Ecology in Context: A Conceptual Model for Analyzing the Significance of Context in Ecological Research

3.1 Identifying the Focal Object

This section is concerned with the identification criterion, i.e., the question of how to identify what the ecological context is the context of. In section 2, I phrased this question in terms of what the focal object to which the ecological context typically applies is or should be. I have further raised the question of whether we should think of this along the lines of organism-environment interactions. In this section, I will propose an alternative answer to this question that suggests understanding the focal object in a broader sense, namely, as ecological processes.

Nevertheless, approaching the concept of context in analogy to the concept of environment seems initially promising. The ecological literature very often phrases the problem of context dependence in terms of dependence on environmental context, forging a conceptual connection between the concepts of ecological context and environment. Furthermore, although ecological study is directed at different units of research (populations, communities, ecosystems) and spans several levels of organization, the perspective on organisms in their interaction with each other and with the environment is still fundamental to ecology. It has also been an essential part of its connection to evolutionary biology. For this reason, further clarifying the notion of context requires explaining its relation to the closely related concept of the environment.

A classic analysis of the environment concept comes from Robert Brandon (1990, 1996). He distinguishes between three different concepts of environment: the external, ecological, and selective environment, which differ in how they are measured. The external environment is “the sum total of the factors, both biotic and physical, external to the organism that influence its survival and reproduction” (Brandon 1990, 47). These are factors that exist and can, in principle, be measured independently of the organism. Brandon refers to this conception of the environment as “the operative conception in ecology” (Brandon 1990, 47). In this sense, the external environment is the environment with which organisms interact. The second concept is the ecological environment, which “reflects those features of the external environment that affect the organisms’ contributions to population growth” (Brandon 1990, 49). This environment concept involves a shift of perspective, because the ecological environment concept is not measured independently of the organism but from the organism’s perspective. The aim is to get at “the environment as experienced by the target organism” (Brandon 1996, 164). Accordingly, the factors measured to determine the ecological environment represent a subset of the external environment.³

Using these environment concepts as a template already provides some conceptual resources to unify the different ways in which ecologists use the term “context.” Frequently, when ecologists talk about environmental context, what they refer to are the ecological conditions that are found at a particular location or that are characteristic of a type of environment. These conditions include abiotic factors such as climate indicators (e.g., average precipitation and temperature), soil properties, and nutrients, as well as biotic factors like community structure, composition, and species richness that provide a delineation of the biologically relevant characteristics at a location. These are all very similar to the external environment concept, as they can be measured and recorded in principle independently of any focal organism. At the same time, ecologists are also interested in the effects of specific contextual factors on the magnitude and sign of ecological processes. Understanding context in this sense means understanding how processes vary in response to particular factors. This perspective is similar to the ecological environment in that a contextually relevant factor is a subset of the external environment that has specific effects on the focal object.

The key difference between the environment concepts and the notion of ecological context is that the focal object of the ecological context is not necessarily and indeed not even typically the organism, as in the case of the ecological environment concept. To be sure, this is a difference in perspective more than anything else, but one that is crucial nevertheless. Consider the following example.

In a study on the invasive giant bamboo, Phyllostachys bambusoides, Spake et al. (2021) investigated the effects of forest canopy cover on bamboo occupancy in secondary forests in Japan. Specifically, they were interested in the effect of shading caused by the surrounding forest canopy and its consequences for bamboo invasion potential. Generally, shading caused by the canopies of large trees is a form of interspecific competition, as it limits the amount of light available for understory vegetation. In contrast, physiological studies of P. bambusoides have shown that giant bamboo displays photoinhibition at specific light intensities, meaning that the excess excitation energy caused by higher light intensities inhibits photosynthesis. The shade cast by the surrounding forest canopy may thus not be a competitive but rather a facilitative factor, and could partly explain invasion success. In our case, the shade cast by the surrounding forest canopy may be facilitative because it protects P. bambusoides from excess light intensities, which would otherwise suppress its population growth and thereby promote its invasion potential. However, the study revealed that the situation was even more complex, in that the relation between the percent canopy cover and bamboo occupancy is systematically dependent on mean annual temperature. In warmer regions, and for a broader range of light intensities, canopy shade is in fact a competitive factor and can inhibit bamboo occupancy. However, in cooler regions combined with higher light intensities, shade cast by native tree species turns into a facilitative factor for bamboo invasion. So, while the interspecific interaction between invasive P. bambusoides and surrounding tree species is facilitative in cool temperatures, it switches to being competitive in warm temperatures. This paradigmatic case of ecological context dependence tells us something relevant about the identification criterion.

