Letter
How to Maintain Ecological Relevance in Ecology Roy H.A. van Grunsven1,* and Maartje Liefting2 Ecology has developed from a descriptive science to a scientific field that has fostered general theories such as island theory and optimal foraging. It has become a hypothesis-driven science with scientists around the world working on the same body of theory. Although this has been fruitful, it also poses an increasing challenge for the reviewing process. This is due to not only the growing number of manuscripts and the increasing complexity of experimental design and statistical analyses [1,2], but also the expanding number of species and ecosystems that are studied. Sufficient knowledge of the studied species and ecosystems is often essential to assess the ecological relevance and scientific merit of the paper. Drawing ecological-evolutionary conclusions from studies on species interactions is at the heart of ecology and a solid knowledge of the ecology of the studied species is essential. This is of particular importance in competition experiments because they only have ecological relevance if the species tested (can) co-occur in natural situations and are tested under the conditions in which they (can) cooccur. However, in his bibliography on competition in plant communities [3], Paul Keddy aptly noted that ‘One of the most difficult tasks in reading the literature is sifting through large numbers of experiments in which investigators have haphazardly selected (a pair of) species and grown them in mixture, without adequately justifying their choice of species and study design.’
interaction between plants and insects, adaptation and coevolution are often wrongfully assumed. For example, the model system of Arabidopsis thaliana and Pieris rapae is commonly used to study the interaction between plants and their specialist herbivores. However, P. rapae feeds on a large number of plant species and is irrelevant as a herbivore of A. thaliana because there is a phenological mismatch; A. thaliana grows in winter and early spring, while P. rapae emerges in May [4]. This model may still be suitable to test physiological responses, but interpreting them in an ecological or evolutionary context, assuming coevolution, does not make sense (see [4] for examples). Other examples of fundamental errors cover studies on trophic interactions between species from different continents, tests of scenarios that do not occur under natural conditions, or use of single genotypes (see [5] for examples). Even though the hypotheses might be tested correctly, this does not contribute to insight into the interactions among organisms and their environment in the natural world. Our aim here is not to review ecologically faulty studies but to point out that this is a general issue spanning different research fields. Although research that lacks ecological relevance is a waste of resources, the consequences can reach much further. In a recent study, /-diversity was used to evaluate the success of ecological restoration. The authors concluded that a dragonfly community could be restored within 3 years after rewetting of a drained peatland [6]. However, when community composition is considered, and not merely species number, it becomes clear that most, if not all, habitat specialists have not returned. This puts the conclusions in a very different perspective. There is an obvious irony in studies that claim to consider conservation a priority, while paving the way for dubious mitigation of anthropogenic impact.
However, such issues are not limited to These examples from different fields of competition studies; in studies on the ecology show that poor ecological
knowledge can lead to studies that lack ecological relevance. In any study on the interaction between species or the response of an organism to environmental variation, it is essential that this represents a realistic and suitable scenario. This issue is not likely to resolve itself, especially because of the expanding number of species and ecosystems that are studied. To begin with, we unequivocally stress the importance of knowing your study system, including basic ecology and taxonomy, and this should be reflected in teaching [7,8]. The primary responsibility lies with the researchers, with the review process functioning as a quality control. However, it is not realistic to assume that reviewers or editors have sufficient knowledge of the species studied and their ecology. This means that such problems of oversight are currently not intercepted in the review process. We propose that editors ask reviewers to flag possible issues on ecological relevance and acknowledge when this is beyond their expertise, as is customary with complex statistics. When this is the case, the editor can ask the authors for clarification or consult a specialist with sufficient knowledge of the species concerned. Similarly, such best practice would be beneficial during the evaluation of grant proposals. We hope that more stringent quality control on the ecological realism of papers will force researchers to think more explicitly about the ecological relevance of the hypotheses that they are testing. By doing this, we will not only be answering questions correctly, but also answer the correct questions in ecology. 1
