essentially analogous to the way in which land use risk is managed in Phytopixal.

Table 30. Advantages and disadvantages of Phytopixal in an Irish context.

Advantages Disadvantages
Input data (quality and parameter) selected by user. Land use data is of low resolution in NI/ROI – time intensive to improve
Spatial presentation of results Soil classification map is of low resolution in NI/ROI.
Results can be re-sampled for lower resolution data presentation Digital Terrain Models are of lower resolution in NI/ROI
Data needed is likely to be available at the start of the project Protocol, rather than a GUI so knowledge of GIS required

4. SCIMAP

4.1 Assessment and testing

SCIMAP uses topographic data to model, for each point in a landscape, the probability of overland flow being generated and a pathway to the river network. With connectivity comes a risk that contaminants such as nutrients, sediment, pathogens or pesticides will be entrained and transferred.

Although SCIMAP was originally developed as an open source desktop application for non-commercial use under a Creative Commons license, development effort has now moved to an online format (http://www.scimap.org.uk/, Figure 26) which is not available for catchments outside Great Britain. The older GIS-packaged software remains available and it was this version of SCIMAP that was used in this analysis.

D5.2 fig26
Figure 26
D5.2 fig27
Figure 27
D5.2 fig28
Figure 28
D5.2 fig29
Figure 29

The basic data requirement of the approach is a DTM to which other data sets such as the digitised drainage network (used to hydrologically correct the DTM to remove bridges etc.), land use and rainfall for the area of interest are added.

For the Derg catchment, hydrological connectivity (based on the Topographic Wetness index (TWI) (Beven and Kirkby (1979)) was calculated using a 5m DTM (Figure 27.A) that was hydrologically corrected by ‘burning’ the drainage network (Figure 27.D) into the elevation model to remove bridges and other obstacles to downslope movement of water in the model. The TWI was then calculated based on the slope and flow accumulation from upslope areas to every point in the raster grid (Figure 28 – for clarity only the 10% highest risk areas are shown).

Land use data was drawn from the CORINE (2012) dataset (Figure 27.B). Average rainfall values were taken from Met Eireann (Irish Meteorological Office) long term rainfall averages from 1981 – 2010 (Figure 27.C) and the location of waterbodies was taken from the OSNI and OSi mapping databases. Subsequently, this analysis was repeated with a higher resolution field-scale land use map defined from high resolution aerial imagery (Figure 27 (C)).

Topography is steepest in the headwaters of the catchment in the west and north (Figure 27 (A)) and becomes flatter as the river widens into the flood plain in the lower catchment. Figure 27 (B) illustrates the distribution of land use across the catchment, with risk categorisation for MCPA apportioned to each CORINE land class based on average pesticide applications rates calculated from the Northern Irish pesticide usage surveys (Lavery et al., 2017, Lavery et al., 2016). The highest risk activities, with respect to MCPA usage, are located in areas of poor soil and less intensive agriculture in the centre and west of the catchment, whilst lower risk activities are more common in the peatland and mountain areas in the uplands. Rainfall in the catchment is highest in the mountains in the west of the catchment (~2700 mm yr-1) and approximately 50% lower in the east (Figure 27 (C)).

The model uses the information provided in Figure 27 (B) and Figure 27 (C) to weight the TWI and identify those parts of the catchment that are most likely to act as source of contamination with pesticides – i.e. where source and pathway intersect. An area with a high land use risk, coincident with high TWI poses a greater relative risk to water quality and would be a prioritised target for mitigation.

As previously discussed the CORINE (2012) dataset uses a resolution of 1 km and this is coarser than field scale at which management is undertaken in this part of the island - typical field sizes in Ireland are between 0.3 and 1 ha (Figure 25 (C)). The SCIMAP protocol was re-run with this higher resolution dataset available for the NI section of the catchment and the results (Figure 29) indicate that there are only very small, highly dispersed parts of the agricultural catchment that pose the greatest risk of contributing flow via surface pathways.

4.2 Implementation

Because SCIMAP and Phytopixal are similar in approach their implementation is considered jointly in »section 3.2 above.

Table 31. Advantages and disadvantages of SCIMAP in an Irish context.

Advantages Disadvantages
Input data quality selected by user Land use data is of low resolution in NI/ROI – time intensive to improve
Spatial presentation of results Soil classification map is of low resolution in NI/ROI
Data needed is likely to be available at the start of the project Digital Terrain Models are of lower resolution in NI/ROI
Desktop version bundled with SAGA GIS (Freeware) No explicit mention of pesticides - Expert knowledge needed
  Web version not currently available outside Great Britain

 


Notes:

For full references to papers quoted in this article see » References

Download the full report for the original figures and tables

 

Go To Top