- freely available
Animals 2013, 3(2), 416-441; doi:10.3390/ani3020416
Published: 16 May 2013
In this review paper we discuss the different modeling techniques that have been used in animal welfare research to date. We look at what questions they have been used to answer, the advantages and pitfalls of the methods, and how future research can best use these approaches to answer some of the most important upcoming questions in farm animal welfare.
The use of models in the life sciences has greatly expanded in scope and advanced in technique in recent decades. However, the range, type and complexity of models used in farm animal welfare is comparatively poor, despite the great scope for use of modeling in this field of research. In this paper, we review the different modeling approaches used in farm animal welfare science to date, discussing the types of questions they have been used to answer, the merits and problems associated with the method, and possible future applications of each technique. We find that the most frequently published types of model used in farm animal welfare are conceptual and assessment models; two types of model that are frequently (though not exclusively) based on expert opinion. Simulation, optimization, scenario, and systems modeling approaches are rarer in animal welfare, despite being commonly used in other related fields. Finally, common issues such as a lack of quantitative data to parameterize models, and model selection and validation are discussed throughout the review, with possible solutions and alternative approaches suggested.
The use of models has greatly advanced in many areas of the life sciences in recent decades, but the range, complexity and type of models developed for use in animal welfare is comparatively poor. This is surprising given the similarities between the aims of animal welfare and those of, for example, the closely related field of animal health epidemiology, where there is great interest in using models (mathematical, statistical and assessment) to understand why, where, how, when and who will be affected by disease outbreaks, and to explore potential control strategies (e.g., stochastic simulation modelling [1,2,3]; decision tree modelling [4,5]; and network modelling [6,7]). In the field of animal welfare, our aims are principally to understand why, where, how, when and who will be affected by the multitude of species-specific welfare problems, and what control strategies we can put in place to prevent these problems arising. However, despite this, we do not see the same type of predictive model as is so often used in disease epidemiology being used to make predictions of welfare problems. One possible reason for this might be the relative lack of data. However, predictive disease models are often based on input data collected from largely varying sources, of varying quality and reliability and combined from multiple previously published scientific studies.
Other related areas of academic research which frequently utilise models, are agricultural economics and sustainability. Here, we see a range of models being developed to explore, for example, the potential margins associated with different farm types and practices, often incorporating elements of animal welfare (although these are usually highly simplistic). These large, multi-factor models are again mostly developed using input data from a large range of sources, of varying quality and reliability. For example, the use of seven, predominantly resource-based, parameters to score animal welfare in a model of sustainability  is, perhaps, not representative of what most animal welfare academics would consider the most critical input data for a model of this type.
Given the breadth of modelling techniques available, and the objectives these models set out to meet, it is perhaps a misnomer to talk of “models” as if they represent one approach. In other fields, such as ecology, where modelling is far more commonly used, it is recognised that models really fall into three main types: detailed (“synthetic”), “minimal” systems, and “minimal for ideas” . Here, the models are defined by what they set out to do. Detailed, or “synthetic”, models aim to produce a detailed description of the sub-components that make a system and how they interact to produce the overall working system. “Minimal” models for systems aim to explain certain types of system, but tend to ignore many characteristics of the real life system and, as such, are not designed to produce specific, detailed predictions. Finally, “minimal models for ideas” are for exploring concepts without anchoring it to a particular species or type of system. Again, this type of model, which includes some of the most famous models in ecology, such as the Lotka-Volterra model, is not meant to produce testable predictions or to be applied to real life situations. We see examples of each of these types of model in the farm animal welfare literature and these will be discussed in section two.
Modelling in farm animal welfare has mostly fallen into the frequentist statistical type, in other words, quantitative analysis of an experimentally derived, or longitudinally collected data set e.g., [10,11,12,13,14,15,16,17,18], where the main aim is to determine which abiotic and biotic factors in an animal’s environment are associated with the development of poor welfare, or change in welfare state, and which indicators can be used to determine welfare state reliably. To date, there has been little use of Bayesian approaches (though see [19,20,21]). In recent years, there has been an increasing trend to also use more statistical and engineering type methods in the formation of welfare indicators e.g., [22,23,24,25,26]. Some of these types of studies have been reviewed and discussed previously e.g., . In this review article, we focus on the other types of models being produced in the field of animal welfare—namely, the assessment models, simulation, optimisation, scenario, systems and conceptual models. This paper is not intended to provide an exhaustive review of all modelling approaches. Rather, the models reviewed herein are those considered to be most promising in the field of farm animal welfare science. We will briefly describe each type of model, outline the types of questions they could be used to answer—highlighting the potential drawbacks associated with the approach—and provide examples from the scientific literature of their use in animal welfare research. Finally, we will discuss where the main gaps in farm animal welfare modelling lie, and suggest possible reasons why this may be and how we might go forward. The aim of this paper is not to provide an instruction manual on using the different modelling approaches, but to open discussion between empirical and theoretical researchers, and to forearm empirical researchers with a set of questions that they can aim to answer with theoretical approaches.
2. Models in Farm Animal Welfare
2.1. Assessment Models
2.1.1. Risk Assessment
One of the most common types of model used in animal welfare research, particularly at the science-policy interface, is risk assessment. Risk assessment characterises the probability of a negative event occurring and quantifies the consequences of such an event. Ultimately, it provides a means of comparing different welfare problems both at the level of the affected individual and at the population level, within and between species, based on a number of key factors [27,28,29,30,31]. They are therefore frequently utilised to produce lists of welfare priorities that are based on scientific research, for policy makers to consider. The European Food Safety Authority’s (EFSA) Panel on Animal Health and Welfare (AHAW) has, for instance, published risk assessments of welfare in dairy cows , farmed fish [27,28,33], fattening pigs , beef cattle and calves , broilers  and broiler breeders , in which aspects of housing, husbandry, management, stunning, slaughter and genetic selection, amongst other factors, were assessed.
In risk assessment terminology, welfare problems are caused by a series of “hazards”. The fast growth rate of standard commercial breeds of broiler chicken may be considered a potential welfare hazard with possible consequences including skeletal disorders, sudden death syndrome, ascites, high body mass and muscle disorders . In risk assessment, each identified hazard is characterized based on three factors: intensity of consequences, duration of effect of consequences (either as an absolute value if comparing within a breed or species, or as a proportion of lifetime if comparing between breeds or species; see  for an example) and prevalence (the proportion of affected individuals at any one time). Quantification (even in its simplest form, as categorical variables with groups including “mild”, “moderate”, or “severe”) allows consequences and their impact on the animals experiencing them to be compared. However, in this basic form of risk assessment, specific details of the hazards are not considered. A more accurate risk estimate can be obtained by also including information about the hazard itself in the calculation, such as estimates for the duration and probability of exposure to the hazard. In providing a quantitative, or even qualitative, value for each of these factors, the aim is to produce an objective estimate of risk for a series of potential welfare hazards.
However, as outlined in Collins , there are currently three main issues with welfare risk assessment procedures: (1) An incorrect assumption of independence between one hazard and its consequences and another hazard and its consequences. Although it is almost certainly true that some combinations of hazards will be truly independent, in most cases there is at least some degree of non-independence between hazards and, indeed, some overlap in the types of consequences emerging from different hazards. The effects of non-independence on a risk assessment procedure are a risk score that is either over- or under-estimated, typically as a result of conflated prevalence information, but also, and rather less easily measurable, conflated intensity information. (2) The common use of expert opinion to quantify each of the hazard characteristics, in lieu of purely data-based estimates, is a potential source of bias and unreliability in the risk assessment process. Of course, data-based estimates may also be subject to bias arising from selection, collection, analysis and/or interpretation of data . However, just as expert opinion is gathered from numerous individuals or groups, ranging in field of expertise, data-based estimates should be based on a range of studies and datasets derived from numerous sources. Doing so would reduce the impact of isolated instances of bias. Collating scientific evidence by means of systematic reviews would involve quality assessing studies whereby those at risk of bias would be excluded, and meta-analysis enables the potential effects of biases to be formally examined . (3) Level of uncertainty and variability are frequently not calculated in welfare risk assessments. This makes assessing the information they contain very difficult. Did all experts agree 100% on all the scores for the hazards, or was there a large divide in opinion? Without this information, it is very difficult to assess how confident one can feel in the result. This information would routinely be given in a typical statistical analysis, in the form of standard errors, confidence intervals and sample sizes. The same should be true for risk assessment procedures. However, above all, it is particularly true when quantifying intensity of suffering. This is essentially because the strength and reliability of expert opinion is dependent on the scientific research that has been published on a subject. There is no single welfare indicator that can reliably measure welfare across different contexts and across different species. Instead, we are reliant on multiple indicators that can often give conflicting results regarding affective state. Thus a random group of experts would be expected to differ in their opinions on intensity more than they might on something more easily quantifiable, such as prevalence.
To summarise, at this stage, animal welfare risk assessment is often used as a conceptual modelling tool to allow very different welfare problems to be compared. However, it could be made more quantitative and less subjective using systematic searches and data mining tools, to extract information from the resulting large volume of peer- and non-peer reviewed papers, and meta-analysis of the extracted data.
2.1.2. Welfare Assessment
Welfare assessment is typically the term given to the in situ appraisal of an animal’s affective and physical states, or most commonly, to the group assessment of welfare at the herd or flock level, e.g.,  or at the system level e.g., . Welfare assessment systems can be categorised on the type of information they collect: (i) resource-based; and (ii) animal-based assessment systems. Resource-based systems are primarily interested in the inputs provided for the animals . Input factors can often be assessed with a high degree of reliability , but are perhaps more suited to informing farmers about possible prevention and solution strategies to problems, rather than identifying and assessing the health and mental well-being of the animals . It has been shown that resource-based parameters alone are not sufficient to assess welfare e.g.,  and can be thought of as providing a “risk assessment” (or “housing condition assessment”) as opposed to a welfare assessment [46,47,48]. This could, however, be overcome by using resource-based parameters that have strong links with animal-based measures and can reliably predict welfare at the individual level . Animal-based measures of physical condition and health, disease status, and behaviour e.g., [44,49,50,51,52,53] are also limited when used alone as they cannot conclusively identify the causes of poor welfare .
