Trends and weather events in long-term densities of an insular lizard

Abstract

Lizard populations have been traditionally considered unusually stable. However, there are very few examples to verify this hypothesis on long-term studied populations. Over almost four decades, we have obtained long time series of the abundance of Lilford’s Wall Lizard, Podarcis lilfordi, from all known populations of Menorca (Balearic Islands, Spain). Lizard densities were estimated from line transects data using N-mixture models. Irregular time series were then checked for the presence or absence of a monotonic trend and for density dependence of each population. With multiple regression models we studied the relationship between weather events and the annual abundance of lizards. Density dependence was detected only in two populations: Addaia Petita and Binicodrell Gros. The analysis with EGSS and OUSS models, as well as the results with a Bayesian analysis, showed that most populations seem stable over the years, with a positive and statistically significant trend in five of the 17 populations under study. Only on Sanitja Island the population of lizards exhibits a significantly negative trend. Weather events have a very variable influence on each of these populations. In two small islets, Mel and Ses Mones, the annual number of cloudy days have a negative effect on the abundance of lizards. In Bledas Island, with an exposed geographical situation, we observe a significantly negative effect of all three weather events considered: storms, cloudy days and strong winds. Surprisingly, in Aire and Addaia Gran islands, the annual number of days with storms had a significantly positive effect, probably because the marine subsidies contributed to trophic availability of these two islands by the storms. This work is a first attempt to establish the real abundance of this species, its long-term trend and the influence of weather events on densities of a large set of populations. The relevance of weather events is strongly different on each population and depends on environmental conditions of each population, especially lizard abundance itself, the island surface, geographical location and degree of exposure to weather events.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-34615-x.

Introduction

Understanding natural fluctuations of population size is a central problem of evolutionary ecology and conservation¹. Vertebrate populations can fluctuate through time due to several factors. Even so, populations of some groups, as lizards, have been traditionally considered unusually stable. Schoener^2,3 pointed out that lizard populations can be substantially less variable through time than most of the vertebrate and arthropod species studied at that time. Population stability is particularly important from the viewpoint of conservation strategies because it is widely recognized that more variable and smaller populations can suffer a higher risk of extinction^4–6. Andrews & Wright⁷ considered premature the presumption of stability in lizard populations. These authors studied for 22 years a population of Anolis limifrons on Barro Colorado Island (Panamá) and found that it was far from stability. The fluctuations of the density of this lizard were correlated with the variation and distribution of annual rainfall⁸. It was concluded that wide fluctuations in the density were mainly governed by annual rainfall⁷.

Along annual fluctuations of lizard populations, two different paradigms can be recognized. Density dependence hypothesis assumes that a population tends to return to an equilibrium situation when it is displaced from it⁹. While in other cases, the hypothesis is that density itself would not be a relevant variable and population fluctuations would depend on external biotic (available food resources, predation pressure, competition or diseases) and/or abiotic factors (temperature, rainfall, soil composition or refuge availability⁹). In a meta-analysis on density dependence, Brooks & Bradshaw¹⁰ found evidence of density dependence in about 20 species of ectotherms, out of a total of 1198 species under study.

In a comparison of squamate ectotherms with endothermic vertebrates, we observe that ectotherms are less likely to be limited by the availability of food resources^11–13. Probably the reason is that ectotherms have lower energetic costs¹⁰. In this sense, the abundance of lizards would be typically limited by the availability of habitat features linked with their thermal requirements¹⁴. Habitat characteristics can limit basking sites of a population or any other feature important for thermorregulation^15,16, to overwinter¹⁷ or as anti-predator refuge¹⁸.

It is well-known that climatological variability promotes evolutionary and ecological responses in most organisms^19–21. Ectotherms have metabolic rates closely related to environmental temperatures, so that populations of ectothermic species can potentially be affected to a greater extent by climatological factors²². Adolph and Porter²³ predicted that if two populations of ectotherms live in environments with different thermal regimes, the population of the colder environment will be less active and, consequently, would have a higher survival, but a lower annual fecundity. We can extend this argument in the comparison of the same population along the years, even if it is difficult to make predictions on the changes induced by these climatological factors, especially, in highly unpredictable environments. Anyway, the influence of these abiotic factors on the activity of ectotherms is clear^24–26.

