Texto 4 - Agroecology (...)

Agroecology, scaling and interdisciplinarity
Tommy Dalgaard (a),[*]*, Nicholas J. Hutchings (a), John R. Porter (b)

(a) Department of Agroecology, Danish Institute of Agricultural Sciences. P.O. Box 50. DK-8830 Tjele, Denmark.
(b) Department of Agricultural Sciences, The Royal Veterinary and A.gricultural University, Agrovej 10, DK-2630 Taastrup, Denmark.
Received 9 October 2001; received in revised form 11 April 2003; accepted 15 April 2003

Abstract
Based on a review of its history, its present structure and its objective in the future, agroecology is defined as an integrative discipline that includes elements from agronomy, ecology, sociology and economics. Agroecology's credentials as a separate scientific discipline were measured against the norms of science, defined by Robert King Merton (1973): communalism, universality, disinterestedness, originality and doubt. It is concluded that agroecology meets many of these norms and where it differs, it does so in a way that perhaps anticipates the manner and the direction in which the social position of science is changing.
Accepting agroecology as a separate scientific discipline, the two main issues with which it most contend were considered to be those of scaling and interdisciplinarity. Scaling is a problem because results of agroecological research are typically generated at small spatial scales whereas applications are frequently implemented in larger, administrative units. A frame-­work to convey information from science to decision-makers was proposed and tested in a case study of farm energy use. Interdisciplinarity is a problem because researchers from different disciplines see the world from different viewpoints, use different language, work at different locations and use different criteria to evaluate one another's work. Progress in this area is likely to be slow and driven by the need to justify the value of science to society.
02003 Elsevier B.Y. All rights reserved.

Keywords: Agroecology; Discipline; Food production systems; Hierarchy; Interdisciplinarity; Mertonian norms; Scale; Sustainability

1. Introduction
A major challenge facing the world is how a 21st century population of perhaps 9 billion people will feed themselves in a sustainable manner (Evans, 1998). During the 20th century, a doubled population was fed via the so-called Green Revolution, with its introduction of pesticides, synthetic fertilisers and new high-yielding cultivars. With the reduction in the proportion of hungering people from more than 50% of the total population after World War II to under 20% today (Grigg, 1993), the success of this revolution is indisputable. However, there are still malnourished people and the impacts of intensive agriculture on natural resource degradation and the environment may not be sustainable (Brown el at., 2000). The proposed role of agroecology is to facilitate the design and management of sustainable food production systems (Gliessman, 1998), and to investigate possible synergisms that can help alleviate the above problems (Altieri, 1990). However, agroecology has not fully matured as a scientific discipline. In this paper, the definition and scientific method of agroecology, its credentials as a scientific discipline and the challenges that face it are considered. The intention is to establish a general framework for the integration of information within agroecology, and for the communication of this information to the decision-makers targeted. Here, it is recognised that the rationale for agroecology is currently the need to develop sustainable systems of food production and this requires that knowledge must be effectively delivered to the people who are in a position to take appropriate action.

2. A history of agroecology
The term agroecology was in parallel proposed by German zoologists (Friederichs, 1930), and American crop physiologists (Hanson, 1939) as a synonym for the application of ecology within agriculture. At that time, ecologists had relatively narrow foci but with a trend towards a more integrative view of ecosystems. The early population ecology school of Herny Gleason investigated plant populations seen from the organism's perspective, thereby focusing on the hierarchical levels of the organism (Fig. 1). In contrast, the community ecology school of Frederic Clements viewed plant populations from the landscape perspective, thereby also including higher hierarchical levels than the organism (O'Neill et al., 1986). However, the true roots of agroecology probably lie in the school of process ecology as typified by Tansley (1935), whose worldview included both biotic entities and their environment (Fig. 1). It was from this school of process ecology that the agroecosystem concept emerged (Harper, 1974), and the foundations for modern agroecology were laid.

Fig. 1. The box symbolises the window of agroecology within food production systems. The viewpoints of the different schools of ecology are marked with eye signatures. The classical, scientific disciplines, where some are within the window of agroecology, are lined up in the right column, ordered in a hierarchy with the ‘hard’ disciplines at the bottom and the 'soft' disciplines at the top (Checkland, 1999).


