System Formulation

Examples:

These examples are brief. If you need further information to identify an Input or a Response Information due to a Human Activity or to an Ecological Forcing, don’t hesitate to use further comments in the table.

Examples:

External level (Level 1): When there is no interaction between the Input and the Virtual System, i.e. the input is not changed by changes in the “virtual system or Impact”, it is independent. Certain policy scenarios may of course explore potential changes in this input, if it is necessary for the Policy Issue. For instance:

• atmospheric forcing (weather) is normally considered as an external Input because of the difference in scale with a coastal zone. That means that the meteorology is not changed by the Coastal Zone estuary or land.

• atmospheric deposition is also considered an external Input because it has a larger-scaled origin and thus can be described as being independent of the internal processes of the Virtual System [note:, if the environment that you want to simulate is exposed to a significant production of substances (of NOx, NH4, or polluted aerosols, etc), which enter as key variables or which enter via one of the policy scenarios, then you should consider the local production of these as an internal Input].

• if a policy scenario requests the effects of an improbable event (e.g. 100-yr storm), then a simulated storm can be added to the normal weather input in order to construct this scenario. This input remains as an external one because the storm is not created by any internal dynamic of the system.

External or Internal (Level 2): There is no interaction between the Input and the Virtual System, but there is at least one scenario concerning the manner of influence of the Input on the intensity of the Impact. For instance:

• if a scenario concerns the runoff of a river, then the Virtual System should include the land-use practice, its drainage to the river, and the substance loading. The external Inputs are now those forcing the land runoff to the river. But you must keep in mind that in this case the model would have land-use and river-components and the river outflow would be an internal Input from the river component to the estuary component. On the other hand, if there were no scenarios concerning land-use, the river effluent could be simply specified as an external Input.

External switches to Internal (Level 3): There is a dynamic threshold-interaction, between the Virtual System and an external Input, which converts it to an Internal Input. For instance:

• if the number of tourists is an Input, and the resulting increases in sewage outflow contribute to greater hypoxic conditions (Impact), this would be an external Input. However, if the Social Component is analyzing the negative response of tourists to hypoxic conditions in the beach, then the number of tourists must be considered as an internal Input.

• if the amount of investment capital available is fixed and insufficient to institute more sustainable practice, this amount would be an external Input, i.e. there is no interaction between the system and amount of money available. However, if a deferred payment is allowed, in which the long-term benefits are allowed to payback the initial capital-investment loan, then the investment capital would become an internal Input, i.e. a sum that is calculated during the simulation.

Examples:

• Meteorological data may require daily weather forecasts as inputs for quasi-real time simulations; whereas for long-term forecasts, a statistical composite might be used, e.g. seasonal mean data with typical daily variability superimposed.

• Tourist data would require scenario simulations based on available statistics e.g. growth in number, origin, or type of touristic activity

Examples:

The information you need to assess the relevance of the Inputs to the Virtual System, requires an evaluation of the forcing variables and their connection to the Virtual System. The examples below are illustrative and not definitive. The Inputs that concern mass (or energy) are classified according to their importance by using their magnitude respectively to other Inputs. The Inputs that concern Information are classified according to their importance using the level of the change that is stimulated relative to no such information. This is probably something you should keep in mind as it can be best evaluated later, during the multiple sensitivity analysis tests.

Magnitude:

For the category of Level 1 mass Inputs:

• discharge of water or substance by the Raging River: For a first-order evaluation of the magnitude, you would require the mean values of fluxes of those substances designated as key variables for your Virtual System (e.g. H2O, N, P, etc). As a second evaluation, you would need the sampling interval and the standard deviation of these variables. If they differ in deltaT, you will have to do some data conversion before calculating fluxes. If you do this, you could then make a comparison between the Raging River and another River of the Virtual System making sure, of course, that the different fluxes are dimensionally comparable.

• Perch Catch from a river or lake: It is possible that the catch of these fishes might constitute the Impact of the simulation and one of the outputs of the Ecological Component model. Nevertheless, the catch data are needed for calibration of the model and thereby are identified as an input for the simulation. To evaluate the magnitude of the total catch, you would probably need the commercial, recreational, and illegal catches. The key variable with respect to the Virtual System might be the dry weight of carbon represented by the catch. The sampling time might be longer than the time step used for the simulation (monthly, or yearly) such that a data-analysis method should be used to convert to a shorter time interval, in order to resolve life cycle effects, etc. For the Economic Component, one would need the market value, which also may have some seasonal variability, and would also need other input information as the disposition, the size of the fish, etc.