Our knowledge of the long-term influence of climatological factors as precipitation and temperature is relatively good in some groups. For the Lilford’s Wall Lizard, Podarcis lilfordi (Günther, 1874), we did some preliminary analyses in several populations²⁷, concluding that this influence was extremely variable according to habitat characteristics of each population under study. However, the influence of moderate and extreme weather events in populations of insular lizards is poorly known²¹. At least, in two groups of organisms, butterflies and birds, the extreme weather events (EWE) are driving local population dynamics²⁸. Consequently, one would expect a clear effect on populations in the face of these weather events. This effect cannot be clearly predicted, due to the behavioural or demographic compensatory mechanisms that many populations possess²⁹. In the Mediterranean Basin, storms and droughts are extreme or moderate weather events. Thanks to the records of weather stations, today we have precise information, not only of climatic variables, as temperatures, relative humidity, and rainfall, but also of the frequency of these events over the years. Such weather events have a direct influence on all organisms, but especially on ectothermic animals, such as the small reptiles³⁰.

With moderate weather disturbances the risk of extinction can be directly related with the size of the population³⁰. At least, moderate disturbances are considered factors affecting normal activity cycles of ectothermic organisms. In addition, the onset of daily activity from nocturnal retreats is preceded by basking³¹. Thus, the presence of moderate disturbances as storms, strong winds or fully clouded skies, preclude the onset of activity in diurnal lizards. In fact, reptiles that inhabit temperate climate zones are sensitive to low temperatures and can only face them if they retreat to appropriate thermal shelters³². Our own observations at several insular populations of species of the genus Podarcis (Squamata, Lacertidae), confirm that fully cloudy skys, strong winds, or storms, can be considered moderate or extreme meteorological events that prevent the beginning of daily activity. On days with those events, there is an absence of lizard activity.

As a diurnal lacertid lizard, the Lilford´s Wall Lizard is mainly active during sunny days, where it can reach suitable body temperatures for daily activities^33–35. For instance, in open habitats a wind of more than 1.8 km/h produces a significantly lower activity in Anolis allisoni³⁶. This strong effect of wind on activity could be due to the trade-off between the ability to maintain optimal body temperatures, and the ability to maintain water balance^33,37. However, other studies did not detect any significant influence of wind in daily activity of lizards²⁶. Thus, during years with several moderate or extreme weather events, the opportunities to remain active can be significantly reduced. This reduction in the available days for activity could have a direct effect on the growth rates of individuals, their reproductive potential or even their survival probabilities.

In insular populations it has been shown that moderate weather events can be directly related to annual abundance³⁰. Those years with several days with bad conditions would be years in which lizards have a reduced availability of hours for activity to feed and reproduce that can result in a direct effect on their reproductive potential or their probabilities of survival, in short, on their fitness. Initially, this approach would mean that such events always have a negative effect on populations.

Our aim is to analyse the role of these abiotic factors potentially related with annual variation of lizard densities. We address the study of density dependence and trends, as well as the influence of weather events (EWE) in the annual abundance of the Lilford´s Wall Lizard in Menorca populations. We consider three types of weather events: the annual number of registered storms, the annual number of days that the sky remained fully covered, and the annual number of days with a prevalent wind above 55 km/h (see below). Our main hypothesis is that the annual variation of these extreme weather events has a direct influence on long-term lizard abundances.

Methods

Study period and area

The study was carried out from 1982 to 2024, during spring and summer in all known populations of P. lilfordi from Menorca (Balearic Islands, Spain; Table 1 and Fig. 1). Islets inhabited by lizards have a very variable surface (Table 1), plant cover and human pressure³⁸. Menorca has a Mediterranean climate, with mild temperatures and highly seasonal rainfall, abundant rains in winter and very dry summers.There are frequent and strong winds, principally from the north³⁹.

Table 1.

Time series of density from 17 populations of P. lilfordi under study.

