Landscape Ecology

, Volume 25, Issue 8, pp 1289–1298

Population dynamics under increasing environmental variability: implications of climate change for ecological network design criteria

Authors

    • AlterraWageningen UR
  • Peter Schippers
    • AlterraWageningen UR
  • Anouk Cormont
    • AlterraWageningen UR
  • Marjolein Sterk
    • AlterraWageningen UR
  • Claire C. Vos
    • AlterraWageningen UR
  • Paul F. M. Opdam
    • AlterraWageningen UR
Research Article

DOI: 10.1007/s10980-010-9497-7

Cite this article as:
Verboom, J., Schippers, P., Cormont, A. et al. Landscape Ecol (2010) 25: 1289. doi:10.1007/s10980-010-9497-7

Abstract

There is growing evidence that climate change causes an increase in variation in conditions for plant and animal populations. This increase in variation, e.g. amplified inter-annual variability in temperature and rainfall has population dynamical consequences because it raises the variation in vital demographic rates (survival, reproduction) in these populations. In turn, this amplified environmental variability enlarges population extinction risk. This paper demonstrates that currently used nature conservation policies, principles, and generic and specific design criteria have to be adapted to these new insights. A simulation shows that an increase in variation in vital demographic rates can be compensated for by increasing patch size. A small, short-lived bird species like a warbler that is highly sensitive to environmental fluctuations needs more area for compensation than a large, long-lived bird species like a Bittern. We explore the conservation problems that would arise if patches or reserve sizes would need to be increased, e.g. doubled, in order to compensate for increase in environmental variability. This issue has serious consequences for nature policy when targets are not met, and asks for new design criteria.

Keywords

Climate changeEnvironmental variabilityPopulation viability analysisKey patchExtinctionDesign criteria

Introduction

Climate change will alter both the ecological functioning of landscapes and the public demands for landscape services. This necessitates societal responses. An important issue is the necessity of new paradigms, policies and criteria for ecological network assessment and design (Pressey et al. 2007; Nassauer and Opdam 2008; Opdam et al. 2009). Three interconnected aspects at interacting scales matter. Locally, the hydrology of landscapes may change due to changes in precipitation and temperature regimes, affecting plants and animals living there (Bowers and Harris 1994; Thomas and Lennon 1999; Warren et al. 2001; Hill et al. 2002). At the biogeographic scale, climate change has been correlated to observed changes in species distributions (Parmesan et al. 1999; Thomas and Lennon 1999; Hill et al. 2002; Parmesan and Yohe 2003; Hickling et al. 2006), whereas the impacts of more frequent extreme weather events (Easterling et al. 2000; IPCC 2007) have received little attention (McLaughlin et al. 2002). At the ecological network or metapopulation scale, it is expected that trends in temperature and precipitation play a role, but also does the increased unpredictability of the environment, causing extra stochasticity in population dynamics which leads to elevated extinction risk (Shaffer 1987; Soulé 1987). Obviously, this is expected to aggravate the impacts of habitat fragmentation (e.g. Opdam and Wascher 2004; Vos et al. 2008), causing an increase of local extinction events and disturbing the composition of species communities (Fraterrigo et al. 2009), especially in landscapes which are intensively used by humans. Potentially, such impacts affect the provision of ecosystem services (Diaz et al. 2006). While processes at all three scales call for action, in this paper we focus at the intermediate scale of functional ecological networks (sensu Verboom and Pouwels 2004) and the role of environmental stochasticity or variability. We assume that society will understand the need for adapting conservation strategies and measures (Opdam et al. 2009), and consider how scientific tools have to be adjusted to include this new knowledge.

