Avian extinctions induced by the oldest Amazonian hydropower mega dam: evidence from museum collections and sighting data spanning 172 years

Study area

This study was conducted in the Tucuruí Hydroelectric Reservoir (THR) influence area in the State of Pará, northern Brazil (3°43′ to 5°15′S; 49°12′ to 50°00′W). The Tucuruí Dam, which impounded the eastern Amazonian Tocantins River in 1984, included a 4,000 MW installation capacity power station in the first phase, 72-m above sea level (Tucuruí-I) (Fearnside, 2001). In the second phase from 2002, the installed capacity increased to 8,370 MW, 74-m above sea level (Tucuruí-II). Raising the water level from 72 to 74 m increased the inundated forest area from 2,247 km² in 1984 to ∼2,918 km² in 2002 (Eletronorte, 1989; Fearnside, 2001). This resulted in the creation of some 2,200 forest islands on higher elevation terrain (Fig. 1).

The regional climate is markedly seasonal, with a rainy season from December to May and a dry season from June to November (Alvares et al., 2013). Mean annual precipitation and temperature are 2,354 mm and 27.5 °C, respectively (Alvares et al., 2013). The vegetation is typical of Amazonian terra firme forests, containing 80–90% forest cover and an understory dominated by several palm species (Ferreira et al., 2012).

In 2002, the Tucuruí Lake Environmental Protection Area (APA Lago de Tucuruí, in Portuguese) was implemented spanning an area of 568,667 ha. According to Brazilian Protected Areas legislation (SNUC, 2000), an Environmental Protection Area (APA) is a sustainable-use reserve (IUCN category V) designated as a mosaic to meet both the interests of local communities and wildlife conservation. Around 72% of the forested area has been lost from 1984 to 2017 within this APA, mostly due to forest conversion into cattle pastures (Velastegui-Montoya, De Lima & Adami, 2020).

When the area was flooded in 1984, thousands of terrestrial and arboreal vertebrates were rescued and released into two specific areas of remaining, named Zones of Wildlife Protection (ZPVS 3 and ZPVS 4), where fishing, hunting and human settlements were banned, as enforced by environmental agents from ELETRONORTE, the ELETROBRAS subsidiary company for public power services in the northern Brazil. ZPVS 3, on the left bank, is an area of 37,672 ha where 28.5% is flooded, 60% of which has native vegetation. ZPVS 4, on the right bank, spans an area of 22,731 ha, 59% of which is flooded and 39% is native vegetation remnants (Fig. 1).

We carried out bird inventories along most of the reservoir margins. The forest avifauna was intensively sampled in forest areas located on 39 islands ranging in size from 3.4 to 2,551 ha (mean ± SD = 169.3 ±488.8 ha) in both ZPVS 3 and ZPVS 4 using mist-netting and point count (Fig. 1). Waterbirds, birds associated with riparian habitats, and conspicuous species such as parrots, eagles and falcons were sampled in most of the reservoir area by boat censuses (Fig. 1).

Avian species inventory

Avian species inventory

Our assessment in the THR influence area resulted in 479 species representing 74 families Appendix S1. Of this total, 355 species already had at least one voucher specimen deposited in museum collections for the lower Tocantins River, including 299 species with voucher specimens for the THR influence area. During our 2005–2007 field campaigns, we recorded 404 species and provided our own digital vouchers for 275 species (68%), 140 of which currently have archived vouchers in either WikiAves (82 species represented by images) or xeno-canto (84 species represented by sound-records). Seventy-five species were also photographed following captures by mist-netting. Only 26 species that we recorded are not currently represented by a museum specimen, image, or sound-record; 23 of which were recorded during our quantitative surveys (point counts, mist-netting, or boat census); and 3 species recorded only by either visual or acoustic observations: Panyptila cayennensis, Threnetes leucurus and Micrastur mirandollei. We decided to retain these species in the final checklist because of their high reliability of occurrence considering our experience and their known geographic and ecological distribution. Another two recorded species are considered as hypothetical due to inadequate level of documentation: Jabiru mycteria and Hydropsalis maculicaudus (Table S1). We recorded 107 species for the first time for THR avifauna. Most of these species are residents, 47 species are classified as migrants and 23 with unexpected occurrences may be vagrants, occur as low-density migrants or are expanding their ranges in Amazonia following deforestation (Appendix S1).

According to the IUCN Red List (IUCN, 2019) and the Brazilian Threatened Species List (ICMBio, 2018), a total of 44 (9.2%) threatened species were confirmed as part of the THR avifauna, 34 of which still persist in this area. Most of these threatened species are endemic to the Amazonian biogeographic provinces known as the Belém Endemism Area or the Xingu Endemism Area, east and west of the Tocantins River, respectively. Crax fasciolata pinima (Belém Curassow) and Phlegopsis nigromaculata paraensis (Black-spotted Bare-eye) were not recorded during the 2005–2007 inventory. These are both subspecies from Belém Endemism Area red-listed as Critically Endangered and Vulnerable, respectively. As a replacement for the Belém Curassow, the subspecies Crax fasciolata fasciolata (Bare-faced Curassow) was recorded on both banks of the THR influence area, suggesting that some left bank individuals had been translocated to the THR right bank during the rescue operations as the reservoir filled. In a similar way, we obtained three photos from two records of Ortalis superciliaris (Buff-browed Chachalaca) at the Tucuruí municipality on the left-bank. It is unclear if these records represent an ecologically viable population from translocated individuals or a natural expansion of the geographic distribution.