Nature Conservation and Plant Ecology, Wageningen
University and Research Centre, Wageningen, The Netherlands 2 Animal Ecology, VU University Amsterdam, Amsterdam, The Netherlands *Correspondence: [email protected] (van Grunsven, Roy H.A.). http://dx.doi.org/10.1016/j.tree.2015.07.010 References 1. von Wehrden, H. et al. (2015) A call for statistical editors in ecology. Trends Ecol. Evol. 20, 1–2 2. Székely, T. et al. (2014) Errors in science: the role of reviewers. Trends Ecol. Evol. 29, 371–373
Trends in Ecology & Evolution, October 2015, Vol. 30, No. 10
563
3. Keddy, P.A. (2012) Competition in Plant Communities, Oxford University Press 4. Harvey, J.A. et al. (2007) Nutritional suitability and ecological relevance of Arabidopsis thaliana and Brassica oleracea as foodplants for the cabbage butterfly Pieris rapae. Plant Ecol. 189, 117–126 5. Harvey, J.A. et al. (2015) Integrating more biological and ecological realism into studies of multitrophic interactions. Ecol. Entomol. 40, 349–352 6. Elo, M. et al. (2015) The effect of peatland drainage and restoration on Odonata species richness and abundance. BMC Ecol. 15, 1–8 7. Bortolus, A. (2008) Error cascades in the biological sciences: the unwanted consequences of using bad taxonomy in ecology. AMBIO 37, 114–118 8. Warren, J. et al. (2015) Save field biology skills from extinction risk. Times Higher Education 26 February
Letter
Spatial Sorting Unlikely to Promote Maladaptive Hybridization: Response to Lowe, Muhlfeld, and Allendorf Ben L. Phillips1,* and Stuart J.E. Baird2 Lowe et al. [1] suggest there is a paradox afoot: how can hybrid invasion occur when hybrids are less fit? They then make the fascinating argument that spatial sorting may resolve this paradox by accelerating hybrid invasion. We question both arms of this argument. Is there really a paradox and, if so, can spatial sorting resolve it?
Where Is the Paradox? Lowe et al. claim that hybrid invasion despite lowered hybrid fitness (outbreeding depression) is paradoxical. There is no paradox. It is well understood that when outbreeding depression is caused by genetic incompatibilities, hybrid zones will move, following density gradients downhill [2]. Thus, if the invasive taxon achieves higher density than the native taxon (as many invasive species do), the invasive genome will
564
tend to invade the range of the native genome. This will happen regardless of levels of outbreeding depression; in the extreme case, where F1s are infertile (i.e., there is hybridization but no introgression), the process becomes competitive exclusion – the invader simply outcompetes the native, despite outbreeding depression. Moving ‘tension zones’ of hybridization will similarly track dispersal differentials downhill; if the invasive species is, on average, more dispersive (as many invasive species are), the hybrid zone again tends to invade the native species. So, for reasons that are common in nature, invasion will occur despite lowered hybrid fitness. Lowe et al. muddy their argument by conflating introgression with hybridization. Hybridization is when individuals of different taxa mate and produce offspring; introgression is the movement of DNA into a new genetic background. This distinction matters. Lowe et al. define all offspring of hybrids as themselves hybrids, conjuring the ‘dark power of the genomic ratchet’ to make hybrid invasion inevitable. This is sensational, and wrong. In the circumstances imagined for this ‘ratchet’ all traces of an invading genomic region are, in fact, certain to (eventually) be lost [3]. Backcrossing F1 hybrids for g generations against a ‘pure’ native type, the proportion of ‘invader’ genome declines as 1/2g. After 20 generations (when less than one millionth of the genome is of invader origin and presumably outbreeding depression has ceased to be an issue), Lowe et al. are still happy to call the resultant individuals hybrids and a conservation problem. In fact, the invader genome is now represented by only tiny blocks that are easily lost during ploidy reduction or crossing over at meiosis. These blocks are, however, distributed over multiple individuals – slowing the loss of any trace of invasion at the population level.
