Ethics Statements

The capture permit was authorized by the Ministère de la Région Wallonne (Division de la Nature) and issued on 19th February 2008 on the basis of ethics approval of the field study on newt ecology by Conseil Supérieur Wallon de la Conservation de la Nature.

Study Area and Sampling Procedures

This study was conducted in Pays de Herve, an agricultural area of eastern Belgium at the border of Germany and the Netherlands. The sampling areas were chosen from the known distribution of the crested newt [22]. We surveyed 74 ponds during the newt reproductive season (March–June 2008) (Figure 2). The surveys took place in the 12 ponds historically inhabited by this species (1990–2004), in all ponds within a 500-m radius around these ponds, and also within a 500-m radius around these new sets of surrounding ponds (Figure 2). Surrounding ponds were located using recent topographic maps (Institut Geographic National [IGN], 1∶20,000, published in 1999–2000), colour orthoimages (IGN: Direction Générale de l’Agriculture, 0.5-m resolution, 2006), and by field observations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Location of studied ponds in Pays de Herve (Belgium).

Blue circles: ponds, yellow patches: studied localities (based on historical presence of Triturus cristatus): 1, En Géliveau; 2, Hansez; 3, Haute Rafhay; 4, Stoki; 5, Margarins; 6, Vogelsang; 7, Blanc Baudet; 8, Gut Benesse Hof; 9, Hof Krompelberg; 10, Gemerhet; 11, Corney; 12, Harbenden. Geographic coordinates: Belgian Lambert Grid (expressed in km).

https://doi.org/10.1371/journal.pone.0062727.g002

To obtain an index of newt abundance, each pond was sampled by dip-netting (40×35 cm dip-net with a 1.8 m pole) [32]. All ponds were visually screened for newts at the time of dip-netting but no more species were found in this way. The small size of ponds allowed covering the entire surface of the ponds several times. Deep zones were additionally sampled with large nets (5 to 10 m² seines). This sampling design was particularly adapted to reduce escape possibilities during netting. Censuses ended after several unsuccessful nettings in various areas of the ponds, including open and vegetated parts. This method has the advantage of sampling all micro-habitats and thus gives comparable values across ponds since all ponds were surveyed similarly. Previous studies showed that the detection probability with this type of removal estimate is very high [33] and that using dip-nets is an adequate method for sampling European newts [34]. Because previous studies highlighted that amphibians, including newts, can leave water temporarily during the breeding season [35], [36], our method did not aim to determine the total adult population, but rather to approach the size of the aquatic adult population as closely as possible at any given time. By sampling all habitats, including pond vegetation and banks, hidden newts can also be captured. Previous studies have shown that blind dip netting such as done in the present study does not give lower abundance estimates during daytime than during night-time [37]. Although it is possible that one species was missed in some ponds, the absence of a record indicates that the species is very rare in that pond or suggests that this would at best be a “sink” or transient habitat [38], [39]. Previous studies have shown that newt abundance is correlated with habitat quality [32]. We therefore believe that our approach is sufficiently robust to evaluate the association between species and environmental variables. We took into consideration only adults because we sought to determine the index of abundance similarly across ponds. There were no ponds in which we found larvae and no adults (qualitative checks were carried out at other times during the study period). All amphibians were handled with wet gloves during sampling. All material was washed and disinfected after every visit to a pond.

Pond and Landscape Traits

We measured several environmental variables representing five major processes that can determine newt distribution (Table 2). Five variables describing pond features were measured in situ during the newt census (Table 2). The maximum water depth and macrophyte cover were assessed in the field, whereas the pond surface area was assessed either in the field or obtained through aerial photo interpretation. The presence of introduced fish (both native and exotic to Belgian fauna, but all outside their natural habitat) was determined by dip-netting and seining, visual observations, and interviewing local owners. These fish are often locally invasive. Oxygen was measured with an oximeter (Hach Lange Multi HQ40d). To evaluate water pollution, a specific visit of each pond was made to gather water samples, which were preserved at 4°C and directly transported to the laboratory for chemical analyses. Sampling took place within 2 days in June. The concentration of three nutrients, mostly caused by water pollution (orthophosphates, nitrites and ammonium) was evaluated through colorimetric analysis using, respectively, blue Molybedne, de Griess and Berthelot reagents.

The number of ponds within 100 and 500 m of the focal pond was recorded as a measure of the pond’s present-day connectivity (Table 2). These values were chosen because radio-tracking studies showed that 50% of movements occurred within 100 m, whereas 500 m is usually considered to cover most movements [20], [40]. Connectivity was available for the present-day period only, because not all ponds were recognizable on historical maps. To evaluate landscape composition, shape files were drawn from recent orthoimages (IGN – DGA, 2006) and historic aerial images (IGN, 1947–1954) in ArcGis 9.3 (Esri, Redlands, CA, USA) to represent land cover limits (i.e. forest, croplands, gardens, buildings and ponds). First-order polynomial functions were adjusted using ground control points selected on the scanned historical aerial images and on the 2006 reference orthoimages. Topographical maps and field visits helped create the landscape layers.