Assessment models have been developed for cattle e.g., [41,46,47,51], pigs e.g., [42,52], poultry e.g., [48,53,54], farmed mink and foxes , and to evaluate enrichment materials for pigs (RICHPIG e.g., ). Other researchers have modified existing models for use in their studies. For example, Munsterhjelm et al.  modified the ANI-35-L for use in commercial pig production systems in Finland, and Stott et al.  modified the Welfare Quality® criteria for use in their study on extensive sheep farms.
Some models focus on resource-based indicators (e.g., ANI-35-L ; FOWEL ; SOWEL , but see Bracke ), others on animal-based measures (e.g., Welfare Quality® project models ) and others on both direct and indirect welfare indicators e.g., [46,49]. Aerts et al.  proposed a more “holistic” welfare assessment framework where the housing system (environment), the stockholder, and “the whole animal” are assessed; the former through the ANI-35-L  and the latter through Free Choice Profiling (FCP [63,64]). The proposed model does not focus on ranking attributes or on the overall welfare score but, rather, on how welfare can be practically improved at the farm level . While clearly advantageous, this focus, together with the qualitative “whole animal” assessment (FCP ), might render the model unsuitable for categorising farms for food labelling systems and welfare certification schemes.
Both the quality of stockmanship and human-animal interactions influence animal welfare and are, therefore, important parts of an overall welfare assessment [47,49,59,65]. However, obtaining direct and objective measures on-farm may be difficult [47,48,59]. Stockmanship has previously been measured indirectly through an assessment of the animals’ or the environment’s cleanliness or condition; for example, scoring plumage or hoof condition, cleanliness of feeders, etc., e.g., [47,48]. The Welfare Quality® models generally assess the human-animal relationship (HAR) through animals’ “fear of humans” (i.e., the assessor) on-farm (e.g., using the avoidance distance test in broilers and laying hens) [51,52,53]. Although an informative measure  that quantifies a qualitative phenomenon, direct observation of farmer-animal interactions (potentially scoring aspects of the farmers’ animal handling as in Welfare Quality®’s  scoring of “coercion”, i.e., use of electric goad, stick, etc., in the handling of cattle at slaughter) may also be required to help identify the specific causes of animals’ fear of the assessor and to provide specific advice to farmers that might improve the HAR. Another approach to incorporate human-animal interactions was developed by Stott et al. , who applied an adapted Service Quality Model (SQM), which is typically used in management science, to assess the interaction between profit and welfare on extensively housed sheep farms. The SQM considered the impact of farm management on sheep welfare, including stockperson “empathy” and “knowledge and experience” ; factors which are rarely included in welfare assessment models. The SQM focused on the gap between expectations and observed performance, where animal welfare was considered to be a function of quality of service provided . The results from this were then compared with results using an adaptation of Keeling and Veissier’s  Welfare Quality® criteria (Qualitative Welfare Assessment, QWA). However, although the two methods were reported to give “complementary results”, this does not appear to have been explicitly tested.
Some, if not most, welfare assessment models provide an overall welfare score (e.g., the Integrated Diagnostic System Welfare model ; SOWEL e.g., ), which involves weighting the multiple input factors. Weighting can be based on expert opinion, but this may introduce bias  and may not reflect the views of other stakeholders . Drawing experts from a range of different fields is important in order to balance viewpoints . However, if experts are brought together to choose and/or weight criteria, the group’s consensus view may actually be the view of a persuasive minority . On the other hand, there can be considerable variation in the weightings applied by experts  calling into question the value of this approach. Measuring and reporting variation in opinion, alongside experts’ confidence in their weighting of assessment criteria, as in Bracke et al. , would enable potential users to assess the validity of a model based on expert (or other) opinion. Conjoint analysis, frequently used in market research , has been used to weight welfare parameters based on opinion [74,75]. Here, participants evaluate “welfare profiles” rather than rank or weight individual attributes, which more closely resembles “real world” decision-making .
One approach to objective weighting of criteria is semantic modelling (SM), which was used in the development of the SOWEL , FOWEL  and RICHPIG models  and to assess the importance of wallowing for pig welfare e.g., . Bracke et al.  also demonstrated how animal welfare risk assessment might benefit from employing SM methodology. In SM, assessment criteria weightings are based on “scientific statements” (i.e., statements extracted from the literature denoting empirical observations of some aspect of welfare under particular conditions) e.g., [42,54] and, in later models, include a measure of uncertainty (i.e., strength of each statement) . The validity of the resulting model can then be tested through, for example, comparison with expert opinion e.g., [72,78,79], experiments designed to determine the importance of assessment criteria from the animals’ perspective e.g., , and sensitivity analysis .
It has been suggested that welfare assessment models and their individual criteria should be treated as diagnostic tests and evaluated in terms of their sensitivity and specificity in identifying an animal’s welfare status, as employed by Nyman et al. . Of course, unlike diagnostic tests for the presence of particular pathogens (such as bovine tuberculosis testing in cattle), welfare assessment models are not testing an individual for the presence or absence of a single parasitic entity existing within, or on, their being. Instead, it may be more accurate to consider welfare assessment models as being akin to diagnostic tests for non-parasitic conditions, such as certain mental health disorders in humans, which require triangulation of clinical signs and symptoms to reach a diagnosis [81,82,83]. The optimal, validated assessment model would include the most sensitive and specific criteria, with established cut-off points for “good” and “poor” welfare . It would be important for the model to highlight those animals close to a cut-off point, so that they can be closely monitored. Furthermore, as the good welfare of one animal cannot ethically offset the poor welfare of another within a herd (at least from a deontological viewpoint), a measure of variation is necessary when scoring welfare at the farm level . An alternative approach is to base the assessment on the lowest 25% of welfare scores within a herd [47,48], although this then gives little indication of overall herd welfare. For example, when assessing welfare problems such as lameness, mastitis or foot rot, where prevalence within a herd can be high, the bottom 25% (and more) could all be considered to have compromised welfare.
Although a range of welfare assessment models have been developed, their widespread use in other areas of research is not apparent. Animal welfare is considered an important factor when assessing the social sustainability of farming systems . However, the assessment of animal welfare within agricultural sustainability research varies widely from the use of a limited few measures e.g.,  to the use of a developed model supplemented by additional criteria  to behavioural observations and physiological indicators . Clearly, there is need for multidisciplinary work in this area, and for increased collaboration between farm animal welfare scientists, economists and researchers in sustainability and food security.
2.2. Simulation Modelling
The term simulation modelling is extremely broad and, in actuality, could refer to many of the kinds of model that will be described in the following sub-sections. However, in the more selective sense, a simulation model is one that seeks to recreate patterns observed in real-life with the input of selected, often simplified, variables that are thought to be responsible (at least in part) for the production of the resultant pattern. These models are typically described as “stochastic”, for models that include one or more parameters with random values drawn from identified probability distributions; or “deterministic”, for models that include no random variables and no randomness. Simulation models can also be described as “dynamic”, if they include time as a variable, or “static” if time is not included. Dynamic models are typically represented with differential or partial differential equations. Further to this, dynamic models can be described as “continuous” or “discrete”, depending on whether changes in time are represented with a continuous interval, or discrete time steps (events), respectively [88,89]. Models may further be described as “top-down”, where the developers start with a general overview of a system (the “big picture”), but do not have details of the component subsystems. This type of model tends to be formulated with a large number of differential equations. By comparison, many modern models are “bottom-up”, such as those used in agent-based modelling (ABM) where individual units (e.g., animals, though one could potentially also consider pens, houses or farms as separate units) are programmed to behave according to a set of probability-based rules and can interact with other units (though ABM can also be data-based  or use deterministic rules ). The resulting emergent patterns at the global level of the model can then be compared with patterns observed in real life. Bottom-up, generative models such as ABMs have been used in many different fields, including disease dynamics e.g., [92,93], evolutionary and social processes e.g.,  and to describe financial markets e.g., .
In farm animal welfare, simulation modelling has been used to estimate the costs of welfare improvements for commercial pigs  and the economic and welfare impacts of foot disorders [96,97] and foot disorder interventions  in dairy cattle. Den Ouden et al.’s  model simulated all stages of the production chain (farrowing, fattening and slaughtering) as well as transportation between stages and time spent in lairage. However, effective models need not be complicated. Waterhouse  presented a simple model to estimate the time and effort required to supervise ewes during lambing, and the provision of this care, at different stocking densities. Febrer et al.  simulated social attraction, aversion and indifference in broiler chickens housed at different stocking densities, and compared these simulations to observed spatial distributions of real chickens, to investigate if broilers prefer personal space or the closeness of other birds. Likewise, and as with the Febrer et al.  model, based on just two parameters (in this case, numbers of birds at the trough currently and 5 seconds previously), Collins and Sumpter  showed that broiler chickens demonstrate a Mexican wave-like property in their feeding dynamics around a trough at different stocking densities. Both models [100,101] suggested that social dynamics and clustering occur independently of stocking density.
The development of more complex models may be somewhat constrained by a lack of available quantitative data, which, for example, prohibited the modelling of some intervention measures and interactions between foot disorders in Bruijis et al.’s  study. However, gaps in the literature are to be expected, and existing models can be updated as scientific research progresses. At the same time, we can make good use of existing scientific knowledge by conducting meta-analyses to produce parameter estimates—this could prove particularly effective for studies aiming to investigate the welfare effects of different environmental conditions, or in different breeds, for example, which have not been explicitly compared in a single experiment, but have been partially investigated across several studies.
Modelling the impact of various measures on animal welfare can be difficult due to the range of factors that can interact to influence welfare and the difficulties in estimating their effects . Farmer attitude and knowledge, for example, are complex factors to model . Furthermore, some types of welfare issue may be more straightforward to simulate than others, particularly if they can be reliably quantified. For example, those relating to physical health and disease, where clinical signs can be observed (or tested for) in a group with known estimates of specificity and sensitivity, or behavioural indicators that are validated and quantifiable, such as performance of abnormal behaviours, or social dynamics and clustering.