‘accelerate’ invasive hybridization. We point out above that spatial sorting is not necessary to explain the drive part; will it accelerate invasive hybridization? We think not. Although Lowe et al. explain spatial sorting quite well, they fail to appreciate that it occurs only across density gradients [4]. Highly dispersive individuals will be over-represented at the low-density end of a gradient and because of this accumulation in space we see assortative mating by dispersal ability. Thus range edges (both stable and moving) will see spatial sorting, as will metapopulations in which density differentials emerge. Lowe et al. make the mistake of conflating density gradients (a necessary condition for spatial sorting) with trait gradients (such as clines in gene frequency). A trait gradient alone will not cause assortative mating by dispersal ability, so spatial sorting will simply not play out across trait gradients unless these gradients are also associated with a density gradient. Linked trait/density gradients may exist in cases of hybrid invasion, but it is important to be clear that it is the density gradient, not the trait gradient, that needs to be watched.
In making their argument for accelerated hybrid invasion, Lowe et al. also fail to appreciate that spatial sorting can lead to both increases and decreases in dispersal. If the hybrid zone itself causes (or is trapped by) a region of low density, populations abutting the hybrid zone will see a net loss of dispersive genotypes (dispersers leave, and because they landed in a hybrid zone and suffer lower fitness their genes tend not to come back). Thus, spatial sorting may well lead to lower mean dispersal rates around hybrid zones. Spatial sorting will of course cause more dispersive genotypes to accumulate in the middle of the hybrid zone, but these more dispersive genotypes will be sampled from the already low-dispersal phenotypes around the hybrid zone. Thus, the likely situation is one in which the dispersal rate around Spatial Sorting and Accelerated hybrid zones evolves to be lower than Introgression Lowe et al. switch between arguing that average. Given that this dispersal rate will spatial sorting will ‘drive’ and will affect the rate at which the zone moves,
Trends in Ecology & Evolution, October 2015, Vol. 30, No. 10
How to Maintain Ecological Relevance in Ecology Roy H.A. van Grunsven1,* and Maartje Liefting2 Ecology has developed from a descriptive science to a scientific field that has fostered general theories such as island theory and optimal foraging. It has become a hypothesis-driven science with scientists around the world working on the same body of theory. Although this has been fruitful, it also poses an increasing challenge for the reviewing process. This is due to not only the growing number of manuscripts and the increasing complexity of experimental design and statistical analyses [1,2], but also the expanding number of species and ecosystems that are studied. Sufficient knowledge of the studied species and ecosystems is often essential to assess the ecological relevance and scientific merit of the paper. Drawing ecological-evolutionary conclusions from studies on species interactions is at the heart of ecology and a solid knowledge of the ecology of the studied species is essential. This is of particular importance in competition experiments because they only have ecological relevance if the species tested (can) co-occur in natural situations and are tested under the conditions in which they (can) cooccur. However, in his bibliography on competition in plant communities [3], Paul Keddy aptly noted that ‘One of the most difficult tasks in reading the literature is sifting through large numbers of experiments in which investigators have haphazardly selected (a pair of) species and grown them in mixture, without adequately justifying their choice of species and study design.’
interaction between plants and insects, adaptation and coevolution are often wrongfully assumed. For example, the model system of Arabidopsis thaliana and Pieris rapae is commonly used to study the interaction between plants and their specialist herbivores. However, P. rapae feeds on a large number of plant species and is irrelevant as a herbivore of A. thaliana because there is a phenological mismatch; A. thaliana grows in winter and early spring, while P. rapae emerges in May [4]. This model may still be suitable to test physiological responses, but interpreting them in an ecological or evolutionary context, assuming coevolution, does not make sense (see [4] for examples). Other examples of fundamental errors cover studies on trophic interactions between species from different continents, tests of scenarios that do not occur under natural conditions, or use of single genotypes (see [5] for examples). Even though the hypotheses might be tested correctly, this does not contribute to insight into the interactions among organisms and their environment in the natural world. Our aim here is not to review ecologically faulty studies but to point out that this is a general issue spanning different research fields. Although research that lacks ecological relevance is a waste of resources, the consequences can reach much further. In a recent study, /-diversity was used to evaluate the success of ecological restoration. The authors concluded that a dragonfly community could be restored within 3 years after rewetting of a drained peatland [6]. However, when community composition is considered, and not merely species number, it becomes clear that most, if not all, habitat specialists have not returned. This puts the conclusions in a very different perspective. There is an obvious irony in studies that claim to consider conservation a priority, while paving the way for dubious mitigation of anthropogenic impact.
