Remote sensing vs. Individual-based ecology
Summary:
An effort to assess the relationship and potential synergies of individual-based ecology and remote sensing, and to identify some of the specific challenges of gathering remote-sensing data to develop individual-based ecological theories.
Context:
This was originally written for an Introduction to Complex Systems course I took at U. Michigan’s Center for the Study of Complex System. It was written primarily between the hours of 8pm and 5am the night before it was due, and consequently is on the ranty and ragged side, but it nonetheless contains ideas I find interesting and potentially significant.
The challenges outlined in this paper motivated my thesis research project.
Other Versions
In addition to the paper below, I delivered a short presentation covering a subset of the content:
A .pdf version of the paper is available here.
Content:
- Individual Based Ecology
- IBMs and IBE
- Mechanism in modeling
- Novelty and mechanism.
- Prediction and generalizability.
- Focus on the individual.
- Remote Sensing and Ecology
- Data challenges in ecology.
- The potential for remote sensing.
- Remote Sensing and Individual-Based Ecology
- Sensors, Systems, and Scales
- Choosing a sensor and a system.
- Challenges: spatial resolution.
- Challenges: temporal resolution.
- Challenges: temporal scope.
- Challenges: resolvable system characteristics.
- Challenges: cost and access.
- Challenges: spatial scope.
- Case Studies
- Case study one: treeline development.
- Case study two: intertidal dynamics.
- References
Individual Based Ecology
IBMs and IBE.
In their book Individual-based Modeling in Ecology (2005) veteran individual-based modelers Grimm and Railsback propose a new name for a newish discipline. “Individual-based ecology” (IBE) is to be a complementary arm of the ecological effort, focusing explicitly on how individual-level dynamics produce, in aggregate, ecosystem-level dynamics. This new discipline would be an effort to formalize and extend methodologies and theories which have their ad hoc roots in the development of individual-based models (IBMs). IBMs, in turn, are a class of models defined by their explicit encoding of the behaviors of individuals. They commonly utilize simulation in addition to, or in place of, mathematical analysis. In both their focus on individuals over systems and in simulation instead of purely analytical investigation, they differ from many decades of traditional ecological modeling work.
Ecological modeling dates back to the “golden age” of theoretical ecology, when Lotka and Volterra had their coincident epiphanies about the potential for math to guide insight into natural systems (Lotka 1932, Volterra 1927). Since then substantial effort has been spent on attempting to explain or predict populations and communities of organisms in natural settings. These models have commonly referenced the individuals organisms in those settings only to the extent that they might use whole numbers to describe their quantity (in some cases in fact, the specific mathematic structure of ecological equations allows for fractional or even negative individuals to exist). While such systems-based modeling has contributed to many exciting conceptual triumphs, ecosystems remain only occasionally explainable, and largely unpredictable. The scale of human-induced change in the natural world has made such explanation and prediction increasingly desirable. Both are necessary contributors to rational policy development, and conservation policy will only become more critical as modification of natural systems continues its seemingly inexorable increase in speed and scope. Individual-based ecology represents a fresh and provocative addition to ecological thinking, and a potential contributor to ecological explanation and prediction. IBMs have grown in parallel with the development of complex-systems theory, a wildly cross-disciplinary effort to study systems as the consequence of their interactive parts. IBMs have both drawn from and contributed to the field of complex systems, and complexity and complex adaptive system (cas) theory represents a deep and deepening well of ideas, motivation, and synergy for IBE. (Hartvigsen et al 1998, Levin 1999, Hirsch and Gordon 2001, Green et al 2005).
Mechanism in modeling.
Any model of a real system can have several criteria, and often success in one of these criteria can only be had by sacrificing in another. Prevalent in these trade-offs is the tension between prediction and explanation (Brown et al 2005, 2006). A model may statistically predict the outcome of a system without making any reference to the real components of that system. Alternatively, a model may faithfully reflect each of the relevant entities which build a system in real life, but may reproduce the behavior of the system only qualitatively, without precision. In practice, the nuances of model building compel the modeler to focus on one or the other of these goals. The advantage to predictive modeling is obvious: prediction is often our immediate goal in studying the world, so that we may develop appropriate plans to attain or respond to a probable scenario. On the other hand, the advantage to an explanatory model is that, by representing the system as we truly believe it to be, we can advance and test our theories about the actual mechanisms of that system. Any improvement in the predictive capacity of an explanatory model represents a possible increase in our knowledge of how the real world functions, whereas increasing the precision of a black box predictor only tells us what is going to happen better. If we are driven only by serendipitous statistics, then our capacity to improve our models is limited by our capacity to gather more training data and to improve our statistical capacity to correlate events. While we may rely to a degree on both of these developments, a more fecund hope is that the human capacity for creative pattern recognition will produce radical improvements in mechanistic understanding of a system. Predictive models are in any case typically built upon explanatory understanding. So while predictive models represent potentially lengthy branches off the tree of knowledge, explanatory models extend the trunk itself and provide the possibility of higher level predictors to come.