2.1. "Hard" agroecology
According to Hecht (1995), the hard branch of agroecology (physical-analytical and natural science based) was initiated by works such as "Silent Spring" (Canson, 1964). "The Population Bomb" (Erlich, 1966), "Tragedy of the Commons" (Hardin, 1968) and "The Limits to Growth" (Meadows et al., 1972). The gloomy predictions of these and similar polemical writings have largely not come to pass, mainly because the speed of technological developments was underestimated. However, hard agroecology has shown that badly managed agriculture can lead to the degradation of agricu1tural land (Waldon et al., 1998), undesirable changes in semi-natural ecosystems (Lambert et al., 1990) and the depletion and pollution of natural resources (e.g. de Molenaar, 1990). Consequently, the focus of agricultural science has changed over the past 20-30 years from the maximisation of food and fibre production towards understanding the mechanisms linking costs (nutrient losses, loss of biodiversity and landscape degradation) to the benefits of agriculture (production, wealth generation and landscape maintenance). To understand these linkages required a combination of ecology, agronomy and economy (Reintjes el al., 1992) that may be considered "hard" agroecology. Such hard systems thinking, integrating various disciplines within natural sciences and economy, was significantly developed during the 1980s and 1990s, but remains the approach of an engineer or a classical economist (Checkland, 1999). This means that the resources entering and leaving agricultural systems are considered to be finite capital measured in physical or monetary units (Pearce, 1996). Furthermore, the position of the observer and scientist are thought of as external to the systems under study, which as we will see is not necessarily the case of soft agroecology.

2.2. "Soft” agroecology
There has been a debate whether hard system optimisation of agriculture alone could solve the problem of feeding an expanding world population. It is increasingly felt that this is not the case and that a much broader view of the Structure, function and role of agroecosystems is called for (Conway, 1987). Such a vision addresses hard issues such as the follows of energy and matter through agroecosystems but also includes the role of human and society, and the empowerment of citizens for developing their own food systems, and thereby feeding themselves. The exploration of the interaction between these human activity system and the hard agroecosystem is here defined as soft agroecology. According to this soft system thinking (Checkland, 1999), the capital entering and leaving agricultura1 systems is not only measured in physical units but also includes cultural knowledge, human experiences, potentials for technological development, etc. In contrast to hard capital, this soft capital is flexible (Pearce, 1996) and can even to some degree substitute hard capital. For example, knowledge of traditional farming systems inherited from their forefathers may help future farmers to save physical inputs (Gliessman, 1990a). However, a major problem is that the disciplines of rural sociology and economics, which deal with this area of soft agroecology, tend to operate at higher hierarchical levels than the hard disciplines of agronomy and ecology. This means that the soft disciplines often work at the farm or the regional level, while the hard disciplines often work at the plot or the field level. Furthermore, some soft systems researchers work as accomplices to the farmer, both giving and receiving knowledge, unlike their hard systems colleagues who work as external observers to the system under study. This is a consequence of the inclusion of interactions between humans within the window of agroecology (Fig. 1). They argue that all people dealing with agricultural production systems, including scientists, are intimately and subjectively involved in the activities of the growing of food and that to study this process is to become a part of it (Longino, 1990).

2.3. Where is agroecology now?
Recognising that agroecology is still developing, a survey of the published literature was conducted to establish its current status. The survey was conducted by interrogating electronic databases (CAB, 2001; AGRICOLA. 2001; ISIS-SCI, 2001; SSCI. 2001; ECONLIT, 2001), reading literature reviews (Carls, 1988, 1989, 1990) and visiting The Agroecology Library, University of California, Santa Cruz. In agreement with Carroll et at (1990), most references were related to natural sciences within the fields of agronomy and ecology (e.g. the work in Gliessman. 1990b). However, references were also found within the social sciences (e.g. Francis and King. 1997; Thomas and Kevan, 1993), economics (e.g. Allen, 1999; Rosset, 1996), or in combination of two or more areas (e.g. Edwards et al., 1993; Van Latesteijn, 1997). To quantify this distribution, the number of references to "agroecology" or "agro-ecology" (with a hyphen) in literature databases of natural sciences (ISIS-SCl, 2001), resulting in 94 references, social sciences (SSCI. 2001) and economics (ECONLIT. 2001) were compared. The majority (66%) of the references were only found in the natural science databases, with 13% only in social science database, and 5% only in economic literature. No references were in the databases from all three fields of science. The remaining references were found in two out of the three fields, with 2% in social and natural sciences, none in social sciences and economics, and 16% in a combination between natural sciences and economics (Fig. 2).
Compared to the total number of references in the searched databases, relatively few referred to agroecology. For example CAB (2001) refers to 1195 abstracts including the term agroecology out of the more than 2 million references in total. In comparison, more than 300,000 references referred to animal nutrition. Using the definition of agroecology stated in the next section, we could have redefined a number of additional and often earlier studies as agroecological, even though the authors chose not to describe them as such at the time. However, the point of the survey was not to determine what work was being done but rather whether the scientists involved considered the studies to be agroecological.



