Scope 47- Long-term Ecological Research, An International Perspective

13.1 A HIERARCHICAL DESCRIPTION OF SYSTEMS

13.1.1 THREE DIMENSIONS OR ASPECTS OF COMPLEX SYSTEMS

Problems in ecosystems research and management can have three dimensions:

1. A dimension of details (for example, the spatial distribution of characteristics such as soil types);
2. A dynamic dimension, of development and change; and
3. A strategic dimension, concerning, for example, the role of heterogeneity or disruptions and surprise for sustaining life.

For many problems all of these dimensions have to be considered together. These dimensions are used for a three-layer view of systems, and different methods are appropriate on each layer. Results from one layer provide inputs for the other layers, and a combined, 'multifaceted' method is thus defined to deal with 'multifaceted' problems in complex systems. The word 'multifaceted' is due to Zeigler ( 1979). 'Ecosystem' is used to include human ecosystems.

13.1.2 THE BOTTOM LAYER IN THE THREE-LAYER VIEW  OF A SYSTEM

The bottom layer in the three-layer view of a system described above is the layer of simple data and other details within one system. These data are often precise, usually numerous, and simple to measure and evaluate. One important class of such data are spatial distributions of features. Examples of such features are the soil type or orographic features, such as aspect, slope, and elevation, or other factors such as microclimate, location of rivers, boundaries of properties, and administrative boundaries. Other details may be those due to diversity of species, differences within populations, and other forms of heterogeneity. Most of these latter details do not pose scientific or management problems other than the fact that they are numerous.

The bottom layer structures are simple, obvious, and often linear. Problems on this layer are well defined, and lend themselves to fast reactions to problems and the solution of problems, frequent needs in industrial production or in metabolism. Dynamics of the system on this layer are also simple.

13.1.3 INTERMEDIATE LAYER OR LAYER OF COMPLEX  DYNAMICS

The intermediate layer in the system is the layer of such important aspects of ecosystems as their structures and their dynamics. Some examples of dynamics are seasonal changes, successions, actions by man, migratory patterns, or adaptations. Ecosystems both generate dynamics and react to changes of their environment.

Problems on this intermediate layer deal with aggregated characteristics, such as the complex structures which generate dynamics in, for example, predator-prey relationships, the structure of a company and its customers in determining the liquidity of the company, or the global systems structures which drive weather and climate.

Uncertainty is often high on this layer because of variable impacts from the system's environment. Climate variations and human behavior are both examples of variable impacts, and are both important causes of uncertainty. The preservation of liquidity in the company is an example of many factors as a cause of increased uncertainty. New orders may create the need for prefinancing, and thus temporarily decrease liquidity; customers may disappear from the market, and new customers may emerge.

Feedback loops are appropriate as reactions to many such changes. As long as a feedback reaction is appropriate to deal with a change, the structure of the system can remain unchanged.

13.1.4 HIGHEST OR STRATEGIC LAYER

Structures can and do change. Since models on the intermediate layer depict structures, they usually become invalid if the real structure changes. Change can be caused by rare events, by destruction, by learning, or by chance; an ecosystem seems to use as many different strategies to bring about change as it uses to cope with it. Some of the coping strategies are diversity, variability, system architecture, resilience, combinations and decoupling of systems, and replacement of systems (Ashby, 1959, 1960; Simon, 1962; Grossmann, 1978; Holling, 1978).

The results of the likelihood and form of changes from the strategic layer are important for evaluations on the intermediate layer. If the structure is changed, the resulting dynamics will usually change also. However, dynamics can change without change of the structure, as in the cáse of change in modes of behavior (Thom, 1975; Haken, 1978), and conditions which lead to change of the behavioral mode are, therefore, evaluated on the highest layer. Not all structural changes result in changes of the system dynamics, as a changed, and therefore different, structure can continue to fulfil the same purpose. Factors important for the evaluation of the behavioral mode and for the assessment of structural changes are called 'strategic factors' or 'strategic criteria'. These criteria are usually qualitative, imprecise, often subjective (as, for example, in portfolio-analysis), and highly aggregated. Changes of structure and of mode, and the reasons for change, characterize a third aspect of ecosystems. On this layer, the uncertainty is very high.

