Drastic Population Size Change in Two Populations of the Golden-Striped Salamander over a Forty-Year Period—Are Eucalypt Plantations to Blame?

Abstract

Over the last half century the Iberian peninsula has seen the large scale planting of exotic gum trees (Eucalyptus sp.) therewith reducing space for native wildlife. An additional effect of the gum tree plantations may be the lowering of the water table in adjacent streams, to which amphibian species with a larval niche in the running sections of small streams would be especially susceptible. In northwestern Iberia that niche is occupied by the Golden-striped salamander, Chioglossa lusitanica. I here report on the demographic trends of two C. lusitanica populations over a forty-year period, in two areas of one mountain range near Porto in northwestern Portugal. In both areas advantage was taken of the migration pattern of C. lusitanica to sites for aestivation and breeding in summer and fall. The area of the Silveirinhos brook was transformed in a plantation of gum trees shortly after the research started, while the area of Poço do Inferno remained virtually unaffected. At Silveirinhos the adult C. lusitanica population declined by one or two orders of magnitude, from ca. 1500 individuals to less than 50 at present. Demographic models that operate under a uniform larval mortality yielded population sizes that are compatible with field observations, including the late onset of the decline at 14 or more years after the planting of the gum trees and the near-extinction at year 32. An alternative reason for the relatively recent population collapse of C. lusitanica may have been disease, but no sick individuals or corpses have become available for clinical investigation. Conversely, the control population at Poço do Inferno increased in size by a factor of five or more. These data support the hypothesis that gum tree plantations have a strong negative effect on C. lusitanica. The population size increase at Poço do Inferno is probably attributable to the installation of wastewater treatment in the adjacent town of Valongo, with a discharge in the Simão river and closely connected to Poço do Inferno. This result suggests that the Simão river contributes to the Poço do Inferno population and that medium-large streams may constitute prime C. lusitanica habitat. Larval mortalities estimated for stable and declining populations are remarkable close (0.69 and 0.73, respectively), but dissimilar to the larval mortality at ca. 0.51 of an increasing population such as at Poço do Inferno. Suggestions for further research include (i) the continued monitoring of the Silveirinhos population with inspections at 5–10 year intervals, and (ii) checking for the persistence of C. lusitanica populations in and around gum tree plantations where the species was reported decades ago.

1. Introduction

To document and understand natural population size fluctuations in long-lived animal species such as many vertebrates requires long-term data. To distinguish natural fluctuations from a pattern of change, such as may be expected from large-scale anthropogenic activities, requires even more data. Long-term data are, however, notoriously difficult to obtain. One dramatic change in the landscapes of southwestern Europe has been the planting of exotic gum trees (Eucalyptus sp.) over vast areas and one of the most drastic projects took effect in the last quarter of the 20th century, through the Portuguese Forestry Project with assistance of the World Bank [1,2]. In Portugal, the area with gum tree plantations—mostly the Tasmanian blue gum, Eucalyptus globulus—has increased eight-fold from the mid 1960s to a coverage of ca. 25% of today’s forested area, with a substantial reduction in space for native wildlife as a consequence. An additional effect may be a reduced run-off causing watercourses to dry up early in the season, even if the streambed itself was not affected by the plantation. Such a drying effect would be most detrimental to species that live in the running section of streams [3,4]. I set out to study the possible effect of the large scale planting of gum trees in an area in the vicinity of Porto, Portugal, where shortly before I had studied the natural history of an enigmatic amphibian, the Golden-striped salamander, Chioglossa lusitanica (Figure 1). The study took place along two headwater streams at different slopes of one mountain range. One area remained unaffected, whereas in 1980 the other area was turned into a gum tree plantation. The planting involved the rip-plowing of the Serra de Santa Justa, except for a ca. 20 m wide buffer zone covering the Ribeira de Silveirinhos and its margins, as required by Portuguese law [5]. With data going back to 1976, the control population has increased in size, whereas the population at the gum tree plantation has decimated.

Figure 1. Golden-striped salamander, Chioglossa lusitanica: (A) adult, drawing and copyright Bas Blankevoort © Naturalis Biodiversity Center, reproduced from [6] with permission; (B) embryos in various stages of development, Photo M. Sparreboom; and (C) first-year larva, Photo JWA.

Figure 1. Golden-striped salamander, Chioglossa lusitanica: (A) adult, drawing and copyright Bas Blankevoort © Naturalis Biodiversity Center, reproduced from [6] with permission; (B) embryos in various stages of development, Photo M. Sparreboom; and (C) first-year larva, Photo JWA.

