This study focuses upon the entry level stewardship (ELS) as it is both very widespread, incorporating 5 M ha of English Farmland (Natural England 2013a), and has many options that are applicable to other UK and European agri-environment schemes (AES). ELS allows enrolled participants to select from a suite of management options, each with a point value. Participants must select 30 points worth of options per hectare of their enrolled holding and are paid £30 per hectare in return. These payments total £163 M per annum as of January 2013 with a further £1.4 M spent on monitoring (Natural England 2013a). Due to the timing of the expert survey, this study focuses upon the third edition of ELS (Natural England 2010), although a fourth edition is now in use (Natural England 2013b).
Estimating habitat benefits
To evaluate the potential benefits of each option for providing good quality habitat for pollinators, an expert panel survey was conducted. As primary ecological data on the responses of pollinators to ELS management options is limited to a few options and focal pollinator taxa (e.g. Potts et al. 2009; Pywell et al. 2011; Carvell et al. 2007), an expert panel was used to evaluate the relative benefits of each ELS option to pollinator habitat. Similar methods have been used to assess pressures (Kuldna et al. 2009) and model habitat suitability (Lonsdorf et al. 2009) for pollinators. Experts were academics with at least three publications on pollinator ecology and non-academics recommended on the basis of 10 or more years’ experience in UK bee or hoverfly ecology. In total 35 experts were approached in March 2010. Delphi panel and Bayesian models (Czmebor et al. 2011) were considered but not pursued due to the difficulty in eliciting multiple responses and limited primary data available for modelling outcomes.
Experts were surveyed via e-mail, following a small pilot survey, with reminders sent to non-respondents after 2 and 4 weeks. Respondents were asked to rate each option on providing good quality habitat (i.e. suitable nesting or forage resources) for a wide range of wild pollinators (bees and hoverflies) in farmed landscapes across the UK on a scale from 0 (no benefit) to 3 (great benefit). This simple scale was selected due to the volume of options under consideration potentially increasing respondent fatigue. Experts were also asked to report their confidence in their response on a four point scale from (0) not confident to (3) very confident. From this the Pollinator Habitat Benefit (PHB) values, weighted by expert confidence, of each option were calculated as:
$$PHB_{i} = \frac{{\sum_{e = 1}^{E} {(H_{ei} } \times C_{e} )}}{{\sum\nolimits_{e = 1}^{E} {C_{e} } }}$$
(1)
where H
ei
is the habitat quality score allocated by expert e to option i and C
e
is expert’s self-reported confidence. To avoid respondent fatigue, only one confidence measure was taken for all options. To control for the effects of between expert variation (Czmebor et al. 2011) this was then divided by the total confidence values to produce an average across all experts within the original 0–3 scale.
Redistributing ELS options
Using the expert PHB weights, a series of models were developed to redistribute the 2012 composition of ELS options in a manner which reflected their relative benefits for providing habitat for insect pollinators. These models allocated units of each option based upon the benefit they provided to pollinator habitats relative to other options within specific categories; with the most beneficial option allocated the greatest number of units and the least beneficial allocated the least units. This method was chosen over optimisation models for the sake of methodological simplicity, particularly given the high number of variables involved, and to avoid scenarios dominated by high benefit and/or low cost options. The changes in costs and habitat benefit (measured as the sum value of PHB) were then appraised for each model. The number of units and total ELS points generated by each option as of December 2012 were obtained from Natural England databases (Cloither 2013, Pers Comm) excluding options that are no longer available (e.g. EM1-4) or those that relate only to historic or built features (e.g. ED1-5) and water bodies. Mixed stocking (EK5) was also excluded to avoid double counting as this option can be combined with other grassland options. Options relating to severely disadvantaged areas (EL1-6) and ELS variants, (organic and upland ELS), were not included to reduce respondent fatigue and maintain model simplicity by only considering broadly applicable options.
The remaining options were grouped into categories based upon their management units (hedge/ditch options, managed in metres/hectares; further subdivided into grassland and arable, and plots/trees) and the area and points values of options within each category were summed to produce a baseline estimate (Table 1). For option EC4, which could be present in both grassland and cropland, the area and points were distributed proportionate to the relative area of the two groups; 24 % cropland and 76 % grassland (DEFRA 2013).
Table 2 Weighted and unweighted mean PHB scores attributed to 2010 ELS options
Table 3 Number of units of each ELS option after redistribution
Table 4 Total private and public cost:benefit changes under the three ELS redistribution models
Table 5 Total units of each option type under the three ELS redistribution models
Table 6 Changes in total costs to producers and total pollinator habitat quality benefits under the three sensitivity analyses
For each option a habitat quality (HQ) score was calculated as:
$$HQ_{i} = PHB_{i} \times ELS_{i}$$
(2)
where ELS
i
is the ELS points value (and therefore farmer payment) attached to each unit of option i. This weights the quantitative metric of option quality relative to the scale of their implementation as a single hectare of habitat will typically provide a substantially greater total resource than a single metre of habitat. How ELS points are derived is presently unclear as although EU rules state they must be based upon their costs, including income foregone, earlier and recent revisions taking into account the biodiversity benefits of options have moved away from this initial approach (Natural England 2012, 2013b). As such ELS points largely represent relative general biodiversity benefit, which is then weighted by the expert PHB scores. To give a measure of the value of each option relative to all other options with the same unit category (c), proportional habitat quality (pHQ
ic
) values are then estimated as:
$$pHQ_{ic} = \frac{{HQ_{ic} }}{{\mathop \sum \nolimits_{i = 1}^{C} HQ_{ic} }}$$
(3)
The pHQ score for option i therefore represents its benefit to pollinator habitat relative to all other options within category c. pHQi scores are therefore always between 0 and 1 and the sum of all pHQi scores for a given category of c always equal 1. Using these pHQ values, three variant analyses were conducted to redistribute the overall composition of options towards a composition which reflects the relative benefits of the options for providing good quality habitat for pollinators. Model A generates a mix of options that redistribute the absolute area of ELS options currently utilised to reflect their relative benefits to pollinator oriented habitat. It thus redistributes the composition of options based upon the total utilised area of options within each category (i.e. the most beneficial option will take up the greatest number of units and so on). The area of different option categories is maintained to reflect current uptake patterns and preferences. This model allows the total number of ELS points, and therefore the total area of English farmland enrolled in the scheme, to expand, however no additional area of land is taken out of production.