One of the advantages of Individual-based models, and individual-based ecology more broadly, is that they are particularly suited to mechanistic understanding. Mechanistic fidelity comes from a tightening of the relationship between the entities of a model (either a formal or purely conceptual model) and the respective entities in nature. By contrast, we can also imagine a purely statistical and predictive relationship between models and nature. In this case, the content of a model is irrelevant. It might be an equational black box or it might have narrative components, yet be entirely dissimilar to the components of the natural system the model is meant to predict. As long as the model predicts what the natural system is going to do, its internal structure is moot. Product over process. A mechanistic model identifies what the most relevant entities are in nature and tries to describe and/or mimic those entities as part of its structure.
What the most “relevant” parts of a natural system are will inevitably be guided by our specific and subjective goals. Specific entities are (we are likely to claim) causally linked to specific dynamics. The success of individual-based approaches to date, and the compelling insights offered by complexity research in general, suggests that many of the specific dynamics we might be interested in are most closely causally linked to the entities we think of as ‘individuals’ or ‘agents’. We shouldn’t think that we are trying to obtain absolute granularity in our mechanism. As Volker and Railsback point out, “everything that organisms are and do emerges from interactions among their genes, their neurons, and their environment” (p. 118), and hence if complete mechanistic fidelity were our benchmark, then we would have intractably complex models. But as our models become increasingly individual-oriented, they are likely becoming more mechanistic (Grimm 1999, DeAngelis and Gross 1992). Some drivers in ecosystems truly are system level: energy and nutrient flows tied to climate, soils and hydrology will carry on regardless of the behavior of one organism or another. Those interactions can potentially be encapsulated in a individually-mechanistic model. Many biotic dynamics are the result of individual-level interactions, and those can only be captured by individually-mechanistic models.
Novelty and mechanism.
I suspect that one of the advantages of mechanistically realistic models is that they have a robustness in the face of novelty. Statistical—predictive—models are composed of inferential relationships between the data on hand and the statistics we use to tie them together. As a product of numbers, they are especially vulnerable to overfitting, and can only be trusted to function within the range of conditions for which training data was available to parameterize them. Mechanistic–explanatory–models have a special capacity. If they are good mechanistic models and reflect the true entities in the world, then they will behave like those entities, even if they are presented with perturbations beyond the ranges of the data used to construct them. Hence, they may predict the outcomes of real-world scenarios which we have not yet experienced. Given that human impacts on the ecosystem are modifying system drivers beyond the ranges over which they have previously varied, having models which may predict real behavior outside of the ranges of observed data may be a very useful thing. Most strikingly, global climate change is likely to produce sets of patterns in ecosystems at many scales, patterns which may often be new to human experience (e.g. Niemann and Visintini 2005). Having models robust to such novelty should probably then be a goal for ecology (e.g. Maytin et al 1995). Thus mechanism should be a goal, and thus individual-based ecology seems yet more relevant.
Prediction and Generalizability.
In his A Critique for Ecology, Peters (1991) blames his criticism that ecology is not gaining predictive power on too great of a focus on mechanism. He takes the positivist view that assigning causes to effects is a vain pursuit, at least in the face of the massive number of potentially hidden actors in ecological systems, and suggests that we must focus our efforts on correlation over causation. Ecology has long struggled to produce relationships which are both precisely predictive and also generalizable from spatial location to spatial location, or across scales or among organisms. A science which requires a specialist for every species and region in order to offer prediction is not a very useful science. According to Peters, a research focus on statistical inference will lead to greater statistical inference, and statistical inference is what ecology is lacking. Explanation he likens to a glorified natural history approach.