2. Materials and Methods

2.1. Study Areas and Fieldwork

The two study areas, called “Ribeira de Silveirinhos” and “Poço do Inferno”, are located at the northern flanks of the Serra de Santa Justa, immediately south of the town Valongo at ca. 10 km east of Porto, Portugal. The Silveirinhos brook runs in southerly direction, has a length of 4 km and discharges in the Rio Ferreira. The Poço do Inferno brook runs in northeastern direction, has a length of ca. 600 m and discharges in the Simão river. The distance between the study plots is 1340 m “as the crow flies”, while the “as the salmon swims” distance along the Ferreira and Simão rivers is ca. 10 km (Appendix 1 with Figure A1).

The Ribeira de Silveirinhos is a small semi-permanent mountain brook in slate and quartzite terrain. In a zone of up to 30 m wide the valley has a dense vegetation of Alnus glutinosa, Eucalyptus globulus and Acacia sp., with an undergrowth of brambles, Digitalis purpurea, Pteridium aquilinum and gum tree and Acacia shoots. The search plot is situated close to the Silveirinhos source and includes a dry stone, V-shaped built structure of ca. 720 m³—presumably constructed from the debris of a local excavation—spanning the valley (41.17330°N, 8.50080°W). This “dam” of 28 m length, up to 7 m high and up to 22 m wide has a tunnel for the stream to run through (see Figure 1 in [7]). The age of the dam is unknown, but judging from the size of the trees growing on top of it, already in 1977, it must have been old. The tunnel, the dam wall, the brook and the vegetated zones at either side over the 200 m long downstream section of the brook were regularly surveyed by night for post-metamorphic C. lusitanica. The open maquis vegetation at further distance from the brook was initially inspected also, but no golden-striped salamanders (Chioglossa) were found and searching outside of the densely vegetated zone was abandoned. By daylight the selected stretch of the Silveirinhos was dip-netted for larvae of C. lusitanica and other amphibian species. In the 1979–1981 period the Serra de Santa Justa and adjacent mountain ranges were rip-plowed and transformed from semi-natural open maquis (see Figure 4 in [8]) into gum tree plantations almost entirely (details for the Valongo municipality are in Figure 8 in [9]), but the planting on newly created terraces stopped short of the densely vegetated zone at ca. 10 m from the Silveirinhos brook. The area upstream of the dam remained unchanged. The area is rather left alone and I have observed no management of the gum tree plantations other than the clear-cutting of sections.

The “Poço do Inferno” is a small impermanent mountain brook in schistose/quartzite terrain rich in iron oxide and with a vegetation consisting of Pteridium aquilinum, Cytisus striatus, Calluna vulgaris, Erica sp. and Frangula azorica and stands of gum and pine trees on its banks (see Figure 3 in [7]). Several narrow and more or less deep corridors with entrances close to the brook were built for gold-digging in Roman times (locally called “minas” [10]). However, bore holes from modern machinery witness a more recent human activity. The research was carried out in two mines, the “Upper mine” with the entrance at 41.18064°N and 8.48816°W at a distance of 260 m from the Simão, and the “Lower mine” with the entrance at 41.18185°N and 8.48677°W at 80 m from the Simão. The investigated area includes an 80 m section of the brook in between them, upstream from the dirt road crossing the brook.

Both mines consist of a main gallery and two side galleries, roughly laid out in the shape of a cross (Appendix 2 with Figure A2). Galleries are up to 2 m wide. The height allows straight up walking except at several places deeper in the mines that are littered with heaps of rock debris. The Upper mine has shallow but permanent standing water for the deeper part of the main gallery and a water level of ca. 25 cm in one short side gallery. This mine has an additional gallery starting near the entrance. This stretch of ca. 25 m length is fairly dry, rarely visited by C. lusitanica and, with a few exceptions, was not included in the searches. The entrance of the Upper mine is located at a short distance from the Poço do Inferno brook with no observable discharge, except after periods of very heavy rain when the mine’s water level rises markedly as to overflow the heap of soil at the entrance (see Figure 4 in [7]). In the Lower mine water drips down the walls and the ceilings at many points and a small stream runs through the mine permanently and at the mine entrance discharges in the Poço do Inferno brook. The main gallery is wetter than the two arms at either side. Both mines were prospected over their length of ca. 105 m (Upper mine) and 236 m (Lower mine). Adult salamanders and mated pairs were counted along with a rough count of the number of eggs (or, actually, embryos) (Figure 1). A further search by turning stones from the heaps of stone debris indicated the presence of many more adult and embryonic C. lusitanica, but this was not carried through, because of the danger of harming salamanders and the work load involved.