Improved understanding of the spatial interactions across landscapes stimulated a shift in the focus of strategies of nature conservation from individual sites to ecosystem networks (Jongman and Pungetti 2004). This shift in focus was paralleled by the development of impact assessment tools (Verboom and Pouwels 2004; Mortberg et al. 2007) and design-oriented tools (Opdam and Steingröver 2008; Gordon et al. 2009). These tools, based on metapopulation theory, facilitate decision making in nature conservation policy and its implementation at regional and local scales. The ecological network paradigm has led to a focus on large, interconnected protected areas at various scales such as the Natura 2000 network in Europe and the National Ecological Network in the Netherlands (Jongman and Pungetti 2004). A major challenge in developing network assessment tools was how to incorporate population viability theory (Soulé 1987) in a multi-species based tool that can be handled in spatially explicit decision making on land pattern change. Here, we use the key patch approach as a population viability assessment (PVA) tool, derived from metapopulation theory (Verboom et al. 2001; Verboom and Pouwels 2004) and based on the relationship between habitat network pattern and population processes ( Vos et al. 2001; Opdam et al. 2003). Key patches are large patches with a stabilizing role in habitat networks. Habitat networks that contain a key patch require less total network size, compared to habitat networks that consists of smaller patches only (Verboom et al. 2001), and key patches may serve as stepping stones for species with shifting ranges (Vos et al. 2008).The key patch approach was used successfully in the Netherlands and Europe in assessment studies of terrestrial biodiversity (Reijnen et al. 1995; Foppen et al. 1999; Groot Bruinderink et al. 2003; Snep and Ottburg 2008) and conservation oriented landscape design (Van Rooij et al. 2003; Vos et al. 2007) and is a paradigm in spatial planning in The Netherlands.

The key patch approach however is based on 20th century data on population dynamics. Hence, when applied in the context of climate change—specifically, increased weather variability such as an increased inter-annual variation in temperature and rainfall—the key patch approach, and all other population viability assessment methods (Soulé 1987; Lande 1988; Remmert 1994; Traill et al. 2007), possibly overestimate the viability of (meta)populations and the sustainability of networks. We hypothesize that survival in habitat networks under climate change will require larger patches, holding larger populations, to compensate for the increase in environmental stochasticity.

Therefore, we propose that both for assessment and design tools, the current basis of population viability analysis have to be reconsidered.

In this paper, we explore the need for new paradigms, policies and landscape design criteria, both from a scientific and a societal point of view. We review the current minimum area requirements calculations of the key patch approach in the light of increased environmental variability, and present modeling evidence that PVA design criteria are indeed sensitive to assumptions about this variation. We then discuss consequences and knowledge gaps.

Conceptual overview

The key patch approach to ecological network assessment

Methods that assign biodiversity ratings to scenarios on the basis of landscape pattern are a valuable tool for landscape assessment and design (Opdam et al. 2003). While different methods were developed for this purpose in different parts of the world, the key patch approach (Verboom et al. 2001; Verboom and Pouwels 2004) has been found suitable for terrestrial animal species in human-dominated landscapes with islands of natural habitat and metapopulation type species dynamics (Vermaat et al. 2008). We use the key patch approach here as an example. Our conclusions apply to other assessment tools, such as the general rules-of-thumb of Thomas (1990) and Jones and Diamond (1976).

In order to find the relationship between landscape structure and biodiversity potential, one has to scale down to individual species and their requirements (Opdam et al. 2003). In other words, the landscape is analyzed from a species perspective. This yields so-called ‘ecologically scaled landscape indices’ (ESLI, Vos et al. 2001). Instead of using real species, such as policy target species, one can use so-called ‘ecological profiles’, hypothetical species with certain area requirements and dispersal characteristics, and habitat preferences, representative for a group of species sharing these requirements, characteristics and preferences (Vos et al. 2001). The landscape is then treated as a set of specific habitat patches and ecological networks and their ecologically scaled indices are compared to specific criteria for sustainable networks (Verboom et al. 2001; Verboom and Pouwels 2004). Key patches (Verboom et al. 2001) play an important role in determining whether ecological networks are sustainable. Eventually, from sustainable habitat networks for specific species (e.g. target species) one can scale up to landscape metrics for biodiversity (Opdam et al. 2003). This approach of using ecologically scaled landscape indices, the key patch approach, and landscape cohesion assessments is used in the nature policy process for assessments, scenario study and ecological network design in the Netherlands (resulting in the National Ecological Network) and other European countries (Van der Sluis et al. 2003; Van der Sluis and Pedroli 2004).