“In practice, research into causal connections is not the same as the search for predictive power and causal research programs assume some very different characteristics. The crux of this difference is that causalist research seeks first to place the effect in an explanatory model which will also provide predictions. As a result, causalist research seeks first to build a causal web, relegating an evaluation of predictive power to such time as the web is described. Instrumentalist science strives for predictive power from the first” (Peters 1991 p. 133)
Given his focus on predictive power, Peters would probably be uncomfortable with individual-based approaches to ecology. IBE, as with other complexity-based disciplines, draws our attention to the ubiquity of non-linear dynamics, and the consequent path-dependant sensitivity of those systems (Solé et al 1999, Levin 2002). Consequently we appear less, rather than more, likely to accomplish prediction. In particular, our conclusions seem to become less generalizable (Batty and Torrens 2001). Any given mechanism may yield significantly different results from one place to the next, or even from one moment to the next. This would seem to be an anathema to Peters, an almost dialectic divorce from his intended direction for ecological science.
While I agree with Peters that ecology must improve, or at least sophisticate (Brown et al 2005), its predictive ability in order to be relevant to conservation planning, I am swayed by the contrasting arguments of Volker and Railsback. They suggest that the unalloyed desire for generalizability which has developed in sympathy with Peters has cramped ecology into a posture inappropriate for its task.
Ecological science in the 20th century was marked by tension between traditional, descriptive, ‘natural history’ approaches and those who entered the field because they believed that they could productively apply their experiences in the hard sciences of physics and math. In what strikes me as both a response to Peters specifically and a rallying cry for the unity of the two ecological schools, Grimm and Railsback reflect that
“Intuitively, we think of “general” theory as being independent of specific contexts. Theory in physics is very general in this way, but physics deals with matter and forces, which are indeed independent of history and context. Seeking the same sort of generality in ecology has not proved to be very productive….Organisms are not atoms, and the “forces” of ecology are not fundamental properties of matter and space but emerge from interaction among individuals and their environment. In ecology, therefore, useful theories are likely to be context-specific and the search for generality must include the search for the limits of this generality”. (2005 p. 56 -57)
Coming to grips with the limits of prediction suggested by complex dynamics remains a major barrier to individual-based ecology, but one which may be revealed rather than created by it. Already we are developing strong answers to questions about generalizability.
Focus on the individual.
By putting “individual” in the proposed name of the discipline, Volker and Railsback are going further than saying that it will be based on the methodologies of complexity, or of cas. Many processes from other disciplines, which are identified as complex adaptive systems because of their focus on the emergence of system properties from the interactions of agents, do not necessarily have as their ‘agent’ an individual. In some cases this is because they operate at sub-individual scales, i.e. in physics or neurology, and in some instances, systems which include individuals are modeled using some aggregate level of organization as the agent. For instance, In human ecology, the ‘household’ is commonly the unit (e.g. Clarke et al 1997) , and in some existing complexity-based ecological models, ‘grid cells’ full of individuals (Moloney and Levin, 1996), metapopulations (Guichard 2005), or whole species (Nuetel et al 2002) may make collective choices for themselves. The constraint to individuals is intended to leverage the most robust theory in biology: evolution through natural selection. If we are to choose a mechanism which is most reliable as an explanatory factor in biology, that is the one. Thus, for optimum mechanistic robustness, Volker and Railsback suggest we focus at the level upon which natural selection acts. Notwithstanding the possibility that the individual is not that level (Dawkins 1989), the individual is certainly the level at which we can most tractably study selection, and thus this is the foundation they recommend for developing a new domain science.
Remote Sensing and Ecology
Data challenges in ecology.
Peters is not being alarmist when he asserts that ecology has problems. Most sciences have as their domain isolatable subjects. Physics studies forces and types of matter which retain their properties even when in isolation from each other. Although chemistry is more fundamentally about interaction of components, those interactions can be realistically captured in the confined and controllable space of a test tube. Other branches of biology may be in part concerned with the interactions of their subjects with the outside world or each other, but generally not as the primary focus of research. Ecology on the other hand is specifically about what happens in the world, and the world is a highly variable and interactive place. A phenomenon of interest may be driven by biotic, abiotic, or a combination of forces. Potential abiotic drivers, such as temperature, humidity, chemical composition, shade or wind, may vary at the scale of meters or less, and some or all may be interactively incident on a subject. Biotic drivers– other organisms–are even more abundant and are also potentially variable and interactive. Establishing any entity as causal with any given effect can be a challenge. There is always the lurking possibility that a third, unmeasured factor indirectly correlates what you thought were a causally linked pair. Laboratory studies can limit variance, but at the cost of explanatory power: just because flour beetles behave a certain way in the lab, doesn’t mean other beetles necessarily do the same in the crazy real world, even if it is suggestive of it. For any perturbation or observation experiment, gathering enough data to convince yourself that all drivers not under study have been averaged out can be a significant challenge. There is often a serious tension between the need for high sample sizes in the face of substantial environmental variance and the amount of time, resources, hip waders and undergraduates on hand to gather the data. Time is another significant factor. Often ecological systems cycle over the course of seasons or even decades. Studying a classic predator-prey relationship such as wolves and moose for instance, may yield apparently random results at the scale of years or even generations, when real cycles do exist and can be observed if data is recorded for many decades. Such long-term research programs exist, but are rare due to the obvious resource and personnel challenges they face.
