Population decline in the Critically Endangered Musschia isambertoi (Campanulaceae) endemic to Desertas Islands (Madeira Archipelago) calls for urgent conservation management

Macaronesian islands display a high degree of plant endemicity and correspond to some of the most threatened biodiversity hotspots due to several human driven impacts such as habitat loss, invasive species, overexploitation or climate change. The conservation status of the largest of the two known population of Musschia isambertoi M. Seq., R. Jardim, Magda Silva and L. Carvalho (Campanulaceae) was assessed recently. Field work was done in June 2018 in “ Porto das Moças ” Deserta Grande (Madeira Archipelago, Portugal), the locus typicus , where in 2006 close to 30 plants were detected with a high variation of age/size between individuals comprising both fully flowering and non-flowering adult plants as well as seedlings


Introduction
The Macaronesian archipelagos, due to their high degree of plant diversity and endemicity are included in the Mediterranean Basin hotspot (Myers, Mittermeier, Mittermeier, Fonseca, & Kent, 2000), being also among the most threatened due to several human driven impacts (e. g. Caujapé-Castells et al., 2010;Médail & Quézel, 1997). Historical documents refer to an early and rapid destruction of the vegetation in the Madeira Archipelago, which comprises Madeira Island, Porto Santo Island and Desertas Islands. The main drivers of the elimination of the pristine forest were: wheat and other cereals cultivation (early 15th century), sugar cane plantation and production (15th and 16th centuries), followed by the expansion of agriculture and tree felling for wood exploitation (Menezes de Sequeira, Jardim, & Capelo, 2007;Moore, 2009Moore, , 2010. Simultaneously, the introduction of non-native herbivores such as goats (Capra hircus), pigs (Sus scrofa) and rabbits (Oryctolagus cuniculus) (e.g. for the introduction of rabbits in Porto Santo Island see Rocha et al. (2017) and references therein), was the main driver of landscape change from as early as the 15th century.
In the Madeira Archipelago several introduced herbivores such as rabbits, goats and pigs became feral (Menezes de Sequeira, Jardim, & Capelo, 2007). The arrival of these invasive species occurred in the 15th century soon after or simultaneously with the first human settlers. In the Portuguese settlements that occurred in most Atlantic islands, the same common behaviour was followed, and included the introduction of mammals, later used as food supply during long ship voyages (e.g. Chynoweth, Litton, Lepczyk, Hess, & Cordell, 2013;Grove, 1995). According to the accounts given by Diogo Gomes, goats were introduced in the Selvagens possibly before 1460 by order of the Infante D. Henrique (Pereira, 1899;Zino & Biscoito, 1994). In any case the introduction of goats in the Desertas was even earlier, due to the proximity to Madeira Island and to the references, by Gomes Eanes de Azurara in his "Chronica do Descobrimento de Guiné" (1452-1453) and Valentim Fernandes in the 1508 manuscript "Das Ilhas do Mar Oceano" (Baião, 1940), to the orders given by the Infante D. Henrique (1394-1460) to introduce goats in the Desertas Islands. The presence of goats in large numbers in the Desertas Islands in the 16th century is referred to by Nicols Nichols (1583) and by Arditi (1567), the latter author even describes the presence of shepherds, and apart from goats, other animals such as cows, sheep, mules, etc.
The largest known population of Musschia isambertoi corresponds to the holotype locality, "Portugal, Madeira: Ilhas Desertas, Deserta Grande, perto da Fajã Pequena, Porto das Moças, 16-V-2006, M. Silva, L. Carvalho, C. Viveiros and P. Gouveia 868 (MA 751556)" (Fig. 1a, b, c). "Porto das Moças" (possibly less than five individuals were ever found outside the range of this location in one single location). Images taken in 2006 show an age-complex population with several fully flowering/ fructifying plants and other mature plants not in flower along with many younger individuals and seedlings (Fig. 2a, d). Nevertheless, the species was classified by Menezes de Sequeira, Jardim, Silva and Carvalho (2007) as Critically Endangered (CR, C2a(i,ii); D), mainly due to "the scarce number of populations and the reduced occupancy and occurrence […] and also due to the grazing effects through the introduction of goats".
In order to re-evaluate the conservation status of Musschia isambertoi, this study assesses population dimension, age, plant size and genetic diversity, as well as floristic diversity, therefore evaluating possible changes in species dominance, life form dominance change or percentage of endemics versus native non-endemic or non-native plant taxa.