Selecting and validating the best simulation model can sometimes prove difficult. In an ideal world, all simulations would be tested using sensitivity analysis, and validated against an independent dataset after development. In practice, this is often not performed prior to publication if the model is based on the only available, existing dataset for a particular problem at that time. However, even if this is the case, statistical methods such as cross-validation, sub-sampling and bootstrapping could be utilised to create alternative datasets for model validation [102,103,104]. Of course, a simulation is highly unlikely to make predictions with 100% accuracy. Where agreement between the model output and a dataset is relatively mediocre (measured with goodness-of-fit), one could draw conclusions from the differences between the conditions in the model and the dataset, rather than from model output per se . An alternative approach may be to improve goodness-of-fit. Goodness-of-fit is often improved by including additional parameters in a simulation, but this can often make interpreting the model output difficult  or lead to overfitting . Furthermore, goodness-of-fit doesn’t tell us if the model itself is any good—it only tells us how well (or rather, how badly) the predicted values match those in an independent dataset, and overall fit can be good even if there are some areas where the fit is very bad .
There is great potential for simulation modelling to be used to a much wider extent in animal welfare research. For example, given the expected depletion in oil reserves [109,110,111,112], changes in fossil fuel [111,113] and electricity prices , projected increases in demand for meat  and the potential effects of changes in climate [115,116,117,118] over coming decades, a long-term evaluation of the possible effects of these factors on farm animal welfare could prove extremely valuable for welfare scientists, policy makers and industry, particularly if potential intervention or amelioration strategies could also be investigated.
2.3. Optimisation Models
As the name suggests, optimisation models are developed to find the optimal solution to a defined problem and typically take the form of linear and dynamic programming in livestock science . Linear programming (LP) models reveal the optimal set of variables that maximises (or minimises) a particular function under specified constraints [58,119]. Dynamic programming (DP) models are also based on mathematical optimisation, where a larger problem can be broken down into multiple smaller subproblems, which can in turn be broken down in a recursive manner. Two classic examples include Dijkstra’s shortest route algorithm , and the Tower of Hanoi recursive solution. Farm animal welfare research employing optimisation models have principally focussed on the economics of animal welfare. Unlike simulation models, which can be used to simulate the behaviour of a system, animal or disorder and estimate the impacts of manipulating various input factors, optimisation models are designed to solve a specific problem, optimising a particular function and identifying the best possible strategy or outcome .
Langford and Stott  used DP to maximise dairy farmers’ economic gain through determining the optimum (financial) decision to keep or replace a heifer at each parity over a 20-year cycle. Modelling different farm scenarios (high and low rates of infertility, mastitis and lameness) allowed estimates of the long-term effects of improving cow welfare. Several other studies have also used these models to maximise financial return at the production level. For example, LP models have been used to minimise the costs of improving pig welfare , to determine the most profitable body condition at which to maintain sheep  and to estimate the maximum profit potential of individual sheep farms . Vosough Ahmadi et al.  used LP to investigate trade-offs between economics (profit) and sow welfare under different farrowing systems, and to provide a framework for designing economically feasible, high welfare systems. Other studies have used LP to estimate the price of pork produced under high welfare farrowing systems , to compare dairy farming systems and to estimate the impact of management changes on economics and animal welfare, amongst other factors .
There appears to be great potential for the use of optimisation models in future research, where animal welfare science and economics combine to find win-win scenarios for farmer and animal . However, future LP models might also be used to examine potential trade-offs between animal welfare and environmental impacts e.g.,  and aim, for example, to minimise GHG emissions under animal welfare constraints. One of the main difficulties with developing such optimisation models, as with all other models discussed in this paper, is in pinpointing the welfare components for inclusion and in determining the complex effects that different changes would have on the animals’ well-being . Changes perceived to be beneficial by the public, like banning gestation crates for sows and battery cages for laying hens may, in practice, improve some aspects of the animals’ welfare and reduce other aspects [69,128], which further emphasises the need to include multiple animal-based measures in models of animal welfare and, where possible, to base each relationship on scientific evidence rather than on opinion.
2.4. Scenario Modelling
Scenario modelling is not really a separate type of modelling used in farm animal welfare research, so much as a method that often overlaps with several of the approaches already discussed (simulations, optimisation models). It may offer an alternative approach to forecasting whereby, rather than trying to predict future events, it compares a variety of alternative futures (or potential solutions), asking the “what if…?” question . The effects of different scenarios on the model outcomes are estimated and then compared with the basic simulation (often reflecting the current situation). For example, in one study, three alternative future scenarios for organic dairy farming in Denmark (focussing on profit, animal welfare, and environment) were modelled and evaluated in terms of their economic and environmental impacts. Model simulations and expert knowledge were used to parameterise the scenarios . Alternative cow welfare and pig performance scenarios have been investigated, using DP and LP, to estimate the effect on farmer income  and on the price of pork produced under different farrowing systems , respectively.
Although farm income has risen over recent years, farming is still not a highly profitable business . Therefore, the balance of economic viability and good animal welfare is a tight line to tread. Scenario modelling provides a means of comparing alternative solutions (e.g., increased retail price of animal products, additional Common Agricultural Policy (CAP) Single Farm Payments, introduction of a national tax, complete shift to large-scale farming) in terms of their impacts on the utility (i.e., well-being) of all stakeholders (i.e., animals, farmers, retailers, consumers, citizens) and on the national economy. Scenario modelling might also be used to compare alterative future solutions that would enable the livestock industry—particularly in countries with already highly-intensive production and low welfare standards (e.g., China )—to meet the increasing demand for animal products .
Comparing multiple different possible futures may, however, be hampered by the fact that scenario modelling does not incorporate any element which could determine which of the futures is most likely, or optimal. Rather, it simply outlines what the different possible conditions could be under each of the different scenarios. It is then left open to interpretation which of the possibilities would be most ideal or most probable under current circumstances. This facet means that scenario modelling is an excellent choice for making value-free judgements within the model construct, although it also means that judgements must be made post-hoc, perhaps rather more subjectively than would be the case with other modelling methods.
2.5. Systems Modelling
Systems modelling is the analysis of complex systems, and investigation of how functionally different sub-processes within a system integrate and interact to produce a coherent system [132,133]. Unlike most other types of modelling discussed in this paper, systems modelling is almost by definition interdisciplinary, and is based on the underlying principle that to fully understand a system, one must understand it at different organisational levels  such as at the molecular, cellular, organismal and species level, all within one model. To date, the majority of systems biology studies have focussed on, for example, drug discovery , forecasting and diagnostics in plant, animal and human disease , and the design of bio-products such as bio-fuels .
This approach has the potential to make major contributions towards the development of more sustainable farming systems . For example, the Sustainable and Integrated Management Systems for Dairy Production (SIMSDAIRY) model brings together existing models, equations and “score matrices” that reflect economic, environmental, ecological and social factors relevant to farm sustainability (including animal welfare), involving simulation, optimisation and scenario modelling at the farm level [138,139]. The SIMSDAIRY model has been used to compare the potential of improving nutrient management with plant and animal genetics in enhancing the overall sustainability of UK dairy farms . In another study, the SIMSDAIRY model was used to compare three simulated organic systems with two intensive systems (differing in their use of nitrogen fixing plants and use of fertilizers and pesticides) and compared, for example, the estimated GHG emissions, farm income, biodiversity and animal welfare of simulated farms . Neither of these two studies focussed on animal welfare, but the model incorporates an animal welfare score matrix . The factors assessed to calculate animal welfare were mainly resource-based. There is clearly much scope for further development of this sub-model , perhaps by integrating an animal-based welfare assessment model (e.g., the Welfare Quality® model for cattle ).
Systems modelling facilitates our understanding of how different parts of that system interact and enables a combination of changes, within different parts of the system, to be evaluated . Developing systems models can be very time consuming and the end-product is highly specific. Nonetheless, adopting a systems (or, perhaps, a network) approach could provide valuable tools to improve our understanding of how animal welfare interacts with other measures of sustainability and, particularly: (1) to estimate how specific improvements in farm animal welfare would impact on the environmental, ecological and societal sustainability of the farm; (2) to estimate how improvements in the economic, environmental, ecological and/or societal sustainability of the system would impact on animal welfare; (3) to identify specific improvements in animal welfare that, if implemented, would improve overall sustainability of the farming system (leading to win-win scenarios that are more likely to be adopted); and (4) to identify combinations of factors within different parts of the system that could be changed to optimise animal welfare and overall sustainability .
2.6. Conceptual Models
Conceptual models allow consideration of the fundamental activities of a system without being tied to details of the physical reality of that system. Checkland  defined a conceptual model to be “a statement of what is logically and necessarily implied by the [root] definition. It is not a recommendation of what ought to exist nor of what does exist in the real situation.” Thus, economic models exemplify this type of approach. Conceptual models identified in the farm animal welfare scientific literature are diverse in subject matter, including human and animal willingness-to-pay (WTP) for improved animal welfare , motivation for sucking behaviour in calves , stereotypy development and maintenance under feed restriction in pigs , a quantitative genetic model of animal learning  and a model to aid artificial selection for enhanced welfare in pigs . Other conceptual modelling studies have considered the impacts of farmers’ perceptions, attitudes and behaviour on their choice of whether to implement welfare improvements , a matter which has generally been overlooked in other models of animal welfare. Other researchers have applied a conceptual socio-psychological model (the “theory of planned behaviour”) to understand farmers’ decisions, and underlying motivations, to change husbandry practices, such as group housing for pregnant sows , alternatives to mulesing in sheep , and to identify interventions that might encourage farmers to implement such changes .
Lusk  conceptualised the construction of a distinct market for animal welfare, separate from the trade of meat products (i.e., the development and trade of animal well-being units [AWBU]). Simple algorithms to calculate AWBU on a farm were proposed, building on existing welfare assessment models. To establish a successful market, clearly AWBU need to reflect consumer preferences  and this is where an assessment model of farm animal welfare based on public perceptions e.g.,  could be very useful. This model may seem idealistic, and it’s realisation improbable. This is recognised by Lusk  who exemplifies the existence of such a market. While it is unlikely that this idea will overtake the drive for welfare labelling of animal products, it gives us food for thought on additional ways in which animal welfare improvements could be realised in the future.
