Landscape-scale simulations as a tool in multi-criteria decision making to support agri-environment schemes

Highlights

• Species are impacted differently by various management practices across different landscapes.

• The effect of a particular management strategy on a species is dependent on landscape context.

• Landscape models help predict effect of management on various species across landscapes.

• Landscape simulations can provide an integrated approach to multi-criteria decision-making.

Abstract

Increasing concerns over the environmental impacts of agriculture in Europe has led to the introduction of agri-environment schemes (AES) into the Common Agricultural Policy to help mitigate biodiversity loss by encouraging farmers with subsidies for implementing environmentally-friendly farming techniques. However, effectiveness of AES has been mixed and only partially successful in achieving desired outcomes. To improve effectiveness and reduce costs, multi-criteria decision analysis (MCDA) can help support decision-making and determine the most effective management action. Although MCDA has great potential for evaluating policy measures, it rarely considers the context-dependency of species responses to management practices across different landscapes. Landscape simulations can, therefore, be valuable for reducing the uncertainties when predicting the consequences of management actions. A potential suitable simulation system is the Animal, Landscape, and Man Simulation System (ALMaSS), a mechanistic simulation with can improve MCDA with the automatic integration of landscape context and a species ecology and behaviour. The aim of this study was to demonstrate the effectiveness of ALMaSS in evaluating AES management practices across different landscapes and estimate their ability to achieve the proposed conservation outcomes in three species of conservation interest. In this study, the effect of a particular management strategy on a species was dependent on the landscape context, in our case, a combination of landscape structure and the type and distribution of farms, and varied depending on the metrics being measured. We demonstrate how simulations can be used for MCDA to select between management strategies with different costs. Despite the complexity of ALMaSS models, the simulation results provided are easy to interpret. Landscape simulations, such as ALMaSS, can be an important tool in multi-criteria decision making by simulating a wide range of managements and contexts and provide supporting information for filtering management options based on specific conservation goals.

Introduction

The Common Agriculture Policy (CAP) was introduced in the 1960s to support European farmers and increase agricultural productivity to ensure availability of quality food. The CAP is consistently evolving based on current needs and the increasing concerns over the environmental impacts of agriculture has led to major reforms in the last decades, with a larger focus on environmental conservation and measures to support and manage biodiversity in agricultural landscapes (Lazíková et al., 2019). The current 2014–2020 reforms promote several tools to mitigate the environmental impact of agriculture with one of the main policy tools agri-environment schemes (AES), which financially compensate farmers for implementing measures that benefit wildlife and help protect the environment (Kleijn and Sutherland, 2003). Because certain farming practices are particularly favourable for the environment, some examples of common AES actions include crop rotation, introducing buffer strips, managing livestock, rotational set-aside, enhancing habitats for wildlife, and organic farming. Implementation of AES has already led to an increase in the abundance and diversity of a wide range of animal and plant species in areas which may have otherwise not have been afforded such protections (Science for Environment Policy, 2017). However, research suggests that these schemes may only be partially successful in their desired objectives and refinements are necessary to improve effectiveness and increase the cost-effectiveness of various AES measures (Kleijn et al., 2001; Berendse et al., 2004; Kleijn et al., 2004, 2006; Batary et al., 2015; Żmihorski et al., 2016).

Although the success of any management action will depend upon the goal of management or the policy implemented and the metrics used, there are often multiple objectives, and balancing these with limited budgets can be difficult. Fortunately, AES are adaptive and continually revised since they are typically multi-annual projects persisting during CAP periods. The adaptability of these schemes allow for the refinement of techniques and strategies using decision-making strategies to reduce uncertainties and maximise success. In one example, monitoring corn bunting (Emberiza calandra) populations identified new management options (i.e., late silage-cutting) which helped refine AES options targeting the key species (Perkins et al., 2011). Besides monitoring, a modern and more cost-effective approach is multi-criteria decision analysis (MCDA), a knowledge synthesis methodology which explores potential outcomes of multiple actions according to several criteria (e.g. see Belton and Stewart, 2002; Davies et al., 2013). MCDA's strength is its ability to handle and analyse the qualitative and quantitative data involved with complex multidimensional environmental decision-making. So far, it has been used to evaluate AES effectiveness in the UK (Park et al., 2004; Westbury et al., 2011), assess long-term environmental impacts of AES on land use in Denmark (Vesterager et al., 2012), and ex post evaluation of AES in the EU (Finn et al., 2009; Vesterager et al., 2012). Although MCDA is a useful tool in providing the foundation for policy decision-making, due to the scarcity of ecological information and the lack of recognizing the relationship between management action and environmental effect, as a relatively new method it use has so far been limited in scope.

