Robert Muetzelfeldt

The Simile visual modelling environment

Abstract Simile is a visual modelling environment that has been developed to overcome the problems involved in implementing agro-ecological simulation models using conventional programming languages: problems such as the effort and skill needed to program the models, the lack of transparency in models implemented as programs, and the lack of re-useability of models and submodels. It combines the familiar System Dynamics (compartment-flow) paradigm with an object-based paradigm, allowing many forms of disaggregation to be handled, as well as spatial modelling and individual-based modelling. Its visual modelling interface makes it accessible to non-programmers, at the same time allowing models to be largely self-documenting. Models can be run very efficiently as compiled C++ programs, and users can develop new visualisation tools for displaying model results. Simile has been used in international research programmes, including the modelling of Mediterranean vegetation dynamics and modelling the interaction between households and land at the forest margin in developing countries. Simile has been developed in a spirit of open standards for model sharing. Models are saved as a text file in a structured format, with a view to enable model sharing with other modelling environments and to encourage others to develop additional tools for working with models.

Multiscale digital Arabidopsis predicts individual organ and whole-organism growth

Significance Plants respond to environmental change by triggering biochemical and developmental networks across multiple scales. Multiscale models that link genetic input to the whole-plant scale and beyond might therefore improve biological understanding and yield prediction. We report a modular approach to build such models, validated by a framework model of Arabidopsis thaliana comprising four existing mathematical models. Our model brings together gene dynamics, carbon partitioning, organ growth, shoot architecture, and development in response to environmental signals. It predicted the biomass of each leaf in independent data, demonstrated flexible control of photosynthesis across photoperiods, and predicted the pleiotropic phenotype of a developmentally misregulated transgenic line. Systems biology, crop science, and ecology might thus be linked productively in a community-based approach to modeling. Understanding how dynamic molecular networks affect whole-organism physiology, analogous to mapping genotype to phenotype, remains a key challenge in biology. Quantitative models that represent processes at multiple scales and link understanding from several research domains can help to tackle this problem. Such integrated models are more common in crop science and ecophysiology than in the research communities that elucidate molecular networks. Several laboratories have modeled particular aspects of growth in Arabidopsis thaliana, but it was unclear whether these existing models could productively be combined. We test this approach by constructing a multiscale model of Arabidopsis rosette growth. Four existing models were integrated with minimal parameter modification (leaf water content and one flowering parameter used measured data). The resulting framework model links genetic regulation and biochemical dynamics to events at the organ and whole-plant levels, helping to understand the combined effects of endogenous and environmental regulators on Arabidopsis growth. The framework model was validated and tested with metabolic, physiological, and biomass data from two laboratories, for five photoperiods, three accessions, and a transgenic line, highlighting the plasticity of plant growth strategies. The model was extended to include stochastic development. Model simulations gave insight into the developmental control of leaf production and provided a quantitative explanation for the pleiotropic developmental phenotype caused by overexpression of miR156, which was an open question. Modular, multiscale models, assembling knowledge from systems biology to ecophysiology, will help to understand and to engineer plant behavior from the genome to the field.

HIERARCHICAL APPROACH TO FOREST ECOSYSTEM SIMULATION

Abstract Predictions of impact of environmental change on forest ecosystems need to utilise available data collected at a range of organisational scales: cell, leaf, tree, and forest. The model FORDYN is designed to scale up biochemical and physiological data to make predictions about forest growth and development over periods of many years. There are four hierarchical levels within the model. Level 1 represents a set of individual trees within the ecosystem, and deals with the establishment, growth and mortality of individual trees. Level 2 deals with the daily allocation of biomass to above- and below-ground parts of each tree, which is determined according to the Thornley transport-resistance method. Level 3 permits physiological data to be drawn into play: it includes hourly calculations of photosynthesis, respiration according to empirical relationships available from the literature, as well as the transpiration rate calculated from a form of the Penman-Monteith equation. Level 4 incorporates the Farquhar model of C3 photosynthesis, which provides the basis for simulating the impacts of temperature and CO2 on the ecosystem. An important feature of this model is that the user can decide which of the four organisational levels are to be used in the simulation. In mode 1, only level 1 is called. This mode may by appropriate for use by a forest manager. In mode 2, levels 1 and 2 are called; in mode 3, levels 1, 2, and 3 are called, and in mode 4, FORDYN uses all 4 levels of organisation. Linkage between levels is discussed. Generally, results from a low level provide an input to the next higher level (scaling up by integration, whether by summing or by multiplication). Results from a high level frequently feed back to the lower level, by setting a constraint (usually, a state at the higher level influences the rate at the lower level).

THE USE OF PROLOG FOR IMPROVING THE RIGOUR AND ACCESSIBILITY OF ECOLOGICAL MODELLING

Abstract We introduce three concepts that offer considerable benefit to the process of ecological modelling: the descriptive representation of models; the explicit representation of knowledge about how to model; and the development of knowledge-based systems that can help ecologists construct models. Prolog, a computer language based on formal logic, has much to offer in realising these ideas. We introduce the concept of a ‘model blueprint’, a complete, formal specification of the structure of a model, and show how a blueprint can be represented as a Prolog program, basing our analysis on system dynamics models for simplicity. We consider ways in which the Prolog interpreter can be used selectively to retrieve information about the model, to check for errors in the formulation of the model, and to evaluate the model mathematically. However, there are drawbacks with this approach, so we discuss ways of overcoming these by implementing - also in Prolog - programs which buffer the user from the difficulties of working at the level of the Prolog interpreter. These include the generation of descriptions of model structure, and the development of a program to help in the construction of simulation models.

