The Carpish Mountains represent montane forest in central Peru which is rich in endemic species of plants and animals. The Carpish Mountains are in the Department of Huánuco, accessible from the main highway (“carretera central”) connecting the cities of Huánuco and Tingo María. Botanical expeditions in this area have occurred since the 18th century with the Spanish explorers Ruiz and Pavon (Beltran & Salinas, 2010); later expeditions occurred in the 1930’s and 1940’s (Weberbauer, 1945; Goodspeed & Stork, 1955). Ferreyra (1950) described the vegetation and was the first to propose that Carpish be designated a national park, due to its high level of endemic species. As Beltran & Salinas (2010) explain, a single classification for the vegetation system is not possible, so these botanists and the ornithologists Parker & O’Neill (1976) differentiated the following vegetation zones: premontane forest (1 600 – 2 300 m), montane forest (2 300-2 800 m), elfin forest (2 800 – 3 600 m) and puna grassland (over 3 600 m).

The Carpish Mountains was not a protected area at the end of this study. The area has been disturbed by cattle grazing and fire (Franke et al., 2005) and the main central highway allowed human colonization that caused severe deforestation (Dourojeanni et al., 2010). Part of the landscape below 2 000 m was already modified in the 1950’s and 1970’s (Goodspeed & Stork, 1955; Tallman, 1974), but original vegetation is still present in higher and inaccessible areas, which function as a natural refuge (Sylvester et al., 2014). There remain important portions of elfin forest at the treeline that harbor endemic bird species (Wege & Long, 1995). An assessment of the montane forest in Peru lists this area as a priority for conservation (CDC-UNALM & TNC, 2006).

There have been ornithological studies in Carpish early in the past century (Zimmer, 1930), besides inventories and a description of mixed species flocks (Vuilleumier, 1970). Carpish has been the object of interest for Louisiana State University through several field expeditions (Tallman, 1974; Parker & O’Neill, 1976). They discovered new species of endemic birds there (Blake & Hocking, 1974; Lowery & Tallman, 1976). Carpish has been a preferred place for birdwatching in Peru (Valqui, 2004) because of the high richness of endemic birds, however tourism has not been frequent due to past problems (years 1980 – 2000) with security in the area. Terrorist groups and drug-dealers have dominated here (Fjeldsa et al., 2005), something that hinders the visit of birdwatchers and researchers (Puhakka et al., 2011). Despite this social problem, Peruvian researchers have completed inventories of plants (Salinas, 2005) and amphibians (Rodríguez, 2007) in the lower areas.

Unchog (3 800 m, 09°42'32.3" S, 076°09'39.1" W), where we did our surveys, is communal land of the Comunidad Campesina Cochabamba of Huánuco. There were three elfin forest patches with continuous vegetative cover (Unchog, Magma and Simeompampa), with ~8 ha each and no more than 2 km apart from each other (Figure 1). The predominant trees are several species of the Weinmannia genus, which is common in the Andes above 2 000 m (Rapp et al., 2012). Between these elfin forest patches, there were native shrubs and pastures.

Figure 1.
Location of the elfin forest of Unchog in Peru The forest patches sampled are marked in red.

Unchog forest was visited during different periods from May 2011 to August 2014, over a total of 134 days; 17 days within May-July of 2011, 24 days within February, July-November of 2012, 41 days in January-July, September, November of 2013 and 52 days in January and May-August of 2014. For more details in sampling effort, see Gonzalez & Loiselle (2016). This forest has a dry and rainy season, and some bird species are more prone to be present or plants blossoming depending on the season. We clumped our observations of both seasons during the entire research time to increase the chance to see all birds and plants of this network. Oscar Gonzalez identified nectarivorous birds (hummingbirds, flowerpiercers and conebills) walking along the forest edges and inside forests mainly in the morning with binoculars. In addition, mist-nets in different areas of the three forests patches were used (2 399 net-hours) and we recorded songs with a Marantz PMD-620 voice recorder and a Sennheiser shotgun microphone. This methodology of bird evaluation followed the protocol of Bibby et al. (2000), mist nets were preferred over point counts as a better method for detection due to the cryptic behavior of small birds in the cloud forest (Martin et al., 2017). We recorded all birds captured in mist nets but for the aim of this study, we are reporting just the nectarivorous. Passerine birds were banded with unique color band codes. Identification follows Schulenberg et al. (2007). The names for the birds were taken from the South American Classification Committee list for Peru (Remsen et al., 2016) and Avibase (Avibase, 2003).