13.1.5 HISTORY OF THE THREE-LAYERED DESCRIPTION  OF SYSTEMS APPLIED HERE

The systems view outlined above also has hierarchical features. Hierarchical or stratified perceptions of complex systems have a long tradition. Simon(1962) provided examples demonstrating why a hierarchical architecture increases the probability of reliable functioning; Bertallanffy (1969) gave a hierarchical description of the cosmos; Mesarovic et al. (1971) gave a detailed mathematical description of specific hierarchical systems; Bossel and Strobel (1977) combined a dynamic model with strategic criteria; Vester and von Hesler (1980) suggested a six-level scheme for regional planning; hierarchical descriptions were applied to ecosystems by Allen and Starr ( 1982); and implications of cybernetics for socio-ecological systems have been investigated by Rappaport ( 1979).

The view of systems presented here is partially based on this material, but foremost, it is based on experience derived from problems in manifold projects (MAB 1 project, Adisoemarto and Brunig, 1978; Grossmann, 1978; MAB 11 project, Vester and von Hesler, 1980; ARP-Project IIASA, Grossmann, 1983; MAB 1 project, Brunig et al., 1986; pollution abatement projects, Grossmann and Grossmann, 1985; Grossmann and Orthofer, 1987; and projects on forest die-back, Grossmann, 1988a). The view was developed to applicability by the author for the MAB 6 Project Berchtesgaden (Grossmann et al., 1983), and was refined in several case studies within this project (Grossmann et al., 1984; Grossmann and Schaller, 1986; Haber et al., 1984; Haber, 1989a, b) and other projects.

13.1.6 A RELATIVISTIC VIEW OF SYSTEMS

The hierarchical view of systems is relativistic. A specific ecosystem usually is part of another hierarchy, the hierarchy of its lower and higher ecosystems, and, eventually, of its environment, even if this is no longer an ecosystem. The highest layer of one system may represent a simple datum on the lowest layer of a higher system in another hierarchy.

13.2 APPROPRIATE METHODS TO DEAL WITH THE DIFFERENT PROBLEMS ON THE DIFFERENT  LAYERS

In the three-layer view of systems given above, there are differences between the layers with respect to

Methods chosen to deal with problems must be appropriate for these characteristics.

The decision of which methods are appropriate can be based on the characteristics of the data.

13.2.1 METHODS FOR THE BOTTOM LAYER

Geographical Information Systems (GIS) are an important tool for storing, evaluating, depicting, updating, and processing spatial data. A GIS should store the topology of spatial relationships, i.e. which polygon is adjacent to which others, adjacent to which line features, and so on. This task is difficult if maps have to be combined (so-called map overlay, for example, of maps of the soil, microclimate, slope, and aspect, to determine the ecological conditions of the site) and is offered by only a few GISs. In our projects, we use ARC/INFO, which combines software for spatial evaluations with a relational data bank to store the topology. Data banks are, in general, appropriate on this layer, as are statistics, spread sheets, and real-time control for management.

13.2.2 METHODS FOR THE INTERMEDIATE LAYER

On this layer, the complex structures and their dynamics must be depicted and evaluated. Multiloop feedback models are one appropriate method for this task, and often exhibit dynamics comparable to those of reality if they mirror real structures. However, all methods that either model real complex structures or lead to complex or aggregated dynamics are appropriate. Object-oriented programming can be used to model dynamic processes. Some AI methods can be used to model and evaluate structures of decision making.