I explored for the presence of an underground connection from the Upper mine to the Poço do Inferno brook. In spring 2011 an aqueous solution containing 60 g Fluorescein sodium salt (Sigma F6377) was released in the standing water section deep in the Upper mine. Water samples of 50 ml were taken at the Poço do Inferno brook at 20 m upstream (negative control), aside and at 170 m downstream of the entrance of the Upper mine, at eight-hour intervals over a three day period. The pooled samples were measured for Fluorescein with a Perkin Elmer LS 45 Fluorescence Spectrometer under FL WinLab software and a detection limit of 0.013 µg Fluorescein/L.

2.2. Pattern of Annual Migration

The dam serves as an aestivation site for adult Chioglossa salamanders that in fall, winter and spring reside along the brook [8,11]. Both mines are places of C. lusitanica shelter, mating and egg-deposition. In both areas, observations on C. lusitanica were made on a regular basis in the spring, summer and autumn of 1977 and irregularly ever since. Most visits were during the autumnal period of reproduction. Egg numbers for 1976 were estimated in early spring 1977, by a count of molded embryos and the empty gelatinous shells left behind by hatched embryos. In the course of 1977 and in October 2010, adult salamanders in the dam were marked by toe-clipping. The adult population size was calculated with Begon’s weighted mean method [12], based upon the recapture of marked salamanders. Additional data have become available from colleagues and from the literature.

2.3. Natural History of Chioglossa lusitanica and Unknowns

The Golden-striped salamander is an exceptionally slender and long-tailed salamander of the family Salamandridae, endemic to the northwestern corner of the Iberian Peninsula. While described 150 years ago [13], C. lusitanica remained virtually unstudied until a broad survey in 1957 [14] and my own ecological work in the late 1970s [15,16]. More has become known about the ecology of the species [4,17,18] and about important life history parameters such as demography and reproductive cycle [19,20]. The species distribution is by now well documented [21,22,23] and we have some knowledge on morphological variation [24,25,26] and phylogeography [27,28,29,30]. By far the most prominent feature of C. lusitanica is its dependency on conditions of moisture, which is reflected by (i) a distribution confined to areas of high precipitation (>1000 mm annually); (ii) a drought-induced pattern of migration along the brooks, for which distances of >350 m overnight have been recorded; and (iii) a period of aestivation followed by mating and egg-laying in the rainy season [16]. Still, many questions regarding the life history of C. lusitanica are open, especially on the early stages of life, the use of underground habitats and dispersal.

2.4. Population Modeling

The population demography of C. lusitanica was modeled with Vortex 10 [31]. The variables required for modeling fell into one of four categories: (i) inferred from general amphibian knowledge (e.g., breeding system); (ii) known or inferred from the literature, including present paper (adult age distribution, sex-ratio, clutch size, adult mortality, population size at the start of the experiment, etc.); (iii) virtually unknown and modeled over a wide range of possible conditions (carrying capacity, K); and (iv) the parameter of interest as hypothesized to be affected by the planting of gum trees (larval mortality, q_l). Scenarios in which larval mortality was not uniform over time (e.g., “continuous change” or “broken stick” models) were not carried out because of the lack of data over the 1996–2009 period. The life table produced by Vortex indicates that the size of the breeding population is typically 1/5th–1/6th of the total population. Vortex was run for 1000 iterations per set of variable combinations. For details on how the demographic parameters were determined see Appendix 3 along with Figure A3. In the below Ň refers to population size estimates, Ń refers to census population size and Ñ refers to modeled population size and q_l and q_b are larval and breeding adult annual mortality rates, respectively.

3. Results

With data from the literature and colleagues included, estimates on the C. lusitanica adult population size (Ň_b) in the dam were made on seven annual occasions (Appendix 4). Larval counts (Ń_l) for the Silveirinhos brook were made on five annual occasions. Counts on the numbers of adults (Ń_b) and embryos (Ń_e) in the Poço do Inferno mine galleries were made on 12–16 annual occasions. The size of the adult C. lusitanica population in the Silveirinhos dam was estimated at Ň_b = 1324 with a standard error (SE) of 265 in 1977, at Ň_b = 1477 ± 207 in 1992 and at Ň_b = 113 ± 56 in 2010. In 1977, the number of larvae dip-netted in the adjacent 200 m long section of the Siveirinha brook was maximally 81 first-year larvae and eight second-year larvae. In 2015, these numbers were five and three, respectively. The maximum number of adults observed in the Upper mine was Ń_b = 78 in 1977 and 279 in 2014. In the Lower mine the numbers were 100 in 1977 and 336 in 2014. The embryos numbers counted in the mines rose gradually from Ń_e ~650 in 1976 to ca. 3550 in 2014. The most telling data points are probably those of 1977 and 1995 when accumulate counts of freshly deposited eggs were made over 16-25 visits spanning the entire reproductive period. These counts gave Ń_e ~ 550 in 1977 and ca. 5000 in 1995. In 1977, the ratio of juveniles to adults was Ń_j/Ń_b = 0.42 along the Silveirinhos brook (751 salamanders in 17 visits) and 0.55 along the Poço do Inferno brook (116 salamanders in 11 visits). No juveniles were observed inside the dam or in the mines. Over the 39 years of study, the pattern is that of a sharp population decline at the Silveirinhos and a substantial increase at the Poço do Inferno (Figure 2).