Material and methods
Field work (7 June 2018) took place at "Porto das Moças", Deserta Grande island, the locus typicus where on May 16th, 2006 many plants were detected with a high variation of age/size between individuals (Fig. 2a, d).

Site description
The Madeira archipelago comprises three groups of islands, Porto Santo (about 18 Mya old), Madeira and Desertas (>7 Mya old), that share a common geological origin (see Ramalho et al., 2015 and references therein). To the southeast the Madeira archipelago is continued by the Desertas sub-archipelago composed by three small islands. The northernmost Ilhéu Chão (ca. 0.4 km 2 , 1.6 km long 0.5 km wide), is the smallest rising only 98 m above sea level. The largest, Deserta Grande (ca. 10 km 2 , 11.7 km long and 1.9 km wide), has a maximum elevation of 479 m above sea level. Finally, Bugio (ca. 3 km 2 , 7.5 km long 0.7 km wide), the southern island, and the most orographically complex, has a maximum elevation of 388 m above sea level. According to the model proposed for Madeira Natural Potential Vegetation (Capelo, Menezes de Sequeira, Jardim, & Costa, 2004) Desertas islands should have possessed a microforest of Madeiran oleaster (Olea maderensis) (0− 200 m a.s.l.), a microforest dominated by Sideroxylon marmulano (200− 300 m a.s.l.) and a mediterranean laurel forest dominated by Apollonias barbujana (300− 800 m a.s.l.). The Desertas' flora includes many known floristic elements of these two communities (Table 1, based on Jardim & Menezes de Sequeira, 2008), nowadays only growing in crevices of deep ravines. No native mammal terrestrial herbivores were present on the archipelago of Madeira and, as described for Madeira island and Porto Santo, deforestation rate must have been fast and complete in the Desertas, due to the early introduction of several herbivores (a common practice in the discovery see Grove, 1995 and references above) and the reported use for agriculture (Silva & Menezes, 1946;Menezes de Sequeira, Jardim, & Capelo, 2007). Rabbits were eradicated from the Desertas (Bell, 2001) but goats are still present. Musschia isambertoi was found growing near to seashore, at the base of the unstable cliff of Porto das Moças, within a colluvial deposit mainly composed of boulders with poor soil development.  Menezes de Sequeira, Jardim, Silva, & Carvalho, 2007) Tall rosetted monocarpic unbranched plant, up to 2 m. Young leaves (1st year) up to 6 cm, adult leaves up to 40 cm, with indistinct petiole, herbaceous, shallowly undulated, with bi-serrate margins, becoming senescent but attached to the stout stem, plants producing 4-5 leaves yearly, flowering after 4-5 years growth. Inflorescence up to 150 cm long, unbranched except for the terminal part, bracteate (bracts gradually shorter towards the top); flowers densely crowded, green, yellowish with reddish-brown tones. Sepals green with reddish brown apex and veins, sometimes yellowish towards the base. Corolla bright green, yellowish towards the base, lobes narrowly triangular-lanceolate, acuminate, pollinated by lizards (Lacerta dugesii mauli, = Teira dugesii mauli).

Population age structure
Musschia isambertoi population age structure in 2006 and 2018 was compared assuming, the following age categories, seedlings (plants with no adult leaves, i.e. large leaves ca. 7 cm), young adult plants (with adult leaves but no evidence of senescent leaves), mature adult plants (with adult leaves and clear evidence of senescent leaves), flowering plants. The 2006 plants were counted using several pictures taken during the original visit (Suppl. Mat. Images 1-4), seedlings must have been underestimated due to the fact that they are small and grow amid other plants and between rocks, whereas the 2018 data correspond to direct observation.

Floristic diversity
A list of all vascular plant taxa was prepared based on plants identified in the field, collected, and further identified based on Press and Short (1994). Nomenclature and native status follow Jardim and Menezes de Sequeira (2008). Plant specimens corresponding to all collected plant taxa are kept at the herbarium of the University of Madeira (UMad). Abundance-dominance for each plant taxa was registered using the coefficients defined by Braun-Blanquet (1964), percent cover was obtained following van der Maarel (1979).