However, such issues are not limited to These examples from different fields of competition studies; in studies on the ecology show that poor ecological
knowledge can lead to studies that lack ecological relevance. In any study on the interaction between species or the response of an organism to environmental variation, it is essential that this represents a realistic and suitable scenario. This issue is not likely to resolve itself, especially because of the expanding number of species and ecosystems that are studied. To begin with, we unequivocally stress the importance of knowing your study system, including basic ecology and taxonomy, and this should be reflected in teaching [7,8]. The primary responsibility lies with the researchers, with the review process functioning as a quality control. However, it is not realistic to assume that reviewers or editors have sufficient knowledge of the species studied and their ecology. This means that such problems of oversight are currently not intercepted in the review process. We propose that editors ask reviewers to flag possible issues on ecological relevance and acknowledge when this is beyond their expertise, as is customary with complex statistics. When this is the case, the editor can ask the authors for clarification or consult a specialist with sufficient knowledge of the species concerned. Similarly, such best practice would be beneficial during the evaluation of grant proposals. We hope that more stringent quality control on the ecological realism of papers will force researchers to think more explicitly about the ecological relevance of the hypotheses that they are testing. By doing this, we will not only be answering questions correctly, but also answer the correct questions in ecology. 1
Nature Conservation and Plant Ecology, Wageningen
University and Research Centre, Wageningen, The Netherlands 2 Animal Ecology, VU University Amsterdam, Amsterdam, The Netherlands *Correspondence: [email protected] (van Grunsven, Roy H.A.). http://dx.doi.org/10.1016/j.tree.2015.07.010 References 1. von Wehrden, H. et al. (2015) A call for statistical editors in ecology. Trends Ecol. Evol. 20, 1–2 2. Székely, T. et al. (2014) Errors in science: the role of reviewers. Trends Ecol. Evol. 29, 371–373
Trends in Ecology & Evolution, October 2015, Vol. 30, No. 10
563
3. Keddy, P.A. (2012) Competition in Plant Communities, Oxford University Press 4. Harvey, J.A. et al. (2007) Nutritional suitability and ecological relevance of Arabidopsis thaliana and Brassica oleracea as foodplants for the cabbage butterfly Pieris rapae. Plant Ecol. 189, 117–126 5. Harvey, J.A. et al. (2015) Integrating more biological and ecological realism into studies of multitrophic interactions. Ecol. Entomol. 40, 349–352 6. Elo, M. et al. (2015) The effect of peatland drainage and restoration on Odonata species richness and abundance. BMC Ecol. 15, 1–8 7. Bortolus, A. (2008) Error cascades in the biological sciences: the unwanted consequences of using bad taxonomy in ecology. AMBIO 37, 114–118 8. Warren, J. et al. (2015) Save field biology skills from extinction risk. Times Higher Education 26 February
Letter
Spatial Sorting Unlikely to Promote Maladaptive Hybridization: Response to Lowe, Muhlfeld, and Allendorf Ben L. Phillips1,* and Stuart J.E. Baird2 Lowe et al. [1] suggest there is a paradox afoot: how can hybrid invasion occur when hybrids are less fit? They then make the fascinating argument that spatial sorting may resolve this paradox by accelerating hybrid invasion. We question both arms of this argument. Is there really a paradox and, if so, can spatial sorting resolve it?