Figure 2. Numbers of Chioglossa lusitanica observed and population sizes estimated in the Serra de Santa Justa in northwestern Portugal over the 1976–2015 period. (Top panel) Embryo (round symbols) and adult numbers (squares) counted in two mine galleries at Poco do Inferno. The 1977 and 1995 cumulative embryo counts are shown separately with solid round symbols. (Bottom panel) First-year larvae counted in the Ribeira de Silveirinhos (crosses) and adult population size estimates from capture–mark–recapture in the “dam” (squares, Ň_b solid symbols, Ň_b ± SE open symbols). For numerical detail see Appendix 4.

Figure 2. Numbers of Chioglossa lusitanica observed and population sizes estimated in the Serra de Santa Justa in northwestern Portugal over the 1976–2015 period. (Top panel) Embryo (round symbols) and adult numbers (squares) counted in two mine galleries at Poco do Inferno. The 1977 and 1995 cumulative embryo counts are shown separately with solid round symbols. (Bottom panel) First-year larvae counted in the Ribeira de Silveirinhos (crosses) and adult population size estimates from capture–mark–recapture in the “dam” (squares, Ň_b solid symbols, Ň_b ± SE open symbols). For numerical detail see Appendix 4.

Following the application of Fluorescein in the Upper mine the concentrations in the Poço do Inferno brook were 0.26 µg Fluorescein/L adjacent to the mine entrance and 1.49 µg Fluorescein/L at 170 m downstream. The control sample (upstream of the Upper mine) was under the detection limit.

I let Vortex determine the values of q_l at which a hypothetical Chioglossa population would retain a 10,000 initial population size at year 100 with demographic parameter values as in Appendix 3. A stable population (rate of increase r_stoch = 0 [31]) is obtained at q_l = 0.69 when q_b = 0.22 and at q_l = 0.62 when q_b = 0.28 (Figure 3). A ten-fold population size increase, such as observed at Poço do Inferno, is associated with q_l = 0.51 when q_b = 0.22 and with q_l = 0.38 when q_b = 0.28. Demographic modeling further indicated that the recent low population size at Silveirinhos may be associated with a uniform larval mortality since the planting of gum trees in 1980 of q_l = 0.76–0.79 (2010, 30 years) (Figure 4). The figure also shows that K has little effect on such a declining population, in particular if q_l > 0.70.

Figure 3. Population size development in a hypothetical population of Chioglossa lusitanica as a function of adult mortality (q_b, vertical axis) and larval mortality (q_l, horizontal axis) modeled with Vortex. The curved line indicates parameter combinations at which population size is stable (r_stoch = r = 0 [31]). Higher and lower q-values result in declining and increasing population sizes, respectively. The situation in which adult mortality is higher than larval mortality (parameter space shaded, q_b > q_l) is considered unrealistic. For details see text and Appendix 3. Round symbols refer to parameter combinations determined for Poço do Inferno where population size is increasing and square symbols refer to parameter combinations determined for the declining Ribeira de Silveirinhos population (symbols as in Figure 2).

Figure 3. Population size development in a hypothetical population of Chioglossa lusitanica as a function of adult mortality (q_b, vertical axis) and larval mortality (q_l, horizontal axis) modeled with Vortex. The curved line indicates parameter combinations at which population size is stable (r_stoch = r = 0 [31]). Higher and lower q-values result in declining and increasing population sizes, respectively. The situation in which adult mortality is higher than larval mortality (parameter space shaded, q_b > q_l) is considered unrealistic. For details see text and Appendix 3. Round symbols refer to parameter combinations determined for Poço do Inferno where population size is increasing and square symbols refer to parameter combinations determined for the declining Ribeira de Silveirinhos population (symbols as in Figure 2).

