An ecosystem is a biological environment consisting of all the living organisms or biotic component, in a particular area, and the nonliving, or abiotic component, with which the organisms interact, such as air, soil, water and sunlight.
Ecosystems have become particularly important politically, since the Convention on Biological Diversity (CBD) - ratified by 192 countries - defines "the protection of ecosystems, natural habitats and the maintenance of viable populations of species in natural surroundings" as a commitment of ratifying countries. This has created the political necessity to spatially identify ecosystems and somehow distinguish among them. The CBD defines an "ecosystem" as a "dynamic complex of plant, animal and micro-organism communities and their non-living environment interacting as a functional unit".
With the need of protecting ecosystems, the political need arose to describe and identify them efficiently. Vreugdenhil et al. argued that this could be achieved most effectively by using a physiognomic-ecological classification system, as ecosystems are easily recognizable in the field as well as on satellite images. They argued that the structure and seasonality of the associated vegetation, or flora, complemented with ecological data (such as elevation, humidity, and drainage), are each determining modifiers that separate partially distinct sets of species. This is true not only for plant species, but also for species of animals, fungi and bacteria. The degree of ecosystem distinction is subject to the physiognomic modifiers that can be identified on an image and/or in the field. Where necessary, specific fauna elements can be added, such as seasonal concentrations of animals and the distribution of coral reefs.
Several physiognomic-ecological classification systems are available:
§ Physiognomic-Ecological Classification of Plant Formations of the Earth: a system based on the 1974 work of Mueller-Dombois and Heinz Ellenberg, and developed by UNESCO. This classification "describes the above-ground or underwater vegetation structures and cover as observed in the field, described as plant life forms. This classification is fundamentally a species-independent physiognomic, hierarchical vegetation classification system which also takes into account ecological factors such as climate, elevation, human influences such as grazing, hydric regimes and survival strategies such as seasonality.
§ Land Cover Classification System (LCCS), developed by the Food and Agriculture Organization (FAO).
§ Forest-Range Environmental Study Ecosystems (FRES) developed by the United States Forest Service for use in the United States.
Several aquatic classification systems are available, and an effort is being made by the United States Geological Survey (USGS) and the Inter-American Biodiversity Information Network (IABIN) to design a complete ecosystem classification system that will cover both terrestrial and aquatic ecosystems.
From a philosophy of science perspective, ecosystems are not discrete units of nature that simply can be identified using the most "correct" type of classification approach. In agreement with the definition by Tansley ("mental isolates"), any attempt to delineate or classify ecosystems should be explicit about the observer/analyst input in the classification including its normativerationale.
According to Billings (1978) the term ecosystem is "an energy-driven complex of a community of organisms and its controlling environment" [CBD, 2010]. It means also "a dynamic complex of plant, animal and micro-organism communities and their non-living environment interacting as a functional unit" (Article 2 of the Convention)[CBD, 2010]. A structure is the construction or the composition of a thing. In this context the meaning of construction is less important because ecosystems where not constructed by following a (human) plan like it is the case by the construction of a car for example. The natural world is fundamentally different from the world built by humans. In this case the meaning of composition is more crucial: The composition of organisms and natural systems, the visual pattern which we are able to realise (e g. the anatomy of an organism), provides fundamental information as basis for the study functions and a framework where dynamic action takes place [Golley, 2000, p. 21]. The problem is that nature does not provides us with a visual pattern of ecosystems. Ecosystems are rather based on a human definition. Therefore it is necessary to visualize their structures artificially.Golley [2000, p. 31] merges the terms ecosystem and structure to ecosystem structures: "An ecosystem structure is a network of interaction between components of the system". These components are the so called "entities" which are defined by Golley and Keller  as "objects which are bound in space and/or time and which are internally coherent". Fredrick Ferré, a philosopher with affection to metaphysics identified six kinds of ecological entities [Ferré, 1996]:
1. Aggregate entities are physical bodies with boundaries, e.g. mountains and lakes; they form a background for the ecological play.
2. Systematic entities include ecosystems. These retain structure and function under continually changing environmental conditions.
3. Formal entities are based on the knowledge added to them, e.g. a species;
4. Organic entities are living organisms. They are creative because they generate unique, new forms of life.
5. Compound entities are those which have strong internal relationships but are without apparent internal system dynamics, e.g. water or sodium chloride (chemicals).
6. Fundamental entities are of little use for ecologists (Metaphysics needs such a category).
Aggregate, organic and compound entities (1, 4 and 5) are interpretable as structural components which can interact with each other and which are embedded in systematic entities: ecosystems.
Construction of ecosystem structures
By visualising ecosystem models it is possible to emphasize on the components, the interaction between components or both in various levels of abstraction. Three different approaches with different purposes are shown in the following: One of the first people who operationalised the ecosystem concept (ecosystem approach) was Raymond Lindeman in 1942. He and his wife sampled the biota and environmental factors of the small "Cedar Bog Lake" for about two years. They and classified several feeding groups and arranged them into a trophic pattern, the food cycle. Ooze as sink for organic matter is placed at the centre of the diagram. The conceptual model of the Cedar Bog Lake ecosystem (Fig. 2) contains internal cycles of elements among the system components and an outflow from the system to the larger environment in which it is nested.
Figure 2. Food cycle diagram based on Cedar Bog Lake, Minnesota, [Lindeman 1941]
Lindeman stayed within the biological paradigm where biologically defined components (organisms and species) play a dominant role. In 1957 Lindeman's circular diagram was replaced by the more abstract energy flow diagram developed by Howard T. Odum (Fig. 3). Odum generalised concrete species and organisms to functional taxa. This allows to reduce the biological component to a node and emphasise the connections between the nodes. Hence, the dominant structures are the relationships, not the biological properties.
Figure 3. Energy flow diagram with units in kilocalories per square meter per year in the Silver Springs community, Florida. The letters indicate herbivores (H), carnivores (C), top carnivores (TC) and decomposers (D). [Odum, 1957]
In 1967 Bormann and Likens presented their ecosystem model which consisted of storages of chemical elements and transfer flows (intra system cycle) between compartments, linked to the earth' biosphere by inputs and outputs (Fig. 4).
Golley noted that there is not only one correct or most useful approach to develop an ecosystem structure. It is common to compile a list of species present in the ecosystem (like Lindeman does) and then a list of interactions between species. Depending from the data which can be rather qualitative or more quantitative it is possible to construct quite different maps or diagrams of the interaction network. In order to find a good compromise between abstraction and reality Golley suggests to focus on local ecotopes, the smallest systems which are homogenius for the properties of interest [Golley 2000, p. 31] .
Figure 4. Relationships of the geological, meteorological and biological components within the ecosystem, connected through inputs and outputs to the biosphere [Borman and Likens 1967]