2.4. A definition of agroecology
In this paper, agroecology is defined as "the study of the interactions between plants, animals, humans and the environment within agricultural systems". Agroecology as a discipline therefore covers integrative studies within agronomy, ecology, sociology and economics (Fig. 1). Most authors acknowledge agroecology as a discipline of integration, but define it in other terms, for example as 'the application of ecological science to design and management of sustainable agroecosystems' (Gliessman, 1998). Thereby, the upper part of the window of agroecology in Fig. 1 is excluded. Clearly there is still not one, finally acknowledged definition of agroecology, indicating the ongoing development within the discipline.
The historical development of agroecology shows that it began originally as a part of crop physiology, agricultural zoology, and ecology but the term was adopted by a movement which wished to promote the development of sustainable agriculture through the integration of ideas and methods from other disciplines (Altieri, 1980). Now agroecology departments exist at a number of universities across the world but particularly in the USA and Europe. This implies that at least some people think that agroecology has made the transition from a proposition to a separate scientific discipline. In the next section, the case for considering agroecology as a separate scientific discipline is examined.

3. Agroecology as a separate, scientific discipline

3.1. A separate discipline
To be considered a separate discipline, agroecology must be distinguishable from existing disciplines. The argument is that agroecology is distinguished from its parental disciplines of agronomy, ecology and socio-economics by its integration between these disciplines and across scales. The agroecology-related studies found in the literature survey were characterised by an integrative approach, where information from single disciplines was collected and combined to solve problems at a higher scale. An additional indication that agroecology is a separate discipline is that the numbers of references to agroecology have increased in recent years, indicating that more scientists feel that their work lies sufficiently far from the existing scientific disciplines that an alternative term is necessary.

3.2. A scientific discipline
The assessment of agroecology as a scientific discipline was made using the norms of science as defined by the sociologist Robert King Merton (Merton, 1973). This approach was inspired by a recent attempt by the physicist John Ziman (Ziman, 2000) to define science in terms of what it is and what it means.
The first Mertonian norm of science is communalism, meaning that the outcomes of academic science are delivered to the public in the broadest sense, including other scientific colleagues and the wider public. Scientists differ in the weight they assign to the importance of these dissemination routes. These can vary from scientific papers in specialist journals to popular television programmes. Agroecology values communalism. It is probably the case that many agroecologists place as much emphasis on sharing results with society as with their scientific colleagues.
The second norm is that science should be universal and open to contributions from all, irrespective of race, gender; nationality, religion, etc. The only things that should wither, and be excluded from science, are ideas and theories not meeting with experimental verification or observation. Agroecologists would try to maintain the norm of universality in the Mertonian sense, as is evidenced by the papers refereed to in Section 2.3. However, universality in agroecology can often be very broad and may deliberately include other stakeholders, so that agroecology sometimes borders on being a socio-political movement. Agroecology faces a paradox when moving its focus to higher and thereby more aggregated levels of hierarchy (Figs. 1 and 3). At the highest hierarchical levels, local context can so swamp generality that research in, for example, rural sociology often ends as a series of case studies from which it is impossible to draw general conclusions.