Figure 4. Adult population size of Chioglossa lusitanica at the Ribeira de Silveirinhos modeled with Vortex (log Ñ_b median value, vertical axis) as a function of larval mortality (q_l, horizontal axis) modeled with Vortex under a carrying capacity of K = 33,000. The year is 2010, i.e., at 30 years after the widespread planting of gum trees. The dots drawn in for low q_l values represent the modeled variation in carrying capacity of, from top to bottom, 100,000, 70,000, 50,000, 20,000, 10,000 and 7000. Note that the size of the breeding population is about 1/5th–1/6th of the total population; for details see Figure A3 in Appendix 3. The horizontal lines represent the adult population size as estimated in 2010 (logŇ_b ± SE) and the interrupted vertical lines indicate the values of larval mortality at which modeled and observed data coincide, at around q_l = 0.78 (0.77 < q_l < 0.80).

Figure 4. Adult population size of Chioglossa lusitanica at the Ribeira de Silveirinhos modeled with Vortex (log Ñ_b median value, vertical axis) as a function of larval mortality (q_l, horizontal axis) modeled with Vortex under a carrying capacity of K = 33,000. The year is 2010, i.e., at 30 years after the widespread planting of gum trees. The dots drawn in for low q_l values represent the modeled variation in carrying capacity of, from top to bottom, 100,000, 70,000, 50,000, 20,000, 10,000 and 7000. Note that the size of the breeding population is about 1/5th–1/6th of the total population; for details see Figure A3 in Appendix 3. The horizontal lines represent the adult population size as estimated in 2010 (logŇ_b ± SE) and the interrupted vertical lines indicate the values of larval mortality at which modeled and observed data coincide, at around q_l = 0.78 (0.77 < q_l < 0.80).

More precise figures are derived from plotting the proportion of modeled population sizes (Ñ_b) as falling inside or outside the Ň_b ± SE range, with q_l varying over the 0.5–1.0 range (Figure 5). The 1984 (4 year) and 1994 (14 years) population size estimates do not help to secure q_l estimates, presumably because the population had little opportunity to already decline (1984 [3]) or because the precision was poor (1994 [32]). The fairly precise 1992 estimate (12 years) is incompatible with q_l < 0.58 and q_l > 0.83 and the estimate of near-extinction in 2012 (32 years) is incompatible with q_l < 0.70. By its own the 2010 (30 years) Ň_b estimate is the most informative with a range of 0.70 < q_l < 0.86 and a best estimate of q_l = 0.78. Considering all data points, the best estimate for q_l is in the 0.70–0.83 range (Figure 5). Under a higher adult mortality (from q_b = 0.22 to q_b = 0.28) the q_l estimates decrease by ca. 8% points (e.g., from 0.73 to 0.65; Appendix 5 along with Figure A4). New and additional information on this project may be placed at http://www.repository.naturalis.nl/record/569285 under the sole responsibility of the author.

Figure 5. The probability of modeled population sizes of Chioglossa lusitanica at Ribeira de Silveirinhos falling within the population size range estimated in the field (Ň ± SE), as a function of larval mortality (q_l, horizontal axis). Populations were modeled with Vortex for 4–32 years following the widespread planting of gum trees in 1980. The results were fitted to a Gaussian distribution with Prism 6 software [33] (r² > 0.95) and then log-transformed. Support is maximal for q_l = 0.73 (purple bar, p > 0.10). The grey rectangle indicates the 0.70 < q_l < 0.83 range over which none of the Ñ - Ň combinations are implausible (p > 0.05). The green bar highlights the q_l = 0.78 value already retrieved for 2010 in Figure 4. Adult mortality q_b is set at 0.22; for results with q_b = 0.28 see Appendix 5.

Figure 5. The probability of modeled population sizes of Chioglossa lusitanica at Ribeira de Silveirinhos falling within the population size range estimated in the field (Ň ± SE), as a function of larval mortality (q_l, horizontal axis). Populations were modeled with Vortex for 4–32 years following the widespread planting of gum trees in 1980. The results were fitted to a Gaussian distribution with Prism 6 software [33] (r² > 0.95) and then log-transformed. Support is maximal for q_l = 0.73 (purple bar, p > 0.10). The grey rectangle indicates the 0.70 < q_l < 0.83 range over which none of the Ñ - Ň combinations are implausible (p > 0.05). The green bar highlights the q_l = 0.78 value already retrieved for 2010 in Figure 4. Adult mortality q_b is set at 0.22; for results with q_b = 0.28 see Appendix 5.

