6.	IMPLEMENTING THE GOOS END-TO-END APPROACH

6.1	THE C-GOOS DESIGN PROCESS (Thompson)

The ultimate goal of C-GOOS is to provide the basis, in observations and models, for assessing the effects of human activities and for predicting change in coastal waters. At its first meeting, the C-GOOS panel developed a procedure for the design of end-to-end observing systems that link bottom-up (measurement programmes) and top down (user needs) perspectives. Critical links between these "end members" include precise definition of the attributes to be predicted or described, determination of acceptable time lags between observation and the delivery of products, identification of models that are to be used to link measurements to products, and the definition of model inputs and outputs. The process begins with the identification of operational categories (preserving and restoring healthy ecosystems, sustaining living marine resources, mitigating natural hazards and safe and efficient marine operations) and related environmental issues (Table I). Subsequent steps are as follows:

1. Final Prediction:
   Define the final form(s) of the prediction. It is recognized, for example, the coastal managers do not need predictions about the possible occurrence of a red tide in the form of a complex model output. A straight forward alert may suffice. On the other hand, a coastal engineer designing flood defenses may need a precise confidence interval for the probability that a critical level will be exceeded. The term prediction is not used simply in the sense of forecasting the future, but also in the sense of estimating by interpolation a quantity which is not observed directly; it may include, for example, inferring the present biodiversity of an ecosystem from measurements made at a small number of observing stations. It also includes the spatial extrapolation of return times of extreme sea-levels from a tide gauge with a long record to a coastal site with little or no sea-level data. Examples of predictions include: frequencies of flooding and extreme waves and currents; optimal shipping routes; extent of potential loss of habitat; probable effects of oxygen depletion in bottom water.
2. Lead Time:
   This is the acceptable time lag between measurement and prediction. For cases involving straightforward spatial interpretation this may be zero (e.g., the probability of a specified sea-level being exceeded at a site without a tide gauge). On the other hand, useful storm surge forecasts are required hours to days ahead while land use management decisions might be based on GIS products that require days-months to produce.
3. Identification of the Types of Models to be used:
   This may range from conceptual models, GIS, and simple regression models (based on empirical relationships) to sophisticated, coupled ocean-atmosphere and hydrodynamic-ecosystem models based on theory and empirically derived parameters.
4. Model Variables (Outputs):
   This describes the quantity predicted directly by the model. It might be, for example, time-varying fields of currents or productivity, linear trends of sea level over recent decades, or ice distribution. In many instances this will differ from the final form of the prediction provided to users which will commonly be a highly reduced version of the raw model output.
5. Model Inputs:
   The are the observations needed to make the predictions. Many are common to several issues, e.g., winds, air pressure, sea-levels and currents, sea surface temperature and salinity, and concentrations of nutrients, chlorophyll-a, oxygen, and suspended particulate matter.
6. Feasibility:
   The feasibility (cost and the availability of acceptable technologies and techniques) of the approach or method is ranked high, medium or low.
7. Cost-Benefit Analysis:
   This is the ratio between the cost of the measurement programme and the benefit of making the prediction. This is arguable the most difficult step, but a relative ranking of high, medium or low may suffice in the first instance. When completed, this final column should be used to order the measurements (model inputs) in terms of the cost of measurement vs the impact of the input data on the model output. The step is explored in greater depth in section 6.2.

Table I. Globally ubiquitous indicators of environmental change in, and human uses of, coastal waters. This is a modification of Table I in the report of C-GOOS-I. It has been modified to distinguish between causes and consequences. Indicators of change are the consequences of either natural or anthropogenic sources of variability, or both.

OPERATIONAL CATEGORY	INDICATORS OF CHANGE
Preserve & Restore Healthy Ecosystems/Manage Resources for Sustainable Use	declining living marine resources
	oxygen depletion (hypoxia, anoxia)
	increased in phytoplankton biomass
	harmful algal blooms
	fish kills
	habitat loss (e.g., wetlands, sea grasses, coral reefs)
	diseases in marine organisms
	growth of nonindigenous species
	loss of biodiversity
	temperature & salinity distributions
Mitigate Coastal Hazards	loss of property and human life
	lack of economic stability
	higher insurance rates
	sea-level rise
	coastal erosion
Safe & Efficient Marine Operations	loss of property and human life
	spills of hazardous materials (oil, chemicals, radio-isotopes)
	introduction of nonindigenous species (ballast water)

A working group was charged at C-GOOS-I to perform this analysis for all of the issues (indicators of change in the new table) listed in Table I. This proved to be a difficult task, in part because measurements of input variables differ in the extent to which they are operational (measured routinely in a timely fashion with known precision and accuracy), i.e., observing systems for climate that require inputs of physical variables (e.g., wind, temperature, salinity, currents, sea surface height) are operational or close to being operational. This cannot be said for observing systems for ecosystem health or the management of living resources in that many biological and chemical variables cannot be measured at this time in an operational sense. Thus, the working group found that separate analyses were needed for those issues that relied solely on physical quantities and for those that relied on multidisciplinary inputs.

In addition to differences in operational status, this reflects the reality that physical variables such as sea-level, currents and water temperature are not affected strongly by biological variables while biological and chemical variables interact strongly and are affected by the physical environment.