13.2.3 METHODS FOR THE HIGHEST LAYER

General or 'strategic' criteria are appropriate to evaluate reasons and forms of structural changes, or to develop policies on how to make systems more viable or how to replace them with systems that are more suitable to prevailing or emerging conditions. Several such criteria were mentioned above. Markowitz's Portfolio Analysis (Markowitz, 1959) and its extensions help in the economics to determine a tradeoff between risk and profit by appropriate diversification of shares, resources, customers, etc. The subjectivity in decisions on this layer was mentioned above. In natural ecosystems no conscious decision making is done; however, similar criteria seem to be applicable to natural ecosystems. Scenarios may be developed from the evaluation of strategic criteria; for example, the 'let it burn' policy in some United States forests is based upon strategic evaluations of the consequences of too-rigid fire control.

Long-term ecological research and long-term monitoring are adequate approaches to problems on the strategic layer. Table 13.1 is a summary of systems characteristics and appropriate methods.

Table 13.1 Summary of characteristics of the three layers and of appropriate methods

13.2.4 COMBINATION OF METHODS IN THE THREE-LAYER APPROACH

Implementation of the results of a strategic evaluation is usually difficult; this is because the analysis is 'soft' and the implementation often has to be done on the layer of details. This became very obvious in the Man and the Biosphere (MAB 11) project 'Lower Main' (MAB program area 11: 'Human Settlements'), where the planners could not translate the results of strategic evaluations into maps. The planners agreed with the outcomes of the strategic analysis (for example, to increase the diversity of the means of transport) but could not derive from this recommendation where to build what. Regional plans often have to be precise to the centimeter, as there is a considerable gap between the outcome of strategic analyses and the details of implementation. The method outlined here helps to narrow or even bridge this gap.

13.2.4.1 Combination between the strategic layer and the layer of  complex dynamics

The evaluation of strategic criteria usually allows the derivation of different scenarios on change, including structural change, of the system on all layers. These scenarios are then evaluated with models on the intermediate layers. The structure of the models may have to be adapted to accommodate the changes, or their mode of operation may have to be changed. Models may even have to be replaced. The results of strategic analyses thus translate into corresponding model outcomes. The models are in turn used to check the strategic analyses, because on the intermediate layer far more and more precise knowledge is available. One well-known feature of models is counterintuitive behavior (Forrester 1969), which is highly effective in testing and refining scenarios. This top-down and bottom-up linkage between the strategic and the intermediate layer helps to build reasonable scenarios that are more suitable to implementation, and to translate them into model structures and model dynamics. This is one step to implementation of strategies, but not a sufficient one.

13.2.4.2 Combination of the layer of complex dynamics and the bottom layer

In a combination of the intermediate and bottom layers, the resulting dynamics are translated into maps using the GIS. The precise and numerous data on the bottom layer help in testing the model dynamics and the underlying strategic analyses, and in improving analyses and models. The whole procedure can be interactive. The resulting maps can be used to guide actions, allowing planners to depict where to do what and when. The combination was applied to projects on forest die-back (Grossmann et al., 1983,1984; Grossmann and Schaller, 1986; Grossmann, 1988a) and to the development of new agricultural strategies (Haber et al., 1989a). In forest die-back the procedure was

1. To select one hypothesis on forest die-back and the spatial variables that are important to evaluate this specific hypothesis. Important variables for all hypotheses are the age and species of trees and quality of the soils. Important additional variables for the soil acidification hypothesis are at least the pH-value and buffering capacity of the soils and isolines of H-deposition. Important additional variables for an ozone hypothesis are isolines of durations of the different ozone doses. Usually a map is composed of different polygons such that within one polygon the value of the specific variable is roughly constant (for example, in a soil map there is one type of soil within one polygon).
2. To overlay all maps selected for one specific hypothesis. This is done with the Geographical Information System (GIS). Overlay proceeds like having all maps on transparencies and putting them on top of each other. The resulting map has many more polygons than each individual map. It is called the 'smallest common geometry', SCG. A polygon in the SCG as derived here has the property that all variables of the maps that are combined in the SCG have roughly a constant value within this polygon: within one polygon there is only one tree species (or one type of stand), the trees are in the same age class, the pH-value (in the case of the soil-acidification hypothesis) is the same, and so on. The GIS automatically creates a new data bank for the SCG.
3. To attribute an ordering to the polygons of the SCG based on the evaluation of the risk according to the different variables. This ordering is qualitative. It gives the liability for something to happen (for example, the risk that the corresponding areas in a forest are damaged). In regional planning the outcome is the probability of use of the different polygons for housing or other construction, if suitable criteria such as availability, accessibility, steepness, etc., are used. Such a SCG is therefore called a 'risk map'.
4. To connect the risk map (bottom layer) with the dynamic model (intermediate layer). If for a specific year forest area containing 3% of severely damaged trees (damage class 5) has to be located, take the polygons with the highest risk until 3% of the forest area is found and attribute damage class 5 to these polygons. If 7% of damage class 4 is to be found, take the polygons with the next lower risks until another 7% of the forest area is found and so on.
5. To plot the resulting map using the GIS.
6. To repeat this procedure for all points in time for which maps are wanted.