4. Discussion

4.1. Population Development at Ribeira de Silveirinhos

The adult population size of C. lusitanica in the Silveirinhos dam was in the order of Ň = 1300–2200 in the 1977–1994 period and ca. 100 in 2010 (Figure 2). Observations at 11 monthly visits in 2012–2013 yielded ten sightings and just two salamanders were caught, so that a capture–mark–recapture estimate was impossible to make [34]. The data indicate a decline in C. lusitanica adult population size of one or two orders of magnitude. However, some assumptions are implicit to this conclusion and for interpreting the data it is important to understand the spatial dynamics of C. lusitanica at the Silveirinhos brook and dam system.

The schist and quartzite structure of the Serra de Santa Justa does not provide deep shelter and the thin layer of soil dries out quickly. The Silveirinhos dam serves as a summer hiding site for the adult population that at wetter periods of the year is distributed along the brook. The decision to migrate to the dam or stay put depends on the quality and depth of the shelter the individual salamander is in [11]. The total length of the Silveirinhos over which Chioglossa salamanders migrate to the dam is not known, but distances of >700 m have been recorded [7]. In summer 1994, the population size of the dam was estimated at Ň_b = 1974 and the density of adult C. lusitanica in winter 1994–1995 was estimated at 11.4 ± 2.16 per m of brook [32]. Arntzen concluded that 60% of adult salamanders in the 200 m stretch downstream of the dam used the dam for aestivation, so that a length of ca. 330 m of the brook would account for the numbers estimated for the dam [8].

Is the decline real? One explanation for the low numbers of adult C. lusitanica in the dam in recent years is that migration for aestivation is no longer required, for example because the newly created gum tree terraces provide deep and sufficiently moist shelters year round. Another possibility is that low numbers were observed in exceptionally wet years in which migration for shelter from drought was not required. However, these explanations do not hold since the low number of adults in the dam is paralleled by low numbers of larvae in the brook (Figure 2, Appendix 4). In addition, adults were rare along the brook in winter 2014 and in spring 2015. I conclude that the current C. lusitanica population size at Silveirinhos is one or two orders of magnitude down relative to the 1977–1994 period. At this low size, the population may be especially prone to stochastic events, eventually leading to local extinction [31,35]. Is the population collapse at Silveirinhos to be attributed to (i) the gum tree plantations created in 1980? Other, not mutually exclusive, explanations are (ii) destruction of the sites that C. lusitanica uses for egg-deposition; (iii) increased competition or predation; and (iv) an outbreak of disease.

4.2. Potential Reasons for the Decline

Gum tree plantations: The streambed of the Silveirinhos, along with a vegetated zone on either side of the brook, stayed outside the plantation and were not structurally altered. The salamander decline is thought to operate through a reduced aquatic run-off causing the stream to fall dry earlier in the season and over a larger section of the valley. In rip-plowed plantations, gum tree transpiration rates remain high even when water supply from the soil would be dwindling [36]. This would in first instance affect the running sections of the brook, which is the larval niche of C. lusitanica [8,37]. Drought would therewith affect C. lusitanica more than the other local amphibians. Considering a generation time of ca. 6 years ([19], Appendix 3), a gradual decline in C. lusitanica population size would be expected from 1980 onwards. While the precision of the dam population size estimates is low, especially for the crucial year 1994 (Figure 2), the larval count for that year confirms that the population size of C. lusitanica was yet substantial. A sharp decline in population size appears to have taken place only after 1994, which is 14 or more years after the creation of the gum tree plantation. For the population demographic modeling exercise to shed light on this issue, see below.

Destruction of egg-deposition sites: At the Ribeira de Silveirinhos, the places where C. lusitanica lays the eggs are unknown. At Poço do Inferno, oviposition takes place in the deeper parts of the mines—and even there, frequently hidden from sight. This suggests that normally the deepest available underground sections of the stream will be used or, at Silveirinhos, perhaps at wet and moist sites deep inside the dam. The gum trees taproots do normally not exceed a length of 3 m [38] but roots grow quickly towards aquatic resources and may then be reaching a depth of 14 m [39]. Gum trees are reported to exploit aquifers through which more water is consumed than is available from precipitation [40] and could therewith destruct Chioglossa egg-laying sites. I have, however, no information that would suggest the physical destruction or drying up of underground habitats to explain the Silveirinhos population collapse.

