The problems of prediction and scale in applied ecology: the example of fire as a management tool


E-mail:; Tel. +44 (0) 1865 271274; Fax: +44 (0) 1865 271249


  • 1Predicting the large-scale consequences of novel management practices requires that we are able to accurately predict how such management will affect ecological landscapes based on short-term datasets. In this profile we highlight the lessons learned from four studies into the large-scale effects of fires as management tools.
  • 2Ecological interactions are key to understanding the effects of fires. Species respond differentially to processes such as competition and herbivory, as well as to fires. This means that community composition may change substantially with changing frequency of fires.
  • 3The studies highlighted in this profile reinforce the role of fire in modulating the frequency of invasion of alien species in a range of ecosystem types. In two of the studies, alien species were superior competitors to native ones in the absence of fires, but unable to regenerate following fires.
  • 4In the future, it will be necessary that the results of short-term experiments are integrated with long-term observational data and with theoretical models to predict long-term dynamics. We discuss how this may be done.
  • 5Synthesis and applications. The studies of the management of ecosystems using fires as a tool are prime examples of the application of ecology. In addition to measuring the proximate effects of such management using experiments, it is necessary to understand in detail the effects of fire on the underlying interactions between species, and feedbacks between different components of ecosystems.


The scope of the Journal of Applied Ecology is wide and includes the management of a huge range of systems, recently including studies on intensively managed agricultural systems (Grundy, Mead & Burston 2003; Wickramasinghe et al. 2003), through to monitoring (Brooks 2003; McNeil & Waddington 2003; Tessier & Raynal 2003), managing (Hirst et al. 2003) and restoring natural ones (Dorland et al. 2003; Pretty et al. 2003; Pywell et al. 2003; Smith et al. 2003). The range of approaches employed in the Journal is also extensive, including experimental (Le Duc, Pakeman & Marrs 2003), observational (Clarke et al. 2003) and modelling studies (Freckleton et al. 2003; Gremillet et al. 2003; Norris & McCulloch 2003; Rowcliffe, Cowlishaw & Long 2003; Stillman et al. 2003). The common problem that faces all of these studies is one of scale – the scales (temporal and spatial) at which we aim to predict the dynamics of ecological systems are far larger than those at which we can possibly observe them.

Solving the problem of scaling is especially difficult when trying to predict the effects of novel management practices. The only way to test effects is through an experimental approach. For example, the recent debate over the possible biodiversity consequences of GM crops has used a wide-scale experimental evaluation of the effects of these on biodiversity (Firbank et al. 2003; Perry et al. 2003) and this has fed directly into policy decisions (Gray 2004). The question, of course, is whether it is valid to scale up from the results of short-term studies of this kind.

A second complication with scaling from short-term studies to predict the consequences of novel management at landscape scales is that ecological interactions play a key role in determining the structure of many systems. This means that it is not enough to understand the proximate effects of novel management. The way that management interacts with other components of the system also has to be understood. Moreover, in the long term feedback loops exist between components (Augustine 2003), and these also need to be understood.

In this Special Profile we are highlighting four papers that deal with the use of fire as a management tool in ecology (Fuhlendorf & Engle 2004; Gillespie & Allen 2004; Parr et al. 2004; Paynter & Flanagan 2004). What these four papers have in common is that they show that the effects of fire as a management tool can be predicted only when the interaction between fire and other ecological processes are understood. These interactions are of key importance to the problem of scaling from experimental studies to landscape scale management. We highlight that in terms of scaling a key element for success lies in accurately characterizing the impacts of novel management on all elements of a system, not just the net outcome of management.

fires and ecosystem structure

Fires can be both a constructive and destructive component of an ecosystem. The short-term destructive power of fires is well known, and annually fires wreak havoc over huge areas in North America, the Amazon, Asian rainforests, Australia, and recently in Europe (Moretti et al. 2002). However, fire can be an extremely potent force in structuring ecosystems, and indeed has long been used as a tool in ecosystem management, to the point that the structure of ecological communities has become reliant on anthropogenic fires. In North America, for instance, fires are far less frequent than they were two centuries ago, and this has had important knock-on effects for plant and animal communities (e.g. Gallant et al. 2003; Whitlock, Shafer & Marlon 2003).