Variation, extremes and empirical evidence of their ecological impacts

Climatic models project that increasing atmospheric concentrations of greenhouse gases will result in changes in daily, seasonal, inter-annual, and decadal variability (IPCC 2001, 2007). Global warming is likely to lead to greater extremes of drying and heavy rainfall and increase the risk of droughts and floods in many different regions. The duration, location, frequency, and intensity of extreme weather and climate events are likely to change, and would result in mostly adverse impacts on biophysical systems. Climate change is likely to result in changes in the variance and frequency of extremes of climatic variables (see Fig. 1). Increases in mean temperature and variance together will lead to increases in frequency of heatwaves, with fewer frost days and cold days. Overall, there will be less change for cold weather compared to warm weather. The amplitude and frequency of extreme precipitation events is expected to increase over many areas and the return period for extreme precipitation events are projected to decrease (IPCC 2001, 2007). In several regions of the world, indications of changes in heatwaves, droughts and floods have been observed (IPCC 2001). Secondary effects of extreme weather events include altered fire and flooding regimes.
https://static-content.springer.com/image/art%3A10.1007%2Fs10980-010-9497-7/MediaObjects/10980_2010_9497_Fig1_HTML.gif
Fig. 1

Schematic diagram showing the effects on extreme temperatures when the mean increases, leading to more record hot weather, and also the variance increases, leading to much more record hot weather (drawn after IPCC 2001). For precipitation, it is unclear whether the mean will increase or decrease; this will differ from place to place. An increase in the occurrence of both extremes (droughts and severe precipitation events) is expected

These extreme events are expected to have profound effects upon ecosystems (Easterling et al. 2000). Several studies show evidence of effects of weather fluctuations on population dynamics. For example, Grotan et al. (2008) show an effect of variation in winter climate (temperature and snow depth), and precipitation in early summer on annual changes in population size of Swiss ibex (Capra ibex). Their results suggest that fluctuations affect either mortality of individuals (tub hypothesis, Lack 1954; Saether et al. 2004) or fecundity (tap hypothesis, Saether et al. 2004). A study of Great tit (Parus major) populations shows that various variables, including inter-annual variation in temperature, explain fluctuations in populations size (Grotan et al. 2009). Morris et al. (2008) analyzed the sensitivity of long-term population stochasticity to changes in the means and standard deviations of vital demographic rates in response to climate variability and found a difference between short- and long-lived species. The results of this study highlight the potential vulnerability of short-lived species to increased weather fluctuations. Piessens et al. (2009) found for the Small blue butterfly (Cupido minimus) that local extinctions after the extreme summer heat wave of 2003 were correlated with small population size. This result supports our hypothesis that survival in habitat networks under climate change will require larger patches, holding larger populations, to compensate for the increase in weather variability.

Den Boer (1986) studied the population dynamics of carabid beetles in a period with alternating wet and dry years. He demonstrated that Pterostichus versicolor, a generalist living in heterogeneous habitat with independently fluctuating subpopulations, showed less population fluctuations compared to Calathus melanocephalus, a specialist living in spatially homogeneous conditions.

While increased variation and extreme events are expected to have adverse effects on populations on average, the recorded and expected lower frequency of cold and frost days (see Fig. 1) might have a positive impact on some species (cf. Drake 2005). Some preliminary results show a decrease in inter-annual variation in certain bird species (Cormont, Schippers, Foppen, Van Turnhout, unpublished data).

Environmental stochasticity in population viability assessment methods

Population viability is dependent upon population size and three levels of stochasticity: demographic and environmental stochasticity, and random catastrophes (Fig. 2, after Shaffer 1987). Demographic stochasticity results from random events in the reproduction and survival of individuals. Populations regulated mainly by demographic stochasticity have high viabilities at relatively low size. Environmental stochasticity results from unpredictable changes like in temperature and rainfall with important consequences for food supply. Populations regulated mainly by environmental stochasticity have a viability that is less sensitive to population size (slope is less steep); larger sizes are needed to obtain the same level of viability. Random catastrophes are events such as floods, wildfires and droughts for which there are no safe numbers, as even large populations are at risk of becoming extinct. Real populations will experience a mix of the three types of stochasticity and therefore the relation between population size (or protected area) and viability will be somewhere between the two extremes.
https://static-content.springer.com/image/art%3A10.1007%2Fs10980-010-9497-7/MediaObjects/10980_2010_9497_Fig2_HTML.gif
Fig. 2