Genetic diversity
To characterize genetic diversity within the population, total genomic DNA was extracted from silica gel dried leaves using the method of Pich and Schubert (1993) with minor modifications. The extracted DNA was evaluated by loading 1 μL on a 0.8 % agarose gel containing ethidium bromide. Gels were digitally acquired under UV light (DigiGenius, Syngene, UK). Densitometric analysis of gel lanes were performed through ImageJ software (http://rsbweb.nih.gov/ij/) after which DNA solutions were diluted to 10 ng μL − 1 .

Results and discussion
Only ten Musschia isambertoi plants were found, corresponding to a coeval population of young plants less than 15 cm high, with the number of leaves ranging from 4 to 8 (Fig. 3d, e), these plants can be considered as seedlings or, at most, young plants in their second year of growth. Neither adult plants nor remains of adult plants were found in the study area (Fig. 3b, c). Based on the images taken in 2006 and on field work in 2018 the population status can be compared for both years by the number of seedlings, adult plants (with or without senescent leaves) and flowering plants as shown in Fig. 4. These results stress the lack of age structure in 2018 population.  (2007) refer to a second population based on herbarium data (MADM 485, 1992) this population was not found (in a 2016 survey conducted by the IFCN, for the study later published by Menezes et al. (2018)), it is presumably extirpated, in any case we were unable to observe the location during 2018 field work and therefore we have calculated both the EOO (8 km 2 ) and AOO (8 km 2 , if based in 100 × 100 m then AOO 0.03 km 2 ) assuming two populations still occur, therefore without any changes regarding the evaluation by Menezes de Sequeira, Jardim, Silva and Carvalho (2007).
ISSR results clearly show that all plants sampled (9 out of 10, destructive sampling was avoided) are genetically identical (Fig. 5a, b) and therefore are presumably the offspring of one self-pollinated plant. Musschia isambertoi being an example of a monocarpic plant with insular woodiness, mature plants are not expected to survive seed production. However, no remains of any adult plants were found during fieldwork, possibly indicating that several years have passed since the last seedsetting event.
Ecological data and direct herbivory evidence suggest that perennial plants are subjected to elimination during the short summer season by the feral goat population. Musschia isambertoi habitat is a nonchasmophyte habitat that corresponds to the mesic conditions observed in Porto das Moças. Table 2 summarises the floristic/synecological data (see also Fig. 3b, c), which clearly shows: (1) the absence of phanerophytes, (2) near absence of chamaephytes (except chasmophytes; see Table 2), (3) the reduced number of hemicryptophytes, (4) the prevalence of therophytes. From a total of 24 chamaephytes and 19 phanerophytes referred to the Desertas Islands, as natives or endemics (Table 1), only eight chamaephytes were detected (Table 2), corresponding to a very low frequency and cover (less than 5 %, Fig. 6). These results and the fact that the low number of chamaephytes observed corresponded to young plants support the occurrence of a limiting factor blocking succession. Cubas et al. (2019) refer to the high palatability of island endemic plants, which could also be the case for Musschia plants   (Mesquita, Capelo, & Sousa, 2004), annual plants tend to complete their life cycle before July, therefore the only summer green plants would be chamaephytes and phanerophytes, even if they correspond to young plants, as is the case for M. isambertoi and another Desertas islands endemic, Sinapidendron sempervivifolium also observed in the studied area.
Feral goats in Deserta Grande are a very well-known threat to local flora and their control was the aim of several conservation projects (LIFE95 NAT/P/000125, see also Silva & Menezes, 1946;Bell, 2001 andIFCN, 2017), in fact the actual landscape corresponds to highly disturbed plant communities with almost no tree or shrub cover.