A full discussion on economic models and their contribution to farm animal welfare research is beyond the scope of this paper, but would be extremely worthwhile. However, McInerney’s  model is noteworthy here. This conceptual model illustrates a theoretical relationship and, specifically, conflict between farm animal welfare and productivity (i.e., animal vs. human benefit). Here, improvements in productivity and animal welfare go hand-in-hand up to a certain point, after which further increases in productivity will reduce the welfare of livestock. The question is; where, and how, do we strike the right balance between the well-being of humans and that of the animals we eat ? There is great potential for mathematical modelling to help us answer this question, drawing on both economics and animal welfare science . Such work has already begun with a number of models outlined above (e.g., the optimisation models developed by Langford & Stott , Kingwell  and Vosough Ahmadi et al. ), which identified win-win scenarios for farmers and livestock at farm, or system, level. Such interdisciplinary, bio-economic models have the potential to make important and practical contributions towards improving farm animal welfare.
Thus, conceptual models can provide a theoretical basis to help guide future research, propose potential solutions, collate empirical evidence and illustrate ideas of how different factors might interact. There is, therefore, scope to develop conceptual models within all areas of animal welfare research. However, unlike the other models discussed in this review, conceptual models, not being rooted in physical reality, can often not be explicitly tested.
The aim of this paper was to review the use of non-statistical models in farm animal welfare research. Most of the research to date has described the development, or use, of conceptual, risk and welfare assessment models. The use of mathematical models has been limited, despite their potential to assist us in: (1) predicting when and where welfare problems are likely to arise and who they are most likely to affect; (2) determining how different components of the farming system interact to influence welfare; (3) identifying the best control strategies that we can apply to prevent welfare problems from developing; and (4) bringing together large bodies of evidence from different fields to establish links between animal welfare, economics, environmental and social sustainability in livestock farming. Here, mathematical modelling could assist us in identifying, and tackling, any existing or impending conflicts between, for example, farm animal welfare, farmer livelihood, future food security and environmental legislation.
In a review of the literature, de Boer et al.  identified some potential effects of greenhouse gas (GHG) emission mitigation strategies on animal welfare, human health, emissions, land use and other sustainability factors; but highlighted that all is far from clear. They called for an amalgamation of life cycle (sustainability) assessment (LCA) and simulation models that exist across disciplines and that reflect different levels of the farming system to fully comprehend the consequences of instigating GHG mitigation strategies . The result would be akin to a complex systems model, in which component interactions (i.e., cause-effect) would be modelled throughout the entire production chain at farm, crop and animal levels . Such a model could prove influential in policy makers’ decisions. To ensure that animal welfare science can contribute to its development and that the animal welfare impacts of GHG mitigation are more fully understood, we need to use our existing knowledge to model the links between farm animal welfare and other aspects of sustainability, following on from the models that already exist in animal welfare literature. In doing so, gaps in knowledge will be highlighted; directing future research in this area.
Although numerous studies reviewed herein examined the economics of farm animal welfare, few incorporated consumer willingness-to-pay (WTP) for higher welfare standards as a factor in their models, which would be required for a complete economic evaluation of welfare improvements [75,122]. This may be because of the difficulties incurred in accurately estimating WTP e.g., [153,154], or due to a lack of available data e.g., . However, research into consumer WTP for welfare improvements is increasing e.g., [155,156,157,158]. The values estimated in these studies could be used in future models, and to refine existing models, to identify potential win-win scenarios for farmer and animal. Given that it is ultimately the farmer’s decision whether or not to improve the welfare of animals above that required by legislation, it may also be in our interests to better understand farmers’ decision-making processes, and how the major supermarkets influence these. Here, agent-based modelling (previously used in agricultural e.g.,  and consumer behaviour studies e.g., ) may be useful .
Insufficient quantitative data can cause difficulties when developing valid mathematical models  as the accuracy of model output depends largely on the reliability and validity of input data . Input data can be collected or estimated through direct measurement, literature reviews and expert opinion. While considered the “holy grail” in medicine, meta-analysis is rarely employed in animal welfare science but, where possible, could provide more reliable estimates of input data, and of relationships between parameters, than direct measurement in a single, all-encompassing, study. As randomly controlled trials tend not to be used in animal welfare research, however, differences between studies (housing conditions, breeds, management, etc.) and study limitations must be identified and taken into consideration in the analysis.
Thus, there is no need to wait for the collection of “perfect” data sets to build or parameterise models, which may be one of the reasons behind the limited use of mathematical models in farm animal welfare research to date. There are, already, many scientific papers reporting experimental data on which data mining and meta-analysis techniques could be used to parameterise a model (here, we are referring to all models, including risk assessment). Over 40 years of research has led to a wealth of experimental findings collected under various housing conditions and farming systems, with a range of species and breeds, using various welfare measures and reporting different welfare statuses. Indeed, it would be a shame not to make the most of this data.
It must be noted that modelling is not a precise science and, in some ways, is quite subjective. Parameterising the model with “real” (observed) data will increase objectivity; however, it will also increase the risk of error in model output, and reduce its applicability to other datasets, if the dataset used to parameterise the model is not valid, or contains erroneous values that aren’t typically observed . Increasing model complexity by increasing the number of parameters may result in overfitting, which can reduce the model’s predictive value . Thus, there is a trade-off between bias and variance, whereby bias decreases, and variance increases, as parameter numbers grow .
It could also be said that there is a trade-off between precision and manageability , which will certainly apply to the dataset used to parameterise the model. However, as shown by Mackay and Lee , simpler models can also have the best predictive value when tested against other datasets.
Model validation is vital, otherwise, the model is an untested hypothesis. Validating the model using an independent dataset (i.e., an entirely new dataset to the one used for parameterisation) is generally considered the best approach . However, this is often not possible if all available data is required to build the model. In this case, techniques such as cross-validation, bootstrapping and sub-sampling can be employed [102,103,104].
Assumptions are generally used to simplify the analyses  and are typically based on real data or expert opinion e.g., . It is important to consider the evidence behind model assumptions when interpreting model output and drawing conclusions e.g., . Models that measure animal welfare using resource-based parameters are based on the assumption that welfare can be improved through changes to the animals’ environment, or other resources bestowed on them. While this is likely true, we must also be careful not to fall into the “anthropomorphic trap”. For example, while high stocking density is considered a risk to welfare in broiler chickens , research suggests stocking density per se is generally of less importance than housing management . This illustrates the necessity to base model assumptions on scientific evidence (i.e., establish links between resource- and animal-based parameters ) and to incorporate other factors that might interact with environmental parameters to impact on the animal’s welfare, such as quality of stockmanship and the human-animal relationship [47,49,59,65]. However, where such evidence-based assumptions cannot be made, this should not discourage model development as the model can be strengthened as more data becomes available. In such cases, theoretical concepts can be combined with empirical data to build a model, and the concepts can then be used to guide future data collection to support or refine those assumptions.
We should be working towards determining the sensitivity and specificity of individual welfare indicators and whole assessment models, and establishing non-arbitrary cut-off points for good/poor welfare. Alternatively, we could adopt and adapt methods used in formulating diagnoses in human psychiatry e.g.,  whereby criteria are grouped and individuals must meet a certain number of these criteria to be diagnosed with a particular condition. In animal welfare, this could take the form of, for example, individuals being scored against sets of criteria for good welfare, compromised welfare and severely compromised welfare (e.g., severe feed restriction , mobility score 3 (i.e., lame) ).
Modelling within animal welfare research has been largely focussed on the development and use of conceptual and assessment models. While there is great scope for progressing animal welfare science through integrating our existing knowledge in the development of mathematical models, we found that the use of such techniques has been limited to date. The development of “whole systems” models will require inter-disciplinary collaborations with systems biologists, economists, and sustainability and food security experts. In return, these large-scale models may have scope to influence decision-makers and, certainly, to improve our understanding of how, and where, animal welfare improvements fit into the wider context of sustainability and food security. Preliminary work towards the development of such complex models has already begun with the more specific simulation, optimisation, scenario and assessment models that have been outlined in this paper.
Chérie E. Part is funded by the Department for Employment and Learning, Northern Ireland. We would like to thank three anonymous reviewers for their valuable comments on this paper.
Conflict of Interest
The authors declare no conflict of interest.
- Van der Gaag, M.A.; Mul, H.F.; Huirne, R.B.M. Food safety and control programs in the Dutch pork production chain. In Proceedings of the Fourth International Conference on Chain Management in Agribusiness and the Food Industry, Wageningen, The Netherlands, 25–26 May 2000; pp. 139–145.
- Keeling, M.J.; Woolhouse, M.E.; Shaw, D.J.; Matthews, L.; Chase-Topping, M.; Haydon, D.T.; Cornell, S.J.; Kappey, J.; Wilesmith, J.; Grenfell, B.T. Dynamics of the 2001 UK foot and mouth epidemic: Stochastic dispersal in a heterogeneous landscape. Science 2001, 294, 813–817, doi:10.1126/science.1065973.
- Backer, J.A.; Hagenaars, T.J.; Nodelijk, G.; Van Roermund, H.J.W. Vaccination against foot-and-mouth disease I: Epidemiological consequences. Prev. Vet. Med. 2012, 107, 27–40, doi:10.1016/j.prevetmed.2012.05.012.
- Tomassen, F.H.M.; de Koeijer, A; Mourits, M.C.M.; Dekker, A.; Bouma, A.; Huirne, R.B.M. A decision-tree to optimise control measures during the early stage of a foot-and-mouth disease epidemic. Prev. Vet. Med. 2002, 54, 301–324.
- Milne, C.E.; Dalton, G.E.; Stott, A.W. Integrated control strategies for ectoparasites in Scottish sheep flocks. Livest Sci 2007, 106, 243–253, doi:10.1016/j.livsci.2006.08.010.
- Vernon, M.C.; Keeling, M.J. Representing the UK's cattle herd as static and dynamic networks. Proc. R. Soc. B 2009, 276, 469–476, doi:10.1098/rspb.2008.1009.
- Tinsley, M.; Lewis, F.I.; Brülisauer, F. Network modeling of BVD transmission. Vet. Res. 2012, 43, doi:10.1186/1297-9716-43-11.