Population	Time series	Surface (ha)	EGSS density	OUSS density	μ (EGSS − REML) (95% CI)	λ	τ (p-value)	L (p-value)
Addaia Gran	19	5.55	1299.32	930.42	− 0.0116 (− 0.0389 − 0.0068)	0.9841	− 0.0322 (0.0598)	− 1.8107 (0.514)
Addaia Petita	10	3.29	292.64	270.0	0.0367 (− 0.3409 − 0.4144)	1.0374	0.491 (0.0423)	− 684.30 (0.007)
Ses Aguiles	9	0.25	3785	4609.96	0.0221 (− 0.0898 − 0.1341)	1.022	0.0556 (0.9170)	− 0.2759 (0.33)
en Carbó	10	0.14	498.5	1062.41	0.0885 (− 0.1397 − 0.3167)	1.0925	0.491 (0.0423)	− 6.0166 (0.331)
ses Mones	9	0.62	586.2	916.57	0.0945 (− 0.2359 − 0.4249)	1.099	0.278 (0.3480)	− 0.3681 (0.37)
Colom	21	51.14	720.9	787.16	0.0037 (− 0.0116 − 0.0190)	1.0037	0.0571 (0.7398)	− 0.2489 (0.415)
Mel	9	0.27	732.94	982.16	0.0732 (− 0.0868 − 0.2331)	1.0759	0.278 (0.3480)	− 0.9786 (0.435)
Aire	34	29.8	710.68	1701.82	0.0256 (0.0649 − 0.1162)	1.0259	0.109 (0.3737)	1.5598 (0.105)
Binicodrell Gros	17	0.29	514.18	707.37	− 0.0125 (− 0.1439 − 0.1187)	0.9875	0.162 (0.3870)	− 675.48 (p < 0.0001)
Binicodrell Petit	7	0.02	1025.56	2529.97	0.2341 (0.1331 − 0.3351)	1.2638	0.81 (0.0163)	− 15.1584 (0.866)
Bledas	10	4.35	1010.78	1302.05	0.0212 (− 0.1856 − 0.2280)	1.0214	0.667 (0.0165)	− 8.7355 (0.402)
Rey	14	4.08	1783.75	993.05	− 0.0141 (− 0.1223 − 0.0941)	0.8687	0.0769 (0.7426)	0.1701 (0.074)
Sanitja	25	8.59	2218.56	1245.98	− 0.0234 (− 0.0885 − 0.0417)	0.9769	− 0.287 (0.0471)	− 3.2416 (0.366)
Sargantana	14	2.3	808.76	899.64	0.0065 (− 0.0165 − 0.0295)	1.0065	0.0769 (0.7426)	− 0.9417 (0.459)
Tosqueta	14	0.29	1178.23	1306.18	0.0062 (− 0.0450 − 0.0575)	1.0063	0.154 (0.5022)	− 0.0523 (0.271)
Rovells	9	0.24	585.99	1115.81	0.1638 (0.0798 − 0.2477)	1.1779	0.611 (0.0286)	− 5.6518 (0.147)
Porros	9	0.05	405.66	1303.07	0.0165 (− 0.1623 − 0.1953)	1.0167	0.167 (0.6022)	− 738.1579 (0.087)

The first column indicates the name of the population (see Fig. 1), then, we show the sample size of annual density estimations for each population, the surface in hectares of each population, as well as densities estimated by EGSS and OUSS models (individuals/hectare). We also include the μ value and its 95% confidence intervals calculated from the model EGSS-REML, the corresponding λ value, the results of Mann–Kendall test (τ and corresponding p-value), and the results of PBRLT test of density independence (L and corresponding p-value). Statistically significant p-values of τ and L are in green color when there is a positive trend and in red color in negative trends (see more details in the text).

Fig. 1.

Geographical situation of known populations of the Lilford’s Wall Lizard, Podarcis lilfordi, in coastal islets of Menorca (Balearic Islands, Spain).

Species under study

The Lilford´s Wall Lizard, Podarcis lilfordi (Squamata, Lacertidae), is a medium-sized lizard with a maximum snout-vent length of 81 mm in males and 75 mm in females^40,41. P. lilfordi is a member of the first post-Messinian faunal assemblage reported from Menorca Island^42,43. During the Holocene, around 2,000 years ago, the species became extinct on the main islands of Mallorca and Menorca, probably because of the introduction of terrestrial predators by Romans⁴⁴. It now survives only on the small islets found around the coasts of Menorca, Mallorca and Cabrera Archipelago.

Podarcis lilfordi is structured into four main clades^45–47. Employing the nested clade phylogeographic analysis (NCPA), Terrassa et al.⁴⁷ proposed that lizard populations of Menorca, forming an isolated clade, were largely panmictic prior to the extinction in the main Menorca Island. We surveyed all known populations of Balearic Islands²⁷. However, in most of these populations, we obtained irregular time series of short length. For this reason, in this study we focus our attention only in populations from Menorca coast (Fig. 1), a well-defined clade from which we have longer time series, to study trends, density dependence and the influence of abiotic factors.

Lizard abundances

Field counts of lizards were performed by one of the authors (VPM) during the first years of study and then, from 2009, together by the two authors. Abundance estimates of lizards were carried out with the line transect methodology. Line transects covered almost the entire surface of the islands with vegetation cover or the same paths each sampling year²⁷. In all cases, we did line transects with a bandwidth of 2 m on each side of the line, estimating the perpendicular distance of each lizard to central line of the transect^48–50. Densities were estimated with N-mixture models⁵¹ using the “unmarked” package within R environment. N-mixture models can provide accurate abundance estimations of small vertebrates^52,53. We used the model of abundance from distance sampling with the fitting function “distsamp” that fits a multinomial Poisson model to distance sampling data⁵⁴. The probability of detection for lizard densities was modeled as a function of the perpendicular distance (d) to line transect using the half-normal detection function⁵². For each analysis, the density, in number of individuals per hectare, and the standard error (SE) is provided (Additional file 3). For lizard transects, we also calculated the value of the probability of detection, g(x), given by “unmarked”⁵².