Analysis on the strategic layer

The diversities of forest sites affected by forest damage and of damaged tree species are both high. This led the group in 1983 (Grossman et al., 1983) to believe that biological factors such as diseases or pests are not the main factor of the damage. Due to the same reason, climate factors should also be only a part of the problem. The concentration of most pollutants, such as SO₂ or NO_x, is also highly diverse. Only a few pollutants in particular, ozone pose continental problems during extended periods of time. However, the observed symptoms of tree damage are usually different from ozone damage. Therefore, other secondary substances beside ozone (e.g. hydrogen peroxide, formaldehyde, formic acid, and nitric acid) were also more closely investigated because their concentrations correspond on the average to ozone concentrations and should hence also temporarily pose a continental problem (Grossmann, 1988).

Many other combinations of methods are possible. They seem to work best if they are made according to the three-layer description of complex systems given above. Several projects that adopted different, promising approaches sooner or later ran into difficulties and gave disappointing results. Analysis of the shortcomings will be reported in Haber (1989b).

13.2.4.3 Soft coupling; the local expert

Some variables are not always available. However, certain variables can be calculated; for example, the distribution of fog using elevation of the sites and temperature, humidity, and other climate data from a climate station. Most of the calculations on the distribution and frequency of fog are correct; although some are not, and the map of calculated fog distribution must be discussed with people who know the area, the so-called local experts, who in this particular case are farmers and foresters. The map supports communication.

An understanding of data is important. One value of a variable may have a totally different meaning depending on the context in which it is used. For example, one subspecies of spruce has naturally downward-bent branches. But this shape may also be one of the early symptoms of forest damage. The local people must be asked what the data mean in their area. Large central data banks often lead to disaster in understanding when the source and the definition or meaning of the data are not known (Jeffers 1978). The numbers must be translated into information or knowledge which gives the actual meaning, These corrected data can then become the contents of maps from which time series of maps can be derived. This translation process, using knowledge and judgment of local experts, is called 'soft coupling' (Grossmann, 1983).

13.2.4.4 Software for combining maps and models

Software (the DYS-ARC package, devised by the author) is now available to evaluate models and translate their dynamics into dynamic maps, using a SCG. This software asks which risk map to choose, which time series of data to read, and which type of combination of the time series and the risk map to select. The software then asks for the points in time for which to make the maps, checks the times, and then starts ARC/INFO to produce the corresponding time series of maps. This software can also handle more complex types of risk maps with different classes of risk, where no transition between classes is allowed. It can also process dynamics from sources other than models.

13.3 APPLICATIONS OF DYNAMIC MAPS

The maps on forest die-back were produced without knowledge of the actual distribution of the damage, and thus could be compared with the actual situation. In this comparison, the forest scientists acknowledged the shortcomings of their maps showing the distribution of damage based on on-site evaluations, and rated the dynamic map as superior to their evaluation. Actually, both types of map had weaknesses. The two methods, of on-site inspection and of dynamic maps, can be used in conjunction with each other, to complement and check each other.