Two of the papers in this profile deal with the effects of fire on North American grassland ecosystems, although with slightly different perspectives. Gillespie & Allen (2004) consider the effects of fire on restoration of grassland following invasions by aliens. The key driving forces in this system are (i) differential regeneration of native and alien species following fires, and (ii) differential competitive ability of native and alien species. When fires are rare, alien species are able to outcompete the native species. This leads to progressive elimination of the natives. However, fires can prevent seed set by invasive species, whereas the natives are able to set seed, and the seed survives to germinate the following year. This means that the frequency of native species increases as the frequency of fires increases.

The paper by Fuhlendorf & Engle (2004) takes a different perspective on a related system, the Great Plains of North America. In the system they look at there is the additional complication of grazing: grazing animals prefer to feed on recently burned areas. This leads to a high intensity of grazing with the consequence that plant productivity is depressed in these areas, fuel loads are hence lower, and further burning is unlikely. Also, there are effects on species composition: particularly the frequency of large grass species is reduced. Some of these species are exotic so, as in the previous example, increasing the frequency of fires tended to reduce the abundance of alien invasive species. The most important conclusion from this study is that the landscape has to be maintained as a mosaic of burned and unburned patches: unburned patches are areas that will burn in the future and provide future grazing.

The third paper in this profile also considers the interaction between fire and trophic interactions, in this case how fire interacts with biocontrol to reduce the density of a problem weed (Paynter & Flanagan 2004). Conventional wisdom is that the effectiveness of biocontrol agents is reduced when other control measures are applied. However, the experiment conducted by Paynter & Flanagan suggest that this is not the case, and that a combination of control measures, including fire, can be more effective than when applied singly. Again, the main conclusion is that the effects of fire interact with other ecological factors, and do not simply act in addition to them.

The final paper in the Special Profile (Parr et al. 2004) explores some of the wider ecosystem consequences of using fire as a management tool in southern African savannas. Specifically, whilst many experimental studies of the use of fire as a management tool focus on the management of specific desirable or undesirable species, Parr et al. (2004) look at the effects on non-target species. In the system they study, fire is used to manage vegetation composition. As they point out, the management used is based on the results of short-term, small-scale experiments, but applied over wide areas for long time periods. Scaling is thus the prime focus of this work. Parr et al. (2004) show that in their system the ant communities appear to be highly resilient to the effects of fire, and there is an interaction between the effects of burning and rainfall. The detailed nature of the burn regime was not important compared to whether a burn had been applied or not.

interactions and interacting effects

In terms of scaling the results of short-term studies to larger spatial and temporal scales, each of these four studies shows that the consequences of fires have to be understood in terms of the way that they interact with other ecological processes. In one sense fires can be thought of as a form of disturbance, and it has long been known that disturbances can promote biodiversity through a range of mechanisms (Connell 1978; Nee & May 1992). Moreover, the issue of scale is important since the short-term effects of disturbances may take many generations to become manifest (Tilman et al. 1994). The implication from these four experiments is that when used as a management tool, fire cannot be thought of simply as being a disturbance, but that differential effects of fire on different species are important, as well as the response of herbivores and the spatial distributions of fires.