The potential for remote sensing.
Given these fundamental challenges, remote sensing offers some hope as a new, different, and potentially resource-effective source of data. Remote sensing (RS) is very broadly the use of sensors to record characteristics of subjects which are not physically contiguous with the observer. In practice that usually means satellite and sometimes plane-based imaging of landscapes, although other means exist. In contrast, traditional “ground-sampling” typically consists of recording observations at a series of precise points over an area of interest. Those observations may themselves be as precise as time and methodology allows: tree rings may be cored, streams may be chemically assayed, florets may be counted. But there is almost always more space in between the measurements than the measurements themselves represent. Quadrats are usually spread out. Remote sensing by contrast measures ubiquitously but fuzzily. In a remotely sensed image, every place on the ground has a corresponding pixel. It may however share that pixel with a substantial parcel of the surrounding ground, depending on the spatial resolution of the sensor. Depending on the sensor there may also be serious limitations as to what information has been recorded about that patch of ground. ‘Panchromatic’ sensors can only record levels of visible spectral reflectance of a canopy, not the number of rings in the trees below. ‘Multi-spectral’ sensors can record the presence of a stream, but probably not much about its chemical composition.
Even given these data-quality limitations, remote sensing should at the very least represent a new type of data, and consequently represent a new channel of ecological exploration. Further, many of the dimensions of data limitation are being pushed back, in part by technological improvement in sensors, and largely by improvements in data interpretation. Careful analysis of spectral reflection at the level of the leaf or needle (Stimson at al 2005) may scale to measurement of subtle physiological status at the level of tree, canopy or landscape (Panek and Ustin 2004), and investigation of the interaction of light and fluid may open up streams and seabeds to aerial inspection (Goodman and Ustin2003). Experience with the integration of imagery from multiple sensors to produce common data sets, and the analysis of spatial in combination with spectral pattern, generates synthetic resolutions previously unresolvable (Greenberg et al 2006). Consequently, the trade-off of ubiquity of data versus precision of data seems to increasingly weigh in favour of remote sensing as an ecological data-gathering tool.
Temporally, remote sensing can cover days, seasons, years, and even in some cases decades. Satellites are typically the platform with the best coverage of an area in time, The best may revisit a study site daily, the worst monthly, or only when requested. The longest-term ecological monitoring project that can be conducted currently stands at about 40 years, beginning in 1963 when the American CORONA spy satellites began recording black and white images of the world and dropping film canisters out of low orbit. While this hard limit may at first seem like a mark against RS, it can compare favorably with ground-sampling, which is limited to the present and future and which becomes expensive if even monthly revisits to a study site are required.
Despite these ecological-research-friendly properties, actual use of remote sensing data in ecological research seems to be limited. Although widely used for mapping and monitoring of conservation concerns (for example, Greenberg et al 2005, Stimson et al 2005b), RS data is much less likely to be used to test ecological hypotheses. In instances when RS data does transcend mere mapping and monitoring, it is typically to be incorporated into dynamic modeling, but only to provide a “before” map for the execution of established spatially predictive process models. It is less commonly used for the testing of mechanisms that may ultimately be built into new predictive process models. This may be due to the relative novelty of the tool, or the traditional predominance of engineers in its use, or the cross-disciplinary demands of integrating technical knowledge with ecological knowledge. As remote sensing data proliferates and improves, and the technical know-how to interpret it gradually proliferates and improves, the potential for its use as a driver of ecological theory seems apparent.