Disinterestedness in the reporting of science is the third norm. Science reporting is unusual principally because of its impersonal manner, conveying an impression of non-prejudice and disinterestedness from the reported work.. Thus, the impersonality and care taken in reporting science stems from the knowledge that results and conclusions are likely to be challenged by others. It is thus part of a scientist's duty to facilitate this examination in the interests of the wider scientific enterprise. With respect to disinterestedness, agroecology does not differ from any other scientific discipline. Thus, experiments are reported, models can be verified, and social and economic analyses are sometimes but not always repeatable. An important interaction between these norms is that the commuona1ism of science acts as a control on science's disinterestedness (Ziman, 2000) - the value of an objective scientific observation or experiment assessed via the social process of peer review. Thus, science and agroecology are disinterested attempts to search for objective truths that are paradoxically mediated by socially constructed controls and evaluation processes.
Originality is the fourth norm. The tried and tested route to making an original scientific contribution, in the sense of a 'new' piece of knowledge, is to plough the furrow ever deeper. Thus, it is a rational scientific response to focus on ever more detail in the hope of developing a fragment of the scientific story for oneself. In supplement, agroecology's originality also stems from synthesis as well as from thinking outside the commonly accepted thesis of the existing knowledge base. Marching under the twin banners of synthesis and interdisciplinarity, agroecology, in line with discip1ines like anthropology, psychology and sociology, is at odds with what is commonly termed 'basic', natural research with its clear defined boundaries for research, theoretical framework and sense of coherence. However, science is perhaps moving towards the agroecological model, where the constructed and the objective aspects of science are both recognized. For example, disciplines such as climatology and some aspects of geosciences appear to becoming more integrative and less reductionist. This trend is evident, for example, when the activities of humans are seen as within the system of study rather than external to it. An example would be the role that human activities play in land use change or as drivers of biogeochemical processes such as the global carbon and nitrogen cycles. Nowadays, these originally natural cycles cannot be studied and understood without understanding and integrating the human role. In this 'post-academic’ science (Ziman, 2000) the cultural and social context of science as a process of knowledge creation is explicit.
Doubt is the restraint on originality in science and its application, via scepticism, is the fifth Mertonian norm. This enters in at least two stages in the scientific process. New ideas and theories are evaluated against a sceptical starting point - the null hypothesis. Having successfully cleared this obstacle, a new piece of scientific knowledge is then subjected to further doubt by anonymons referees who act on behalf of the scientific community. In agroecology, the first of these steps sometimes differs from the above scientific norm. Some agroecological studies do not start with a classical null hypothesis but include semi-quantitative surveys, rapid rura1 assessments and studies closely linked to agricultural development. These can be and are subjected to the second level of doubt. However, no method of collecting data should preclude the need for an explicit underlying hypothesis, question or assumption that is being tested.
In summary, agroecology meets many of the Mer­tonian norms of science and where it differs it does so in a way that perhaps anticipates the manner and the direction in which the social position of science is changing. Having concluded this, the next two sections consider two of the main issues that face agroecologists; scale and interdisciplinarity (Marceau, 1999).

4. Scale
The issue of scale means that there is a gap be­tween the scale at which most agroecological information is currently generated and the scale at which most decisions concerning agricultural systems are made (Dalgaard, 2001). The results of agroecological stud­ies, generated on the plot, field or farm level, cannot always readi1y be generalised to the regional, national or global level relevant for decision-makers. Because of this gap, the results are often misinterpreted or not used in the decision-making process (Lerland et al., 2000). Scaling issues have been addressed for many years in sciences such as physics (Crutchfield, 1994) or economics (Cropper and Oates, 1992). Until recently, there has been relatively little focus on methods to convey information between scales in the environmental sciences of ecology (Rastetter et al., 1992) or agronomy (Bietkens et al., 2000; Stein et al., 2001), although within theoretical ecology there are some ref­erences from the 1970s and 1980s (e.g. O'Neill et al., 1986). Consequently, agroecologists tend to use scal­ing procedures that are too simple (Grace et al., 1997) and that are poorly suited to global problems e.g. green house gas emissions (Flavin and Dunn, 1998). How­ever, recent advances in scaling have responded to the need to translate environmental and socio-economic indicators from the scale of observation or collection to that of individual operator or national policy. This has led to several new statistical developments, and the application of geostatistics in particular (Riley, 200 1).

4.1. Hierarchy and scale
Shown below are the classical examples of the hierarchy within natural (1) or agricultural systems (2, 3) (Odum, 1971), where: the lower levels of organization or complexity are to the left, and the higher levels to the right:

1. cell ↔ organism ↔ population ↔ community ↔.ecosystem ↔landscape
2. plot ↔ field ↔farm ↔watershed ↔ region ↔ nation ↔ union ↔ globe
3. cell ↔ organ ↔ animal/plant ↔ herd/field ↔farm ↔ region

These hierarchies represents levels of organisational complexity ranked by category or class, and are the basic structural units of the system investigated (Whyte et al., 1969). Often hierarchical levels are nested, so that high level units consist of lower level units (Fresco, 1995: Fig. 3). The boundary between hierarchical levels may be visible, such as the skin of an organism or the shoreline of a lake, or intangible in the case of for instance of populations and species. There are two dimensions of scale: spatial and temporal (Fig. 3). Consequently, the term scale relates to space and time period (e.g. a regional scale study of a 100 km2 area in 4 years). In this paper, the colloquial definition of scale is used (Curan eI 31., 1997), meaning that large scale studies cover large areas and/or time spans and small scale studies the reverse.