For the category of Level 2 Mass Inputs:

• Raging River Watershed: If a scenario requires a comparison of two types of land-use in the watershed, you must expand your Virtual System in order to include the water drainage from at least the portion of the watershed indicated by the scenario. Now your external Inputs are the rain and the agricultural practices. The Raging River then becomes an internal Input of which a certain, calculated part derives from the selected land-use.

For the category of Level 3 Mass Inputs:

• Foreign Tourists: For your application, the mass of tourists entering or leaving the Virtual System it will probably not of concern; however their water consumption and waste products often are. In addition, their behavioural interaction with the Virtual System is often also of concern. If you need to consider their mass Input to the Virtual System, you need to convert the number of tourists into organic matter input using the available sewage system. If the tourists do not enter into the Social or the Economic Components and if they contribute to a significant portion of the population or to its variability, then you should treat the contribution as an external Input. If they do enter dynamically into your simulation analysis, then they become an internal Input to the Virtual System and the Input representing their contribution must be more carefully calculated. In this case, their relevance to the loading will be a function of their contribution, i.e. whether it is significant to the background loading or not.

Information:

For the Interaction of Level 1 information Inputs:

• Nutrient Variability: If the variability of an Input is significant, then the information about the variability is necessary as part of the Input. This is because using mean values over long periods obscure the information about variability of the loading. Pulses of nitrogen loading due to sudden, intense rainfall events can produce a significantly different response in an estuarine system, as compared to the same estuary exposed to the same mean value but with little or no variability. Often the response of the system to the variation of the input is more important to the simulation of the Impact, than the response to the mean value of the Input.

• Nutrient Ratio: Likewise, the ratios between nutrients constitute an information Input, which can be an important control, e.g. on phytoplankton speciation. Additionally, these ratios can be used to identify and label different nutrient sources (as can N-isotope information).

For the Interaction of Level 2 information Inputs:

• Market Price: The price of a harvested commodity is an information Input. If the price of a harvested commodity is determined at a larger scale than the Virtual System, it is an external Input. However, if a scenario wants to investigate the effect of the illegal harvest of this commodity, which is determined at scale of the Virtual System, then the illegal price becomes an internal Input and must be simulated as a function of internal parameters.

• Governance Restriction: Policy directives made outside of the Virtual System jurisdiction (e.g. the local policy makers) would constitute an external information input parameter. If policy directives can be changed because of the internal function of the system i.e. because of decisions made inside the Virtual System, it would be considered an internal Input function.

For the Interaction of Level 3 information Inputs:

• Public Awareness: If the public acceptance of some aspect of the virtual system is held constant (e.g. FAO exposure level), it would enter as an external control parameter in the simulation analysis. However, in some cases a scenario may require the public reaction to increased or decreased toxic levels, then the toxic level and the public acceptance would vary in the simulation. The acceptable exposure level would then become an internal Input function.

Examples:

Omission: An omitted parameter often occurs because it was not included in the original data monitoring plan for the human activities. In this case, you can choose a data substitution (see below) or you can leave it out, depending on its relevance.

Dimensions: A variable may have inconvenient dimensions or units. This can happen in several aspects: deltaT, spatial, geochemical, biological, partial representation, etc. More specifically:

• If the time-scale of the variable is compatible with the simulation, usually the units can be converted rather simply.

• If the time-scale of the variable is different than that desired, you can apply an interpolation scheme thus conserving the basic statistics of the series. The same applies when you need to fill minor gaps in the time series. Larger gaps should be specially noted, as they will possibly increase the calibration error.

Transformations: Non-conservative substances are changed between the location of the Input source (where data are available) and the boundary of the Virtual System. In this case, you can decide either to estimate the transformation empirically or to include the transformation process as a component in the simulation model. More specifically:

• The meteorological data may be scarce relative to a watershed. If your needs are more specific, you can use spatial combinations of meteorological stations or you can get spatially gridded data.

• The point source data is located upriver from the estuary. You can use estimations of the river uptake and denitrification or you can include them in the simulation model.