We compared our observations to bird lists from previous expeditions to the area, which were primarily conducted by Louisiana State University, Field Museum of Natural History, American Museum of Natural History and Museo de Historia Natural of the Universidad Nacional Mayor de San Marcos. O. Gonzalez visited the bird collections of these institutions. These expeditions did intense collection of birds in the Unchog area mainly in the summer season (May-August) to survey Carpish mountains. We also used eBird records (eBird, 2015) to document observations of birds in Unchog forest up to July 2014. The eBird data are records of bird species presence taken by professional and amateur birdwatchers (Sullivan et al., 2014), displayed in bird checklists of specific locations evaluated in different ways. We focused the location as “Unchog” in our search.

The survey of flowering plants visited by birds was done by setting up two Gentry plots (Gentry, 1982) which were five transects of 100 x 2 m inside the forest patches and occasional collecting by Bernie Britto and Camilo Díaz. This occasional collecting consisted of recording the plants that were not seen inside the plots but were into our study area. Plant identification was done by specimens collected and by photographs. The vouchered plant samples were deposited to the herbarium of the Universidad Peruana Cayetano Heredia and the specimens were compared with others in the herbarium of the Museum of Natural History of the Universidad Nacional Mayor de San Marcos. As authorities for plant names we used the works of Mabberley (1987), Brako & Zarucchi (1993) and The International Plant Names Index (IPNI, n.y.).

For the visualization of the assemblage of nectarivorous birds and their visited plants, we collected the available records on bird-plant visitation of Andean elfin forests from the literature, looking for the species or genera that we recorded in Unchog. Then we ordered the assemblage in a two x two matrix; by species that have most of the interactions at the top left in a nested structure (Jordano et al., 2009). We called this the theoretical network. Then we ordered our observations on the nectarivorous birds visiting flowers of the plants in the same way, but here we had a qualitative matrix, measured by how many times a bird species visited a plant species. Then we applied the function visweb of the bipartite package in the software program R (Dormann et al., 2008), which is useful to visualize interaction networks in a grid with different colors or tones that represent the strength of the interaction. This quantitative assemblage was called the observed network. We compared the theoretical network with the observed network with parameters for qualitative networks, using the function networklevel in bipartite (Dormann et al., 2008). In order to do this, we converted the quantitative network in a qualitative network for our observed assemblage. Connectance and nestedness were chosen as reliable indicators to compare qualitative networks (Jordano et al. 2009). Connectance measures the proportion of interactions, by dividing the total number of species interactions by the total size of the matrix (Birds x plants). Nestedness measures the order of interactions in the matrix, being zero a random distribution and 100 perfect nestedness; here we used the indicator NODF (Dormann et al., 2008).

The assemblages, or networks of nectarivorous birds and their resources provide an important framework to study ecological principles such as diversity, distribution, adaptation and energy flow (Rojas-Nossa, 2013; Bennett et al., 2014; Abrahamczyk & Kessler, 2015; Gonzalez & Loiselle, 2016). The information gathered from nine sources on nectarivorous birds that visited plants in the Andes gave us a limited view of the real interactions in the elfin forest, because there were fewer taxa than in the observed network. The connectance of the observed network was higher, but the nestedness lower. This could indicate that the assemblage of plants and nectarivorous birds could be more fragile than expected because nestedness is considered an indicator of resistance of communities to perturbations (Bascompte et al., 2003). However, we have the limitation that this theoretical network is a compilation of data from different areas, even though from tropical Andean forests, the endemism of plants and animals is high (Stattersfield et al., 1998) and the preferences of birds on specific plants might switch. Then, it is important to analyze in some detail the members of the assemblage.

Nectarivorous Birds

We recorded the nectarivorous bird species that were reported by Tallman (1974) and Parker & O’Neill (1976), who did intensive search for birds in the Carpish area; but not all of the species in the eBird record (2015) were seen in Unchog forest or interacting with plants (Table 2). The species of this potential network that we did not record in the interactions were Colibri thalassinus, Colibri coruscans, Chalcostigma ruficeps, Chalcostigma stanleyi, Metallura phoebe, Agaleactis castelnaudii, Amazilia chionogaster, Conirostrum cinereum and Conirostrum sitticolor. However, during our field study (2011-2014) almost all the records for eBird (2015) were in the dry season; there were no records in February, April and September.