During field work, goats (a small herd, one male and two females) were seen in cliffs above the site at approximately 1 km (Fig. 3a). Although accessibility from the cliffs seems to be complex, reports of goats at the site and evidence of herbivory observed directly (damage on Tolpis succulenta, Sinapidendron sempervivifolium and Musschia aurea, Suppl. Mat. Image 5), in addition to a predominance of seedlings of perennial plants including Musschia isambertoi (Fig. 3d, e), support goats as the main drivers of ecological shift and plant diversity loss. Donlan et al. (2002) refer to a differential effect of herbivory on perennial/endemic plant taxa, which is the case for Musschia isambertoi. However, the preference could simply be the result of phenology versus foraging by goats throughout the year, and in fact early October/November rains result in a fast production of biomass by therophytes, which, together with hemicryptophytes, must constitute the most important part of the goats' diet, until late spring. However, these earlystage (pioneer) plant communities result in extreme food shortage during late spring, summer and early autumn. Therefore, chamaephytes and in fact all perennial plants must be under high herbivore pressure during, at least, late spring to autumn. Palatability could in this case play a secondary role, at least during drought months (June to September), the only plants keeping green leaves and stems being endemic perennials.
The early introduction of goats in the Desertas Islands (Arditi, 1567;Baião, 1940;Azurara, 1452Azurara, -1453Nicols Nichols, 1583), certainly led to a massive landscape change such as the one reported for Porto Santo (historical data for Porto Santo support the presence of a primeval forest similar to the one observed in Madeira). Deforestation led to loss of the primeval forest cover possibly at a very fast rate as described for all Madeira archipelago islands (Menezes de Sequeira, Jardim, & Capelo, 2007).
In the Canary Islands Garzón-Machado et al. (2010) tested the effects of herbivory by several feral mammals (barbary sheep, Ammotragus lervia, goat, and European rabbit (Oryctolagus cuniculus) and concluded that herbivory was responsible for the extremely low diversity patterns found in key ecosystems.
Several authors correlate the evolution of secondary woodiness in island endemics as related to the absence of herbivory (e.g. Carlquist, 1974;Dulin & Kirchoff, 2010). The detection of, exclusively, Musschia isambertoi young plants (presumably seedlings) is coherent with a pioneer plant community dominated by annuals where perennials (hemicryptophytes or chamaephytes) are grazed yearly during summer drought. The shift towards a therophyte dominated community can also be confirmed by the observation of Fig. 2d (2006 Fig. 6). Similar communities can be found in the extreme east of Madeira at Ponta de S. Lourenço corresponding to areas highly disturbed and grazed. This shift to annual plant dominance, as well as a reduction in cover and diversity of endemics (Fig. 7), is a common pattern in plant communities affected by herbivores, namely goats (e.g. Mueller-Dombois & Spatz, 1975).
Therefore, both the ecological shift, and the age and lack of genetic diversity of Musschia isambertoi, support continuous seed recruitment on a limited seed bank that probably originated from a single parental plant. Other perennial chamaephytes formerly dominant were also observed as small young plants resembling the observations on Musschia isambertoi. The feeding on seedlings is reported by Mueller-Dombois and Spatz (1975) and explains the lack of adult flowering chamaephytes (or even phanerophytes).
Results clearly suggest that Musschia isambertoi, on the edge of extinction in 2007, was almost effectively extinct in nature during our field work that led to this current publication. In any case imperative conservation measures are needed and should include: 1. The immediate fencing of the population site; 2. Fresh leaf material collection for tissue culture; 3. The elimination of feral goats from the Deserta Grande; 4. Translocation of some seedlings into conservation gardens. The translocation of M. isambertoi to both the Bugio and ilhéu Chão islands is admissible, although there are no records, actual or historical, of the presence of M. isambertoi in these islets. Koutsovoulou, Daws, and Thanos (2014) report the successful  Campbell and Donlan (2005) on their revision of goat eradication (see also Global Invasive Species Database, 2020) report that the eradication from large areas (>150 km 2 ) is scarce (it would be the case of Madeira Island with close to 736 km 2 ) but Deserta Grande (10 km 2 ) falls in the range of success reported by these authors (islands with less than   150 km 2 ). The eradication of rabbits was completed with success from Deserta Grande (Bell, 2001), but the attempt to eradicate feral goats failed, although the project final report states the opposite (https://ec. europa.eu/environment/life/project/Projects/index.cfm?fuseaction=se arch.dspPage&n_proj_id=38). The failure to eradicate goats from Deserta Grande corresponds to the public opposition cases reported by Genovesi (2005) in his revision of the eradication of invasive species in Europe, including the eradication in Parco Naturale di Portofino in Italy.