In principle, there would be several possibilities for measuring contextual factors in this case. We could either focus on the organism of interest, in our case P. bambusoides, subsume the interaction partner, i.e., native tree species, together with other biotic and abiotic factors, under the ecological environment, and then measure their joint influence on the organism’s contribution to population growth. This would make sense since the interaction partner in an ecological interaction, whether positive or negative, likely affects this contribution. However, although this makes sense from an evolutionary perspective on organism-environment interactions, with an eye on fitness contribution and selection pressures, it does not make much sense from the ecological perspective of studying the mechanisms of species interactions and their variable contribution to community structure and dynamics in the environmental context. In this case, the focal object is not so much the organism and its relation to the environment, but rather the process of the interspecific interaction in and of itself, including its relation to the broader ecological context. Collapsing the interaction partner into the category of biotic factors measured from the perspective of a particular organism does not make sense in this respect, since it does not allow us to ask how processes vary in response to contextual factors. So, while Brandon’s concept of the ecological environment points us in the right direction, we should understand the second aspect of context differently from the ecological environment – namely, as those factors of the external environment that influence the focal process and thereby change its outcome.

Conceptual models of organism-environment interactions do not provide the same kind of information as models of ecological processes and their interactions with the ecological context. Whereas organism-environment interactions record phenomena at the organism level, process-context models are more suitable to record phenomena at different levels of organization of interest to ecologists. In the case at hand, the process-context view can be applied at the level of interacting populations of organisms of different species and be used to ask how contextual factors constrain the population growth of one of them. It could also be used to raise questions at the community level, in asking how the environmental context influences community structure by shaping interspecific interactions.

Process-context models, as suggested here, are thus more flexible in the sense of being applicable at different levels and to different understandings of ecological units (Jax 2006). They help us understand how context shapes ecological processes, such as interspecific interactions in ecosystems. In this sense, both types of models are suitable for answering different kinds of questions, which harks back to the point above that the key difference here is one about perspective. Ecological communities, understood in their broadest and most metaphysically neutral sense as interactions of species with each other and with their environment at a particular point in space and time, can be studied from different perspectives and with different explanatory goals. Depending on these interests, standard organism-environment models or process-context models, as I am suggesting in this paper, may be more or less suitable to achieve these epistemic goals. Because of its connection to the processes, the notion of ecological context can add a distinct dimension that is not covered by the concept of the environment. For this reason, I argue that the ecological context should be understood relative to a focal process, thereby providing an answer to the identification criterion.

3.2.1 Demarcating Types of Factors

The usual characterization of context in terms of biotic and abiotic factors provides a first narrowing-down of factors to consider. However, some questions remain. As we have seen, there is the problem of how to account for more abstract drivers of context dependence that are also frequently mentioned in the literature, like space, time, or geographic location. For example, Bradley et al. (2020, 995), in their study of the habitat function of mangroves, mention spatial and temporal factors, namely “landscape configuration, such as isolation” and “periodicity, e.g., flooding regimes” of the surrounding environment that can influence the habitat suitability of mangroves at different locations. It is well known that such spatial and temporal patterns can affect ecological processes at different scales of observation (Fletcher and Fortin 2018; Frühstückl 2025). However, while the spatial and temporal structure of the abiotic and biotic factors at a location is, therefore, part of the ecological context, there are also good reasons to differentiate between the factors themselves and their spatial and temporal structure at a location.

Variation in space and time is how context dependence typically manifests. This makes ecological relationships variable across space and time without thereby making space and time causal factors themselves, however (Catford, Jansson, and Nilsson 2009). Rather, ecological processes are influenced by environmental factors, and these factors themselves are spatially and temporally structured. Thus, to the extent that space and time are relevant contextual factors themselves, this is arguably only indirectly through spatially or temporally structuring biotic and abiotic factors at a location. For this reason, while environmental factors can be direct causal factors for ecological processes, space and time matter only indirectly.