Remote Sensing and Individual-Based Ecology
Spatially-explicit models are a common feature of IBM, and as with remotely sensed data, that spatiality is typically of 2 dimensions. Although individually-based dynamics may alternatively be represented as, for instance, spatially implicit topologies, a survey of the sample IBMs reported in Grimm and Railsback (2005 ch. 6.2-6.5) yields 10 out of 16 models which represent space as a 2 dimensional plane with non-varying scale. There is an obvious potential for the interaction of remotely sensed imagery and the output from such spatially-explicit IBMs (or for the use of RS imagery as an input of IBMs). There are extant computer science challenges for the actual integration of RS data into IB models (Brown et al 2005b) but these are in the process of being identified and addressed and are certainly surmountable.
A key issue in the integration of remote sensing and specifically individual-based ecology is the capacity to resolve individuals. Less challenging but also necessary is the ability to encompass the system which their aggregate behavior is expected to characterize. Application of sophisticated remote sensing techniques may well be necessary to process and produce spatial scenes for incorporation or comparison with spatial IBMs.
More generally, the importance of spatial pattern in ecology is well established, and spatial ecology is a growing field. If IBMs can be made more spatially robust, either through the tempering effect of empirical data testing or other methods, they have the potential to shape and be shaped by an exciting and dynamic subdiscipline of ecology (Grimm et al 2005, With and King 2004, Camarero et al 2005, Brown et al 2006, Watt 1947, Levin 1992)
Related to the ‘natural history’ v. ‘hard science’ divide in ecology is the divide between field ecologists and theorists. It may not be far from factual to stereotype field ecologists as dismissing theorists as only vaguely relevant and vice a versa. This divide may in part be related to the difficulty of collecting statistically significant quantities of data. Grand theory tends to remain grand and somewhat remote in the absence of ready testability. A platonic air of speculation founded only on better-established speculation seems to pervade theoretical biology. As it is so hard to actually check if a grand-scope ecological theory is true, mathematical elegance eclipses worldly verisimilitude as the highest practical attainment. Individual-based ecology should not be allowed to drift in this direction. While the spiritual and aesthetic pleasures of mathematical theory may be real and positive, the demands of conservation planning are too pressing to allow abstract elegance as a priority for the discipline. The availability of real data in a form neatly applicable to the models would help to keep the focus on mechanistic representation and accurate prediction.
To this end, there is a clear path for integration of remotely-sensed data into IBE. Although nonlinearities may render it unlikely for a model to reproduce the exact patterns of a real landscape, they may well be able to reproduce the more general patterns. Grimm and Railsback devote a chapter (2005 ch. 3) to the methods and goals of pattern-oriented modeling. Although they don’t dwell substantially on methods for gathering data for pattern-based comparison to model predictions (hint: remote sensing) they do point out that by refining IBMs based on comparisons of their output to observed spatial data, individual-based modeling can become a component of the cycle of observation-hypothesis-testing-observation generally associated with the inductive scientific method.
Sensors, Systems, and Scales
Choosing a sensor and a system.
Commonly, remote-sensing research programs are guided by the research lineage of the individual doing the research. Choice of sensor and/or system is typically the consequence of the sensors and systems the researcher’s laboratory colleagues have pursued previously. This kind of path-dependence in research program is not necessarily a bad thing: the deeper and more subtle a researcher’s understanding of the usage tricks of a given imagery type, and the deeper their understanding of the biological realities they are attempting to sense, the more likely their efforts will bear real fruits. In any case, often the relationships established between data providers and data users make it such that a given type of imagery is most easily attainable for a given user.
Given the laundry-list of hurdles to the remote sensor trying to resolve individuals however (see below), such a researcher may have to be willing to cast a wider net. Matching a sensor and a system will be a victory in itself, and it may be necessary to submit oneself to serendipity of scale overlap, regardless of preference or experience. Consequently, a rigorous cataloguing of the relevant scales of sensors and systems may represent a useful or even necessary precursor to choosing a specific research topic (Stimson 2007).
Challenges: spatial resolution.
