Prepare the ESE Models for CouplingHere we evaluate the ESE components and utilizes them to answer certain functional questions particular to each Component before coupling them into the Simulation Model (SM). At this stage, the parallel activities, e.g. other types of models, modifications based on other analyses, can be merged into the ESE Component Models.
Activity ExplanationsThis activity should be conducted concurrently with the following activity, "Conduct ESE Interpretive Analyses". The first three subtasks are primarily reviewing what has been done in the FS. The most important subtask is that of conducting analysis on each of the ESE models in order to ensure that their structure and results can be run independently of each other. Linking the ESE components may be a problem, in particular if ecologists, sociologists and economists developed their models independently and did not intercommunicate well enough. Since the different ESE components often have different temporal and special resolution a special focus should be on converting this before linking the models. Several model blocks such as integrators and mean/variance blocks can be used to facilitate the conversion. The availability of a consistent library of EXTEND model blocks should facilitate the linking task. As noted above, you may take the approach of implementing a single integrated model rather than separate models for ecology, economy and sociology. It is also possible that the sociological element may not be included as a model per se.
Integrate any links to other models or analyses
This task is completely dependent on the extent to which other models have been used to complement or supplement the intended Simulation Analyses. Most fo the examples in this document use the Extend modelling package for the Simulation Model however other packages are avaiable and may be used. There are three points of entry (discussed elsewhere): as auxiliary models provide input at the Formulation Step; as a separate component of one of the ESE Components that becomes integrated into the Simulation Model in this activity; or as a completely independent assessment that generates supplementary information for the Output Step.
Systems involving ecological, economic and sociologic processes may be very complex, not only in terms of the state variables and the processes involved, but also regarding spatial-temporal scales of application. In many cases, it is necessary to link two or more different models. This linking may be “hard”, in the sense of involving step by step of interchange of variables or parameters between the submodels, or “soft”, when the output file of one submodel is used as the input for another submodel. In this case, it is necessary to assure that shared variables have consistent dimensions and scaling. Many modelling programs provide software for importing and exporting data with other applications. Due to frequent changes in the software versions of the different submodels, soft linking is in principle a better solution for long-term exploitation of the linked model.
An important case of external connection may be the linking between system dynamics and spatial applications. Model packages such as Extend are especially well suited for allowing the development of system dynamics applications without requiring students and researchers to master basic programming languages (such as Visual Basic, Fortran, C++ or Java). In principle, EXTEND (or similar packages) could also be used to incorporate spatial applications, but coupling between system dynamics and spatial applications may be facilitated by software development initiatives aiming at linking systems dynamics with geographic information systems. For example components and methods can be stored in an EXTEND library and use PCRaster for the spatial analysis of these models.
An example of soft-linking between different modelsOne example of “soft-linking” approach is that used in the Millennium Ecosystem Assessment (MA, Carpenter et al., 2006). In this project, storylines were developed for different scenarios and a team of modellers was organized to quantify the scenarios. Five global models covering global change processes were selected, based on criteria such as global coverage, publications in peer-reviewed literature and relevance in describing the future of ecosystem services. Linkages among models were adjusted and test calculations were carried out using preliminary driving force assumptions. The results of these tests were used to clarify the procedures of linking the different models.
Consistency between the calculations of the different models was achieved by “soft-linking” them, in the sense that output files from one model were used as inputs to other models. For example, computations of food supply, demand and trade from the IMPACT model were aggregated for the various world regions and animal and crop types and used as input to the IMAGE land cover model. The changes in irrigated areas computed in IMPACT were entered in the WaterGAP model and used to compute regional irrigation water requirements.Run ESE Models for Interpretive analyses
The ESE Component models that are delivered to Appraisal should already have passed the validation tests and the hindcast calibration runs. However, the calibration runs used may not coincide with the intervals needed for the scenarios requiring some adaptation.
In these cases, you could convert data to the lowest common denominator and lose accuracy, or extrapolate upwards from existing data and increase error margins. In general, the first option would probably be better.
Thus, the main emphasis of the WT is to individually run these ESE models to obtain results that can be presented scientifically and to complement some of the Interpretive Analysis specific to each particular ESE Component.
Conduct ESE Interpretive AnalysesThe Interpretive Analysis Task has two major focuses: that concerning the Component Models specific to their ‘disciplinary’ objectives and that concerning the Simulation Model, which combines these ESE Components, specific to the response of the CZ system corresponding to the Policy Issues (scenarios). Analysis and assessment in each of the ESE Components must be completed at the beginning of the Appraisal Step, because their results will bear on the final simulations and interpretations. The scope and methods for these analyses were decided in the Design Step and prepared during the Formulation Step. Due to time restrictions, some aspects of them might need to be initiated in the Formulation Step.