In addition, in section 2, I have raised the question of how to treat anthropogenic factors concerning ecological context. While the notion of environmental context as consisting of biotic and abiotic factors might already seem to be all-encompassing, factors that originate in human activity are typically not conceptualized as belonging to the ecological context. If considered at all, they are more often seen as non-natural factors that are external to the “natural” ecosystems that are the object of scientific interest. Inkpen (2017) gives a detailed discussion and interpretation of this tendency to treat humans as “disturbing factors” in terms of model-idealization. Humans are seen as disturbing conditions because their impact prevents a proper understanding of the target ecosystems. However, as Inkpen argues, this practice of idealizing away anthropogenic factors has rarely been adequately justified.

In contrast, proponents of research into social-ecological systems have long pointed out that ecosystems’ structure and processes are rarely independent of human activity and have thus argued for an integrated study of the interactions between human and natural systems (Steel 2014; Pataki 2019; Liu et al. 2021). Sarkar (2005) mentions a pertinent example of this. To promote bird biodiversity, human-induced grazing was banned in Keoladeo National Park in Rajasthan, India, in the 1980s. However, contrary to initial expectations, the ban had adverse effects on biodiversity. The reason was that different species of the Paspalum genus of grasses and other weeds, previously controlled by grazing, could now spread through the wetlands thereby suffocating shallow bodies of water. As a result, fish species declined, which was followed by a decline in bird populations (Sarkar 2005, 42).

There are many cases like this, where the causal effect of an intervention in a natural system depends on the level of an anthropogenic factor, like agricultural practice or other forms of cultivation. For this reason, human-nature interactions have become an important component of ecological studies at several scales. Very often, however, they are dealt with in specialized sub-disciplines. In contrast, the concept of context could serve as a unifying framework that includes anthropogenic factors in the consideration of the ecological context in general.

While anthropogenic drivers are thus certainly part of the ecological context and, thus, generally not to be treated as exogenous variables, I think there is a similar distinction between direct and indirect drivers to be made at this point. To the extent that anthropogenic factors influence ecological processes, at least some of the time, this does not have to be directly relevant to the focal process – it can also be through their effects on other abiotic and biotic drivers. For instance, in the example above, human-induced grazing of grasslands changes the species composition of the ecosystem, with further downstream effects on community dynamics. In this case, as in the case before, we do not have to account for the anthropogenic factor directly, but can instead measure its influence on the abiotic and biotic features of the system.

Accounting for indirect drivers like spatial and temporal structure and anthropogenic effects in this way reveals that for some epistemic aims, considering the ecological context in terms of abiotic and biotic factors will suffice for arriving at a sufficient understanding of the structure and processes of a system. However, in other cases, considering more remote factors, such as spatial and or temporal structure or the anthropogenic influence on some biotic or abiotic factors, can be relevant for getting an adequate picture of the causal relationships, possible pathways of intervention within the studied system, and, in particular, questions about transferability of causal relationships to novel systems. What this suggests is that the context of ecological relationships has a nested structure with factors at the center that have a direct influence on the observed relationship and more remote factors that influence the direct factors in turn. I will come back to the question of what this means for individuating different contexts in section 3.3. Before that, I want to address the second problem of demarcation, which is about relevance.

3.2.2 Demarcating Relevant Factors

Considering the context of an ecological relationship so far boils down to considering several biotic and abiotic factors that describe the conditions under which an ecological relationship is observed. However, figuring out these conditions is only the first step toward analyzing processes in context. Not every factor that is part of the ecological context will also be responsible for context-dependent effects. This raises the question of how we can distinguish between those factors of the ecological context that are responsible for context dependence and those that are only potentially relevant. In this section, I will argue that a criterion should be defined in terms of the causal relationships between contextual factors and the focal process within a broader causal structure. I will also sketch how a structural causal modeling (SCM) framework can be used to systematically test for the influence of contextual factors.