Probably the single greatest challenge facing the individual-based remote sensor is the ability to resolve the spatial dynamics of individuals. Given that the maximum current resolution of satellites is around the 1/2m scale (and that is only available for the simplest, panchromatic imagery), and to really resolve a subject the subject must be larger than the minimum pixel size, the vast majority of species on earth are necessarily precluded from conventional remote sensing. There are at least two responses to this problem: one is to accept that satellites can only be used for medium and large scale vegetation (e.g. trees and bushes—groundcover is also resolvable at this scale but not at the truly individual level). Another is to use plane-based sensors, in particular aerial photography, which may have resolutions as low as centimeters, but which suffer from high cost and low repeatability. These would make smaller scale vegetation and sessile organisms a possible subject. A third is to construct purpose-built instruments, which have a high front-end cost but potentially lower incremental costs, at least compared to plane-based instruments, and which are less easily repeatable than satellites but more so than plane-based sensors. Unlike satellites (and to an extent plane-based sensors) they cannot collect data in the past. Like plane-based sensors, purpose-built sensors provide the possibility of smaller-scale vegetation and sessile organisms (e.g. Guichard et al 2000 and 2003, Guichard 2005).
Given that spatial resolution seems to make medium-and-large scale vegetation the most likely subject for individual-based remote sensing (‘IB-RS’), it is encouraging to note that effort and skill is being deployed in using hybrid image processing techniques to resolve satellite imagery to the level of individual tree crowns (e.g. Greenberg et al 2006). This represents a potential ‘prime target’ for IB-RS. Consider, for example, the data needs of Solé and Manrubia 1995. Another consideration is a focus on ‘high-contrast’ individuals, those which are more easily resolvable due to a high differential of their reflective characteristics from the background matrix they exist in. These are commonly organisms in biotically stressed environments, such as krummholtz trees at the tree line (Alftine and Malanson 2004) or woody shrubs and trees in semi-arid ecosystems (Stimson et al 2005). Such systems tend to have biotically-driven dynamics which are not necessarily generalizable to more common abiotically-constrained systems. One of my personal research interest is in the relative prevalence of positive feedback in biotically constrained systems and negative feedback in abiotically constrained systems, and as such these high-contrast systems are of interest to me.
Challenges: temporal resolution.
Temporal resolution interacts with spatial resolution in providing some of the hardest constraints on individual-based sensing. No satellite and, to my knowledge, no other existing sensor observes any given area continuously. Without continuous observation, and with the exception of very large individuals indeed, it would be impossible to identify individuals from time sample to time sample except with regards to their position. So any investigation must be of dynamics which occur on a temporal scale equal or greater than the revisiting time of the sensor. This would seem to preclude any non-sessile animals (except perhaps sloths) and place the focus purely on vegetation. The only exceptions might be for purpose-built, personally deployed sensors. Another possible avenue of progress (not much considered here due to my lack of familiarity with it) is the rapid progress being made in widely deployed ‘spatially embedded sensor networks’ of the ‘smart dust’ variety (for these and other exciting frontiers, see Green et al 2005). Perhaps some of those sensors could be image-based, potentially providing coverage of a medium or small-scale landscape nearly ubiquitously in time and at fine spatial resolution.
Challenges: temporal scope.
One of the wonderful things about satellites is that they’ve already collected the data. Although it is possible to acquire the freshest data, it is equally easy to use data from any past season. Many (though not all) satellites record constantly and globally and archive comprehensively. Consequently, if a match can be found between sensor and system, then the data is a done deal. Field work will inevitably be required, but research is contingent more on the known and documented past sensing programme than its future funding and maintenance. The challenge is not so much to collect the data as to make it relevant.
This is not so true of non-satellite RS data. In some cases, the commissioners of expensive plane-collected imagery will be reluctantly willing to release it for further use, but the chances of the locations and times of previous research being relevant to new research are reduced compared to the more comprehensive back-catalogs of satellite data. Individually built and deployed sensors, to the extent that they exist (e.g. Guichard et al 2000), offer a slightly different set of challenges again. In these cases, there will be a very limited set of back-cataloged data available. Individual willingness to release that data will likely depend on individual idiosyncrasy. Perhaps more importantly, the chance that a prospective researcher will know that someone else’s individually-collected data exists is even less than for plane- or satellite-based collection. The campaign histories of research planes are often documented in some public way, and satellite data typically has elaborate data access tools on their respective websites. Individually-collected imagery is primarily discovered through individual papers in the literature or through personal acquaintance.
Challenges: resolvable system characteristics.
Many IBMs focus substantially on presence/absence, or state variables derivable from repeated presence/absence measurement, such as age (Greenberg et al 2005) . For these characters, the limiting sensor characteristics are likely spatial and temporal resolution. However, there is substantial possibility of gathering more subtle, even physiological measurement using sensed landscape imagery (Panek and Ustin 2004), and these characteristics might be used as inputs for or tests of analogous state variables in IBMs. In such cases, the imaging characteristics of sensors, such as multi-spectral, hyper-spectral, LIDAR, and radar, become substantial factors. Although these sensor-characteristic concerns represent the bulk of what worries remote sensors, they would be highly study-dependant, and for IB-RS would likely become a limiting factor only once less subtle constraints of scale have been resolved.