• Non-point nitrogen data for a river outflow is available only in annual averages. If an important scenario involves, for example, land-use or agricultural practices then the relevant land-use should be included in the Virtual System. If the loading is necessary only as part of the river input to the estuary, without specifically entering into a scenario, then the input can be simulated empirically at the needed time step and matching the known long term mean. The short-term variability can be correlated with the drainage forcing, i.e. as with the rainfall or irrigation patterns.

Examples :

Substituting for gaps: Observational data often have gaps. There are numerous options for filling in data gaps. Regardless of the method, some scientific judgment should be exercised here: the bigger the gap, the more judgment. The following are just suggestion and are not intended to be comprehensive. More specifically:

• in some simulation software you can find features that perform interpolation and data fitting (e.g. EXTENDSim).

• small gaps in time series: if the data can be copied into statistical software, you can make a simple fit and recalculation.

• larger gaps in time series: if the original series is long enough in order for you to determine the spectrum, then you can compute large gaps using Fourier, or similar, analysis.

• vertical space gaps: often the nutrient values are sampled at a vertical resolution different from the electronic sensor sampling (CTD). These should be vertically averaged relatively to the depth of the water column that you need to represent in the model.

• horizontal space gap: You will often need horizontal integrations in order to represent the area modelled. You can use objective analysis, or similar, interpolation schemes to make these integrations in an efficient way.

• if hydrodynamic models already exist for the area of interest, you can use them to acquire vertical and horizontal integrations.

• you can also use GIS data, when available, to generate horizontal integrations on land.

Adapting similar data : Processes are less unique than systems, and therefore, you can more easily justify the use of similar data deriving from alternative sources. When calibration data are totally unavailable in the process level, you should consider using data coming from a different area, but it would be more hazardous to do this in the functional component level and not acceptable at the ESE or system model level. In any case, when you use data deriving from a different area you must make sure that the functional range of validity is similar and that the controlling constants and parameters are not case dependent. More specifically:

• a model for the resuspension of sediments, might use the same data for calibration when used for two sea-areas, assuming that the roughness-length and drag coefficients are similar and the kinetic energy is of the same order of magnitude.

• tourist preferences might be considered similar within the same region and same tourist resources, as bathing beach, sailing conditions, recreational fishing, etc. in the Mediterranean, North Sea, Baltic, Atlantic areas.

Using a proxy variable : The raw Input information may not always match the Input needs for the simulation analysis, but most of the times this is what you need. You should take care so that you will do the conversion in a defendable mode and also include any additional information when needed. More specifically:

• commonly, the chlorophyll concentrations are used as a proxy for phytoplankton biomass. Sometimes even double proxies are used in order to convert satellite data at the chlorophyll wavelength and then to estimates of plankton biomass.

• the freshwater balance and external salinity data can be used to calculate the estuarine circulation and flushing.

Constructing empirical data : When the observational data you have are insufficient, particularly on the process level, empirical data for calibration can be generated from the experimental data already existing or from the literature. More specifically:

• often the results of lab experiments can contribute to points on the calibration curve. Calibrations against literature values can be used in the same way. These actions support the credibility of the process representation.

• usually, illegal or recreational fish catch are not available. You could use estimations from other, similar costal zones or put it into the model as a parameter that can take values in a pre-known, determined range.

• sometimes sunlight data are not available. You could use an empirical relationship depended on a solar constant and the latitude, adding the cloudiness determined from meteorological data.

• often, discharge estimates of toxic chemicals from industrial human activities are not known. You could use average values from the literature.

Simulating Input Data: As stated earlier, you have two main options regarding simulation. You can either simulate an input function or variable (external simulation), or you can change the boundaries and include it to the Virtual System (internal simulation). More specifically:

External Simulation

• If a scenario requires evaluating how touristic preferences would change concerning a given Impact, e.g. removal of a seafood species from the market, increased turbidity in bathing beaches, adding day-care facilities, etc. you can simulate the original level of preference and simulate the change based on a statistically reasonable assessment of preference change for the scenario. This is a good example of using ‘change’ as an indicator of policy direction.

• If a scenario requires the response to a large event, as a probable risk factor, which is possible but is not in the data, you can do that by adding a simulated event to the existing data stream.

• At the beginning of the Formulation Step, much of the input data may not yet be available. In this case, you can simulate an input function directly within your selected simulation software.