Dynamics are not only derived from dynamic models. Experts can provide projections based on their (subjective) expectations. As the resulting maps are very precise, and represent the obvious expression of expert expectations, they provide a graphic display by which to judge the validity of the expert opinions. 

    Other applications of dynamic maps are

1. To test scientific hypotheses, as described for forest die-back. Different hypotheses on the cause of forest die-back lead to different dynamics in the models, and to different SCGs in the GIS. Hence different dynamic maps result from different hypotheses. There exist about 200 different hypotheses on the causes of forest die-back. Comparison with the present real situation and its past and future development is a very effective test of the different hypotheses.
2. To update information in the GIS. This means progress towards Automated Geographical Data Banks. As a GIS usually stores large amounts of data, it is time consuming and expensive to update these data. But priorities for update can be determined if dynamic models are combined with the static maps to find out where changes are most likely or most important.
3. To improve classification of remote sensing data. Dynamic maps can be produced to show the same area for the same time as the remote sensing data, but based on different sources of information. Therefore, a comparison of differences can help to classify the remote sensing data, in particular to reveal new and unexpected developments. Pattern evaluation often works even when statistics fail.
4. To provide support for environmental monitoring through the comparison of computed maps to maps based on remote sensing data. The expected development (from any vehicle of anticipation) is translated into dynamic maps and compared with the actual ongoing development (from on-site inspection, false-color infrared photographs, or remote sensing data). Differences may reveal unexpected developments at a very early stage.
5. To give support for environmental early warning. An unexpected development is more easily revealed if the expected development is presented as explicitly and visibly as possible. Dynamic maps are excellent for this purpose.
6. To give support in detection of rare events, slow dynamics, and other somewhat hideous developments, thereby showing that dynamic maps can effectively support long-term ecological research.
7. To guide and schedule management actions in ecosystems (e.g. where to plant or to harvest what at a specific time).
8. To allow evaluation and implementation of strategic criteria. 

The combined method is also helpful in modeling to

9. Assess the applicability of models, and to find errors in structures and data.

10. Calibrate, (in-)validate, or improve models. One data base, on the intermediate layer, allows calibration; the  other, quite different and         on the bottom layer, allows (in- )validation. Besides historical 'validation' about 30 more tests exist for feedback models (Forrester and Senge, 1980); many of these are more powerful if combined with dynamic maps.                                                                 

 13.4 TEST CASE: REASONS FOR THE CHANGES OF AGRICULTURAL LAND USE AND TESTING OF  OPTIONS FOR CHANGE

One case study in a five-year project (MAB Project 6, Berchtesgaden) was on the development of new agricultural policies (Haber, 1989a). The study is briefly summarized below.

The problem concerned farming in Berchtesgaden, Germany, which is especially unprofitable due to unfavorable climate and poor soils. The majority of the farmers have second jobs to subsidize their own farming, and many offer 'holidays on the farm'. Tourists enjoy the landscape, which is characterized by high mountains, lakes, and farms with mountainous pastures. Farming in the area is slowly decreasing as farmers concentrate on tourism. As a consequence, the attractiveness of the landscape is decreasing as the number of cow herds decrease, and natural forest growth takes over abandoned mountain pastures.

An aim of the study was the development of new options for the region, and an indication of whether concentration on tourism will pay or will be counterproductive. Alternative sources of income must be compatible with the climate and the ecological conditions of the region, and to its very low accessibility. Options should be environmentally beneficial.

Strategic analyses

The project group evaluated data on the natural potential, accessibility for traffic, numbers of inhabitants and composition of the economy, new technologies, and other sources of possible innovations in order to assess the options. The history and attitude of the local population were also evaluated. The outcome of the evaluation showed that neither farmers nor the local economy have choices other than tourism. However, since agriculture seems to be important for the viability of tourism in the area, one option is the production of high-quality agricultural products ('organic farming') and direct marketing by the farmers to the tourists. The outcomes of the strategic analyses were used to define three scenarios: abandoning of agriculture and concentration on tourism, continuation of the present behavior, and development of agricultural high-value production and internal marketing.