The ECO program construction system: ways of increasing its representational power and their effects on the user interface

There is a growing interest in programs which help users with little experience of computing to construct simulation models. Much recent development work on such systems has utilized comparatively simple mathematical methods (such as System Dynamics) to facilitate the development of a friendly user interface. The problem with these simple modelling languages is that they assume that users have preconceived ideas of the simulation models which they want to build. In the EC0 project, which involved the construction and testing of programs to help ecologists build simulation models, it became clear that users could not always adapt their ideas to fit into these mathematical frameworks. They required a more expressive input language in which to describe their modelling problems, rather than being forced directly to specify the programs which solved those problems. However, we found that as the input language became more sophisticated the complexity of the user interface became disproportionally larger. We attempt to clarify the reasons for this phenomenon by comparing the various systems which we built to try to solve this problem. This comparison is facilitated by the use of a sorted logic as a lingzur franca for the various formalisms used in each system. Our analysis centres around a small number of key characteristics which we use to highlight the strengths and weaknesses of various dialogue techniques. Many ecologists would like to construct computer simulations of ecological systems, but are discouraged by the complexity of current simulation languages. They must learn how to program in an available language (a considerable effort); or describe their model to a sympathetic modelling expert who will then implement their model on their behalf. Our goal is to relieve ecologists of this handicap by providing a computer system which is easy to use and which helps them transform their view of a problem from ecological terms, ultimately to working computer programs. A key part of our research has been the design and testing of various mechanisms for conducting dialogues between humans and computers in order to obtain specifications for working programs. This paper contains an analysis of these dialogue systems. Our central thesis is that, although it is relatively easy to provide interface methods for users familiar with the mathematical formalism upon which a program construction system is based, it is much more difficult to scale up these techniques to

ZimFlores: A Model to Advise Co-management of the Mafungautsi Forest in Zimbabwe

ZimFlores (version 4) is the outcome of a participatory modelling process and seeks to provide a shared factual basis for exploring land-use options for the communal lands surrounding the Mafungautsi forest. The ZimFlores experience underscores the importance of a sharing a common problem and a common location in which all participants have an interest. Participatory modelling has proved an effective way to consolidate a diverse body of knowledge and make it accessible. Results demonstrate the importance of model outputs that are diagnostic, and which offer insights into the issues under consideration.

Development of a crop knowledge base for Europe

This paper describes the development of a Crop Knowledge Base System for use in crop modelling and yield forecasting. The inherent limitations of databases can cause problems for dealing effectively with geo-referenced data sets. The system described here includes all the normal database operations but also handles richer types of data. Values of queried attributes can be deduced through the use of heritability within hierarchies and by the application of rules that summarise sets of empirical observations. The system is able to reason about similar locations, so that where no information exists in the knowledge base for a particular location, then an alternative location, subject to similar meteorological and agricultural conditions, will be sought.

Modelling decision-making in rural communities at the forest margin

The FLORES simulation model aims to capture the interactions between rural communities living at the forest margin and the resources that they depend upon, in order to provide decision-makers with a tool that they can use to explore the consequences of alternative policy options. A key component of the model is simulating how decision-making agents within the system (individuals, households and the whole village) go about making their decisions. The model presented here is based on an anthropological description of the rules and relationships that people use, rather than on the assumption that people behave in an economically optimal fashion. The approach addresses both short-term decision-making (primarily the allocation of labour to various activities on a weekly basis), and long-term strategic land-use planning, taking into account the variety of tenure and inheritance patterns that operate in real communities. The decision-making sub-model has been implemented in the Rantau Pandan (Sumatra) version of FLORES, using the Simile modelling environment.

Reasoning with direction and rate of change in vegetation state transition modelling

Abstract Better integration in land planning and management can be supported through the use of suitable model-based tools. Vegetation state transition models have been noted as being useful in this context, providing a simple, useful means of capturing available ecological knowledge. We describe a simple ‘proof of concept’ rule-based system developed to contribute methodologically to management-oriented modelling of vegetated landscapes. The system is based upon a clear separation of direction from rate of change and the use of a general temporal reasoning system, a feature that facilitates modelling of situations where environmental change occurs causing an increase or decrease in rate without affecting direction of vegetation change. To ease model development and use the system represents vegetation dynamics in a way that has a close correspondence to the structure of understanding communicated by vegetation ecologists. A test model is described and run under different conditions to demonstrate the system. The results show that although the rule-based system and in particular the temporal reasoning system used operate successfully, there are a number of deficiencies in the modelling system as currently implemented. Future development possibilities are detailed along with a broader discussion regarding the needs of management-oriented modelling and the utility of state transition approaches.

Model transformation rules and model disaggregation

Ecosystem models vary considerably in their degree of disaggregation: the extent to which they ‘lump’ or ‘split’ model components. However, the criteria for the degree of disaggregation in a model are frequently not clear. The development of such criteria would be enhanced if the degree of disaggregation in a model could be easily changed, permitting a ready comparison of the alternative versions. In this paper, a framework is proposed for representing model transformation rules. Each rule indicates how a particular model component or set of components can be replaced by more- or less-disaggregated components. These rules have the potential to automate the process of generating alternative versions of a model differing in degree of disaggregation, and provide a framework within which modellers can express their expert opinion on the legality, costs and benefits of particular lumping or splitting decisions. The approach is wholly dependent on the symbolic representation of model structure, since each transformation rule is in effect a symbolic re-write rule, expressing how the set of symbols defining one model can be replaced by another set defining an alternative model. It is proposed that the logic programming language Prolog is suitable both for representing model structure in symbolic form, and for representing the model transformation rules.