Clearly, more work will be required for this approach to become fully functional as a guide to the design and implementation of C-GOOS. Nevertheless, the Panel feels that the approach is a powerful tool for designing end-to-end observation systems.

The Panel noted that some biological and chemical measurements are operational, for example ocean colour data are routinely produced and used as guides by the fishing industry, and fish statistics are collected regularly. Nevertheless, the Panel agreed that the lack of knowledge of how perturbations are propagated through coastal ecosystems to cause changes such as those listed in Table I is limiting to the design of fully operational observing systems at this time. Recognizing the requirement for additional ecosystem level research, the Panel concluded that, in some cases, the immediate purpose of observing systems will be to document the spectra of variability that characterize coastal systems, i.e., to quantify the temporal and spatial dimensions of patterns of variability that are relevant to indicators of change (Table I). In this regard, the Panel further recognizes the importance of research programmes (e.g., LOICZ and GEOHAB) as the means by which ecosystem models will be developed that will satisfy the requirement for models that link the measurement of properties to outputs that have applied uses.

In terms of measurement programmes, the panel will place high priority on identifying those properties and process that must be measured to document and predict indicators of environmental change (Table I). This includes specifying the time and space scales on which measurements should be made, the precision and accuracy required, and assessment of the need for and impact of new technologies (e.g., HF radar for surface currents; remote sensing for salinity).

(Top of page)

6.2	COST-BENEFIT ANALYSIS OF MEASUREMENTS (Hall)

An important step in the design process described above is the cost-benefit analysis of measurements. The approach taken by the HOTO Panel (IOC, 1996) was adopted for this purpose, prioritorize properties to be measured in terms of their impact (e.g., importance to decision making or as an input variable to a numerical model) and the feasibility (cost, difficulty) of making routine measurements. In an x-y plot of impact versus feasibility, properties fall into one of three categories:

1. the property is easily measured (routine) and has a high impact;
2. the property has a low impact and is difficult to measure (not routine or the technology does not exist); and
3. the property has a high impact and is difficult to measure.

Properties that fall into category (iii) should be the subject of active research and development efforts to move them to category (i).

Three impact versus difficulty diagrams were presented for discussion, one each for physical, chemical and biological properties. Three sources of information were used to assign the level of impact: (i) Tables 2, 3, and 4 from the Miami Coastal GOOS Workshop report, which presented a set of variables considered in terms of the use to particular user sectors (impact defined as how many times a particular variable was mentioned as desirable); (ii) The results of a preliminary issue-specific design analysis conducted during intersession and presented for discussion under 6.1; and (iii) Information from the paper by Costanza et al. (Nature, 387, 253-260,1997), which provides more information on the potential impact of particular variables.

Feasibility of measurement was based on the research experience of the working group and was assessed as follows: (i) low - easily measured with basic knowledge and equipment; (ii) medium - moderate levels of expertise and equipment required; and high - expert knowledge and high-tech equipment required.

Considerable debate followed concerning the impact-feasibility ranking of each property that involved the following issues: directly measured vs a derived estimate (e.g., ocean color vs chlorophyll concentration) and the current lack of understanding of the structure and function of ecosystems often makes it difficult to assign a meaningful level of impact (need more rigorous analyses such as sensitivity analysis of ecosystem models). The suggestion was made that, given the current state of knowledge, as many properties and processes as possible should be measured. This was countered by the fiscal reality of "measuring everything" and the argument that the minimum number of core variables (that are relevant to many indicators of change and satisfy multiple user needs) must be identified that will define the basic skeleton of an integrated observing system and provide the means of comparing different systems, interpolating among systems, and extrapolating to systems that are beyond the range of observation. Additional variables can then be added on a case by case basis depending on the issues being addressed.

6.3	INTEGRATING REMOTE AND IN SITU MEASUREMENTS (Kiefer)

A critical feature of coastal ocean observing systems is their ability to quantify the time-space dimensions of pattern. This is especially challenging given the broad spectrum of variability that characterizes coastal waters. Geographic information systems (GIS) are data analysis tools that reference diverse kinds of data to their position in space. Spatial referencing systems that can move with the water will be important tools for transforming observations in time and space into useful visualizations of time dependent changes in property fields. The Environmental Assessment SYstem (EASY) is one such system. This software integrates data collected on different time (e.g., in situ measurements) and space (e.g., remote sensing) scales to show how property distributions (e.g., chlorophyll, the plume of an effluent, an oil spill) change through time. It also has the capability of running numerical models within the context of the spatial data sets, a feature that has proven useful for both fisheries management and pollution risk analysis.

To illustrate the power of integrating synoptic spatial observations with high resolution time series measurements, a GIS software package was demonstrated. Test applications were shown for the southern California Bight (development and movement of effluent plumes following storms) and the northern Adriatic (Po river outflow and coastal eutrophication). The system works in a PC Windows environment and can run on a PC or on the Internet. It is modular, with discrete files for each project; it can provide 4- dimensional representation so that the evolution of pollutant plumes and other phenomena can be visualised over a specified time period. In California, the users are the general public, who are concerned about the levels of pollution on beaches. The data comprise: storm drain data; offshore current meter data; Synthetic Aperture Radar (SAR) satellite images of surface roughness (a good indicator of plumes, even when clouds are present); and plume models.