Mu

Advances in computing technology, new and ongoing restoration initiatives, concerns about climate change's effects, and the increasing interdisciplinarity of research have encouraged the development of landscape-scale mechanistic models of coupled ecological-geophysical systems. However, communication barriers and uneven infiltration of new strategies for data-driven induction persist in the context of simulation model development across disciplines. One challenge is that ecology and the geosciences have embraced different modeling epistemologies, with ecologists historically favoring inductive inference from generalized, phenomenological models and geoscientists favoring deductive inference from detailed first-principles models. Today, many models used for environmental management, particularly for aquatic ecosystems, tend to be highly detailed, with ecological and geophysical components represented in different modules that are linked but often not closely integrated. These observations highlight a need for cross-disciplinary dialogue about landscape-scale modeling objectives and approaches. The philosophies of pattern-oriented modeling in ecology and exploratory modeling in geophysics have yielded advances in theoretical and applied knowledge in both of those disciplines, but they are not comprehensive across all aspects of landscape-scale modeling. Here we define and synthesize the “Appropriate-Complexity Method” (ACME), which builds upon these two philosophies to guide the development of process-oriented models across a spectrum of scientific and management objectives. ACME helps modelers efficiently converge upon an optimal modeling structure through: i) systematic evaluation of the attributes that comprise computational and representational detail, for which we have developed an operational decision tree; ii) iterative adjustment of models based on pattern-oriented model evaluation strategies; and iii) the use of appropriate datasets (where applicable) to build conceptual models and formulate predictions. Decisions about aspects of computational and representational detail are based on the landscape's emergent properties. They are also based on a hierarchy of classes of questions governing model objectives that represent a multi-attribute tradeoff among validation potential, interpretability, tractability, and generality as functions of computational and representational detail. Tradeoff curves, together with model objectives, provide further guidance for determining the “appropriate” level of complexity for representation of processes in models. Once deemed adequate for addressing the original research question of interest, models may be used for projection and scenario testing. They may next undergo expansion that moves them down the hierarchy, where they can then be used to address research questions of higher specificity, detail, and validation potential, though at a cost of lower tractability and interpretability on the tradeoff curves. This practical, systematic procedure provides clear guidance for the design and improvement of landscape models that may be used to address a wide variety of questions relevant to restoration, over a spectrum of scales.

Individual-based modeling is considered an important tool in ecology and other disciplines. A major challenge of individual-based modeling is that it addresses complex systems that include a large number of entities, hierarchical levels, and processes. To represent these, individual-based models (IBMs) usually comprise a large number of submodels. These submodels might be complex by themselves and interact with each other in many ways, which in turn can affect the overall system behavior in ways that are not always easy to understand. As a result, both the validity and credibility of IBMs can be limited. We here demonstrate how a cascaded design of simulation experiments (cDOE) may support the validity and efficiency of the analysis of IBMs and other ecological simulation models. We take a systematic approach that adopts a divide-and-conquer strategy. In a preparatory phase, submodels and their parameters are configured in “subexperiments”. Consequently, the “top-level experiments” of the simulation model can assess the research questions in a more valid and efficient way. Our strategy thus supports the structural realism of individual-based models because both the behavior of their main components and the relationships between these components are explicitly addressed.

Why I like it: A very nicely written example of the use of a machine learning model as a surrogate for a mechanistic model, and the advantages of the combination. In this case, the machine learning model is a self organising map (a type of neural network) and the mechanistic model is an individual based model for species distribution.

why I like it: This is a “must read” for anyone using Ecopath with Ecosim and a “should read” for anyone modelling foodwebs, especially newcomers. It includes excellent advice on how to check that the parameterisation and trophic balance are reasonable, the complexity is appropriate and the model behaves reasonably in standard runs, as well as how to handle uncertainty. 

why I like it: This paper quantitatively evaluates the state of the art in phytoplankton functional type modelling, looking at how the field is growing and how consistently publishing modellers apply accepted best practise in modelling methods, such as validation, reporting of goodness of fit metrics, and sensitivity analyses, as well as how good the results are, in terms of matching observations. An update on the 2004 paper by Arhonditsis and Brett, and important reference point for modellers in this field!