Functional forms of the relationship between population viability (here presented as the mean time to extinction) and population size or patch carrying capacity (adapted after Shaffer 1987). The original relations are the solid lines, where for a certain level of viability (T1), resp. KDS, KMS and KC are needed as population size (or patch carrying capacity) for populations regulated by demographic stochasticity, environmental variability, and catastrophic events. When the variation in weather increases, relations are expected to shift to the right. This is indicated by the dashed lines, which correspond to size K*DS and K*MS. Certain viability levels (T2) cannot be reached when catastrophic events occur as populations can be wiped out in an extreme event no matter their size

The impact of increased inter-annual variation can be demonstrated with a model experiment. Suppose a population grows at an annual rate of R(t):
$$ {\text{N}}\left( {{\text{t}} + 1} \right) \, = {\text{ R}}\left( {\text{t}} \right){\text{N}}\left( {\text{t}} \right) $$
where N(t + 1) is the population size in year t + 1 and R(t) is drawn from a distribution with an average of 1. For a scenario with low environmental stochasticity (ES), such as low frequency of extreme weather events, populations fluctuate the least and under a severe ES scenario they fluctuate most. It can be demonstrated that the more R(t) is allowed to fluctuate, the more the population will tend to decrease. This has already been pointed out by Lewontin and Cohen (1969) and is the consequence of the population responding to the geometric mean rather than to the arithmetic mean of R(t). Real populations behave of course with much more complexity than this simple example, but mortality and reproduction are usually a kind of multiplicative processes (Boyce et al. 2006) while demographic data (such as the mean number of offspring) are usually presented as arithmetic average. Therefore an increased fluctuation in fecundity and mortality rates can have significant consequences for the risk of extinction, even in large populations. Population viability assessment methods usually use data from the past, and thus underestimate the expected future fluctuation of the vital demographic rates. They may therefore miscalculate the viability of (meta)populations and the sustainability of ecological networks.

The evidence upon increased inter-annual variation (including the occurrence of extreme events) suggests that established relations between population size (or protected area) and viability such as the key patch approach need to be reconsidered (see Fig. 2: species-specific relations between population size and population viability are expected to shift to the right).

Exploring the impact of environmental variability on specific design criteria

In order to explore the sensitivity of the key patch criteria to the amount of environmental variability, an individual-based two-sex population model (METAPHOR/METAPOP) was run with on average one immigrant per generation, see Verboom et al. (2001), Vos et al. (2001) and Schippers et al. (2009) for model details. This was done for two ecological profiles, a large long-lived bird (warbler) and a small short-lived bird (Bittern). We varied the standard deviation of reproduction and survival in order to simulate the effect of changes in ES on key patch size required for population viability. The life history events during the year were: immigration, adult survival, reproduction and dispersal. Recruitment (offspring that survive the first year) and immigration were drawn from a Poisson distribution, sex ratio was 1:1. Adult survival and dispersal were simulated as a binomial process. To account for environmental stochasticity, the yearly mean recruitment and survival were regulated by two independent yearly factors, a yearly mortality factor and a yearly recruitment factor, that were drawn from a normal distribution with standard deviation σm and σr (white noise). We did not assume temporal autocorrelations in weather data on a yearly basis (Barnston 1996) but we made the simplifying assumption that σm = σr; low environmental stochasticity corresponds with lowered σm and σr, high environmental stochasticity goes with elevated σm and σr. By varying σm and σr and the patch size (carrying capacity) we could calculate equal probability lines that relate changes in σm and σr to patch sizes with the same extinction probability. See Electronic supplementary material for more details on model and parameter settings.