It is highly likely that the contributing people for the eBird records went outside the limits of the elfin forest to see these species, because the tourists who come to this birding hotspot do not just birdwatch in the elfin forest. They get in the montane scrub of the Huallaga valley that borders with the elfin forest and they can also go beyond this forest to the cloud forest limit; it is possible to see bird species there that do not frequent the core of the elfin forest (Valqui, 2004). The White-tufted Sumbean Agaleactis castelnaudii was recorded in eBird’s record of Unchog by three people (eBird, 2015), two of them said that they saw them in the montane scrub outside the elfin forest (F. Schmidt, G. Schoezler, personal communication to O. Gonzalez). A. castelnaudii with White-bellied hummingbird Amazilia chionogaster and Sparkling Violetear Colibri coruscans are usually seen in the neighboring habitat too (O. Gonzalez, personal observation). The altitudinal limit of A. chionogaster is 3 500 m, and Green violetear Colibri thalassinus is 3 300 m. These are located below our observation areas. A. chionogaster was seen just once at the beginning of the study and was not seen visiting any flower. C. thalassinus is known to make altitudinal migrations in Central America (Levey & Stiles, 1992). Black Metaltail Metallura phoebe usually does not overlap with other species of Metallura (Schulenberg et al., 2007), but it could be confused with M. theresiae which seems black from a distance. The data on eBird can be a good reference to look for distributions (Sullivan et al., 2014), however it is necessary to recheck the data for possible mistakes before using it for comparison in a specific region (Gonzalez & Wethington, 2014). We registered Purple-backed Thornbill Ramphomicron microrhynchum as well as Parker & O’Neill (1976); but this species was not in the eBird’s checklist (eBird, 2015) within the temporal and spatial range surveyed. The inclusion of a conebill as a nectarivorous bird might seem odd, but Moynihan (1963) mentioned that the species of this group are partly nectarivorous.

The tree line of Manu National Park, in the southern Peruvian Andes, has a very similar landscape to Unchog. Lloyd et al. (2012) found out a distinctive community of plants and birds, different from puna grasslands and the cloud forest. Bird pollinators that are expected in tree line zones include a dozen hummingbirds such as Great Sapphirewing Pterophanes cyanopterus, Sword-billed Hummingbird Ensifera ensifera and species of the genus Coeligena, Eriocnemis and Heliangelus, the last two restricted to the elfin forest. We did not observe Eriocnemis in this study.

The Coppery Metaltail Metallura theresiae was the most dominant flower visitor in this ecosystem, so it could be the most effective pollinator (Willmer, 2011). In other tree line ecosystems in the central Andes, the small hummingbirds Tyrian Metaltail Metallura tyrianthina and Amethyst-throated Sunangel Heliangelus amethysticollis are dominant (Gonzalez, 2008; Lloyd et al., 2012; Toloza-Moreno et al., 2014) but in Unchog, M. theresiae has taken over the tree line and these other hummingbirds stay in lower elevations. The diversification of Metallura hummingbirds was recent (Garcia-Moreno et al., 1999), hummingbirds have their beta-diversity associated with elevation (Weinstein et al., 2014). M. tyrianthina is subordinate (Does not compete, it is not dominant) to M. theresiae and other sister species when it is present at higher elevations (Heindl & Schuchmann, 1998).

Hummingbirds of the genus Coeligena visits several species of plants in Andean forests, Schuchmann (1999) reported them visiting Bomarea, but we did not observe Coeligena feeding on Bomarea. The other plants reported by Schuchmann were not in our study area. Schuchmann (1999) also did not report Tristerix as food for Pterophanes cyanopterus as we observed. Gutiérrez & Rojas (2001) in the paramos of Colombia, found that there is a strong dependence of several small billed hummingbirds with Brachyotum. They also found dependence of Pterophanes with Puya; but not with Tristerix which was also present there. Metallura odomae, sister taxa of Metallura theresiae in high elevation montane forest in the north of Peru feeds mainly in Brachyotum and in Berberis (Hesse, 2010). Small-billed hummingbirds rely on Brachyotum, its importance in the network is explained in Gonzalez & Loiselle (2016). Schuchmann (1999) mentions that the Shining Sunbeam Aglaeactis cupripennis does local migrations, Heliangelus prefers shrubs in the ecotone between paramo and forest and Coeligena forages in the forest itself or forest edges.