Where Is the Paradox? Lowe et al. claim that hybrid invasion despite lowered hybrid fitness (outbreeding depression) is paradoxical. There is no paradox. It is well understood that when outbreeding depression is caused by genetic incompatibilities, hybrid zones will move, following density gradients downhill [2]. Thus, if the invasive taxon achieves higher density than the native taxon (as many invasive species do), the invasive genome will
564
tend to invade the range of the native genome. This will happen regardless of levels of outbreeding depression; in the extreme case, where F1s are infertile (i.e., there is hybridization but no introgression), the process becomes competitive exclusion – the invader simply outcompetes the native, despite outbreeding depression. Moving ‘tension zones’ of hybridization will similarly track dispersal differentials downhill; if the invasive species is, on average, more dispersive (as many invasive species are), the hybrid zone again tends to invade the native species. So, for reasons that are common in nature, invasion will occur despite lowered hybrid fitness. Lowe et al. muddy their argument by conflating introgression with hybridization. Hybridization is when individuals of different taxa mate and produce offspring; introgression is the movement of DNA into a new genetic background. This distinction matters. Lowe et al. define all offspring of hybrids as themselves hybrids, conjuring the ‘dark power of the genomic ratchet’ to make hybrid invasion inevitable. This is sensational, and wrong. In the circumstances imagined for this ‘ratchet’ all traces of an invading genomic region are, in fact, certain to (eventually) be lost [3]. Backcrossing F1 hybrids for g generations against a ‘pure’ native type, the proportion of ‘invader’ genome declines as 1/2g. After 20 generations (when less than one millionth of the genome is of invader origin and presumably outbreeding depression has ceased to be an issue), Lowe et al. are still happy to call the resultant individuals hybrids and a conservation problem. In fact, the invader genome is now represented by only tiny blocks that are easily lost during ploidy reduction or crossing over at meiosis. These blocks are, however, distributed over multiple individuals – slowing the loss of any trace of invasion at the population level.
‘accelerate’ invasive hybridization. We point out above that spatial sorting is not necessary to explain the drive part; will it accelerate invasive hybridization? We think not. Although Lowe et al. explain spatial sorting quite well, they fail to appreciate that it occurs only across density gradients [4]. Highly dispersive individuals will be over-represented at the low-density end of a gradient and because of this accumulation in space we see assortative mating by dispersal ability. Thus range edges (both stable and moving) will see spatial sorting, as will metapopulations in which density differentials emerge. Lowe et al. make the mistake of conflating density gradients (a necessary condition for spatial sorting) with trait gradients (such as clines in gene frequency). A trait gradient alone will not cause assortative mating by dispersal ability, so spatial sorting will simply not play out across trait gradients unless these gradients are also associated with a density gradient. Linked trait/density gradients may exist in cases of hybrid invasion, but it is important to be clear that it is the density gradient, not the trait gradient, that needs to be watched.
In making their argument for accelerated hybrid invasion, Lowe et al. also fail to appreciate that spatial sorting can lead to both increases and decreases in dispersal. If the hybrid zone itself causes (or is trapped by) a region of low density, populations abutting the hybrid zone will see a net loss of dispersive genotypes (dispersers leave, and because they landed in a hybrid zone and suffer lower fitness their genes tend not to come back). Thus, spatial sorting may well lead to lower mean dispersal rates around hybrid zones. Spatial sorting will of course cause more dispersive genotypes to accumulate in the middle of the hybrid zone, but these more dispersive genotypes will be sampled from the already low-dispersal phenotypes around the hybrid zone. Thus, the likely situation is one in which the dispersal rate around Spatial Sorting and Accelerated hybrid zones evolves to be lower than Introgression Lowe et al. switch between arguing that average. Given that this dispersal rate will spatial sorting will ‘drive’ and will affect the rate at which the zone moves,
Trends in Ecology & Evolution, October 2015, Vol. 30, No. 10