Increased competition or predation: Along with the gum tree plantation, the Silveirinhos has seen the building of a small causeway spanning the brook at the latitude of São Pedro da Cova, for use by trucks and other forestry equipment (41.15717°N, 8.49575°W, point C in Appendix 1). With this causeway, a pond was unintentionally created of size 17 m × 35 m. The exact date is not known, but pond creation was after 1995 (V. Lima, pers. comm.) and when I found it in spring 2011, the aquatic vegetation was well-developed, indicating an age of several years. Amphibian species, such as the urodeles Lissotriton boscai, Salamandra salamandra, Triturus marmoratus, and the toad Alytes obstetricans, that have larval stages in pond sections of the stream, are more common at Silveirinhos at the present than they were before [41]. This raises the question, has interspecific competition among larvae increased? However, the niche of C. lusitanica larvae is composed of the running and turbulent sections of a brook and only overlaps with the niche of S. salamandra larvae and not with the other species [8,37]. Adult toads (Bufo spinosus) and salamanders (Salamandra salamandra) are mentioned as potential predators [4] but their niches are largely outside the brook margins where C. lusitanica lives and, at Santa Justa, their density is not so high that they could extirpate the C. lusitanica population.

Outbreak of disease: In favor of this explanation is the sudden and massive decline of the C. lusitanica population after 1994, suggesting low versus high mortality rates before and after that date. In 1994, I observed a new arrival (the Green frog Pelophylax perezi) at the Silveirinhos and, in 2011, a small, unidentified species of carp. These vertebrates may have served as a disease vector. Drastic patterns of decline have been reported for a suite of amphibian species in central and northern Spain that were linked to infections with a Chytrid fungus and a Ranavirus, respectively [42,43,44]. The drastic decline of the marbled newts (Triturus marmoratus) in a pond in northern Portugal may have been caused by an Iridovirus carried by introduced Pumpkinseed sunfish (Lepomis gubbosis) [45,46]. Unfortunately, at Silveirinhos no surveys were carried out during the episode of population collapse and no corpses have become available for clinical investigation. However, given that the observed population development of the Golden-striped salamander (Figure 2) is not significantly at odds with demographic models that operate under a uniform larval mortality rate there is, strictly speaking, no need to invoke an ad hoc explanation, such as disease (for a comparison, see, e.g. [47]).

4.3. Population Development at Poço do Inferno

At Poço do Inferno the data for the Upper mine and the Lower mine can be taken together because C. lusitanica migration from one mine to the other is frequent and may occur overnight ([7,48]. At just under one order of magnitude the local increase in C. lusitanica population size is unequivocal and can only to a minor degree be explained by, e.g., inter-observer variation, phenology, or weather conditions. To appreciate the healthy status of this population, it is important to note that the counts refer to visible eggs and visible adults—many more could have been found in the heaps of rock at the deeper sections of the mine galleries that though were left undisturbed. An increase in population size such as observed in the Poço do Inferno population is difficult to associate with a particular value of larval mortality, because of the dependency of q_l on carrying capacity (in expanding populations) and the availability of only count data.

Reproduction in the Lower mine can proceed unhampered. Matings are readily observed and embryos easy to find. Hatched larvae discharge in the stream running through the mine that will eventually assist transport to the outside. Downstream migration along the Ribeiro do Inferno over 80 m would bring the larvae in the Rio Simão. Upstream larval migration seems improbable due to a ca. 4 m high incline adjacent to the Lower mine entrance. The Upper mine also serves for mating and egg-deposition, over the season the larvae getting thinner and less numerous. In March 1977 and October 2010, I observed second-year larvae that may have been feeding on unfertilized eggs, dead embryos or larvae, since no other food sources than C. lusitanica itself appear to be available [49]. The water in this mine is stagnant with no visible connection to the outside, with the exception of periods of excessive rain. The water to overflow the heap of soil at the entrance of the mine has been observed on two occasions (March 1977 [50] and November 1994 [51]). The Fluorescein measured in the lower section of the Poço do Inferno brook indicates the presence of an underground aquatic connection with the Upper mine, but if the openings are large enough for Chioglossa larvae to pass it is not known. All in all, it is difficult to determine to what extent Chioglossa reproduction in the Upper mine contributes to the Ribeiro do Inferno population, but the mine’s continued use for reproduction—quite possibly for 2000 years—suggests that it is unlikely to be negligible. Still, a simple and effective conservation measure would be the removal of the heap of soil at the mine entrance, so that larvae staying behind will be washed out of the mine following periods of rainfall.