$$U_{ic} = \mathop \sum \nolimits U_{c} \times pHQ_{ic}$$
where U
ic
is the redistributed number of units of option i in category c, Uc is the total number of units (meters, hectares or trees/plots) in the category and pHQ
ic
is the percentage of total HQ (calculated as in Eq. 2) in each option represents within the category. As such each option is allocated a percentage of the total units of category c based upon their relative benefit to pollinator habitat.
Model B generates a mix of options that maintains the current ELS budget, allowing the absolute area of options within the four categories to change. This is accomplished by redistributing the percentage of total ELS points in each option category based upon their pHQ scores (i.e. the most beneficial option will account for the greatest number of points within the category and so on). The number of units of each option is then the total points divided by the options ELS points value. Again, expenditure on categories is maintained to better reflect current enrolment and preferences. This allows the absolute area covered by ELS options to vary, however the total area enrolled in ELS, and the subsequent taxpayer payments, will remain the same.
$$P_{ic} = \mathop \sum \nolimits P_{c} \times pHQ_{ic}$$
where P
ic
is the total ELS points accounted by option i in category c, P
c
is the total ELS points produced by options in category c.
Model C also maintains current ELS budget, however, under this model the ELS points of all options are pooled regardless of their category and the redistribution is based upon the habitat quality benefits of each option in relation to all other options, regardless of their category. As such the most beneficial of all available options will represent the greatest percentage of total redistributed ELS points and so on. As with model B, this allows the number of units of each option to change, although now there is a degree of substitution between option categories and which may affect their prevalence in the overall ELS. To prevent the outputs of this model from being dominated by arable and grassland options, many of which are worth several hundred ELS points, the ELS points for hedge/ditch and plot/tree based options were multiplied by 1,000 (assuming 1 m2/unit of hedge/ditch options) and 10 (assuming 100 m2/unit of plot options) respectively to scale points of these options relative to 1 ha.
$$T_{i} = \mathop \sum \nolimits T \times tHQ_{i}$$
T
i
represents the ELS points accounted by option i, T is the summed points value of all ELS options concerned and tHQ
i
is the percentage of total HQ of all options represented by each option.
For each model the total ELS points and number of units for each option were recalculated to compare with the baseline. Once the ELS composition of each model was calculated the total number of units for each option in each model and the baseline were then multiplied by the average per annum costs per unit (See Table 7 in Appendix) using the costs from the SAFFIE (2007) and Nix (2010), following the establishment and management guidelines laid out in each option (Natural England 2010). Many options had low or no cost. In the case of options requiring sown mixes of plants, the average sowing cost per hectare using a range of mixes from various suppliers was used (See Table 8 in Appendix), however mixes with a total cost of >£1,000/ha, most of which are designed for advanced habitat restoration, were excluded to prevent skewing. The ratio of total costs of implementing all options against total ELS payments (a measure of farmer benefit) were calculated as farmer cost:benefit. The total units of each option were multiplied by their respective HQ score to produce an abstract quantitative measure of the overall benefits of the final option mix for pollinator habitat. As pollination services are largely a public good, each models benefit scores can then be compared with the total ELS payments to gain a measure of public cost:benefit.
Table 7 Estimated costs of implementing ELS options
Table 8 Number of species of various plant families occurring within ELS applicable seed mixes
Sensitivity
As with all models utilising expert opinion, there are a number of ways the values used in this study can be biased; foremost, individual expert uncertainty and overconfidence can cause substantial skewing of the results towards certain options. Therefore each model was recalculated by Jackknifing, removing one expert each time before calculating the PHB. The percentage difference in total farmer costs between each Jackknife and the average of all Jackknives was then compared with the version for all experts. Strong effects from this deletion compared with the “all experts model” would indicate that the model is biased by highly polarised expert opinions. Similarly, expert reported confidence may not be a reliable means of weighting the PHB scores—therefore each model was recalculated using unweighted PHB scores to determine the percentage change caused by weighting. Strong changes would indicate that the weighting system creates an inherent bias. Finally, it is possible that using expert opinion to weight ELS points may not produce an option mix which is substantially different from developing a model based on ELS points alone. Consequently each model was recalculated using only ELS points to estimate relative PHB. Strong differences would indicate that the expert weighting has a substantial impact in guiding the redistributions over ELS points alone.