Results

One can choose a viability level, e.g. 5% extinction probability in 100 years. The population extinction risk increased with higher standard deviations corresponding with increased ES, (Fig. 3). However this increase can be compensated for by increasing patch size. Minimum key patch size required for population viability increased exponentially with increasing environmental stochasticity. Small birds with higher turnover rates are more sensitive for changes in ES: with a small change in ES, their key patch size requirements increased strongly, compared to large birds. For example, while for the large bird profile an increase of ca. 25% seems enough to compensate for a 50% increase in standard deviation of the vital demographic rates, for the small bird profile an increase of more than 100% in patch size would be necessary to achieve this goal. The figure also indicates the positive effect of decreased environmental stochasticity, either by natural processes (e.g. some species may benefit from the reduction in extremely cold winters) or by policy measures targeted at reducing population fluctuations.
https://static-content.springer.com/image/art%3A10.1007%2Fs10980-010-9497-7/MediaObjects/10980_2010_9497_Fig3_HTML.gif
Fig. 3

Simulation results: key patch size required for a certain population viability under demographic and environmental stochasticity for the small (a) and large (b) bird species. Population viability is expressed as P, the extinction probability within 100 years (bold line: P = 5%). Standard deviation of reproduction and survival in default situation (value 1 on horizontal axis): small bird sd = 0.2; large bird sd = 0.082, corresponding patch areas are resp. 2 ha (100 breeding pairs at carrying capacity) and 20 ha (20 breeding pairs at carrying capacity). See Electronic supplementary material for details

Exploring the impacts of altered design criteria

The results presented here may have consequences for nature policy. Adjustments in methods and criteria are necessary, since the status of populations (e.g. viable or not) might have changed due to climate change and nature policy targets (like halting biodiversity loss) may be further away than we thought.

The Netherlands Environmental Assessment Agency explored if and where the protection of target species in the Dutch National Ecological Network (NEN) would be weakened, when climate change would indeed require larger populations to compensate for increased weather variability (NEAA 2008; Vos et al., submitted). The NEN, comprising of 750,000 ha of natural habitat, was designed on the basis of the key patch approach and old design criteria (as in Verboom et al. 2001) to counteract the effects of habitat loss and fragmentation on biodiversity (Foppen et al. 2000; Verboom and Pouwels 2004). As it is unknown to what extent population fluctuations will increase in the future, an analysis was performed in order to explore the sensitivity of assessment results to specific design criteria, assuming double area requirements for key patches to compensate for the increase in extinction risk by climate change. Marshland areas, being part of the NEN, show viable populations for 83% of the target species, estimated with current key patch criteria for target species. If key patch criteria were doubled, only 60% of the marshland target species would show viable populations in the NEN. These results imply a considerable loss in the effectiveness of the NEN for biodiversity protection when indeed larger population fluctuations would occur because of climate change. Figure 4 shows how wetland areas decrease in their contribution to biodiversity conservation. Some wetlands remain strongholds within the wetland network, holding key populations for almost all target species, even when individual area requirements are doubled.
https://static-content.springer.com/image/art%3A10.1007%2Fs10980-010-9497-7/MediaObjects/10980_2010_9497_Fig4_HTML.gif
Fig. 4

The predicted percentage of policy target species for wetlands occurring in key patches for: (a) original key patch criteria (Verboom et al. 2001); and (b) doubled key patch criteria (Vos et al., submitted). This figure zooms in on part of the Netherlands

Discussion and conclusions

With increased variation (including extreme weather events), current population viability assessment methods and design criteria are no longer sufficient to support the design of ecological networks. The currently used methods and criteria for sustainability of ecological networks therefore have to be amended. There is an urgent need for such knowledge, tools and decision support algorithms (Pressey et al. 2007; Nassauer and Opdam 2008; Wu 2008). These tools should take into account patch area in ecological networks—the focus of this paper -, but also connectivity, facilitating changes in species’ distributions (cf. landscape cohesion assessment, Opdam et al. 2003) and measures of resilience, the ability to recover after disturbance (Carpenter et al. 2001).