Statistical Analyses

Three out of five “processes” were represented by multiple environmental variables (Table 2), which were strongly correlated to each other. Including correlated variables may bias the regression results; preliminary models including the original variables showed high values of the variance inflation factor, indicating that multicollinearity affected these models. We therefore used principal component analysis (PCA) to summarize variables in a lower number of uncorrelated components. PCA was performed using the correlation matrix and variable scaling. To keep the roles of these five processes distinct, we performed separate PCAs for the variable set representing each one (Table 2). Extracted components were rotated (Varimax rotation) to improve interpretation. The PCA was run over pond features, present-day and historical pond landscapes at both the 100 and 500 m radius. Extracted components explained 67%, 71%, 85%, 72%, and 74% of the total variance. The correlation between the original variables and the extracted components is shown in Table 2.

We used an information-theoretic approach, based on Akaike’s information criterion (AIC), to identify the processes and the spatial scales most likely to affect the abundance index of the two newt species [7], [41]. We analysed relationships between newt abundance and environmental features using generalized linear models (GLMs), assuming a quasi-Poisson error distribution to take into account overdispersion. First, we built GLMs considering all possible combinations of the variables (either environmental variables or PCA components) representing the five processes (Table 2). For each model, we calculated the quasi-AIC corrected for small sample size (Q-AICc) [42]. A model was not considered as a “candidate model” if a simpler, nested model had a lower Q-AICc [43]. Furthermore, for each process, we considered only one spatial scale at a time, meaning if we included connectivity at the 100 m scale, we did not include connectivity at the 500 m scale and vice-versa. For each candidate model i, we then calculated the Q-AICc weight w_i, which is the probability that a given model is the best one, given the set of candidate models considered [44]. We also reported significance values of variables included in the best models, to facilitate the interpretation of the models and of the role played by predictors [45]. Using Q-AIC instead of Q-AICc would not change the results (for both species the best models would remain the same). Errors were not spatially autocorrelated (for all best models, Moran’s I <0.05, P>0.2). None of the candidate models showed multicollinearity (for all models and all variables, variance inflation factor <5). Conditional partial regression plots were built using the visreg package [46]. Finally, we used an unequal variance t-test to compare the features of ponds inhabited by the rarest species or only by the common species [47].

Commonness and Rarity

The analysis of ecological processes involved in the distribution of T. cristatus and L. vulgaris showed that, despite a difference in commonness, both species share similar responses to environmental features. As shown by the best AIC models (Table 4), all the processes tested had a similar influence on both species: their abundance was related to the same environmental variables, and the effect was in the same direction for both species. This indicates that conservation actions should consider multiple factors to adequately protect these species. The suitability of pond features (water quality, macrophytes, water depth), the absence of introduced fish species, the connectivity among sites, and the composition of present-day and historical landscapes were important for newt distribution. These processes are mostly determined by anthropogenic activities and are acting in a direction that is unfavourable for newt persistence, both in the study area and in most modern agricultural landscapes. It can therefore be expected that without action plans, both the rare and the “common” species will quickly decline. These results support recent evidence that not only the rarest, but also the apparently most common amphibians are at risk [10], [11], [17], [48]. Commonness patterns should not be overlooked in ecological research as commonness is only a transient state, which means that what is common today may be rare tomorrow [2].

The use of surrogate species has been a major tool in conservation planning, but has also received criticism as several conditions need to be met for applying conservation actions efficiently. Recent analyses showed empirical evidence for this concept, but more data are needed on amphibians [49]. Both newt species used similar habitats and responded similarly to disturbance, which supports the potential efficiency of the umbrella species concept in amphibians [3]. The similarity within several processes acting at multiple scales argues even more for simultaneous protection of common and rare species. Moreover, the rarest species responded more strongly to environmental changes than the most common species. Newt populations often exist within networks of meta-populations or patchy populations [50]–[52] and, in some cases, observed changes of occupancy may be part of extinction and colonization dynamics that are characteristics of meta-populations [53]; for instance, L. vulgaris was detected in one pond where it was not found 10 years before. Nevertheless, for T. cristatus the decline persists because of the overall loss of suitable habitats and, under situations of continuous loss of habitat quality, colonizations cannot compensate extinctions [17], [51]. Evaluating changes in occurrence across time gave coherent results with these patterns: in the recent surveys T. cristatus was not found again in half the ponds, whereas L. vulgaris was still found in all but two ponds. Such a focus has broad applications as both species are sympatric over a large part of their distribution ranges. However, targeting T. cristatus populations would favour only a small fraction of L. vulgaris populations, as they are six times more numerous than T. cristatus populations in the study area [22]. These results are confirmed with significant trends at a broader scale that showed that T. cristatus is more localized and less widespread than L. vulgaris [17]. In fact, actions on the ponds inhabited by the more common species could also help the recovery of T. cristatus in nearby populations as this species is favoured by a high density of ponds. On the other hand, our analysis considered species of newts typical of open landscapes, while responses to environmental modifications may be even more complex if the whole amphibian community is considered [22], [54], [55]. This means that the umbrella species approach should be used with care only once ecological requirements are sufficiently known.