Layer of complex dynamics

Extensive research on socio-economic factors had been done by one group taking into account the history of the area during the last centuries. Time series for the years 1970 to 1986 of actual data were available for the number of tourists, accommodations, farms, and cattle, the land use in all categories, and several economic parameters. Several models were constructed in interdisciplinary workshops to evaluate the scenarios. These models depict relationships between farmers and their available labor, extent and type of farming, number of cattle, land use in eight categories, aesthetic value of the landscape, supply and demand in tourism, number of tourists in three categories, crowding by tourists, income from farming and tourism, consequences for the regional economy, etc. Extreme values were determined with a linear programming model; dynamics were evaluated with a model using difference equations (Grossmann, 1988b).

It was easy to develop and evaluate this latter model due to new software, the DYS-ARC package, described earlier in this chapter. A run for the years 1970 to 2020 with time steps of 0.1 years with 300 output variables took 1 minute on an AT compatible PC; in earlier case studies a smaller model needed several hours on a VAX using a DYNAMO dialect. The DYS-ARC package allows the user to view interactively, graphically, and numerically all variables in any composition, so that the consequences of changes can be traced through the whole system.

Validation of models is done with up to 11 methods; for example, the models had to reproduce the known time series starting from 1970. Other tests were extreme value tests, discussion of model structures in workshops, etc. Translation of model results into dynamic maps provided further tests (see below).

Results showed that abandoning agriculture was the most unfavorable scenario because it increased competition with nonagricultural offers in tourism and caused slow deterioration of important characteristics of the landscape. In this scenario, the demand for and the quality of the tourism decreased, and supply and competition increased. Establishment of agricultural high-value production and internal marketing needed far less investment than was anticipated, and turned out to be the most profitable option for the whole region. One unexpected result of this latter option was the too-great increase in intensity of land use, which would cause degradation of the land and agricultural pollution. Restrictions on the most intense types of land use would be needed.

Bottom layer

Geographical information was entered into the GIS using scales down to 1:5000 where necessary. Different map scales were used complimentarily. Sources of information were the available maps, aerial photography, and additional field data collection. Maps now exist within the ARC/INFO GIS on soil types, humidity of the soils, detailed land use including roads, other line information such as railroads, polygon information on settlements, forests, location of farms, type of farming area and present use (from aerial photography and on-site visits), vegetation, and other information.

With this information, the agricultural areas were ordered according to suitability for grazing so that any change in land use suggested by the dynamic model could be translated into time series of maps where these changes are most likely to occur. Important information for this was the soil type, steepness, aspect, size, present state, present use, vegetation, and elevation. This translation is similar to the one explained for forest damage, but slightly more complex.

 This translation allows checks on whether predicted changes are possible and reasonable.

Evaluation of patterns

Inhabitants of the area and local experts instantly perceived remarkable developments in time series of maps, and started heated discussions if the developments seemed reasonable and loudly rejected unlikely or wrong maps.

Statistical analyses

Complex maps are composed of many polygons, and for each polygon all data are available in the data bank of the GIS. Hence statistical analyses could be done with reasonable and homogeneous data.

In several studies, statistics showed a poor correlation whereas there was marked similarity in actual and computer-produced patterns. In one of these cases the prevailing wind direction was specified erroneously so that predicted and actual patterns differed by about 45°; in another case an anemometer had malfunctioned during the last 4 years. The team had not used these specific data from the last 4 years because they gave an unlikely pattern, and a long-term time series and checking of this device (one year later) helped to establish that the team had been right.

With these data and methods it is possible to predict further changes of abandoned areas, as detailed information on ecological conditions of all sites and succession types of sites exists. Tourists and inhabitants had been asked about their preferences for different landscapes using photographs. Evaluation of anticipated landscape changes with these preferences allowed judgments on the validity of model outcomes on the layer of dynamics. With a detailed balancing model, possible increased levels of fertilization have been evaluated, taking into account characteristics of the soils, orographic features (slope), vegetation, and location and intensity of land use. These parameters determine the consequences of agricultural pollution of groundwater.