Such interacting effects are crucial because when the effects of various components of a system do not combine in a simple additive manner, the possibility exists for systems to show alternative stable states (May 1977). This is undoubtedly possible with systems in which fire plays an important role. For example, van Langevelde et al. (2003) suggested that in savanna ecosystems the interactions of fire and grazing could lead to alternative stable states. Therefore, understanding interacting effects of this sort is likely to be important not only in predicting the effects of fire as a management tool, but also in understanding ecosystem level dynamics.

spatial dependency

The paper by Fuhlendorf & Engle (2004) highlights that spatial effects are of key importance when considering the interaction between fire and grazing in their system. Similarly, Akcakaya et al. (2004) recently found in a spatially explicit landscape simulation of the dynamics of the sharp-tailed grouse Tympanuchus phasianellus that it was impossible to accurately predict the consequences of changing land management (including fire) of this species without taking into account the details of landscape dynamics. However, the landscape processes on their own could not accurately predict large-scale population changes, and the details of local dynamics were also important. Thus, the common message from both of these recent studies is that the effects of fire as a management tool cannot be predicted without considering the details of spatial dynamics.

future directions and other approaches

Experimental approaches have the advantage that they are able to examine the effects of different factors controlling for unwanted confounding variation. The disadvantages of experiments in ecology are that they are conducted over short time periods and that it is dangerous to simply extrapolate results (Freckleton & Watkinson 2001). This is an important limitation when attempting to scale up from the results of short-term studies of this sort. This criticism is partly overcome in experiments of the kind described here: if an experiment has looked at the effect of management changes on the range of components of a given system, then extrapolation may be reasonably justified since it can be argued that the underlying structural consequences of changing management are well understood. On the other hand, if just the net consequence of management had been measured, for instance that fire led to an x% change in the abundance of a given species, it would not be justifiable to simply extrapolate this change into the future.

As an alternative to the experimental approach, other methods have also been used to predict the effects of fire in management. As noted in the introduction, these may be coarsely classified into two groups, observational studies and modelling analyses. Observational studies use long-term or large-scale data in combination with statistical analysis to infer the long-term consequences of management. There are numerous such studies published in this journal in recent years (e.g. Arseneault et al. 1997; Russell-Smith et al. 1997; Russell-Smith et al. 1998; Roques et al. 2001; Haskins & Gehring 2004). Studies employing long-term data reconstruct the dynamical implications of fire for communities. For example, Roques et al. (2001) showed how the effects of grazing, rainfall and fire interacted with the process of population regulation to determine rates of shrub encroachment. Many such studies take advantage of the increasing availability of Geographical Informatiom Systems (see also Perry, Sparrow & Owens 1999; Read et al. 2003), and it seems likely that links between GIS and ecological management have much to offer in addressing the problem of scaling in ecology.

Modelling studies also attempt to address the issue of scaling, through building up simulations of natural systems, and then exploring the possible range of impacts of new management upon them. Ecological modelling has been applied to a number of problems published in this journal, such as predicting the likelihood of invasion of non-native vegetation as a function of fire frequency (Rees & Hill 2001), and in designing optimal management programmes using stochastic dynamic programming (McCarthy et al. 2001). Modelling studies are less common than other approaches. One reason may be the problem of model validation: if model projections deal with novel environmental conditions then of course the accuracy of predictions cannot be evaluated. However, whilst this limitation is frequently thought of as being a problem with modelling studies, this problem really applies to any prediction of novel management, irrespective of how the prediction is derived. As pointed out above, for example, the results of experimental studies cannot simply be extrapolated to larger spatial and temporal scales. Similarly, observational studies are essentially correlative and as such may not always be safely used to infer causation. Ultimately, the same criticism can be levelled at any of these methods: it seems likely that in the future the key to robust ecological forecasting will lie in integrating these approaches.

synthesis and applications

The catalyst for this profile is a quartet of papers in this issue on the use of fire as a land management tool. The Journal has published numerous studies on the ecological effects of fire over the past few years, and this constitutes an important and diverse body of work. One aim in this editorial review was to highlight the aspects of these papers, and a number of others published recently, that we believe defines the role of ecology in applications to real-world problems. In these papers fire is not treated simply as a factorial management option. Instead, the authors have attempted to evaluate the effects of fire at the same time as examining the ecology underlying their effects. It is this that sets applied ecology apart from other applied disciplines, and this is a key part of our philosophy in defining the scope of the journal.


I should like to thank Steve Rushton and Gill Kerby for their comments and suggestions on an earlier draft.