4.2. Linear, non-linear and hierarchical scaling
Even though hierarchies and scales are connected, so that high level hierarchies are normally studied on larger temporal and spatial scales, they are not synonymous (Allen and Hoekstra, 1992; Fig. 3). The range over which a single level in a hierarchy can extend has consequences for describing the behaviour of higher hierarchical levels because there may be scale-dependent processes present within one or more levels of. hierarchy.
For example the total diesel fuel use Ftotal (I) can be calculated at the hierarchical level of the field (Eq. (1)), and aggregated to the farm level (Eq. (2)). In this very simplified example, derived from the Dalgaard et al. (2001) model, Fn. is the average fuel use per ha on field n, An the field area in ha, and N is the number of fields on the farm.

Ftotal = FnAn (1)

N
F total = ∑ FnAn (2)
n=1

With a linear scaling procedure (also called simple scaling, Grace et al., 1997), Fn. is constant, e.g. Fn = 100 l ha-1 for all fields, and the fuel use is an identi­cal, linear function of both field and farm area. With a non-linear scaling procedure, Fn is a non-linear function of the field area. For example, if Fn = 103 An-An2, when An <> 100 I ha-1, whereas if An > 3ha, Fn < total =" ∑" n="1">

4.3. Framework of hierarchy and scale – an example
To cross the barrier of scale, a general framework: based upon the classical method of natural sciences, involving observation, hypothesis and test is proposed (Table 2). Within this framework, which here is posed for a simplified case that assumes one uniform decision-maker, specific linear, non-1inear or hierarchical scaling functions may be explored and used to support decision-making (Bierkens et al., 2000). In the following, an application of this simplified framework is illustrated with a simple example, with one group of decision-makers. However, in reality there are often many different actors, stakeholders and decision-makers in many hierarchies and scales, making application of the framework more complicated. In the present example, the criteria of the framework of hierarchy and scale are indicated with numbers in brackets. The targeted decision-makers were Danish politicians who after the Rio-Conference in 1992 demanded information on how agriculture could contribute to reduce the Danish energy use and greenhouse gas emissions by the promised 20%. Specifically, they wanted to know whether three different scenarios for conversion to organic farming might help to reduce the energy use (Bichel Committee, 1999). The time scale was a 12-year period (Danish Ministry of Environment and Energy, 1995), and the spatial scale was the 27,000 km2 agricultural area of Denmark (criteria 1 in Table 2). As the existing figures on energy use were sampled on the field and animal housing level (criteria 2), the question was how to upscale these data to the national level. The simplest option would have been a linear scaling procedure, where the average energy uses for different field crops and livestock housings is multiplied with national crop and livestock figures (criteria 3). However, because of scale dependent non-linearities and significant emerging factors, a linear scaling procedure was too simple for the upscaling (criteria 4). For reasons discussed in section a non-linear scaling procedure (criteria 3) was also too simple to predict fuel use (criteria 4), and a hierarchical scaling procedure was needed (criteria 3). In this case, a two-step application of the hierarchy-scale framework was tested. Step one was from the field to the farm level and step two was the final national level generalisation.

4.3.1. Step one
Measurements revealed a 47% deficiency in farm level fuel use compared to field level literature values linearly upscaled to the farm level (Refsgaard et al., 1998). and extended sampling on the field and farm level was initiated (criteria 4). A new model for calcu1ation of farm level fossil energy was made (Dalgaard et al, 2001) including fuel use as a function of the amount of inputs used, yield and the soil type on each field (criteria 3). Also, the above-mentioned emergent factors of fuel use for transport between fields and the farm and between the farm, fodder stocks and feedstuff businesses were included. Finally, the new model was verified (criteria 4) with samples of fuel use, F, and the Fobserved – Fsimulated difference was found to be insignificant.