The study was made possible by the availability of long-term time series and an understanding of the history of the region and the likely behavior of the local population. Nearly all predictions, model construction, and, in all probability, model results would have been wrong without this knowledge of the history of the region and its ecosystems. For example, the inhabitants tend to stay in their area in spite of crowded housing, lack of jobs, and poor climate. Model parameters quite different from those suitable for other regions in Germany had to be chosen. Forests were nearly completely destroyed in the past in the production of salt, and reforestation was often done without consideration of which subspecies are suitable for climate and soil. Catastrophes in forests are partially due to poor choice of subspecies.

On several occasions the models and historical time series were at odds, but the model was more often correct than the data. In some cases, typing errors in the time series had caused the discrepancies, in other cases established knowledge about the area was deliberately ignored due to scientific prejudices, but had to be included in the model so that it could pass the tests, in particular the historical validation. For this, long -term time series were necessary .

The resulting dynamic maps are effective for testing hypotheses of behavior, successional changes, changes in nitrate content of groundwater, etc. These maps are also effective for support of ongoing environmental monitoring. All of these are important problems in long-term ecological research.

13.5 CONCLUSIONS

Dynamic maps have been successful in testing and improving models, in evaluating scientific hypotheses, in finding errors in data and assumptions, and in checking and developing strategies. Many more applications of model- and strategy-driven maps seem possible, and new options for ecosystem research and management seem to emerge through this method.

13.6 REFERENCES

Adisoemarto and Brunig, E.F. (Eds) (1978). Transactions of the Second International MAB- IUFRO Workshop on Tropical Rainforest Ecosystem Research, 21-25 October 1978, Special Report No.2, Chair of World Forestry, University of Hamburg.

Allen, T.F.H. and Starr, T.B. (1982). Hierarchy. Perspectives for Ecological Complexity. The University of Chicago Press, Chicago, Illinois.

Ashby, W.R. (1959). An Introduction to Cybernetics. Chapman and Hall, London.

Ashby, W.R. (1960). Design for a Brain. John Wiley, New York.

Bertalanffy, L.V. (1969) (rev. edition). General Systems Theory: Essays on its Foundations and its Development. Braziller, New York.

Bossel, H. and Strobel, M. ( 1977). Experiments with an intelligent world model. Futures, 10, No.3, 191-212.

Forrester, J.W. (1969). Urban Dynamics. MIT Press, Cambridge, Massachusetts.

Forrester, J. W. and Senge, P. (1980). Tests for building confidence in system dynamics models. In Legasto, A.A. Jr, Forrester, J.W., Lyneis, J.M. (Eds) System Dynamics. North-Holland, Amsterdam, 209-228.

Grossmann, W.D. (1978). The dynamic meta-model - a tentative suggestion how to approach the solution of a pressing problem. In Adisoemarto, S. and Brunig, E.F. (Eds) Transactions of the Second International MAB-IUFRO Workshop on Tropical Rainforest Ecosystem Research. Special Report No.2. Chair of World Forestry, University of Hamburg, 166-185.

Grossmann, W.D. (1983). Systems approaches towards complex systems. In Messerli, P. and Stucki, E. (Eds) Fachbeiträge der schweizerischen MAB-Information, Vol. 19. Bundesamt für Umweltschutz, Bern.

Grossmann, W.D. (1987). Systemprojekt Lehrforst Rosalia: Untersuchungen zum Problem der neuartigen Waldschäden. Final report. Federal Ministry for Science and Research, Vienna.

Grossmann, W.D. (1988a). Products of photo-oxidation as a decisive factor of the new forest decline? Results and considerations. Ecological Modelling, 42, 281-305.