Because weather conditions can affect the probability of lizard detection, all transects were carried out during the peak of morning lizard activity period (between 06:30 and 08:30 GMT) in spring and early summer. However, on some occasions, we had to repeat transects when weather conditions changed abruptly, with the appearance of clouds or wind. Our extensive knowledge of the study populations, which were visited several times a year, has allowed us to minimize the uncertainty in these density estimates, obtaining detection probabilities g(x) always much more higher than g(x) = 0.15, which Buckland et al.⁴⁹ recommend as a lower limit value beyond which the obtained values should be truncated.

Time series analysis

We employed several analytical tools to study the time series of lizard abundance from populations of Menorca. Our data are discrete, with observations of abundance through density estimation, at specific times and spaced by variable intervals (Additional file 3). Except in the last 10 years, we have not been able to undertake annual monitoring in all populations. These are, therefore, irregular time series. We applied the “lowess” function, from the basic R package⁵⁵, which smooths a previously defined time series^56,57, and the Mann–Kendall test^58,59 to study the presence or absence of a monotonic linear trend in a time series⁶⁰. The null hypothesis is that there is no monotonic trend in each population. The Mann–Kendall test is robust and, being a non-parametric test, it does not assume any type of distribution in the time series. It is also not affected by a reduced number of points in the time series, nor it is affected by irregularity in the spacing of the data. One of the assumptions of the Mann–Kendall test is that, when there is no trend, the time series data must be independent. This means that, in our case, the annually recorded abundances should not be seriously correlated. We initially tested this assumption with the “acf” function. For these tests we have used the “Kendall” package of the R environment⁶¹.

State-space models (SSM) are the appropriate framework for time series analysis, because they are a better alternative to trend analysis of a time series than log-linear regression⁶². Exponential growth models usually assume either that there is no process noise and that only observation errors exist (EGOE models), or that there are no observation errors and that only process noise exists (EGPN models). Dennis et al.⁶³ and Humbert et al.⁶⁴ showed that the use of an EGSS model, which jointly considers the existence of observation errors and process noise, provides significantly more reliable estimates of population trends. Thus, the model assumes that population growth is governed by a stochastic process independent of density, so that the observed abundances come from log-normal sampling errors⁶³. These and similar models allow to convert deterministic ecological models into statistical analysis tools, incorporating a series of terms that represent the omnipresent stochastic forces that affect all natural populations. This noise is independent of the errors attributable to the sampling method used to estimate population abundance and that is usually called “observation error” (see above). Precisely, the problem of separately estimating the observation error and the environmental noise process is what has been solved thanks to the use of exponential growth models of state space (EGSS^63,65). The EGSS model allows a reliable estimate of the trend of a population⁶⁴. As usual, to estimate the population trend the parameter µ is used, so that if µ = 0, there is no observable trend in the population. While µ < 0 indicates a negative trend and µ > 0 a positive trend. Alternatively, we can express it by means of the parameter λ, perhaps the most common in the ecological literature, so that: µ = ln (λ). Thus, λ = 1 indicates an absence of detectable trend in the population, λ > 1, a positive trend of population growth and λ < 1, a negative trend of population decline. We present the values obtained for these parameters (Table 1), calculated using the script provided by Humbert et al.⁶⁴ in the R environment (Exponential Growth State Space model, version November 28, 2009^63–65).

Dennis & Ponciano⁶⁶ employ a different approach, a stochastic version of Gompertz’s deterministic continuous-time growth model. By transforming the model process to a logarithmic scale, an Ornstein–Uhlenbeck process is obtained (OU⁶⁶). If an observation error component is finally added to the model, an OU state space model (OUSS model) is obtained. The observation error is the variability associated with the discrepancies between the hidden states of the time series, that is, the real values of abundance, and the observations made by us. Frequently, observation error reflects the imprecision of the sampling methodology used⁶⁶. OUSS models also allow for additional statistical inferences, such as testing the density independence hypothesis⁶⁶. Density dependence assumes that a natural population tends to have growth rates regulated by its own density⁶⁷. The test is based on the ratio of the likelihoods estimated by parametric bootstrap (parametric bootstrap likelihood-ratio test, PBLRT, see more details in⁶⁷ and¹¹) of the EGSS and OUSS models. These analyses have been carried out with the scripts (ROUSSE-1. 0. R) provided by⁶³ and the MASS package⁶⁸ in the R environment⁵⁵. Dennis & Ponciano⁶⁶ point out that the results of the density independence test should be interpreted with caution and always based on the biological characteristics of the population under analysis. Brooks & Bradshaw¹¹ already proposed an open approach to the problem of density dependence, starting from the basis that neither total dependence nor total independence are strictly realistic.