At the Poço do Inferno, obvious environmental change has been absent. This edge of the Serra de Santa Justa remained free of large-scale gum tree plantation and the mines and their immediate surroundings are essentially unchanged. Minor changes in the area involve the remodeling of some of the dirt tracks, the local extraction of wood and the creation of fire break zones. Most pine trees fell victim to forest fires and were taken over by gum trees, but the area stayed free of terraced plantations such as at the Ribeira de Silveirinhos (see Figure 23 in [9]). One prominent change just outside the research area is the cleaning of the Simão stream through the installment of facilities for sewage treatment in Valongo in the early 1980s. Prior to this, the Simão was an opaque and smelling stream with plentiful remains of cloth washing and other household rubbish. The Simão valley saw further environmental improvement through the project “Requalificação Paisagística das Margens do Rio Simão” and at present, the stream makes a clean and biologically healthy impression (http://www.valongoambiental.com). In the absence of other environmental change I attribute the drastic population size increase of C. lusitanica at Poço do Inferno to the improved health of the Simão River. The section of the Simão adjacent to the Poço do Inferno (in between Valongo and Campo) may be particularly amenable to C. lusitanica due to the lining of the river borders with dry stone walls and even the slate roofing over a ca. 190 m long section of the stream, presumably constructed with the stone debris from the Lower mine (Appendix 1).

4.4. Are Eucalypt Plantations to Blame? Suggestions for Future Investigations

Serious environmental problems have been attributed to gum tree plantations, foremost through a substantial reduction of total water yields when the trees are planted in semi-arid water catchment areas [52,53]. Toxins are released through evaporation, transferred to the soil by fog drip and leached out from leaf litter by rain [54]. More controversial is the impact the plantations would have on soil nutrients and erosion [1]. Gum tree plantations are typically associated with low biodiversity, including bird wildlife [55,56] and have a negative effect on ground arthropods and on the many species of vertebrates that depend on them as prey [57]. Others concluded that gum tree plantations have less diverse flora and fauna than indigenous forests [52]. In central Portugal it was reported that streams running through gum tree plantations had lower number and diversity of aquatic invertebrates than streams running through deciduous forests, but the effect is not consistent [58,59]. The effect of frequent logging through clear cutting may drastically affect the aquatic invertebrate and fish communities in adjacent riverine habitats ([59] and references therein). However, the supposed negative effects of gum trees on wildlife are frequently inconsistent or unrepeatable and alternative explanations usually cannot be ruled out. Moreover, effects may be different for different species [60]. As for C. lusitanica, negative effects of gum tree plantations are assumed to be indirect because the postmetamorphic salamanders live mostly in the buffer zone between the rip-plowed stands and the stream courses. The larval niche is likely to have become reduced, but regretfully no data on change in flow rates of the Silveirinhos are available.

The surveying of two Chioglossa populations over four decades has yielded some remarkable results. At the downside is a decline near to extinction of the Silveirinhos population that may be attributed to the monoculture of gum trees since 1980. The decline is late and sharp, but demographic models indicate that a pattern like this might be expected in ca. 12% of replicates. Nonetheless, the pattern of decline is so strong and abrupt as to be at odds with observations that C. lusitanica is regularly present in gum tree affected areas elsewhere [4]. Another example of an amphibian population possibly affected by an introduced plant species is Fowler’s toad [61].

An outcome of the survey at the upside is the drastic and consistent population size increase of C. lusitanica at the Poço do Inferno. This may be attributed to the improved health of the adjacent Simão River. Moreover, this would suggest that C. lusitanica lives around larger streams than currently appreciated—for a description of stream size and types with and without the species see [4]. This hypothesis can be tested by the marking of adult salamanders along the Simão and their later retrieval in the mines. Chioglossa lusitanica living around large streams would also explain the observation of dense populations without local recruitment, such as at the tiny Salto brook adjacent to the Sousa River with ca. 6 adults per m of the brook [14,15]. The main habitat of C. lusitanica in this area could be the adjacent Salto River. I advocate research to be undertaken to test for the regular presence and reproduction of C. lusitanica in high-order riverine habitats. However, large rivers of the size of the Douro are probably not a good habitat, as suggested by the barrier function of that river revealed by a substantial level of genetic differentiation across the river [28].

In search of a definitive answer on the effects of gum tree plantations on C. lusitanica, I advocate the resurveying of localities with and without newly created plantations that were prospected before, such as under the “EU-life” project in northern Portugal [62] and in Galicia [4]. I also recommend that the population size in the dam of the Silveirinhos be estimated at intervals of five to ten years. A population size with reasonable precision may be obtained with ca. one week of work, if carried out during and following rainy summer and autumn nights. The C. lusitanica population may rebound, either from the small number of local survivors, or from new arrivals that would colonize the area along the lower stretches of the Silveirinhos brook. A recovery to the level of the late past century would indicate the absence of strong effects from the gum tree plantations on C. lusitanica. Where C. lusitanica lays its eggs if there are no mines remains unclear. I suggest that the regular places are deep underground sections of smaller and larger streams. However, testing this proposition will probably have to wait for technical innovations to take place.