Ecological Processes: Shared and Specific Patterns across Studies

Previous studies on the ecology of crested and smooth newts considered heterogeneous sets of environmental variables and spatial scales [1], [15], [20]–[25], [39], [56] (Table 1). Each of these studies highlighted determinants of newt distribution and thus improved our knowledge in terms of conservation management. As outlined by Zanini et al. [57], there is geographic variation of underlying ecological processes and thus different results can be found in contrasted regions. For instance, Hartel et al. [21] indicate that traditional management of the landscapes studied in Romania was the basis of the relative unimportance of landscape determinants. The present study was conducted in an area where rarity was more pronounced than in other studies (but see [23]; Table 1). The results confirm previous knowledge on these species but also show that in altered landscapes, multiple processes are acting simultaneously on newt distribution. Indeed, all the processes considered contribute to explaining the distribution of the two species. In addition, considering different spatial scales and water pollution provided new insights into the anthropogenic pressure on natural populations.

Towards Effective Conservation Measures

Although this study targeted areas that were known to be inhabited by a rare newt, both study species were absent from two-thirds of the ponds. The fact that this pastoral landscape hosts a high density of ponds [31] must therefore be balanced by their limited suitability for newt breeding. Although only one pond had disappeared since the last survey period (1990–2004), several ponds showed signs of future disappearance (i.e. shallow depth, eutrophication, partial destruction), suggesting a higher rate of pond loss over the long term, as shown in other studies [26], [58].

In terms of suitability of aquatic sites, ponds need to be restored to maintain a high water depth (i.e., at least 1 m in such agricultural lands) while avoiding fish introduction, as permanent ponds are more likely to sustain fish. Habitat restoration programs, including pond creations for T. cristatus, have proved to be successful and should be followed over wide areas [59]. Fish have been shown to be particularly detrimental to newts [20], [25], [60]–[62]. However, smooth and crested newts could also coexist with fish [1], [63], [64]. Here, introduced fish were found in 16% of ponds. Although most of them could not predate adult amphibians, they can eat eggs and larvae [65]. This could explain why we only found two populations of T. cristatus and four populations of L. vulgaris coexisting with fish. In these cases of cohabitation, newt abundance was very limited (maximum two T. cristatus and 17 L. vulgaris detected). Large fish were not found in coexistence with newts, except in one population of L. vulgaris where fish were not numerous. The highest susceptibility of crested newts to fish may be due to the more pelagic behaviour of its larvae [66], [67], but more work is needed to understand the mechanisms of coexistence between newts and fish [68]. Because amphibian resilience is possible after fish removal [69], this management action should be included in conservation plans. Aquatic vegetation provides support and protection for the eggs and shelter for the newts [15], [25], [70]–[72] and should be favoured, but without excess, as ecological succession would lead to pond disappearance [73]. Sources of pollution should be identified and managed to avoid run-off of pollutants. Too few studies have integrated pollution by fertilizers and domestic run-offs and how they affect amphibians [74], although laboratory experiments have shown their direct effects on larval stages [29]. The high concentration of pollutants found in ponds within this study and their relation to absence or low abundance of newts show that the landscape studied is heavily polluted. Buffer zones should therefore be designed around each pond to reduce the risk caused by the use of fertilizers and building water discharges.

Previous studies have shown that woods or scrubs should be maintained near breeding ponds as they provide an adequate space for vital activities such as feeding outside the reproductive season, but also for aestivation and wintering [28], [75], [76]. The preferential emigration from ponds is often toward forests instead of open landscapes in both T. cristatus and L. vulgaris [77], and the occupied ponds are usually only at a few hundred metres from forests [21], [22], [25]: these arguments further support the importance of forests. However, both T. cristatus and L. vulgaris typically breed in ponds located in open landscapes. The differences between the effect of past and previous landscapes showed that terrestrial processes are complex and would require specific investigations. Telemetry and capture-mark-recapture studies, such as those conducted by Jehle [28] are needed to understand terrestrial requirements in both traditional and modernized landscapes.

Connectivity, often highlighted in newt research [1], [20], is not only important at large scales around core ponds, but also in the vicinity given that the number of ponds within a radius of 100 m had a significant effect on both species in the present study. This confirms results of radio-telemetry showing that most individuals remain very close to the breeding ponds [28]. Viability prediction models highlighted that T. cristatus populations harbouring more than 40 adults may have a lower risk of extinction in case of isolation [78]. The usual number of adults detected in all the populations studied was below this value, emphasizing the importance of maintaining landscapes with high pond density.

In conclusion, commonness should not be neglected in conservation management and the adequacy of surrogate species should be evaluated to ensure that what is common now does not become rare. In the current perspective of amphibian decline, multiple stressors should be considered together to allow efficient conservation programs.