The Bayesian approach to time series analysis follows a different strategy than frequentist analysis. In this case, instead of looking for the “best” possible model, Bayesian analysis builds an average model, as a synthesis of many probable models. This technique is called Bayesian Model Averaging (BMA) and allows analyzing the trend of a time series and detecting the existence of abrupt changes using a Bayesian Estimator of Abrupt Changes and Trend Algorithm (BEAST⁶⁹). From a mathematical viewpoint, BEAST breaks a time series Y(t) into four components: trend (T), seasonal variability (S, not employed in our analyses), abrupt changes (θ_t and θ_s) and noise (ε). A key point is that in several algorithms, model parameters are treated as unknown constants to be estimated, while in BEAST these parameters are treated as random variables⁶⁹. The parameters selected in our analyses were: the minimum and maximum number of trend changepoints allowed within the time series, and the minimum and maximum distance between consecutive trend changepoints. Time was a year (annual surveys), period = 1.0 and season =”none”.

Thus, BEAST is an algorithm that detects abrupt changes and trends through fuzzy detection of the probability of occurrence of a perturbation of the time series at any point in time. To do this, BEAST combines many possible models by decomposing the time series, instead of relying on a single model⁷⁰. The results are presented in the form of a multiple graph in which the general trend of the analyzed time series, the situation of points of abrupt changes and their probability of occurrence, if such changes have been detected and the proportion of trends to stability are represented (additional file 1).

We have therefore employed different tools in our time series analysis, such as trend analysis and three model approaches. EGSS models are particularly suitable for equal sampling intervals, or for unequal intervals resulting from data missing at specific intervals that would otherwise be equally spaced. EGSS models are therefore suitable for populations with more regular sampling. However, for populations with several irregular intervals, OUSS models, incorporating density dependence, process noise, and observation errors, would be preferable. Furthermore, the use of both OUSS and EGSS models allows for a robust test of independence (pblrt test), using the EGSS model as the null hypothesis and the OUSS model as an alternative hypothesis for density dependence⁶⁶. Finally, BEAST allowed the detection of abrupt changes and their probability, as well as an alternative analysis of the general trend of the abundance of a given population.

Weather events

For all years under study, the information of weather events was obtained (additional file 3). On Menorca, a 702 km² island, the airport’s meteorological station is the only available source of information on extreme weather events. Although microclimatic conditions may vary across the island’s coastal areas, we think that annual meteorological events affect all the lizard populations studied similarly. Meteorological information was obtained from public data provided by AEMET⁷¹. We recorded three main annual extreme or moderate weather events (EWE): the total number of storms, the total number of days fully covered by clouds, and the total number of days with winds stronger than 55 km/hour (additional file 3). We recorded this information from what we call “valid” years, that is, the number of these weather events during 12 months prior to April of a given year, as the start of the main period of lizard activity. This approach is similar to the procedure employed by several authors when analysing the effects of rainfall on time series of abundances in animal populations (⁷² and references therein). We analyzed with multiple regression models the possible relationship between these three weather events and the annual abundance of lizards in populations from Menorca. We have built multiple regression models in which abundance (number of lizards per hectare) was the dependent variable, while annual weather events were the continuous explanatory variables. As it is not obvious that the relationship of such explanatory variables is linear, in the saturated model we included these weather events variables, their quadratic terms, as well as all the interactions between explanatory variables. From this saturated model, we eliminated step by step the interactions and non-significant quadratic terms. We employed the ANOVA to compare the models produced by this reduction of terms. For each population, we obtained at the end the minimum adequate model^73,74. Model assumptions were checked with the four diagnostic plots provided by the lm function of R.

Results

We analyzed highly variable and irregular time series of lizard populations from Menorca, from only 7 years (Binicodrell Petit) to 34 years (Aire Island, Table 1 and additional file 3). Densities estimated with EGSS and OUSS models are showed in Table 1. In general, both the densities estimated in 2024 and previous years (additional file 3) by means of line transects and those obtained with the EGSS and OUSS models show their maximum values on the smallest islets, such as Ses Aguiles, Binicodrell Petit or Porros (Table 1). The largest island, Colom, showed a density of around 700 individuals per hectare (Table 1).