A major challenge is that even landscapes that appear simple, such as agricultural lands, consist of complex ecological networks which make it difficult to accurately predict outcomes and link intervention measures with observed results. AES should be selected based on trade-offs since a single solution will rarely encompass all groups of interest and the success of any management scheme may vary between taxa and species (Marshall et al., 2006). For example, although some interventions may benefit a variety of organisms (e.g., species-rich field boundaries can benefit both birds (Vickery et al., 2009) and insect pollinators (Batary et al., 2010), measures aimed at one taxonomic group or species may be detrimental to others within the ecosystem (e.g., delayed mowing may be beneficial to some bird species, while negatively impacting species which prefer short and sparse ground vegetation (Żmihorski et al., 2016). Such conflicting results in AES measures can have direct consequences, as illustrated with the local extinction of a butterfly population in the Czech Republic (Konvicka et al., 2008). Consequently, landscape context can also have large and contrasting effects on management outcomes (Tscharntke et al., 2005; Marshall et al., 2006). For example, in a meta-analysis evaluating the effectiveness of AES, species richness and abundance of pollinator species were found to be primarily driven by landscape context and farmland type, with more positive responses in croplands (vs. grasslands) in simple (vs. cleared or complex) landscapes (Scheper et al., 2013). Understanding how species interact with their environment and the effect of landscape context is thus essential when evaluating policy measures.

Due to such issues, AES have increasingly focused on the landscape scale for improving outcomes, such as providing wide environmental variation in close proximity through habitat heterogeneity (Benton et al., 2003) and corridors which link habitats for species and enhance biodiversity (Renwick and Lambin, 2011; Delattre et al., 2013). However, determining the precise management action to be undertaken becomes increasingly complex since even small differences in details can produce diverse outcomes. This includes in-field management activities in AES often involve creation of habitat enhancements and restoration which may vary in shape, type, and management. For example, although field margin measures and within-field overwintering refuge “beetle-banks” (to enhance populations of beneficial arthropods) are typically linear in form (Collins et al., 2002), created habitats are flexible with the width of field margin a critical factor in determining the number of species it contains. This is partly due to the well-known species-area relationship (e.g. see Ma et al., 2002), but is also a function of the buffering effect against agricultural inputs. In one study, a 3 m buffer strip reduced spray drift to the field margin by 95%, while a 6 m strip removed it completely (De Snoo and De Wit, 1998). Moreover, the influence of management can also vary depending on the metrics being measured. In the case of grasshoppers, although landscape context determined species diversity, grasshopper density in field margins was heavily influenced by type of management (Badenhausser and Cordeau, 2012). Making reliable predictions when comparing management actions therefore requires detailed spatial resolution and mechanistic relationships to be accurately represented.

Simulation modelling can be valuable in filling these knowledge gaps and the limitations of MCDA by reducing the uncertainty in predicting the consequences of management actions (Drechsler, 2004) by directly incorporating factors and interactions in the simulation. In fact, the use of simulation models to evaluate or plan policy implementation has become increasingly more common, ranging from simple models to explain and predict the relationships between policy and effect (e.g. Hailu and Brown, 2007) to complex hierarchical-spatial modelling process approaches (e.g. Rouillard and Moore, 2008). Simulation models have been commonly used in land management issues such as forest planning (e.g. Summers et al., 2015), marine park zoning (Bruce and Eliot, 2006), and to determine ecological and cost-effective solutions for grassland biodiversity (Sturm et al., 2018). However, the wider application of simulation models for policy analysis has been slow due to the conventional use of monitoring and experimental approaches (Parry et al., 2013). For simulation models to be useful as a method to evaluate policy measures, it requires both depth, in terms of considering its applicability with a detailed realistic mechanistic representation capable of capturing the spatiotemporal interactions common in ecological systems; and breadth, to cover the social (farmers decisions to achieve certain goals) and ecological components (geographic areas) where the policy might be applied.

A potential suitable simulation system for evaluating management policy is the Animal, Landscape, and Man Simulation System (ALMaSS), a mechanistic simulation which can improve MCDA by the inclusion of species modelling and landscape context (Topping et al., 2003). ALMaSS has been already been successfully demonstrated and used in a variety of environmental contexts, including assessment of organic farms on wildlife (Topping, 2011b), energy crop production (Gevers et al., 2011), and recently, to evaluate bidirectional feedbacks between farmer decision-making and land use impacts on wildlife (Malawska and Topping, 2018). However, the applicability of simulations has typically been limited by the resource demands of landscape simulations (e.g. hand digitisation of landscapes) and restrictions on the availability of landscape datasets, resulting in using simple and unrealistic landscapes. ALMaSS has recently removed this limitation through the use of readily available GIS mapping data as well as the increasing availability of data from the Integrated Administration and Control System (IACS), and the associated Land Parcel Identification System (LPIS). This data from the EU subsidy support schemes now allows for the rapid creation of highly detailed and accurate agricultural landscapes (Topping et al., 2016). These advancements in geographic data and simulations now allows researchers to easily evaluate the influence of landscape-context to management measures for the first time, and open up the potential of landscape-scale simulation models, such as ALMaSS, to be used as a valuable tool in MCDA and policy decision-making.

The aim of this study was to demonstrate the effectiveness of ALMaSS in MCDA by evaluating typical AES management practices in different landscapes and estimate their ability to achieve the proposed conservation outcomes. We determined the benefits of common in-field management strategies for wildlife management of three typical species of interest in agricultural systems in terms of economic return per unit area and assessed the degree to which the return was dependent on landscape-context.