Challenges: cost and access.
The financial and procedural burden of obtaining remotely-sensed imagery varies wildly depending on the public or private nature of the sensing platform, and the personal connections of the remote sensor. Although difficult to characterize a priori these logistical considerations are likely to frustrate any non-pecuniary assessment of the ideal sensor-system match.
Challenges: spatial scope.
Most sensors represent a likely match between spatial resolution and spatial scope. The spatial scope of purpose-built and airborne sensors tends to be from 100s of meters to kilometers. Those scales seem likely to encompass the dynamics of most systems which I can imagine being resolved at airborne spatial resolutions, which is typically meters to dozens of meters. Lower resolution, satellite-based sensors typically encompass scenes from many to 100s or even 1000s of kilometers across. For systems of greater spatial scope, mosaicing of contiguous images represents greater access costs and potential image-processing road-blocks, but is often still a viable option. Spatial scope is thus a mercifully minor challenge for IB-RS.
Case Studies
Case study one: treeline development.
Alftine and Malanson (2004) conducted an investigation of the patterns of tree establishment at the alpine-subalpine transition zone. They observed that at the treeline edge, the pattern of tree growth was patchy and irregular. Noting that “large-scale properties of vegetation are inextricably linked with the fine-scale mechanisms by which individual plants can affect and respond to their local environment”, they developed a model of tree establishment built on a cellular automata framework, in which per-grid-cell tree establishment was affected by wind exposure. According to this model, the likelihood of a tree growing in a location was increased if other trees were already in place around that location to block the prevailing wind, and enhance the soil quality. Field sampling was conducted to measure 2-dimensional tree presence and absence over plots mirroring their model grids in scale. The model, which represented interactions at the individual level, was tested against a null model of non-interactive tree establishment for its ability to match the observed data on a few simple pattern metrics. The IB version of the model was significantly more realistic in the spatial patterns of tree establishment produced.
This represents an example of an IBM study which successfully incorporated empirical data to test and refine its assumptions. It also represents an example that could have more easily conducted with remotely sensed, or greatly expanded. Although the cost of tree-resolution data can be prohibitive, the lack of necessity for a time series of data means that such a study could easily ‘piggy-back’ on existing data from any existing study. Many high-resolution scenes have been recorded at the alpine-subalpine transition and could be borrowed or begged from fellow researchers (see 4.f). Of the variables field-sampled for the study, only two (tree position and elevation) were collected across the entire sample area. Both variables could have been readily generated using high-resolution imagery. All other state variables in the model (wind direction, soil quality, initial site quality) were developed by point sampling and interpolation based on elevation. Point sampling in a similar number of locations could be part of a RS-based version of this study, and would represent an excuse for field work.
Case study two: intertidal dynamics.
Guichard and colleagues have conducted a series of studies around the dynamics of growth and survival of mussels and other inter-tidal organisms. The research has focused on the interaction of biotic with abiotic forces (such as tidal flow and bed topography) and has included considerable study of individual-level (Guichard 2003) or other complex interactions (Guichard et al 2005). To build on the empirical nature of this work, a remote sensing platform was constructed using a 6m hydrogen blimp and suspended consumer camera , (Guichard et al 2000). By taking stereo images multiple times over a season, a high resolution 3D map of vegetation presence and development was constructed. The initial study focused on relationships between topographic heterogeneity and organism density, and explicitly compared model predictions with observed patterns. With the remote sensing platform and associated protocol established, the authors are well-positioned to develop their models of the relationships of scale and interaction with more robust empirical data.
This represents an example of the ideal integration of remote sensing with theory: a well established system is being studied using a sensor capable of capturing the relevant scales of both individuals and the encompassing system. Of particular interest is that fact this study includes mussels as study organisms: by focusing on a sessile species, this represents remote sensing of non-vegetation, probably the only example I am aware of (or can imagine). Although limited by the labor necessary to deploy the platform in the field for each desired time-sequence, the fit of sensor resolution to system scale yields considerable latitude for ecological exploration.
References.
statistical, and computational challenges. Bioscience. 55(6).