The Moustached flowerpiercer Diglossa mystacalis has a large range that is not limited to the elfin forest (Parker & O’Neill 1976); Black-throated Flowerpiercer Diglossa brunneiventris is common in disturbed areas and non-forested habitat (Becker et al., 2008; Vaicenbacher et al., 2014). Masked Flowerpiercer Diglossa cyanea travels with mixed species flocks in cloud forests (Guevara et al., 2011; Arbelaez-Cortes & Marin-Gomez, 2012) and was uncommon.

Table 2.
Identity and characteristics of the nectarivorous bird assemblage.

The frequency criteria is based on data of individuals captured in 100 net-hours: No data (ND), Very rare (VR) = Less than 2, Rare (R) = Between 3 to 5, Frequent (F) = Between 6 to 11, Common (C) = more than 12. Conservation status: E = Endemic for Peru, I= Elfin forest indicators (Parker et al.,1996). Record: x = This study, e = eBird.

Own.

The ornithophilus plants expected in the elfin forest according to Schuchman (1999) could be woody, vines and shrubs. Those are members of the genus Bomarea, Fuchsia, Lobelia, Passiflora, Brachyotum, Befaria or Datura. We did not find visitors for Lobelia, Befaria or Datura.

As Stiles (1985) described, in high elevation ecosystems like this elfin forest, the birds pollinate mainly shrubs. The Melastomataceae Brachyotum, Meriania and Miconia were cataloged as hummingbird-pollinated by Varassin et al. (2008). The shrub Brachyotum spp. is an important source of nectar of several nectarivorous species in this region; first descriptions on the flowers of this genus noted that hummingbirds pollinate it (Renner, 1989a). However, Stiles et al. (1992) suggested that these plants are adapted to pollination by the Diglossa flowerpiercers. The plant more connected was Brachyotum lutescens which is an endemic species of Peru only found in this region and maybe in critical situation (León, 2006). As for the shrub Hypericum, common here and in the northern Andes paramo (Caceres et al., 2014), the shrub Brachyotum may be a foundational species; they are the first in colonizing a non-woody area and start forest regeneration (Guenter et al., 2009). Furthermore, Brachyotum facilitates movement among forest patches as nectarivorous birds that move between patches may forage on Brachyotum which is common in the non-forested matrix. Similar to Gutiérrez & Rojas (2001), who studied a tree line zone between montane forest and paramo, Brachyotum (B. lutescens and B. naudini) is most visited by hummingbirds and then other plants identified in this study: Fuchsia, Passiflora and Tristerix.

Ollerton et al. (2009) identified the plants that have birds as pollinators in the elfin forest of Urubamba (Cuzco). Plants that have a common genus in Unchog forest are: Bomarea, Fuchsia, Gaultheria, Passiflora, Puya, and Rubus.Wolf et. al. (1976) reported that Centropogon and Fuchsia are common plants used by hummingbirds in the high altitude forest of Central America. Cronk & Ojeda (2008) and Willmer (2011) identified Fuchsia and Passiflora as typical bird-pollinated flowers. Botanists such as Weberling (1989) and Gentry (1993) recognize that Fuchsia plants are adapted for the visitation by hummingbirds. Gentry (1993) also confirmed that Brachyotumand Tristerix are characteristic for hummingbird pollination. In New Zealand, Tristerix is known to suffer the effect of pollination limitation in the seed set (Montgomery et al., 2001), even though is self-compatible. However, it has populations that are gynodioecious (hermaphrodite and female flowers) so that could be an evolutionary advantage in case there would not be pollinators (Arroyo & Raven, 1975; Atsatt & Rundel, 1982; Robertson et al., 1999). In one hermaphroditic species of Fuchsia, its nectar production is assumed related to the evolution of the pollination traits of the flower (Atsatt & Rundel, 1982). The presence of Ensifera ensifera in the montane forest limit with the paramo in Colombia is associated with the presence of blossoming peaks of Fuchsia and Passiflora (Gutiérrez & Rojas, 2001). However, specialization in pollination networks can be misleading (Schemske, 1983). In Unchog Ensifera, was seen occasionally and not visiting Passiflora but Tristerix. Furthermore, Metallura theresiae visited Passifloraflowers illegitimately (Gonzalez & Loiselle, 2016). Benevides et al. (2013) stated that Passiflora is a very important resource for several pollinators, but this plant was scarce in the elfin forest and did not receive many visitors.