Grossmann, W.D. (1988b). Regio: Ein Modellbausatz zur Bearbeitung ökonomisch-ökologischer Probleme im Rahmen eines hierarchischen Systemkonzeptes (REGIO: A Model Kit for Socio-Economic Systems). Final Report Research Area 99, MAB Project 6 Berchtesgaden, Parts 1 and 2. National Park Administration, Berchtesgaden.

Grossmann, W.D. and Brunig, E.F. (1983). Concepts of hierarchy in forest sector modeling. In Seppala, R., Row, C. and Morgan, A. (Eds) Forest Sector Modelling. AB Academic Publishers, 105-117.

Grossmann, W.D., Haber, W., Kerner, F., Richter, U., Schaller, J. and Sittard, M. (1983). Ziele, Fragestellungen und Methoden. Ökosystemforschung Berchtesgaden. MAB- Mitteilungen Nr. 16. Deutsches Nationalkomitee MAB, Bonn.

Grossmann, W.D. and Orthofer, R. (1987). Grobbeurteilung von Kohlenwasserstoffemissionen im Hinblick auf ihre möglichen Beiträge zum Waldsterben. Report OEF2S-A-1121. Austrian Research Center, Seibersdorf.

Grossmann, W.D. and Schaller, J. (1986). Geographical maps on forest die-off, driven by dynamic models. Ecological Modelling, 31, 341-353.

Grossmann, W.D., Schaller, J. and Sittard, M. (1984). 'Zeitkarten': eine neue Methode zum Test von Hypothesen und Gegenmassnahmen beim Waldsterben. Allgemeine Forstzeitschrift, München.

Haber, W. (Ed.) (1986). Mögliche Auswirkungen der geplanten Olympischen Winterspiele 1992 auf das Regionale System Berchtesgaden. Lehrstuhl für Landschaftsökologie. TU München /Weihenstephan.

Haber, W. (Ed.) (1989a). (1): Report on the consequences of different agricultural policies in Berchtesgaden, using a combination of dynamic models and a GIS. (2): An update of Grossmann et al. 1983, taking into account the experiences from 5 years research in the MAB 6 Project Berchtesgaden. In press.

Haber, W. (Ed.) (1989b). Report on the project on forest die back investigated by combining dynamic feedback models and area related balancing. Lehrstuhl für Landschaftsökologie. TU München /Weihenstephan. In press.

Haber, W., Grossmann, W.D., Schaller, J. (1984). Integrated evaluation and synthesis of data by connection dynamic feedback models with a geographic information system. In Brandt, J. and Agger, P. (Eds) Methodology in Landscape Ecological Research and Planning. Proc. 1st International Seminar. Int. Assoc. of Landscape Ecol. Roskilde Univ. Center, Roskilde.

Haken, H. (1978). Synergetics, 2nd edition. Springer-Verlag, Berlin.

Rolling, C.S. (Ed.) (1978). Adaptive Environmental Assessment and Management. John Wiley, New York.

Jeffers, J.N.R. (1978). An Introduction to Systems Analysis: With Ecological Applications. Edward Arnold, London.

Markowitz, H.M. ( 1959). Portfolio Selection; Efficient Diversification of Investments. John Wiley, New York.

Mesarovic, M., Macko, M. and Takahara, Y. (1971). Theory of Hierarchical, Multilevel Systems. Academic Press, New York.

Rappaport, R.A. (1979). Ecology, Meaning and Religion. North Atlantic Books, Richmond. 

Simon, H. (1962). The architecture of complexity. Proc. American Philos. Soc., 106,   December.

Thorn, R. (1975). Structural Stability and Morphogenesis. Benjamin, Reading, Massachusetts.

Vester, F. and von Hesler, A. (1980). The Sensitivity Model. RPU (Regionale Planungsgemeinschaft Untermain), Frankfurt.

Zeigler, B.P. (1979). Structuring principles for multifaceted systems modelling. In Zeigler, B., Elzas, M. et al. (Eds) Methodology in Systems Modeling and Simulation. North-Holland, Amsterdam, 93-135.