- Del Prado, A.; Misselbrook, T.; Chadwick, D.; Hopkins, A.; Dewhurst, R.J.; Davison, P.; Butler, A.; Schröder, J.; Scholefield, D. SIMSDAIRY: A modelling framework to identify sustainable dairy farms in the UK. Framework description and test for organic systems and N fertiliser optimisation. Sci. Total Environ. 2011, 409, 3993–4009, doi:10.1016/j.scitotenv.2011.05.050.
- Roughgarden, J. Ecology. In Encyclopedia of Climate and Weather; Schneider, S.H., Ed.; Oxford University Press: New York, NY, USA, 1996; pp. 268–269.
- Drake, K.A.; Donnelly, C.A.; Dawkins, M.S. Influence of rearing and lay risk factors on propensity for feather damage in laying hens. Br. Poult. Sci. 2010, 51, 725–33, doi:10.1080/00071668.2010.528751.
- Moe, R.; Bakken, M. Anxiolytic drugs inhibit hyperthermia induced by handling in farmed silver foxes (Vulpes vulpes). Anim. Welf. 1998, 7, 97–100.
- Neisen, G.; Wechsler, B.; Gygax, L. Effects of the introduction of single heifers or pairs of heifers into dairy-cow herds on the temporal and spatial associations of heifers and cows. Appl. Anim. Behav. Sci. 2009, 119, 127–136, doi:10.1016/j.applanim.2009.04.006.
- Reefmann, N.; Bütikofer Kaszàs, F.; Wechsler, B.; Gygax, L. Ear and tail postures as indicators of emotional valence in sheep. Appl. Anim. Behav. Sci. 2009, 118, 199–207, doi:10.1016/j.applanim.2009.02.013.
- Smulders, D.; Hautekiet, V.; Verbeke, G.; Geers, R. Tail and ear biting lesions in pigs: An epidemiological study. Anim. Welf. 2008, 17, 61–69.
- Temple, D.; Courboulay, V.; Velarde, A.; Dalmau, A.; Manteca, X. The welfare of growing pigs in five different production systems in France and Spain: Assessment of health. Anim. Welf. 2012, 21, 257–271, doi:10.7120/09627222.214.171.1247.
- Temple, D.; Courboulay, V.; Manteca, X.; Velarde, A.; Dalmau, A. The welfare of growing pigs in five different production systems: Assessment of feeding and housing. Animal 2012, 6, 656–667, doi:10.1017/S1751731111001868.
- Turnbull, J.; Bell, A.; Adams, C.; Bron, J.; Huntingford, F. Stocking density and welfare of cage farmed Atlantic salmon: Application of a multivariate analysis. Aquaculture 2005, 243, 121–132, doi:10.1016/j.aquaculture.2004.09.022.
- Zonderland, J.J.; Van Riel, J.W.; Bracke, M.B.M.; Kemp, B.; Den Hartog, L.A.; Spoolder, H.A.M. Tail posture predicts tail damage among weaned piglets. Appl. Anim. Behav. Sci. 2009, 121, 165–170, doi:10.1016/j.applanim.2009.09.002.
- Turner, S.P.; Roehe, R.; Mekkawy, W.; Farnworth, M.J.; Knap, P.W.; Lawrence, A.B. Bayesian analysis of genetic associations of skin lesions and behavioural traits to identify genetic components of individual aggressiveness in pigs. Behav. Genet. 2008, 38, 67–75, doi:10.1007/s10519-007-9171-2.
- Roberts, S.J.; Cain, R.; Dawkins, M.S. Prediction of welfare outcomes for broiler chickens using Bayesian regression on continuous optical flow data. J. R. Soc. Interface 2012, 9, 3436–3443, doi:10.1098/rsif.2012.0594.
- Oberbauer, A.M.; Berry, S.L.; Belanger, J.M.; McGoldrick, R.M.; Pinos-Rodriquez, J.M.; Famula, T.R. Determining the heritable component of dairy cattle foot lesions. J. Dairy Sci. 2013, 96, 605–613, doi:10.3168/jds.2012-5485.
- Asher, L.; Collins, L.M.; Ortiz-Pelaez, A.; Drewe, J.A.; Nicol, C.J.; Pfeiffer, D.U. Recent advances in the analysis of behavioural organization and interpretation as indicators of animal welfare. J. R. Soc. Interface 2009, 6, 1103–1119, doi:10.1098/rsif.2009.0221.
- Collins, L.M.; Asher, L.; Pfeiffer, D.U.; Browne, W.J.; Nicol, C.J. Clustering and synchrony in laying hens: The effect of environmental resources on social dynamics. Appl. Anim. Behav. Sci. 2011, 129, 43–53, doi:10.1016/j.applanim.2010.10.007.
- Dawkins, M.S.; Lee, H.; Waitt, C.D.; Roberts, S.J. Optical flow patterns in broiler chicken flocks as automated measures of behaviour and gait. Appl. Anim. Behav. Sci. 2009, 119, 203–209, doi:10.1016/j.applanim.2009.04.009.
- Rutherford, K.M.D.; Haskell, M.J.; Glasbey, C.; Jones, R.B.; Lawrence, A.B. Fractal analysis of animal behaviour as an indicator of animal welfare. Anim. Welf. 2004, 13, S99–S103.
- Abeyesinghe, S.M.; Drewe, J.A.; Asher, L.; Wathes, C.M.; Collins, L.M. Do hens have friends? Appl. Anim. Behav. Sci. 2013, 143, 61–66, doi:10.1016/j.applanim.2012.12.003.
- EFSA (European Food Safety Authority) Panel on Animal Health and Welfare (AHAW). Animal welfare aspects of husbandry systems for farmed Atlantic salmon. Scientific Opinion of the Panel on Animal Health and Welfare. EFSA J. 2008, doi:10.2903/j.efsa.2008.736. Available online: http://www.efsa.europa.eu/en/efsajournal/doc/736.pdf (accessed on 28 January 2013).
- EFSA (European Food Safety Authority) Panel on Animal Health and Welfare (AHAW). Species-specific welfare aspects of the main systems of stunning and killing of farmed fish: Rainbow trout. Scientific Opinion of the Panel on Animal Health and Welfare. EFSA J. 2009, doi:10.2903/j.efsa.2009.1012.
- EFSA (European Food Safety Authority) Panel on Animal Health and Welfare (AHAW). Scientific Opinion on the influence of genetic parameters on the welfare and the resistance to stress of commercial broilers. EFSA J. 2010, doi:10.2903/j.efsa.2010.1666. Available online: http://www.efsa.europa.eu/en/efsajournal/doc/1666.pdf (accessed on 28 January 2013).
- EFSA (European Food Safety Authority) Panel on Animal Health and Welfare (AHAW). Scientific Opinion on welfare aspects of the management and housing of the grand-parent and parent stocks raised and kept for breeding purposes. EFSA J. 2010, doi:10.2903/ j.efsa.2010.1667. Available online: http://www.efsa.europa.eu/de/efsajournal/doc/1667.pdf (accessed on 28 January 2013).
- Collins, L.M.; Asher, L.; Summers, J.F.; Diesel, G.; McGreevy, P.D. Welfare epidemiology as a tool to assess the welfare impact of inherited defects on the pedigree dog population. Anim. Welf. 2010, 19, 67–75.
- EFSA (European Food Safety Authority) Panel on Animal Health and Welfare (AHAW). Scientific Opinion on welfare of dairy cows in relation to behaviour, fear and pain based on a risk assessment with special reference to the impact of housing, feeding, management and genetic selection. EFSA J. 2009, doi:10.2903/j.efsa.2009.1139. Available online: http://www.efsa.europa.eu/en/efsajournal/doc/1139.pdf (accessed on 4 March 2013).
- EFSA (European Food Safety Authority) Panel on Animal Health and Welfare (AHAW). Species-specific welfare aspects of the main systems of stunning and killing of farmed seabass and seabream. Scientific Opinion of the Panel on Animal Health and Welfare. EFSA J. 2009, doi:10.2903/j.efsa.2009.1010. Available online: http://www.efsa.europa.eu/en/efsajournal/doc/ 1010.pdf (accessed on 4 March 2013).
- EFSA (European Food Safety Authority) Panel on Animal Health and Welfare (AHAW). Animal health and welfare in fattening pigs in relation to housing and husbandry. Scientific Opinion of the Panel on Animal Health and Welfare. EFSA J. 2007, doi:10.2903/j.efsa.2007.564. Available online: http://www.efsa.europa.eu/en/efsajournal/doc/564.pdf (accessed on 4 March 2013).
- EFSA (European Food Safety Authority) Panel on Animal Health and Welfare (AHAW). Scientific Opinion on the welfare of cattle kept for beef production and the welfare in intensive calf farming systems. EFSA J. 2012, doi:10.2903/j.efsa.2012.2669. Available online: http://www.efsa.europa.eu/en/efsajournal/doc/2669.pdf (accessed on 4 March 2013).
- Collins, L.M. Non-intrusive social preference indicators in broiler chickens. Ph.D. Thesis, University of Oxford, Oxford, UK, 2005.
- Collins, L.M.; Asher, L.; Summers, J.; McGreevy, P. Getting priorities straight: Risk assessment and decision-making in the improvement of inherited disorders in pedigree dogs. Vet. J. 2011, 189, 147–154, doi:10.1016/j.tvjl.2011.06.012.
- Collins, L.M. Welfare risk assessment: The benefits and common pitfalls. Anim. Welf. 2012, 21, 73–79.
- Tripepi, G.; Jager, K.J.; Dekker, F.W.; Zoccali, C. Selection bias and information bias in clinical research. Nephron. Clin. Pract. 2010, 115, C94–C99, doi:10.1159/000312871.
- Borenstein, M.; Hedges, L.V.; Higgins, J.P.T.; Rothstein, H.R. Introduction to Meta-Analysis; Jorn Wiley & Sons, Ltd.: Chichester, UK, 2009.
- Nyman, A.-K.; Lindberg, A.; Hallén-Sandgren, C. Can pre-collected register data be used to identify dairy herds with good cattle welfare? Acta Vet. Scand. 2011, 53, S8–S13, doi:10.1186/1751-0147-53-S1-S8.