In general, we found agreement between the results obtained with the EGSS, OUSS, and Rbeast models. This was at least true for the Bayesian analyses with a good fit (R² > 0.40). Populations with a poor fit of the Bayesian model correspond to stable time series (Table 1 and supplement). In those cases where a significantly positive trend was detected (Addaia Petita, en Carbó, Binicodrell Petit, Bledas, and Rovells), the Bayesian model fit were high, and this positive trend was confirmed. The same is true for the only case with a significantly negative trend detected (Sanitja, see also⁷⁵).

Density dependence and trends

According to the results of the PBRLT test (additional file 3), we only detected a density dependence in the time series of Addaia Petita and Binicodrell Gros (Table 1). In the remaining populations, there is an independence of densities of the time series. In addition, μ and λ values, together with τ values, indicate that most populations are stable over the years, with five populations showing a positive and statistically significant trend (Addaia Petita, En Carbó, Binicodrell Petit, Bledas and Rovells). Finally, only the population of Sanitja Island showed a negative and significant trend (Table 1; additional file 3), that we study in detail elsewhere⁷⁵.

The Bayesian analysis gives similar results from the point of view of stability and trends. In general, positive or stable trends are dominant (additional file 1), even if fitted values of the average model are poor in several cases (ses Aguiles, ses Mones, Colom, Mel, Binicodrell Gros, Tosqueta and Porros). An important negative component was showed in Addaia Gran, Rei, Sanitja and Tosqueta (additional file 1).

Influence of weather events

The number of days with strong winds only had a negative and significant relationship with lizard density in Bledas Island (t = -3.552, p = 0.016). In this population, we have also found a similar relationship in the case of the number of days with storms (t = -3.120, p = 0.026) and the number of days with overcast skies (t = -4.265, p = 0.007). In addition, storms also had a negative and marginally significant effect in the small islet of Binicodrell Petit (t = -2.541, P = 0.051). In Mel we found a negative effect of the annual number of days with overcast skies (t = -3.059, p = 0.018). Surprisingly, in Addaia Gran and Aire islands, the annual number of days with storms had a positive and statistically significant effect (Addaia Gran: t = 3.919, p = 0.001; Aire: t = t = 2.723, p = 0.014), Fig. 2). In nine populations: ses Aguiles, En Carbó, Ses Mones, Binicodrell Gros, Rei, Sanitja, Sargantana, Tosqueta and Porros, we did not detect any influence of the three weather events on annual time series of lizard abundance.

Fig. 2.

Relationship between the density of lizards and the number of annual storms recorded on the islands of Aire and Addaia Gran. For each population, the R-squared value (R²), the F-statistic, its corresponding p-value, and the equation of the linear regression are indicated. Regression lines and confidence intervals (95%) are showed.

Discussion

We can consider that the annual changes in the density of reptile populations are one of the central themes of their natural history, both on islands and on continents. But it is not common to have reliable data with long time series that illustrate such long-term changes. Only a few specific studies exist, since the pioneering work of Heatwole⁷⁶. In this study, two Australian populations of Amphibolurus ornatus and Amphibolurus fordi showed a remarkable stability from year to year, despite significant fluctuations in abiotic conditions such as temperature and humidity⁷⁶.

The existence of very high lizard densities in small islets and medium or low densities in larger islands is a common observation in Balearic Islands^{27,38,41,77–79} in other archipelagoes of the Mediterranean Basin^80–82 and, in general, in several insular ecosystems (e.g.⁸³). Density in ectothermic species appears to be inversely related to body size, with small lizards generally exhibiting high densities⁸⁴. Furthermore, environmental productivity is predicted to positively influence ectotherm density, as is the case of the Lilford’s Wall Lizard, living in places with few or no competitors²⁷.

In P. lilfordi we found examples of both paradigms on the regulation of population density. The paradigm of density-dependence¹⁰ seems to be present in Addaia Petita and Binicodrell Gros, as well as in other populations of the Cabrera Archipelago not studied here²⁷. The mechanistic paradigm of abundance regulation due to factors external to the population⁸⁵ is predominant in the rest of the populations of Menorca. This result is what might be expected since ectotherm populations are usually more affected by external factors. There is a well-known case of density dependence in the Common frog, Rana temporaria⁸⁶ but, in general, density dependence is rare among ectotherms. In their analysis of 22 ectothermic species with time series of at least 10 years, Muniz et al.⁸⁵ included only three Lacertilia, Anolis limifrons, Cyclura cyclura and Sceloporus graciosus. In none of these species did they detect a density dependence. However, Tinkle et al.⁸⁷ found density dependence in their long-term study of Sceloporus graciosus. The populations of the Lilford’s Wall Lizard in which a density dependence seems to be detected are, in all cases, of small extension. We have not found such dependence in large populations such as those of Aire or Colom, where external factors can prevail as regulators of abundance (see below).