Before going into that, it is perhaps helpful to clarify how ecologists typically conceptualize context dependence in standard statistical models. Ecologists measure context dependence in terms of statistical interactions (Duncan and Kefford 2021; Spake et al. 2023). According to this understanding, the relationship between variables X and Y, representing ecological factors, is context-dependent if the magnitude or sign of the effect of X on Y depends on the value of a third variable, Z. In a causal interpretation, this means that the effect of intervening on X to change the value of Y depends at least on some value(s) of Z. Considering X and Y to be the variables that represent the entities of the focal ecological process, Z is not only part of the ecological context from the perspective of the X − Y relationship but also a relevant contextual factor, since Z modifies the effect of X on Y. Thus, Z needs to be accounted for in a causal model of the process by which X influences Y. This accords with Nancy Cartwright’s point about interactive causal factors: “Two causal factors are interactive if in combination they act like a single causal factor whose effects are different from at least one of the two acting separately” (Cartwright 1983, 31). For this reason, context matters not only to get a clearer picture of the kinds of factors that linearly influence a phenomenon or process but also because the contextual factor, through its interaction with the focal process, is part of the causal structure of the phenomenon. Thus, a more adequate model for the context-dependent relationship between X and Y, in contrast to equation (1) from section 2, is given in equation (2), where the term XZ describes the interaction between X and Z.

$𝑌={𝛼}_0+{𝛼}_1{X}_1+{𝛼}_2{X}_2+ ... +{𝛼}_{3}XZ+є$(2)

However, I suggest that we can use the information about the relationship between X and Y with respect to Z that is contained in equation (2) to formulate a general criterion of relevance of contextual factors which I call the causal path criterion:

Let X and Y stand for ecological variables as before, and let e_i (= Z) denote any biotic or abiotic environmental factor out of a set of such factors that are part of the context of the X − Y relationship. Then, e_i) is a relevant contextual factor if it satisfies the following condition:

$𝑃(𝑌|𝑑o(𝑋 =𝑥), {𝑒}_{𝑖}>𝜃) \ne 𝑃(𝑌|𝑑𝑜(𝑋=𝑥),e_i\le \theta )$(3)

In this inequality, θ represents a threshold value for measurements of e_i. As e_i crosses this threshold value, the effect of intervening on X to change Y changes either in magnitude or direction. (In the binary case, where it is merely about the presence of e_i, θ would be either 0 or 1.) In contrast, if this condition is not fulfilled, an environmental factor e_i would still be part of the ecological context, and it could, in principle, also contribute linearly to the occurrence of Y. However, it would not be necessary to include it in a conceptualization of the X − Y relationship. According to the causal path criterion, a contextual factor becomes relevant if it changes the effect of intervening on the independent variable to change the dependent variable. Relevant contextual factors are thus in some sense part of the causal pathways of the focal relationships and, therefore, need to be taken into account when estimating the causal effect of an independent variable X on a dependent variable Y.

A contextual factor relevant to the causal structure of an ecological relationship represents a distinct type of influence. While numerous biotic or abiotic factors can, in principle, affect an observed relationship, they do not all do so in the same way. Omitting or altering many such factors in a causal model of the X − Y relationship may lead to less precise predictions of the outcome, but it will not alter the causal effect of X on Y. In contrast, changes in the factor e_i directly modify the effect of X, making e_i integral to understanding and conceptualizing the X − Y relationship and the corresponding ecological phenomenon.

The difference here is about whether factors contribute additively or interactively to their effect. Consider the effects of a fertilizer and of planting companion plants on crop growth rates as an example. Both the amount of fertilizer applied and the planting of companion plants will have a causal effect on plant growth. However, when their combined effect is additive, the specific effect of fertilizer on plant growth will stay the same even if we don’t plant any companion plants. If, however, the two factors were interactive, omitting any one factor would impact the causal effect of the other factor. In this case, not planting companion plants would reduce the effectiveness of the fertilizer on plant growth.