4.3.2. Step two
The derived model was used in the final national level generalisation of the fossil energy use in the primary Danish agricultural sector. A linear scaling procedure was used, where the estimated average energy use for each crop and animal type was multiplied by the areas of crops and the namber of animals according to national agricultural statistics (criteria 3). The simulated and lineally upscaled energy use embedded in each of the accounted energy carriers was similar to the expected energy use according to national statistics, with a total difference of less than 3% (Dalgaard et al., 2002). In this case, it was concluded that the applied linear scaling procedure was sufficient for the step two generalisation (criteria 4).

5. Interdisciplinarity
Interdisciplinary means working across traditional disciplinary boundaries. For science in general, this can lead to creative breakthroughs, the identification of oversights, and provide more holistic solutions than work within single disciplines (Nissani, 1997). For agroecology, the specific issue is its continued growth from its roots in agriculture and ecology to include relevant aspects in sociology and economics. This development is desirable both because humans are an integral and important part of food producing systems and because it is necessary if decision-makers are to act on the basis of both ecological, social and economic principles (Wood, 1998).
Achieving interdisciplinarity will require the removal of the barriers to the flow of information between the disciplines relevant to agroecology. These barriers include mind set and communication, where science has developed into increasingly specialised disciplines, talking different languages and having different areas of interest. Ideally, one could call upon scientists with a more generalist background to assist in communication between specialists but institutional barriers within modern science mean there is no encouragement for such creatures to flourish. These barriers are both physical and organisational. The physical barrier is that scientists from the different disciplines that interact with agroecology are normally in different institutes or departments, often in different physical locations. With the developments in information technologies and infrastructures this barrier may be less than it once was but the lack of social interaction will continue to be an obstacle to collaboration. The organizational barrier relates to the way in which science reward researchers via the provision of resources and career advancement. This depends heavily on the publication of papers in peer-reviewed scientific journals. Here the researchers within agroecology are faced with new opportunities but also several problems. The challenge of scaling is encouraging the development of novel experimental methodologies, e.g. through the combined use of modelling, observational science and advanced statistical mapping procedures (Riley, 2001). However, for the more reductionist scientists, integration to higher hierarchical levels, e.g. from the field to the farm level, means fewer opportunities for controlled experiments, experiments of greater duration and more time spent communicating with scientists from other disciplines. The experiments are likely to provide results that are more difficult to generalise than is common for more reductionist disciplines. However, as discussed earlier, the norms and social position of science are changing as the presumption that science is sacrosanct withers and scientists increasingly have to argue their value to human endeavour. As agriculture is a mature science, compared to other disciplines such as biotechnology or microelectronics, agroecologists may find themselves in the forefront of this development.