Natural Component Analyses
These activities pertain to the Interpretive Analyses, which are specific to the disciplinary objectives of each of the ESE components. It is particularly important that each component can pass review within its disciplinary area of expertise. Each of the ESE Component models has an intrinsic value with respect to the particular aspect that it is simulating. These need to be analyzed through separate runs of the respective ESE model and described. The reader is cautioned that some aspects of the below explanations may vary depending on yours system.
These analyses and descriptions interpret the simulation results of the Natural Component model and its objectives.
Process Approximations . In general, the simulation model represents the functionality most relevant to simulating the Impact with respect to the chosen scenarios. The way it is represented makes a large difference in its ability to capture the functionality in order that it can serve as a proxy for the behavior of the system to our scenarios. Each of these representations must be justified scientifically; it is important to record all the underlying assumptions to the choices being made. In the Formulation Step, much of this is done in a reductionist mode, process-by-process and component-by-component. This text will enter into the final Scientific Article in the methods section.
Simulation Results. Here the purpose is to describe how the model responds to reasonable variations in its inputs, including any proposed changes in Inputs dictated by the chosen scenario. In addition, some analysis will be needed regarding how the NC model responds to pulses of energy that are beyond the range of the scenarios but within the range of risk factors specified in the Design Step.
Relative to the particular aspect that the NC simulation model addresses, the model may need or offer some scientific description. This will be particularly true if the model output is well correlated with observed data, e.g. in the hindcast runs, and thereby provide useful information on the dynamics of the system. The most likely situation will involve an agreement due to calibration adjustment, such as the use of different representations of processes, inclusion of new (or not well recognized) methods, or simply results that have not yet been reported.
The output of a model must not be a unique value but should be expressed by a triplet (Y, D a Y, a ), which represents the mean value Y of the result given by the mathematical modelling, the imprecision range D a Y of the mean value and the degree of confidence of each value belonging to this imprecision range (example of the non-classical methodology based on the fuzzy set and fuzzy logic theory).
As explained in the Formulation Step, following Morecroft (2007), evaluation of a model implies several categories of tests. Tests of behavior assessed, by visual or statistical means, the fit between the trajectories of simulated and actual data. Tests of structure include questions on boundary adequacy, dimensional consistency, parameter verifications, and robustness of behavior. Tests of learning refer to the comparison between the model results and the mental models and expectations of the public. Recommended approaches to evaluate the impact on model output of using imprecise input information include sensitivity analysis and uncertainty analysis. The latter approach estimates the uncertainty of the solution from the uncertainty of model input parameters (Fresissinet et al., 1999).
One part of the tests of structure is the study of model stability (not only of natural but also of the socio-economic components). This property has been defined in many ways. For example, a system may be considered unstable if an infinitesimal change in system parameters can cause qualitative changes in system behaviour. Food web models including several categories of organisms or functional groups are highly non-linear and can display dynamic behaviours ranging from asymptotically stable equilibrium points to limit cycles and to chaotic oscillations when parameters or forcing variables are changed. Understanding the intrinsic dynamics of these models is necessary for building realistic ecosystem models (Lima et al., 2002). A widely used method for studying stability in a differential equation model is based on the construction of a Lyapunov function (Rosen, 1970). Unstable models were often considered fragile and inadequate representations of real ecosystems (Goh, 1977). However, recent work has pointed out the potential significance of oscillations and chaos in natural situations (Huisman and Weissing, 1999; Benincà et al., 2008).Economic Component Analyses
These analyses and descriptions interpret the simulation results of the Economic Component model and its objectives.
Discussion of Results . Here we would discuss the results of the Economic Component analysis independent of the Simulation Analysis, which will come later. This is to provide a credible basis for our approach to the specifics of the Economic Component, before assessing the interactions with the other Components and the System scenarios.
Simulation. This should describe the results of the Economic Simulation.
- Explain the objectives and what is being quantified in the model in the context of the policy option. Include the rational for the quantification of the Economic processes and the approximations used.
- Explain the results of the quantification in the context of the objectives and what they demonstrate. For example, did they reveal any unexpected results, how do they compare with data, are they useful to the management of the economic enterprise involved?
- Document all underlying assumptions to facilitate the output step.
Supportive. Because this exercise is demonstrative, it is important to discuss its potential value with respect to a more complete exercise; i.e. Illustrate how the value of the exercise might be improved, or how it might be expanded to associated economic policy options.Social Component Analyses
These analyses and descriptions interpret the simulation results of the Social Component model and its objectives.
Scope and Limitations . For the first SAF applications, we are necessarily limiting the scope of the Social Assessments (cf. scaling in the Introduction under Implementation). The example of economic assessment chosen for simulation analysis must be limited to the Virtual System and treat one or more of its scenarios. This is partially pre-determined in the expression of the scenarios.
Assessment in Coastal Zone s. We recommend that you choose a social component, which is relevant to the policy issue from the list below
If this list seems to long, perhaps some kind of decision tree can be made to assist in the choice of component. In particular, we would emphasise the social component of communities relevant to the policy issue.