It is known that Puya is a very important source for hummingbirds in the high Andes (Woods & Ramsay, 2001; Salinas et al., 2007; Hornung-Leoni et al., 2013). Schuchmann (1999) states that Patagona gigas is highly dependent on Puya. Besides this hummingbird species, we observed Coeligena torquata, Metallura theresiae and Diglossa brunneiventris visiting Puya, but the dependence on them was weak.

The blackberry Rubus had more than one species in the network but the shape of the flowers was very similar in all species. Berberis and Tropaeolum are genera that are present also in the Chilean Andes, but the weather here is temperate and they are pollinated by insects (Arroyo et al., 1982). The shrub Dunalia spinosa, related to D. brakiacantha in a premontane habitat also in the Chilean Andes, has flowers rich in nectar and is visited by different birds (Torres-Mura & Hertel, 2006). In a study involving Disterigma in Colombia, small-billed hummingbirds were the most common visitors; with Coeligena and a Diglossa that robbed its nectar as seldom visitors (Navarro et al., 2008). In the sampling we detected Cavendishia bracteata, Pernettya postratta and Fuchsia apetala; the first two flowers were visited by hummingbirds in Colombia (Gutierrez-Zamora, 2008). F. apetala has large and orange flowers, but we did not see any bird visitor to this species. Plants with small flowers such as Pernettya postrata and Gualtheria spp. could be able to self-pollinate (Luteyn, 2002) and did not have a strong dependence with birds. Epiphytes, which are known to be important as nectar resource in montane forests (Kromer et al., 2006) were almost absent here because at high elevations close to the treeline, the abundance of epiphytes drops (Zotz et al., 2014).

It is important to consider the bird composition of this assemblage might be affected by local movements or migrations, nectarivorous birds have been recorded to have local migrations in another montane forest such as in Manu (Merkord, 2010) or Costa Rica (Levey & Stiles, 1992; Boyle et al., 2011). On local movements of the birds in Unchog, it was observed that only one Diglossa mystacalis of 69 banded flowerpiercers (D. mystacalis and D. brunneiventris) switched between forest patches in 8 months. We do not have information on movements of the hummingbirds here. It is known that hummingbirds fly long distances following floral resources (Levey & Stiles, 1992; Hobson et al., 2003; Gutierrez et al., 2004) and usually move between fragments of montane forest and high montane scrub habitats at high elevations (Toloza-Moreno et al., 2014). Even Metallura theresiae is observed all year round and nesting at different times of the year in the elfin forest (O. Gonzalez, personal observation) we cannot discard the possibility that this hummingbird moves to lower elevations. Hobson et al. (2003) proved with isotopes that hummingbirds move along an elevational gradient in a montane forest, among them Metallura baroni, an endemic. However, Metallura theresiae is at the core of the network observed here (Gonzalez & Loiselle, 2016) and appears to be constant throughout time and space (eBird, 2015).

About other vertebrates that are nectarivorous, bats are a possibility in our study area. Even though we did not observe bats visiting flowers, it is probable that nectarivorous bats of genus Anoura, observed in the upper montane forest of Yanachaga-Chemillen National Park (90 km southeast) reach the elevation of the elfin forest (Vivar, 2006); and pollinate Macrocarpaea flowers (Renner, 1989b); this tree has populations limiting the elfin forest. There are records in the Kosnipata valley in Cusco (1 600 m. elevation) of Rubus pollen transported by bats (Maguina et al., 2012) and of Passiflora pollen in a montane forest in Ecuador (Muchhala & Jarrin-V, 2002). Insects visited some plants that birds also visit; in the dry season of 2014 a total of 29 % of the flower visitors of Gentianella violacea were insects (O. Gonzalez, unpublished data).

Unchog forest is a place of high endemism of the tropical Andes, shown by this assemblage of nectarivorous birds and their resources, as well as several other birds and plants. This endemism in elfin forests leads to its importance for conservation efforts (Herzog & Kattan, 2011; Tejedor-Garavito et al., 2014). Unfortunately, the elfin forest is highly fragmented; and habitat loss by deforestation due to logging, livestock grazing or clearance for agriculture could hasten loss of pollinators and pollinator services. Detailed research is needed on the individual responses of plants to deforestation (Renner, 1998). The conservation of the tree line ecosystems, where there are remnants of elfin forests and enclosed watersheds are the resources for nectarivorous birds and also for several endemics, should be a priority for conservation (Ramirez-Villegas et al., 2014).

ogonzalez@ec.edu