The causal path criterion is best used in connection with structural causal models. These models combine structural equations with graphical causal models. While structural equations describe how the value of a variable of interest is a deterministic function of some endogenous variables plus unmeasured exogenous variables (see equation (4)), graphical models such as directed acyclic graphs (DAGs; see figure 1) visualize the assumed causal structure within a system of variables (Pearl 2009; Spirtes, Glymour, and Scheines 2000):

Figure 1:

Directed acyclic graph (DAG) displaying a simple causal structure involving the endogenous variables X, Y, and Z and unmeasured exogenous variables.

$𝑌=𝑓(𝑋,𝑍,U_{𝑌})$(4)

As Arif and MacNeil (2023) argue, using DAGs to model the structure of causal relationships in ecological systems makes it easier to identify biases and spurious associations than statistical models that just list all of the factors that are suspected to be relevant for the phenomenon in question. Furthermore, the DAG visualizes the assumptions about the causal structure within a system that can be based on a combination of already available knowledge (from the literature or expert advice) and statistical or experimental data. The conceptual model of ecological context that I am developing here can be used to assist the construction of a DAG, in that it provides a structured way of thinking about the causal influences on a focal causal relationship (I will come back to this in section 4).

Within this framework of causal inference, applying the causal path criterion amounts to comparing the effects of two different interventions:

$𝑃(𝑌|𝑑𝑜(𝑋=𝑥), 𝑍>𝜃)$(5)

and

$𝑃(𝑌|𝑑𝑜(𝑋=𝑥), 𝑍\le 𝜃)$(6)

and adjusting for Z accordingly when inferring the causal effect of X on Y (Pearl 2009, ch. 3).

Coming back to the question of how to draw the boundary around what is a genuine part of the phenomenon of interest and what is merely contextual about it, we can now give a two-fold answer: On the one hand, the causal path criterion allows us to make some distinction. Contextual factors that satisfy the causal path criterion are not entirely external to the phenomenon of interest. Because of their relatively tight causal integration with the observed process, they are more appropriately treated as part of the process rather than as an external condition of it.

On the other hand, this ties in with the nested structure of the ecological context. Choosing a focal object in accordance with the research aims is the first step. Identifying the context of the focal ecological process is then a stepwise process that starts with identifying the biotic and abiotic factors that can be measured at the location of the process. Depending on the aims and the scale of investigation, this can also include identifying anthropogenic factors that causally influence biotic and abiotic factors, in addition to identifying the spatial and temporal structure of the system. From there, we identify the subset of factors that account (potentially) for context dependence by applying the causal path criterion (supported by drawing a DAG of the causal system). The hierarchically nested structure of the ecological context is also the crucial final piece for addressing the individuation criterion.

4 A Conceptual Model of Ecological Context

Based on the above considerations regarding the explication of the concept of context, I suggest the conceptual model of ecological context presented in figure 2. Depicted at the center of the figure is an interspecific interaction between populations of species i and j. This interaction is the focal process of investigation of the system. The arrow from N_i to N_j is meant to signify that, in this case, the investigation is about the effect of a population of species i on a population of species j. The solid outer perimeter marks the spatiotemporal location, which is also determined by the spatial and temporal scales of observation. The spatiotemporal location includes a number of anthropogenic (a_i) and environmental biotic and abiotic (e_i) factors. The dashed square, separating anthropogenic from environmental factors, signifies that the separation between anthropogenic and non-anthropogenic factors may be permeable and does not in any way represent a sharp boundary. Within the spatiotemporal context, investigation identifies e₅ = Z as a contextually relevant factor, as it modifies the effect of population i = X on population j = Y. The slightly broader dashed ellipse around these elements signifies that by modifying the focal process, e₅ is part of the causal structure of the process.

Figure 2:

Conceptual model of ecological context.

To illustrate this further, we can use this conceptual framework to analyze the case of invasive giant bamboo that I already mentioned in section 3.1. The first step is to identify the spatiotemporal context of the focal object, which is the interspecific interaction between P. bambusoides and native tree species. In particular, Spake et al. considered climatic variables (temperature, precipitation, solar radiation, sunshine duration, snow depth), topographical variables (slope and aspect), stand structure (forest canopy and type), as well as propagule pressure from giant bamboo and anthropogenic factors like distance from roads, which was used as a proxy for the time since management of bamboo forests has been abandoned. (Spake et al. 2021, 1996). The second step is then to identify, among these factors of the external environment, the relevant subset that modifies the focal relationship and thereby accounts for the observed variation in magnitude or sign. In the case of giant bamboo, Spake et al. identified an interaction between temperature, solar radiation, and native tree canopy that could explain variation in bamboo occupancy (cf. Spake et al. 2021, 1998). Because of this causal interaction, the effect of the forest canopy on bamboo occupancy can switch from positive (facilitative) to negative (competitive), which is ultimately responsible for the context dependence of the interspecific interaction between invasive bamboo and native tree species, with the associated downstream effects on transferability and management.