- Bracke, M.B.M.; Spruijt, B.M.; Metz, J.H.M.; Schouten, W.G.P. Decision support system for overall welfare assessment in pregnant sows A: Model structure and weighting procedure. J. Anim. Sci. 2002, 80, 1819–1834.
- Amon, T.; Amon, B.; Ofner, E.; Boxberger, J. Precision of assessment of animal welfare by the “TGI 35 L” Austrian Animal Needs Index. Acta Agr. Scand. A 2001, 51, 114–117.
- Main, D.C.J.; Kent, J.P.; Wemelsfelder, F.; Ofner, E.; Tuyttens, F.A.M. Applications for the methods of on-farm welfare assessment. Anim. Welf. 2003, 12, 523–528.
- Sandøe, P.; Munksgaard, L.; Bådsgaard, N.P.; Jensen, K.H. How to manage the management factor—Assessing animal welfare at the farm level. In Livestock Farming Systems: More than Food Production; Sørensen, J.T., Ed.; Wageningen Academic Publishers: Wageningen, The Netherlands, 1997; pp. 221–230.
- Calamari, L.; Bertoni, G. Model to evaluate welfare in dairy cow farms. Ital. J. Anim. Sci. 2009, 8, 301–323.
- Bartussek, H.; Leeb, C.H.; Held, S. Animal needs index for cattle: ANI35 L/2000—Cattle. Federal Research Institute for Agriculture in Alpine Regions: Irdning, Austria, 2000. Available online: http://www.bartussek.at/pdf/anicattle.pdf (accessed on 21 February 2013).
- Bartussek, H. Animal needs index for laying hens: ANI 35-L/2001—Laying hens. Federal Research Institute for Agriculture in Alpine Regions: Irdning, Austria, 2001. Available online: http://www.bartussek.at/pdf/anilayinghens.pdf (accessed on 21 February 2013).
- Waiblinger, S.; Knierim, U.; Winckler, C. The development of an epidemiologically based on-farm welfare assessment system for use with dairy cows. Acta Agr. Scand. A 2001, 51, 73–77.
- Whay, H.R.; Main, D.C.J.; Greent, L.E.; Webster, A.J.F. Animal-based measures for the assessment of welfare state of dairy cattle, pigs and laying hens: Consensus of expert opinion. Anim. Welf. 2003, 12, 205–217.
- Welfare Quality® Consortium: Lelystad, The Netherlands, 2009.
- Welfare Quality® Consortium: Lelystad, The Netherlands, 2009.
- Welfare Quality® Consortium: Lelystad, The Netherlands, 2009.
- De Mol, R.M.; Schouten, W.G.P.; Evers, E.; Drost, H.; Houwers, H.W.J.; Smits, A.C. A computer model for welfare assessment of poultry production systems for laying hens. Neth. J. Agr. Sci. 2006, 54, 157–168.
- Mononen, J.; Møller, S.; Hansen, S.; Hovland, A.; Koistinen, T.; Lidfors, L.; Malmkvist, J.; Vinke, C.; Ahola, L. The development of on-farm welfare assessment protocols for foxes and mink: The WelFur project. Anim. Welf. 2012, 21, 363–371.
- Bracke, M.B.M. RICHPIG: A semantic model to assess enrichment materials for pigs. Anim. Welf. 2008, 17, 289–304.
- Munsterhjelm, C.; Valros, A.; Heinonen, M.; Hälli, O.; Peltoniemi, O.A.T. Welfare index and reproductive performance in the sow. Reprod. Domest. Anim. 2006, 41, 494–500, doi:10.1111/j.1439-0531.2006.00700.x.
- Stott, A.W.; Vosough Ahmadi, B.; Dwyer, C.M.; Kupiec, B.; Morgan-Davies, C.; Milne, C.E.; Ringrose, S.; Goddard, P.; Phillips, K.; Waterhouse, A. Interactions between profit and welfare on extensive sheep farms. Anim. Welf. 2012, 21, 57–64.
- Bartussek, H. A review of the animal needs index (ANI) for the assessment of animals’ well-being in the housing systems for Austrian proprietary products and legislation. Livest. Prod. Sci. 1999, 61, 179–192, doi:10.1016/S0301-6226(99)00067-6.
- Bracke, M.B.M. Animal-based parameters are no panacea for on-farm monitoring of animal welfare. Anim. Welf. 2007, 16, 229–231.
- Blokhuis, H.J.; Veissier, I.; Miele, M.; Jones, B. The Welfare Quality® project and beyond: Safeguarding farm animal well-being. Acta Agr. Scand. A 2010, 60, 129–140.
- Aerts, S.; Lips, D.; Spencer, S.; Decuypere, E.; De Tavernier, J. A new framework for the assessment of animal welfare: Integrating existing knowledge from a practical ethics perspective. J. Agric. Environ. Ethics 2006, 19, 67–76.
- Wemelsfelder, F.; Hunter, T.E.A.; Mendl, M.T.; Lawrence, A.B. The spontaneous qualitative assessment of behavioral expressions in pigs: First explorations of a novel methodology for integrative animal welfare measurement. Appl. Anim. Behav. Sci. 2000, 67, 193–215.
- Wemelsfelder, F.; Hunter, T.E.A.; Mendl, M.T.; Lawrence, A.B. Assessing the “whole animal”: A free choice profiling approach. Anim. Behav. 2001, 62, 209–220.
- Rushen, J. Changing concepts of farm animal welfare: Bridging the gap between applied and basic research. Appl. Anim. Behav. Sci. 2003, 81, 199–214, doi:10.1016/S0168-1591(02)00281-2.
- Waiblinger, S.; Menke, C.; Fölsch, D.W. Influences on the avoidance and approach behaviour of dairy cows towards humans on 35 farms. Appl. Anim. Behav. Sci. 2003, 84, 23–39.
- Keeling, L.; Veissier, I. Developing a monitoring system to assess welfare quality in cattle, pigs and chickens. In Proceedings of the Welfare Quality Conference: Science and Society Improving Animal Welfare, Brussels, Belguim, 17–18 November 2005; pp. 46–50. Available online: http://ec.europa.eu/food/animal/welfare/sum_proceed_wq_conf_en.pdf (accessed on 4 February 2013).
- Krueger, T.; Page, T.; Hubacek, K.; Smith, L.; Hiscock, K. The role of expert opinion in environmental modelling. Environ. Modell. Softw. 2012, 36, 4–18, doi:10.1016/j.envsoft.2012.01.011.
- Deen, J.; Anil, S.S.; Anil, L. Sow housing: Opportunities, constraints, and unknowns. J. Am. Vet. Med. Assoc. 2005, 226, 1331–1334, doi:10.2460/javma.2005.226.1331.
- Botreau, R.; Bracke, M.B.M.; Perny, P.; Butterworth, A.; Capdeville, J.; Van Reenen, C.G.; Veissier, I. Aggregation of measures to produce an overall assessment of animal welfare. Part 2: Analysis of constraints. Animal 2007, 1, 1188–1197.
- Rodenburg, T.B.; Tuyttens, F.A.M.; de Reu, K.; Herman, L.; Zoons, J.; Sonck, B. Welfare assessment of laying hens in furnished cages and non-cage systems: Assimilating expert opinion. Anim. Welf. 2008, 17, 355–361.
- Bracke, M.B.M.; Metz, J.H.M.; Spruijt, B.M.; Schouten, W.G.P. Decision support system for overall welfare assessment in pregnant sows B: Validation by expert opinion. J. Anim. Sci. 2002, 80, 1835–1845.
- Green, P.E.; Krieger, A.M.; Wind, Y. Thirty years of conjoint analysis: Reflections and prospects. Interfaces (Providence) 2001, 31, S56–S73.
- Angus, L.J.; Bowen, H.; Gill, L.A.S.; Knowles, T.G.; Butterworth, A. The use of conjoint analysis to determine the importance of factors that affect on-farm welfare of the dairy cow. Anim. Welf. 2005, 14, 203–213.
- Den Ouden, M.; Nijsing, J.T.; Dijkhuizen, A.A.; Huirne, R.B.M. Economic optimization of pork production-marketing chains: I. Model input on animal welfare and costs. Livest. Prod. Sci. 1997, 48, 23–37, doi:10.1016/S0301-6226(96)01411-X.
- Bracke, M.B.M.; Spoolder, H.A.M. Review of wallowing in pigs: Implications for animal welfare. Anim. Welf. 2011, 20, 347–363.
- Bracke, M.B.M.; Edwards, S.A.; Metz, J.H.M.; Noordhuizen, J.P.T.M.; Algers, B. Synthesis of semantic modelling and risk analysis methodology applied to animal welfare. Animal 2008, 2, 1061–1072.
- Bracke, M.B.M.; Zonderland, J.J.; Bleumer, E.J.B. Expert judgement on enrichment materials for pigs validates preliminary RICHPIG Model. Appl. Anim. Behav. Sci. 2007, 104, 1–13.
- Bracke, M.B.M.; Zonderland, J.J.; Bleumer, E.J.B. Expert consultation on weighting factors of criteria for assessing environmental enrichment materials for pigs. Appl. Anim. Behav. Sci. 2007, 104, 14–23.
- Bracke, M.B.M. Multifactorial testing of enrichment criteria: Pigs “demand” hygiene and destructibility more than sound. Appl. Anim. Behav. Sci. 2007, 107, 218–232, doi:10.1016/j.applanim.2006.10.001.
- American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders: DSM-IV-TR™, 4th. Text Revision ed.; American Psychiatric Association: Washington, DC, USA, 2000.
- Freudenreich, O. Psychotic Disorders: A Practical Guide; Lippincott Williams & Wilkins: Philadelphia, PA, USA, 2008.
- World Health Organization. The ICD-10 Classification of Mental and Behavioural Disorders: Clinical Descriptions and Diagnostic Guidelines; World Health Organization: Geneva, Switzerland, 1992.
- Van Calker, K.J.; Berentsen, P.B.M.; Giesen, G.W.J.; Huirne, R.B.M. Identifying and ranking attributes that determine sustainability in Dutch dairy farming. Agric. Human Values 2005, 22, 53–63.