Others discuss the impacts of increased variability in vital demographic rates in relation to habitat configuration (Bossuyt and Honnay 2006; Fraterrigo et al. 2009). They conclude that landscape connectivity and successful dispersal are vital to metapopulation survival, especially when variation is taken into account. So climate adaptation should target both patch sizes and landscape connectivity (Vos et al. 2008).

Whereas population viability assessments based upon size may only prioritize landscapes with large homogeneous areas, maximizing carrying capacities of species, the expected increase in inter-annual variation and frequency of extreme weather events might ask for additional adaptation strategies. Rather than relying solely on enlarging patch areas (increasing carrying capacities), landscape design could aim at minimizing population fluctuations (decreasing fluctuations in vital demographic variables) for the same overall effect upon population viability; increasing the internal heterogeneity of patches and landscapes will enhance resilience, dampen fluctuations and spread the extinction risk when extreme events such as drought, fire or flooding occur (Kindvall 1996; Sutcliffe et al. 1997; Piha et al. 2007; Wu 2008). A special case is the creation of climate refugia (Vos et al. 2008).

In addition, we acknowledge that genetic constraints and phenological mismatches should be integrated in population viability assessments. Populations shifting their distribution can lose their genetic diversity, and thus their adaptive capacity is at risk (Hewitt 1996; Grotan et al. 2008). Increased patchiness of the landscape in combination with fluctuations in size (bottlenecks) may enhance this effect (Cobben et al., in prep.). Phenological mismatches appear to be an important mechanism for climate change effects and should be taken into account when designing new methods and criteria, as there is a large variation in climate response between species (Both et al. 2006; Menzel et al. 2006).

The results such as presented here (Figs. 3 and 4) may alter if different assumptions are made for the underlying processes and the parameters. For example, we assumed that on a yearly basis processes are not temporally autocorrelated (white noise), without addressing how temporal autocorrelation may alter population dynamics. Spatial autocorrelation should be considered when more than one patch is taken into account (Robert 2009). Indeed a negative interaction with habitat fragmentation is to be expected, as spatially correlated disturbances will shorten metapopulation extinction time (Akcakaya and Baur 1996). Given that both temporal and spatial autocorrelation might exist, the assumption that immigration is on average the same everywhere and in all years is a simplification. In real life, the Santa Claus effect (Hengeveld and Hemerik 2002), the opposite of rescue effect, might cause populations that are doing well to get more immigrants than populations that are endangered. Empirical data and projected future fluctuations in vital demographic rates and their impacts are not currently available and this is a serious knowledge gap. In spite of this limitation, theoretical explorations as we did by doubling area requirements can provide important insights into the conservation problems lying ahead of us (Saether et al. 2004; Bossuyt and Honnay 2006; Fraterrigo et al. 2009; Robert 2009) and the use of ecological profiles like the warbler and the Bittern are helpful.

We assume that society will understand the need for adapting conservation strategies and measures (Opdam et al. 2009). Modified population viability assessments and additional adaptation strategies together, based upon new knowledge, will assist in improving the network design process prioritizing adaptation measures such as increasing area and/or quality, increasing connectivity, or managing for heterogeneity and gradients. This will result in more cost effective measures and in sustainable ecological networks under climate change. Further research is needed to reconsider generic design criteria (such as: ‘big is beautiful’ and ‘connectivity is good’) and derive new specific design criteria. Until then, we need to realise that in the light of climate change the existing generic criteria are even more important now than they were in the past (area, connectivity) and we propose patch and landscape heterogeneity as a third dimension.

Acknowledgments

We thank Ruud Foppen and Chris van Turnhout of SOVON Dutch Centre for Field Ornithology for fruitful discussion. This research was funded by the Dutch national research program ‘Climate Changes Spatial Planning’ and is part of the strategic research program ‘Sustainable spatial development of ecosystems, landscapes, seas and regions’ (Project Ecological Resilience) which is funded by the Dutch Ministry of Agriculture, Nature Conservation and Food Quality, and carried out by Wageningen University and Research centre.

Supplementary material

10980_2010_9497_MOESM1_ESM.doc (87 kb)
Supplementary material 1 (DOC 21 kb)

Copyright information

© Springer Science+Business Media B.V. 2010