The populations of the Lilford’ Wall Lizard that inhabit the coastal islets of Menorca are, for the most part, in a stable situation, despite the existence of significant annual fluctuations in density (additional files 1 and 3). This is also the situation in insular populations of lizards from other latitudes³. In Addaia Gran we detected a negative, but non-significant, value in the trend analysis (Table 1, additional file 3). Likewise, as an exception, in Sanitja we detected a negative and statistically significant trend (Table 1), probably because of the introduction of a group of goats on the island during several years. Goats caused the impoverishment of trophic resources and a significant lower consumption of plant matter⁷⁵.

The carrying capacity of a population is tied to environmental conditions. Thus, weather events are a density-independent source of mortality¹⁴. On the 1,500 hectares of Barro Colorado Island (Panamá), Andrews⁹ found a weak relationship between fluctuations in the density of Anolis limifrons and rainfall. Also in a continental area, the desert of the Chuckwalla Valley in California (USA), the population size of Cnemidophorus tigris was directly related to rainfall, which in turn was related to primary productivity and arthropod abundance⁸⁸. Apparently, the relationship between this abiotic factor, in turn related to the greater or lesser availability of trophic resources, is only relevant in populations of a certain extension where stochastic factors have less relevance in shaping the abundance of lizards. In this study we do not address the long-term effects of rainfall and temperatures²⁷ but only the influence of moderate or extreme weather events.

The geographical location of the coastal islets of Menorca and their orography seems also to have an important role in relation with the influence of weather events on the abundance of lizards. Only in the case of Bledas, we find a negative and significant relationship between strong winds and density. Bledas is an abrupt island, located on the northern coast of Menorca and exposed to the influence of the dominant north winds. The orography of Bledas promotes a strong soil erosion of the eastern slope of the islet, the only one where lizards are present (Fig. 3).

Fig. 3.

The upper area of Bledas Island seen from the summit and looking south. The most important shrub cover is from Ephedra fragilis. Most of the island’s lizard population is concentrated in this area, with a much lower density on the very steep eastern slope (on the left) and even lower on the almost completely steep western slope.

Moreover, the combined effect of orography and geographical location apparently causes Bledas to be the only population in which there is a negative and statistically significant effect of the three weather events under study: strong winds, storms and cloudy days. The annual number of cloudy days had also a negative and significant effect in the small Mel Islet. Mel Islet is located inside a brackish lagoon, the Albufera d’es Grau, a unique location of P. lilfordi in the Balearic Islands (Fig. 1).

More surprisingly, in Addaia Gran and Aire islands, the years with a higher number of storms are those in which higher densities of lizards have been recorded (Fig. 2). This is not the first known case in which an extreme and apparently negative weather event has a positive effect on a population. Leuenberger et al.⁸⁹ showed that experimentally induced ice storms had positive effects on the abundance of certain generalist bird species, because of the increase in habitat heterogeneity and a better adaptation of such species to weather changes. These effects have also been observed in the face of natural storms, once again in generalist species such as the European Starling, Sturnus vulgaris⁹⁰, the Dark-eyed Junco, Junco hyemalis⁹¹ and the lizard Anolis sagrei from Bahamas Islands^92,93.

Addaia Gran and Aire are highly exposed to storms, especially during winter, because on both islands there are large areas with lower coasts and beaches that are swept away by the waves during storms. Storms temporarily disrupt activity, but at the same time, they are the source of remains and marine organisms to the coasts of the islands. In this dynamic environment of coastal zones, waves deposit marine invertebrates and detach part of the seagrass biomass and transport it to adjacent terrestrial systems. This import of organic matter contributes significantly to the function of biological communities of these coastal habitats⁹⁴. Large banquettes of the Neptune seagrass, Posidonia oceanica, are currently formed during winter and spring storms at several points of the northern coast of Aire Island and western coast of Addaia Gran Island. These wracks of P. oceanica are quickly colonized by several groups of macroinvertebrates in a succession of isopods, beetles, flies, and other arthropods⁹⁵. These groups are currently important components of the diet of P. lilfordi in Addaia Gran and Aire islands^96,97. It has been proposed that flexibility in habitat selection and diet breadth could play an important role in explaining this positive effect of weather disturbances¹⁹. Thus, the positive effect of these events could be related to adaptive resilience and rapid compensation, thanks to the deployment of highly plastic behavioural strategies for obtaining food and to select suitable habitats¹⁹. An adaptive resilience is a typical trait of foraging ecology of the Lilford’s Wall Lizard⁹⁷. We frequently observed lizards actively foraging both on and inside the wracks of P. oceanica (Fig. 4). Also, we made direct observations of lizard eating jellyfishes, Pelagia noctiluca, and mass stranding of the Cnidaria Velella velella (Fig. 5).