• Many times one finds that an input parameter or constant obtained from the literature is not functioning properly in one’s formulation. This is common with dynamic parameters, which are considered as constant for a particular application but actually vary dependent on the case-study. In such cases, you should consider changing the value of the parameternoting this in the model documentation.

Internal Simulation

• If as a result of the scenario discussions with the Reference Group of your area, it becomes desirable to include in the simulation model another portion of the coastal zone, the best solution would be to include this within the Virtual System, if feasible. Land-use considerations are a good example where the scenario would like to include other human activities that were not formally treated. However, in many cases the additional scenario requirement can be approximated and treated as an external Input to the system (i.e. one that changes the simulation output but the human activity Input is not changed by the simulation output).

• Something that can also occur is that the original boundaries selected were not sufficient to achieve the accuracy needed for the simulation, so that what was formerly an Input function becomes an additional component to the simulation.

Combining tricks: Often it is necessary to combine one or more of the above substitution tricks. This is fine if you can reasonably defend it. More specifically:

• the commercial fish catch may be erratically sampled and/or not represent the total fishing mortality in the Ecological Component model. In this case, the economic assessment could be restricted to only the market portion, or it could be extended to include the total catch. In either case, you will still need to estimate the non-market portions by defendable means.

• tourist data might only be available seasonally, from airport arrivals or hotel registrations. If you consider that these data are incomplete, then you might make some estimation to ‘fill in’ the data set and to make it representative and suitable for the model time-step, by using empirical correlations with published data from other similar locations or situations.

• a small watershed may be important for your simulation but there may not be any measured runoff data available. You can add a land drainage component in your simulation model or you can approximate water yield (specific to land-use/soil type) as a fraction of the local rainfall.

Example (from the implementation of the SAF at Mar Piccolo, Taranto, Italy ):

Impact: The reduction of the productivity and the quality of the mussel culture.

Policy Issue: Including Mussel Culture in a Management Plan for Sustainable Use of the Mar Piccolo Resources.

Thematic Scenarios with operational ‘ Baby Scenarios’:

A. Evaluating the environmental conditions controlling Mussels growth.
A1. To what extent would optimal, environmental conditions reduce the costs of mussel culture and increase socio-economic benefits?
A2. What kind of indicators can we use to estimate the mussel growth based on different types of food?
A3. What would be nutrients target ratio BE in order to optimize MP productivity?
A4. To what degree are contaminant substances or organisms inhibiting or endangering the mussel’s growth?

B. Evaluating the measures and costs needed to sustainable Mussels growth
B1. Are there other uses preventing better environmental conditions for mussel culture?
B2. What technological options or policy strategies are available to mitigate these damaging effects?
B3. What are the socio-economic consequences of these options or strategies?

C. Evaluating human health effects from exposure to hazardous levels of contaminants or microorganisms.
C1. What are the implications to human health due to mussel’s uptake of hazardous substances or microorganisms?
C2. What are the health costs projected from exposure to these contaminants?

Example:

Approximations: In order to capture the needed functionality with your models, you can devise mechanisms to represent space for the purpose of the time-simulation. The need must be weighed against the output required by the scenarios and against the need - in terms of resources available - to resolution required. More specifically:

Spatial Resolution
• The need to calculate or visualize changes relative to a geographic grid. You could do that using a GIS adaptation appropriate for the simulation software you use (for ExtendSim, a PCRaster adaption has been developed through the project SPICOSA).
• The need to integrate quantities or compute advective fluxes in aquatic systems with hydrodynamic numerical models. If it is available it could be useful for providing data distributions or calibration checks, as stated in former subtasks.

Parameter Space
• Vertical resolution is often necessary and you can assess it by adding components that represent a depth based on the value of a parameter e.g., ‘light depth’ or mixed layer depth, nephaloid thickness, etc., calculated in your simulation software and used to define a internal functional component.
• Horizontal resolution in the aqueous regime may be necessary in order to distinguish different spatial components again based on a parameter calculated by your model, e.g., effluent plume, tidally mixed volume, etc.

Virtual Space
In the case of land-use or benthic regimes (fixed in space) a virtual space can be defined having the mean characteristics of the selected land-use, e.g. soil type, vegetation, inclination, distance to stream, benthic vegetation, area fished, etc.