- Meul, M.; Van Passel, S.; Fremaut, D.; Haesaert, G. Higher sustainability performance of intensive grazing versus zero-grazing dairy systems. Agron. Sustain. Dev. 2012, 32, 629–638, doi:10.1007/s13593-011-0074-5.
- Mollenhorst, H.; Berentsen, P.B.M.; De Boer, I.J.M. On-farm quantification of sustainability indicators: An application to egg production systems. Br. Poult. Sci. 2006, 47, 405–417, doi:10.1080/00071660600829282.
- Castellini, C.; Boggia, A.; Cortina, C.; Dal Bosco, A.; Paolotti, L.; Novelli, E.; Mugnai, C. A multicriteria approach for measuring the sustainability of different poultry production systems. J. Clean. Prod. 2012, 37, 192–201, doi:10.1016/j.jclepro.2012.07.006.
- Anderson, L.H.; Martinson, B.C.; Hall, K.M.; Duncan, I.G. Critical Review of Stochastic Simulation Literature and Applications for Health Actuaries. Society of Actuaries, 2007. Available online: http://www.societyofactuaries.org (accessed on 14 May 2013).
- Plà, L.M. Review of mathematical models for sow herd management. Livest. Sci. 2007, 106, 107–119, doi:10.1016/j.livsci.2006.09.003.
- Boero, R.; Squazzoni, F. Does empirical embeddedness matter? Methodological issues on agent-based models for analytical social science. J. Artif. Soc. Soc. Simul. 2005, 8. Available online: http://jasss.soc.surrey.ac.uk/8/4/6.html (accessed on 26 April 2013).
- Bryceson, K.P.; Smith, C.S. Abstraction and modeling of agri-food chains as complex decision making systems. In Proceedings of the 110th EAAE Seminar on System Dynamics and Innovation in Food Network, Innsbruck-Igls, Austria, 18–22 February 2008; Available online: http://ageconsearch.umn.edu/bitstream/49773/2/Bryceson.pdf (accessed on 26 April 2013).
- Eubank, S.; Guclu, H.; Kumar, V.S.A.; Marathe, M.V.; Srinivasan, A.; Toroczkai, Z.; Wang, N. Modelling disease outbreaks in realistic urban social networks. Nature 2004, 429, 180–184, doi:10.1038/nature02541.
- Yang, Y.; Atkinson, P.; Ettema, D. Individual spacetime activity-based modelling of infectious disease transmission within a city. J. R. Soc. Interface 2008, 5, 759–772, doi:10.1098/rsif.2007.1218.
- Bowles, S.; Gintis, H. The evolution of strong reciprocity: Cooperation in heterogeneous populations. Theor. Popul. Biol. 2004, 65, 17–28, doi:10.1016/j.tpb.2003.07.001.
- Jefferies, P.; Hart, M.L.; Hui, P.M.; Johnson, N.F. From market games to real-world markets. Eur. Phys. J. B 2001, 20, 493–501, doi:10.1007/s100510170228.
- Bruijnis, M.R.N.; Hogeveen, H.; Stassen, E.N. Assessing economic consequences of foot disorders in dairy cattle using a dynamic stochastic simulation model. J. Dairy Sci. 2010, 93, 2419–2432, doi:10.3168/jds.2009-2721.
- Bruijnis, M.; Beerda, B.; Hogeveen, H.; Stassen, E. Foot disorders in dairy cattle: Impact on cow and dairy farmer. Anim. Welf. 2012, 21, 33–40.
- Bruijnis, M.R.N.; Hogeveen, H.; Stassen, E.N. Measures to improve dairy cow foot health: Consequences for farmer income and dairy cow welfare. Animal 2013, 7, 167–175, doi:10.1017/S1751731112001383.
- Waterhouse, A. Animal welfare and sustainability of production under extensive conditions—A European perspective. Appl. Anim. Behav. Sci. 1996, 49, 29–40, doi:10.1016/0168-1591(95)00665-6.
- Febrer, K.; Jones, T.A.; Donnelly, C.A.; Dawkins, M.S. Forced to crowd or choosing to cluster? Spatial distribution indicates social attraction in broiler chickens. Anim. Behav. 2006, 72, 1291–1300.
- Collins, L.M.; Sumpter, D. The feeding dynamics of broiler chickens. J. R. Soc. Interface 2007, 4, 65–72, doi:10.1098/rsif.2006.0157.
- Bischl, B.; Trautmann, H.; Weihs, C. Resampling methods for meta-model validation with recommendations for evolutionary computation. Evol. Comput. 2012, 20, 249–275, doi:10.1162/EVCO_a_00069.
- Halfon, E. Probabilistic validation of computer simulations using the bootstrap. Ecol. Modell. 1989, 46, 213–219, doi:10.1016/0304-3800(89)90018-5.
- Verbyla, D.L.; Litvaitis, J.A. Resampling methods for evaluating classification accuracy of wildlife habitat models. Environ. Manage. 1989, 13, 783–787.
- Berentsen, P.B.M.; Giesen, G.W.J.; Schneiders, M.M.F.H. Conversion from conventional to biological dairy farming: Economic and environmental consequences at farm level. Biol. Agric. Hortic. 1998, 16, 311–328.
- Browne, M.W.; Cudeck, R. Alternative ways of assessing model fit. In Testing Structural Equation Models; Bollen, K.A., Long, J.S., Eds.; Sage: Beverly Hills, CA, USA; pp. 136–162.
- Mackay, M.; Lee, M. Choice of Models for the Analysis and Forecasting of Hospital Beds. Health Care Manag. Sci. 2005, 8, 221–230, doi:10.1007/s10729-005-2013-y.
- Lehmann, D.R. Validity and Goodness of Fit in Data Analysis. Adv. Consum. Res. 1975, 2, 741–750.
- Aleklett, K.; Höök, M.; Jakobsson, K.; Lardelli, M.; Snowden, S.; Söderbergh, B. The peak of the oil age—Analyzing the world oil production reference scenario in world energy outlook 2008. Energy Policy 2010, 38, 1398–1414, doi:10.1016/j.enpol.2009.11.021.
- Guseo, R.; Dalla Valle, A.; Guidolin, M. World oil depletion models: Price effects compared with strategic or technological interventions. Technol. Forecast. Soc. Change 2007, 74, 452–469.
- Owen, N.A.; Inderwildi, O.R.; King, D.A. The status of conventional world oil reserves—Hype or cause for concern? Energy Policy 2010, 38, 4743–4749.
- Sorrell, S.; Speirs, J.; Bentley, R.; Brandt, A.; Miller, R. Global oil depletion: A review of the evidence. Energy Policy 2010, 38, 5290–5295.
- Shafiee, S.; Topal, E. A long-term view of worldwide fossil fuel prices. Appl. Energ. 2010, 87, 988–1000, doi:10.1016/j.apenergy.2009.09.012.
- European Commission. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions. Energy Roadmap 2050. European Commission: Brussels, Belgium, 2011. Available online: http://ec.europa.eu/energy/energy2020/roadmap/doc/com_2011_8852_en.pdf (accessed on 7 March 2013).
- Thornton, P.K. Livestock production: Recent trends, future prospects. Philos. Trans. R. Soc. Lond. B Biol. Sci. 2010, 365, 2853–2867, doi:10.1098/rstb.2010.0134.
- Bindi, M.; Olesen, J.E. The responses of agriculture in Europe to climate change. Reg. Environ. Chang. 2011, 11, S151–S158, doi:10.1007/s10113-010-0173-x.
- Olesen, J.E.; Bindi, M. Consequences of climate change for European agricultural productivity, land use and policy. Eur. J. Agron. 2002, 16, 239–262, doi:10.1016/S1161-0301(02)00004-7.
- Zachariadis, T. Forecast of electricity consumption in Cyprus up to the year 2030: The potential impact of climate change. Energy Policy 2010, 38, 744–750, doi:10.1016/j.enpol.2009.10.019.
- Throsby, C.D. Some notes on “dynamic” linear programming. Rev. Mark. Agric. Econ. 1962, 30, 119–141.
- Sniedovich, M. Dijkstra’s algorithm revisited: The dynamic programming connexion. J. Control. Cybern. 2006, 35, 599–620.
- Langford, F.; Stott, A. Culled early or culled late: Economic decisions and risks to welfare in dairy cows. Anim. Welf. 2012, 21, 41–55, doi:10.7120/096272812X13345905673647.
- Den Ouden, M.; Huirne, R.B.M.; Dijkhuizen, A.A.; Van Beek, P. Economic optimization of pork production-marketing chains. II. Modelling outcome. Livest. Prod. Sci. 1997, 48, 39–50, doi:10.1016/S0301-6226(96)01412-1.
- Kingwell, R. Sheep animal welfare in a low rainfall Mediterranean environment: A profitable investment? Agric. Syst. 2002, 74, 221–240, doi:10.1016/S0308-521X(01)00086-5.
- Vosough Ahmadi, B.; Stott, A.W.; Baxter, E.M.; Lawrence, A.B.; Edwards, S.A. Animal welfare and economic optimisation of farrowing systems. Anim. Welf. 2011, 20, 57–67.
- Guy, J.; Cain, P.; Seddon, Y.; Baxter, E.; Edwards, S. Economic evaluation of high welfare indoor farrowing systems for pigs. Anim. Welf. 2012, 21, 19–24, doi:10.7120/096272812X13345905673520.
- Van Calker, K.J.; Berentsen, P.B.M.; De Boer, I.J.M.; Giesen, G.W.J.; Huirne, R.B.M. Modelling worker physical health and societal sustainability at farm level: An application to conventional and organic dairy farming. Agric. Syst. 2007, 94, 205–219, doi:10.1016/j.agsy.2006.08.006.
- Oudshoorn, F.W.; Sørensen, C.A.G.; De Boer, I.J.M. Economic and environmental evaluation of three goal-vision based scenarios for organic dairy farming in Denmark. Agric. Syst. 2011, 104, 315–325, doi:10.1016/j.agsy.2010.12.003.
- Savory, C.J. Laying hen welfare standards: A classic case of “power to the people”. Anim. Welf. 2004, 13, S153–S158.
- Kilsby, D. Scenario Modelling 101. Prepared for the Engineers Australia/UNSW Joint Seminar “Planning Sydney’s Transport”. 2 September 2003. Available online: http://www.kilsby.com.au/archive/p0903.pdf (accessed on 24 January 2013).