Fig. 4.

An adult female of P. lilfordi from Aire Island foraging on a wrack of Posidonia oceanica.

Fig. 5.

Southern coast of Aire Island. The blue surface in the foreground corresponds to an extraordinary concentration of Velella velella. Carried by the waves, especially during storms, hundreds of thousands of individuals accumulate on the shores of Aire Island. We observed on several occasions the lizards coming down to the coast and consuming these and other invertebrates (see more detais in the text).

Thus, our provisional hypothesis is that storms can be considered the cause of the so-called resource pulses. A resource pulse is defined as an increased resource availability in space and time, with a combination of its low frequency (rarity), large magnitude (intensity), and short duration (brevity,⁹⁸). For instance, Polis et al.⁹⁹ and Grant et al.¹⁰⁰ described the effects of a well-known resource pulse, the El Niño Event, on terrestrial populations of species living in islands. Yang et al.⁹⁸ described several cases of resource pulses, including storm-driven nutrient runoffs from terrestrial systems into aquatic or marine systems^101–103. In our case, we consider a resource pulse the storm-driven arrival of marine subsidies, as seagrasses or invertebrate species to shores of islets. In central Bahamas, the experimental addition of seaweed to shoreline plots resulted in a fast response of the lizard Anolis sagrei, with a diet shift effect, from local terrestrial prey to marine detritivores. The last effect was an increase lizard density, 90% higher in subsidized plots⁹². This dietary shift implies the foraging behavior and habitat utilization of lizards^93,105.

Lizards inhabiting islands of the Gulf of California show a diet indirectly subsidized by seabird and tidal activity. Marine material is brought by seabirds and tides to the coast of islands and consumed by arthropods that are the main prey types of lizards. In these islands, tides and seabirds are responsible for the contribution of marine subsidies to the diet of the lizard Uta stansburiana^104,105. On several coastal islets of Menorca, seabirds contribute with organic matter that enters directly into the diet of the lizards⁹⁵ or that is consumed by the arthropod prey of the lizards. In Western Mediterranean, tides have a very limited amplitude, about 10 cm between low tide and high tide¹⁰⁶, and the contribution of marine matter would be mainly produced thanks to seabirds and storms. In U. stansburiana, during a survey of eight years, Barrett et al.¹⁰⁴ showed that marine diet subsidies resulted in an increase of lizard abundance. It is probably the case of Aire and Addaia Gran islands, with a significant effect of storms on lizard abundances, thanks to the effect of this cross-ecosystem transport of energy and nutrients^107,108. Only future studies of the effects of storms on the lizard populations of these islets will allow us to establish the true influence of these meteorological events on foraging behavior, the use of trophic resources, and, ultimately, the demography of these populations.

The negative effect of weather events has not been considered a relevant factor in the conservation of lizard populations on small coastal islets. However, our study indicates that it may be a relevant factor, as in the case of the islet of Bledas. In small populations with little vegetation cover and few refuges, it is likely to be an important factor in their decline or even extinction, especially under the current scenario of climate change and the intensity of extreme weather events.

Conclusions

Until now, estimated densities of the Lilford’ Wall lizard were obtained from some years or limited periods and therefore did not adequately represent the average abundance of these populations over time. This work constitutes the first attempt to establish the real abundances of this species (see also²⁷), their long-term trends and the influence of weather events on these densities. The relevance of these events is strongly different in each population and depends on the ecological conditions of each population, especially the lizard abundance itself, the island surface, the geographical location and the degree of exposure to weather events.

Author contributions

A P–C and V P-M obtained field data, analyzed and interpreted them. A P–C and V P-M wrote the manuscript. Both authors read and approved the final manuscript.

Funding

Throughout the last 35 years, several research grants from the Institut Menorquí d’Estudis, allowed the two authors to carry out the studies presented here. The lizard surveys in Menorca were constant during the last 10 years thanks to the support and financing of the Agency of the Biosphere Reserve (Insular Council of Menorca), especially thanks to Félix de Pablo, Juan Juaneda and Irene Estaún.

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable. Line transects do not imply the capture of lizards.