Thus, while processes involving invasive giant bamboo and native tree species can occur in many spatiotemporal contexts, i.e., geographic locations, when studying this interspecific interaction, temperature and solar radiation, and thus context, become an integral part of the phenomenon through their causal interaction with the focal object. What the model contributes to this understanding is a method to bring all these different factors, whether potentially or actually important, into a unifying structure.

Furthermore, this model can also help with addressing the conditions for transferability. Transferability is generally conceived of as an inference from a study to a target system, and the crucial problem lies in justifying or validating these inferences. Transferring knowledge of causal relationships from one ecosystem to another is a complex problem in general. However, it becomes even more problematic when study and target systems are both characterized by a high level of heterogeneity (Steel 2008). In recent years, concerns about the generality of ecological studies have been raised more often, and in many of these instances, context and context dependence are essential components of the problem (Bradley et al. 2020; Spake et al. 2022; Catford et al. 2022).

For instance, Spake et al. (2022) criticize ecologists for often failing to clearly specify the target systems to which their findings are supposed to be transferable:

Ecologists’ statements concerning generality in both primary case studies and syntheses often do not use formal definitions of generality and, in our experience, usually gloss over the assessments required to individuate both the studied context and the target context over which to transfer specific estimands of interest. (Spake et al. 2022, 1820)

They suggest that to remedy this situation, ecologists need to be, first, more precise about the inferences they are making, specifically, whether the inference is from sample to population or from the sampled population to a different population. Second, they also need to be more precise in specifying the study and the target context, i.e., the conditions that potentially influence the observed processes in the study system and the degree to which these are similar in the target system. To this end, they introduce the concept of parameter space (Spake et al. 2022, 1820). The parameter space is an abstract space that is spanned by the dimensions of the edaphic, taxonomic, and climatic variables of a study, such that every point in this space corresponds to a particular distribution of values for each of its dimensions. The context in which individual studies were carried out can then be located within parameter space and compared to the parameters of the target system, thereby providing more precise estimates of transferability.

Bradley et al. (2020) present a similar approach that suggests classifying the ecological contexts of studies into a typology of settings that can be used to support transferability across systems. As they define the term, a setting is a “typology that describes a collection of real world locations that are unified by key aspects of their context and that are therefore likely to share broad similarities in ecological function” (Bradley et al. 2020, 987).

Although the terminology is different, the principle is the same for both approaches. Estimating transferability consists of researching the ecologically relevant conditions for the occurrence of an ecological relationship, which also includes trying to identify those factors that potentially modify that relationship. In a second step, these conditions are then translated into an abstract structure, be that the parameter space as in Spake et al.’s model or settings as in Bradley et al.’s. The outcome is an abstract model of the conditions under which an ecological relationship is expected to be stable. This can then be used to estimate transferability to real-world ecosystems, provided we have adequate knowledge about the distribution of ecologically relevant factors and a good understanding of the causal structure of that system.

The concepts of parameter space and settings are very similar to the concept of ecological context that I have been explicating. Researching the ecologically relevant conditions and singling out those factors that potentially modify the process are analogous to identifying the spatiotemporal context and relevant contextual factors to a focal process. In addition to these approaches, however, the concept of ecological context can also provide a unifying framework that encompasses both conceptions and provides a systematic and structured way of researching the conditions that have to be satisfied for an ecological process to occur. It achieves this by placing different kinds of factors, spatial, anthropogenic, and ecological, within a common structure. In this way, it can help with analyzing processes in complex ecosystems and with addressing the problem of transferability, specifically in conservation or restoration practice.