Why I like it: A clear and interesting presentation of an approach to “analog modelling” - that is, predicting by looking for past conditions that are similar to current observed conditions and looking at what actually happend after those similar conditions in the past. This approach has lots of potential for data-rich circumstances, including when good remote sensing data are available.

Why I like it: An interesting data analysis, relavant to phytoplankton functional group modelling. The conclusion is in the title (local niche characteristics determine phytoplankton community composition in lotic systems), which is good news for generically applicable models.

why I like it: The model presented in this paper demonstrates some intensive thinking about how we can really make models reflect detailed physiological understanding of a system component, and how that can be useful in assessing the state of an ecosystem.

what I like about it: This is a nice overview of trends in estuarine modelling, documenting the increasing resolution and complexity of models, discussing appropriate applications of different types of mechanistic estuarine models, and current challenges in hydrodynamic and ecological modelling. Their recommendations for future work are sound, though rather conservative: I guess that’s inevitable with such a large group of authors. A thoughtful review.

Why I like it: This is becoming more and more relevant to modellers as we handle bigger and bigger datasets and use more sophisticated analysis techniques.

This paper takes a rigorous look at a wide range of model performance metrics (statistical indictors of model performance) and the correlations between them, concluding by recommending two small subsets of metrics, each of which collectively captures most of the information included in the full set.

why I like it: This is a review of progress in numerical weather modelling since the 1980s. We are a little behind this in aquatic systems modelling, but are now taking some similar paths, and there are a few things we can learn from the meteorologists.

Why I like it: This provides a neat, simple alternative to the usual approach of having two or more discrete organic matter compartments, instead using a model with one compartment, a parameter, h, describing how far along the decay continuum it is, and a mineralisation rate that is a function of h. This paper applies it to a soil model, but it is easy to see how it could be adapted to an aquatic systems model.

Why I like it: This paper implements a relatively novel modelling approach, explicitly representing the evolution of plankton and plankton diversity. Over a very long simulation in the complex spatial domain of a global model, they compare this approach (starting with a small number of species and letting them evolve to adapt to niches) with the previously explored approach of seeding the whole environment with the full range of possible species and allowing environmental pressures to select survivors. Cell volumes are used as the key feature that is allowed to evolve, with allometric relationships used to influence other factors, such as grazing pressures. They show that the species distributions resulting from the two approaches are similar, but that the evolutionary approach has two advantages: it is less sensitive to variations in initial conditions and it is computationally more efficient. It is perhaps also more conceptually realistic. The 20-year spin-up time for evolution to arrive at a relatively stable configuration might be a limiting factor in some applications.

Why I like it: This has two points of interest. The first is the detailed treatment of dissolved organic matter in a marine model, paying attention to variable C:N:P. The second is that they use matrix inversion to optimise parameters, which is a very direct but often overlooked approach suitable for simple mechanistic models.

Why I like it: This is a nice implementation of a dynamic energy budget (DEB) model, a type of individual based model (IBM) that focuses on the energy uptake and requirements of individual organisms for growth, development and reproduction. It is clearly described and related well to observations.

Why I like it: This is an easy one to overlook, because the highlights and abstract are not well written and the title is opaque. It is worth persevering, however, because the methods, results and discussion are interesting. For a simple population system, the paper compares a continuous, deterministic model (i.e. a traditional deterministic ODE model) with a discrete, stochastic model (an individual based model or IBM), a discrete, deterministic model (an IDE) and a continuous, stochastic model. They evaluate where the models converge and where they diverge, showing that there are some classes of scenarios where the higher computational performance of the continuous models pays off, while there are other population dynamics cases that the these models are unable to reproduce, which often occur at small population sizes. They advocate for the development of hybrid models to handle situations that may fluctuate between large and small population sizes, or include some large and other small populations. This is an intriguing idea and I guess it is, in a way, the approach taken in the model, Atlantis, that uses traditional mechanistic models to simulate phytoplankton production but agent-based models to represent fish and other animals.