6. Procedures and strategies to address agroecological questions at different scales
It is argued here that studies of how to cross the barriers of scaling and interdisciplinarity should be central issues for future agroecological research projects. For soil sciences, Bouma (1997) drew similar conclusions and appealed for new methods to deal with the issues and proposed a seven-step procedure for research in sustainable management of agricultural soils (Bouma et al., 1998). This procedure was found useful to address agroecological questions (Wagenet, 1998) but problems were encountered in integrating socio-economic information and in the issues of hierarchy and scale (Dumanski et al., 1998). The framework presented here (Table 2) builds upon Bouma et al.'s (1998) procedure and corresponds to the application of the scientific method of natural sciences-observing, meassuring and interpreting - stressed in the introduction to the agroecology book by Carroll et al. (1990). The difference is that the framework presented here was extended to distinguish between hierarchy and scale in the form of the defined linear, non-linear or hierarchical scaling procedures. In addition, the present framework included a test (criteria 4) in the defined scaling procedure and an iterative exploration of scaling functions that proceeds until the error does not exceed a sensible threshold value (Table 2). To develop such scaling functions, comprehensive decision-support systems have been developed (Bierkens et al., 2000) but to date only methods to answer hard agroecological questions have been included, while the integration of information from the soft agroecological ones is not included. The framework of hierarchy and scale presented here is similarly bound to hard agroecology but provides a hint at a starting point for interdisciplinary research projects. This would typically be a common problem of sustainability, investigated by both soft and hard researchers, corresponding to the criteria l problem in the framework. However because the soft sciences do not produce pure quantitative results, it might not merely be a question of currency shifts - for example from the physical units of an agronomic study to the monetary units of an economical study (Squire and Gibson, 1998)- followed by the traditional sca1ing procedure of criteria 2-4. Instead, criteria 2-4 could be interpreted as a communicative process (Bawden, 1995), where statements regarding solutions to the common problem are compared at spatio-temporal scales of relevance for decisions.
One example where the resu1ts from soft and hard sciences differed, and interdisciplinary research would gain knowledge is illustrated in the recent debate concerning the possible benefits of introducing a "golden rice variety" (Schiermaier, 2001). In contrast to traditional varieties, the golden rice is genetically engineered to contain Vitamin A precursors, a deficiency of which causes blindness and other illnesses. One of the inventing plant scientists predicted the introduction of this new variety to solve the "unnecessary death and blindness of millions of poor every year" (Potrykus, 2001). However, nutritionists argue that Vitamin A can only be absorbed by the body if consumed with sufficient oils and fats, which is often lacking in many Third World diets, and socia1 scientists argue that some of the Vitamin A problem is caused by the social status of eating hulled, white rice, with a low Vitamin A content, and that better nutrition could as easily have been achieved by campaigning for the consumption of existing, more wholesome varieties of brown paddy rice (Schnapp and Schiermaier, 2001). This example illustrates that the potential effect of feeding malnourished people with golden rice varieties differs when seen from the perspective of the uni-disciplinary perspective of the plant scientist than from a perspective that also includes nutritional and social knowledge. Estimating consequences of conversion to organic farming is another subject where both ecological (Dalgaard et al., 1998), economic (Hansen et al., 2001), and social driving forces (Trewavas, 2001) are relevant to include. This is because conversion to organic farming is driven both by the ecological potential of this system compared to conventional fuming and by the socio-economic gains for farmers and the society.
A common feature of the problems encountered in the development of sustainable food production is the need for feedback mechanisms between the different research disciplines and between decision-makers and researchers. This is because the decisions to be made must take into account both the functioning of natural ecosystems and the response of humans acting either as individuals or as part of society. Consequently soft and hard science mechanisms interact with one another, and a dialectical approach (Levins and Lewontin, 1985), where both top-down and bottom-up viewpoints are valued, becomes fundamental for the iteratve integration of information from multiple disciplines.

7. Conclusions and perspectives
The current driving force for agroeco1ogy is the need to facilitate the development of more sustainable agricultural systems. This emphasis on sustainability is drawing agroecology up from its roots in agronomy and eco1ogy to include elements of both sociology and economics. This study found that agroecology can currently be defined as the study of interactions between plants, animals, humans and the environmemt within agricultural systems. One of its hallmarks is that it integrates between scientific disciplines and scales.
The first Green Revolution was achieved primarily through the development and application of technology. Whilst, successful in terms of food production, serious questions have been raised concerning the impact of these agricultural practices on the health of the cultivated land (Oldeman et al., 1991)_ Conway (1997) argued that a second Green Revolution is required. which is even more productive than the fist Green Revolution and even more "green" in terms of conserving natural resources and the environment. In addition to the productive and environmental aspects, the social and economic dimensions of agricu1tural systems must therefore also be considerad.
In recent years, significant progress has been made in understanding the issue of scaling and in the development of appropriate techniques. The barriers to interdisciplinarity are mainly cultural and political not technical, and lying deeply embedded in the way science has developed, these barriers present the major obstacle to the development of agroecology.

Acknowledgements
The authors wish to thank the ARLAS research project, and the Danish Research Councils for funding this work as part of the research project Agrar2000, from which this is paper no. 4. Tommy Dalgaard would like to thank Dr. Avaz Koocheki from Iran and Professors Steve Gliessman and Miguel Altieri from the USA for initial discussions during a study visit to The Agroecology Centre, University of California, Santa Cruz and to Chris Kjeldsen, Aalborg University, for his comments on the thesis presented in this paper. Finally, we thank the reviewers of the paper for their many useful comments, which helped significantly to improve our work.

References

[*] Corresponding author. Tel.: +45-8999-1732 - fax: +45-8999.1819.
E-mail address: tommy.dalgaard@agrsci.dk (T. Dalgaard).
URL: http://www.agrsci.dk/jbs/tda/TDAhomepage2002/tommy.html.
0167.8809/$ - see front matter © 2003 Elsevier B.V. All rights reserved.
doi: 10.1016/SO 167-8809(03)00l52-X