- DEFRA (Department for Environment, Food & Rural Affairs). Total Income from Farming 2011—2nd Estimate. 2012.
- Li, P.J. Exponential growth, animal welfare, environmental and food safety impact: The case of China’s livestock production. J. Agric. Environ. Ethics 2009, 22, 217–240, doi:10.1007/s10806-008-9140-7.
- Kitano, H. Computational systems biology. Nature 2002, 420, 206–210, doi:10.1038/nature01254.
- Koch, I.; Schreiber, F. Introduction. In Modeling in Systems Biology: The Petri Net Approach; Koch, I., Reisig, W., Schreiber, F., Eds.; Springer-Verlag: London, UK, 2011; pp. 3–18.
- Butcher, E.C. Can cell systems biology rescue drug discovery? Nat. Rev. Drug Discov. 2005, 4, 461–467, doi:10.1038/nrd1754.
- Hood, L.; Health, J.R.; Phelps, M.E.; Biaoyang, L. Systems biology and new technologies enable predictive and preventative medicine. Science 2004, 306, 640–643, doi:10.1126/science.1104635.
- Rupprecht, J. From systems biology to fuel-Chlamydomonas reinhardtii as a model for a systems biology approach to improve biohydrogen production. J. Biotechnol. 2009, 142, 10–20, doi:10.1016/j.jbiotec.2009.02.008.
- Gibon, A.; Sibbald, A.R.; Flamant, J.C.; Lhoste, P.; Revilla, R.; Rubino, R.; Sørensen, J.T. Livestock farming systems research in Europe and its potential contribution for managing towards sustainability in livestock farming. Livest. Prod. Sci. 1999, 61, 121–137, doi:10.1016/S0301-6226(99)00062-7.
- Del Prado, A.; Scholefield, D.; Chadwick, D.; Misselbrook, T.; Haygarth, P.; Hopkins, A.; Dewhurst, R.; Jones, R.; Moorby, J.; Davison, P.; Lord, E.; Turner, M.; Aikman, P.; Schröder, J. A modelling framework to identify new integrated dairy production systems. Grassland Sci. Eur. 2006, 11, 766–768.
- Del Prado, A.; Scholefield, D. Use of SIMSDAIRY modelling framework system to compare the scope on the sustainability of a dairy farm of animal and plant genetic-based improvements with management-based changes. J. Agr. Sci. 2008, 146, 195–211.
- Checkland, P.B. Systems Thinking, Systems Practice; John Wiley: Chichester, UK, 1981.
- Lusk, J.L.; Norwood, F.B. Animal Welfare Economics. Appl. Econ. Perspect. Pol. 2011, 33, 463–483, doi:10.1093/aepp/ppr036.
- de Passillé, A.M.; Rushen, J. Motivational and physiological analysis of the causes and consequences of non-nutritive sucking by calves. Appl. Anim. Behav. Sci. 1997, 53, 15–31, doi:10.1016/S0168-1591(96)01148-3.
- Lawrence, A.B.; Terlouw, E.M. A review of behavioral factors involved in the development and continued performance of stereotypic behaviors in pigs. J. Anim. Sci. 1993, 71, 2815–2825.
- Dietl, G.; Nürnberg, G.; Reinsch, N. A note on a quantitative genetic approach for modelling of differentiation tasks. Appl. Anim. Behav. Sci. 2006, 100, 319–326.
- Kanis, E.; van den Belt, H.; Groen, A.F.; Schakel, J.; de Greef, K.H. Breeding for improved welfare in pigs: A conceptual framework and its use in practice. Anim. Sci. 2004, 78, 315–329.
- Lagerkvist, C.J.; Hansson, H.; Hess, S.; Hoffmann, R. Provision of farm animal welfare: Integrating productivity and non-use values. Appl. Econ. Perspect. Pol. 2011, 33, 484–509, doi:10.1093/aepp/ppr037.
- de Lauwere, C.; van Asseldonk, M.; van’t Riet, J.; de Hoop, J.; ten Pierick, E. Understanding farmers’ decisions with regard to animal welfare: The case of changing to group housing for pregnant sows. Livest. Sci. 2012, 143, 151–161, doi:10.1016/j.livsci.2011.09.007.
- Wells, A.E.D.; Sneddon, J.; Lee, J.A.; Blache, D. Farmer’s response to societal concerns about farm animal welfare: The case of mulesing. J. Agric. Environ. Ethics 2011, 24, 645–658, doi:10.1007/s10806-010-9284-0.
- Lusk, J.L. The market for animal welfare. Agric. Human Values 2011, 28, 561–575, doi:10.1007/s10460-011-9318-x.
- Vanhonacker, F.; Verbeke, W.; Poucke, E.; Pieniak, Z.; Nijs, G.; Tuyttens, F. The concept of farm animal welfare: Citizen perceptions and stakeholder opinion in Flanders, Belgium. J. Agric. Environ. Ethics 2012, 25, 79–101, doi:10.1007/s10806-010-9299-6.
- McInerney, J. Animal Welfare, Economics and Policy. Report on a Study Undertaken for the Farm & Animal Health Economics Division of Defra. 2004. Available online: http://archive.defra.gov.uk/evidence/economics/foodfarm/reports/documents/animalwelfare.pdf (accessed 26 April 2013).
- de Boer, I.; Cederberg, C.; Eady, S.; Gollnow, S.; Kristensen, T.; Macleod, M.; Meul, M.; Nemecek, T.; Phong, L.T.; Thoma, G.; van der Werf, H.M.G.; Williams, A.G.; Zonderland-Thomassen, M.A. Greenhouse gas mitigation in animal production: Towards an integrated life cycle sustainability assessment. Curr. Opin. Environ. Sustain. 2011, 3, 423–431.
- Kriström, B. Practical problems in contingent valuation. In Determining the Value of Non-Marketed Goods: Studies in Risk and Uncertainty; Kopp, R.J., Pommerehne, W.W., Schwarz, N., Eds.; Kluwer Academic Press: Norwell, MA, USA, 1997; Volume 10, pp. 235–272.
- Sichtmann, C.; Wilken, R.; Diamantopoulos, A. Estimating Willingness-to-pay with choice-based conjoint analysis—Can consumer characteristics explain variations in accuracy? Br. J. Manage. 2011, 22, 628–645, doi:10.1111/j.1467-8551.2010.00696.x.
- Glass, C.A.; Hutchinson, W.G.; Beattie, V.E. Measuring the value to the public of pig welfare improvements: A contingent valuation approach. Anim. Welf. 2005, 14, 61–69.
- Grimsrud, K.M.; Nielsen, H.M.; Navrud, S.; Olesen, I. Households’ willingness-to-pay for improved fish welfare in breeding programs for farmed Atlantic salmon. Aquaculture 2013, 372–375, 19–27.
- Kehlbacher, A.; Bennett, R.; Balcombe, K. Measuring the consumer benefits of improving farm animal welfare to inform welfare labelling. Food Policy 2012, 37, 627–633, doi:10.1016/j.foodpol.2012.07.002.
- Solgaard, H. S.; Yang, Y. Consumers’ perception of farmed fish and willingness to pay for fish welfare. Br. Food J. 2011, 113, 997–1010, doi:10.1108/00070701111153751.
- Castella, J.C.; Boissau, S.; Trung, T.N.; Quang, D.D. Agrarian transition and lowland-upland interactions in mountain areas in northern Vietnam: Application of a multi-agent model. Agric. Syst. 2005, 86, 312–332, doi:10.1016/j.agsy.2004.11.001.
- Roozmand, O.; Ghasem-Aghaee, N.; Hofstede, G.J.; Nematbakhsh, M.A.; Baraani, A.; Verwaart, T. Agent-based modeling of consumer decision making process based on power distance and personality. Knowl.-Based Syst. 2011, 24, 1075–1095.
- Edwards-Jones, G. Modelling farmer decision-making: Concepts, progress and challenges. Anim. Sci. 2006, 82, 783–790, doi:10.1017/ASC2006112.
- Miller, G.Y.; McNamara, P.E.; Singer, R.S. Stakeholder position paper: Economist’s perspectives on antibiotic use in animals. Prev. Vet. Med. 2006, 73, 163–168, doi:10.1016/j.prevetmed.2005.09.018.
- Araújo, M.B.; Pearson, R.G.; Thuiller, W.; Erhard, M. Validation of species—Climate impact models under climate change. Glob. Chang. Biol. 2005, 11, 1–10, doi:10.1111/j.1529-8817.2003.00895.x.
- Yu, L.; Lai, K.K.; Wang, S.; Huang, W. A bias-variance-complexity trade-off framework for complex system modeling. In Computational Science and its Applications—ICCSA 2006. International Conference Glasgow, UK, May 2006 Proceedings, Part 1; Springer-Verlag: Berlin, Germany, 2006; pp. 518–527.
- Padhye, J.; Firoiu, V.; Towsley, D.F.; Kurose, J.F. Modeling TCP reno performance: A simple model and its empirical validation. IEEE/ACM Trans. Network 2000, 8, 133–145.
- Bolin, C.A.; Smith, S. Life cycle assessment of ACQ-treated lumber with comparison to wood plastic composite decking. J. Clean. Prod. 2011, 19, 620–629, doi:10.1016/j.jclepro.2010.12.004.
- Dawkins, M.S.; Donnelly, C.A.; Jones, T.A. Chicken welfare is influenced more by housing conditions than by stocking density. Nature 2004, 427, 342–344, doi:10.1038/nature02226.
- Mench, J.A.; Falcone, C. Welfare Concerns in Feed-Restricted Meat-Type Poultry Parent Stocks. In Proceedings of the 21st World’s Poultry Congress, Montreal, Canada, 20–24 August 2000.
- Bell, N.J. No Lame Cows—Is it Possible? Experiences from the Healthy Feet Project. In Proceedings of the Cattle Lameness Conference, Loughborough, UK, 14 April 2010; pp. 23–29. Available online: http://www.cattlelamenessconference.org.uk/CLC2010proceedings.pdf (accessed on 26 April 2013).
© 2013 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).