In fact recent attempts to mitigate the effect of feral goats in Deserta Grande by the IFCN have faced the opposition of some political and animal rights groups (e.g. https://funchalnoticias.net/2018/10/19/pa n-madeira-insurge-se-contra-o-abate-a-tiro-de-cabras-nas-ilhas-desertas / or https://www.radiocalheta.pt/manifestacao-contra-exterminio-dascabras-nas-desertas-surpreende-partida-do-rali). In Madeira, local laws allow for goat eradication; there was public funding (EU) but there was no public awareness of the procedures undertaken. Moreover, it was the permitted filming of the eradication programme (issued in the News in RTP, the Portuguese National Television, https://arquivos.rtp.pt/conteu dos/operacao-cabra/), sanctioned by the Services of the Natural Park of Madeira, that caused some public outcry due to the methods ("Judas goats" were being used). The suspension of the programme when only a few goats were left resulted in a massive resurgence of the goat population and therefore was a misuse of EC funding. As referred by Keitt et al. (2011), the eradication of invasive vertebrates has proven to be a crucial and highly beneficial action on several islands. In the Madeira archipelago the eradication of both goats and rabbits from the Bugio islet ( Fig. 1) was clearly a success. But in other islets, and in Porto Santo, these actions not only failed to achieve their objectives, and in the case of Porto Santo, the reintroduction of rabbits was actually promoted by the local authorities (IFCN, 2019). Keitt et al. (2011) describe a long history of eradication attempts in islands starting in 1673 and report very high success rate in goat eradication (95.2 %) in a total of almost 170 attempts. As stated by Simberloff et al. (2013), biological invasions should be dealt with in a time frame from prevention to early detection and finally management (including eradication). These strategies result in distinct management costs that increase in time. For Deserta Grande the total cost of the actions, although unknown, can be estimated ranging from 500,000 to 1, 000,000 euros (including an EC LIFE project). The recent opening of both Deserta Grande and Selvagem Grande to tourism (Regional Legislative Decree 15/2017/M, https://dre.pt/application/conteudo/ 107477153) should raise increased concerns on the introduction of alien species.
Goat eradication has proved to be an important conservation tool leading to ecosystem restoration (e.g., Courchamp et al., 2003;Genovesi, 2005;Garzón-Machado et al., 2010;Keitt et al., 2011), in addition to clearly benefitting the biodiversity of islands ; see Keitt et al., 2011, for a review on eradication programs; see Schweizer, Jones, & Holmes (2015), for a review of results on biodiversity).
As proposed by Campbell and Donlan (2005) or Genovesi (2005) and supported by Garzón-Machado et al. (2010) on the Canary Islands, the eradication of feral goats from island ecosystems should be a routine procedure. Our results strongly support eradication procedures being re-started in Deserta Grande and, the immediate fencing of the known largest population must be implemented immediately, in order to prevent herbivory on an already small, coeval and genetically homogenous population of the rarest Madeiran endemic plant species. These actions should be accompanied by vegetation monitoring in order to closely follow succession and early detection of any possible invasions by non-native plants. Due to the usual discontinuous and limited available funding, complete eradication must be implemented instead of limited control actions, as suggested by Cruz, Donlan, Campbell, and Carrion (2005). Campbell, Donlan, Cruz, and Carrion (2004) refer to the financial costs of incomplete eradication attempts. Although they refer to Pinta Island (Galapagos) and to monitoring actions that assumed (wrongly) that the goats had been eradicated, this case can be compared with the attempt to eradicate goats from Deserta Grande. Limited control actions have led Musschia isambertoi to the edge of extinction, or possible extinction, and the island's vegetation to a lower diversity state where pioneer communities dominate most of the landscape.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments
We are grateful to the Instituto das Florestas e Conservação da Natureza that generously supported our visit to Deserta Grande. Also, our thanks to all the "Vigilantes da Natureza", and the crew of the ship Buteo, which transported us to and from the island. Without their help this study would not have been possible. This work was funded by:

Appendix A. Supplementary data
Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.jnc.2021.125955.