Our species accumulation curves are near asymptotic for all sampling techniques (Fig. 2). The total number of estimated species produced by the Chao1 estimator indicates a total sampling completeness of 80–85% for mist-netting, 84–86% for point-counts and 92% for boat censuses (Fig. 2). These three quantitative survey techniques have inherent biases toward some species groups and are affected by habitat characteristics, but their complementary use increased the total richness as indicated by the number of species uniquely recorded by each method: 23, 66, and 99 species were exclusively recorded by mist-netting, point-counts, and boat censuses, respectively. Another 39 species were exclusively recorded by our qualitative effort. This effort resulted in a total of 404 species (316 species on the left bank and 307 on the right bank). Furthermore, recent sighting records made by 47 volunteers who recorded a total of 358 species were searched using online citizen science databases and added to our overall effort.

Possible extinctions on THR influence area

We failed to record 75 species for the THR influence area, 53 of which (11.06% of the THR avifauna) were considered as extinction-prone candidates (Table 1). We excluded two species considered as vagrants (Pterodroma incerta and Stercorarius pomarinus), one exotic species (Passer domesticus) and 12 species with digital vouchers on WikiAves, whose records are the first for the THR avifauna (Appendix S1). We also excluded seven species recorded by volunteers after 1984 with archived vouchers in WikiAves, e-Bird or xeno-canto: Picumnus aurifrons (two records in 2017 and one record in 2018); Pyrrhura anerythra (one record of one group with three individuals in 2020); Sakesphorus luctuosus (three records - 2018, 2019 and 2020); Myrmoborus leucophrys (one record in 2018); Cranioleuca vulpina (one record in 2019); Iodopleura isabellae (one record in 2015); Pachyramphus rufus (four records: 2013, 2017, 2018, 2020). These seven species cannot be considered extinct, but the number of sighting records suggest that their populations are declining and may be locally extinct in large portions of the THR influence area.

We were able to estimate extinction probabilities for 20 species (for which the number of historical records was ≥ 5). A total of 15 species were considered to be extirpated based on Solow’s equation and nine species were presumed to be extant (Table 1). Psophia interjecta, an endemic of the Xingu Endemism Area that was split from Psophia viridis (Ribas et al., 2012), and Pyrilia vulturina, a target species in the Brazilian wild bird trade, are listed as Endangered and Vulnerable in the IUCN Red List, respectively (IUCN, 2019). For eight species the extinction dates ranged from 1980 to 2003 and their extinction events most likely directly result from the implementation of the Tucuruí Dam. Nine species are strictly dependent on forest habitat, including three that are restricted to river-edge or other fluvial habitats that were critically impacted by the dam and the rising floodwaters: Hypocnemoides melanopogon, Nasica longirostris and Automolus rufipileatus. Two other species, Gymnomystax mexicanus and Nemosia pileata, are also found in riverine habitats and, together with Chlorestes notatus, Manacus manacus and Chiroxiphia pareola, are common species in old secondary forests and degraded primary forests.

Another 67 species recorded for lower Tocantins River, either upstream or downstream of the THR influence area, were missing from our surveys (Table S1). These species were distinguished into five groups: (1) species recorded downstream of the THR; (2) species restricted to natural savannah habitats, with low probability of occurrence in THR; (3) species recorded upstream of the THR in the Marabá municipal county; (4) species recorded both upstream and downstream of the THR; and (5) species recently recorded in any of the municipal counties around the THR influence area. We are reasonably confident that most of these highly forest-dependent species once occurred throughout the THR influence area but have now become locally extinct.

Extinction risk

Our extinction vulnerability assessment indicates that 54% of the 236 species classified as vulnerable through a combination of criteria related to rarity were recorded in fewer than seven of the 139 sites on the THR influence area that we surveyed and 30% were recorded in only one or two (Fig. 3 and Table S2). Considering a set of 133 vulnerable species recorded in 39 forest islands with different areas, we visually identified four broad occurrence categories (Fig. S3). First, species restricted to forest islands larger than 1,800 ha, including terrestrial and understory insectivorous species such as Microcerculus marginatus, Sclerurus rufigularis, and obligate ant following Dendrocincla merula and Phlegopsis nigromaculata. This category also included the midstory insectivore Jacamerops aureus and the canopy frugivore Deroptyus acciptrinus. The second category included species occurring on forest islands larger than 100 ha and also understory and midstory insectivores, such as Piprites chloris, Platyrynchus saturatus, the small frugivorous Lepidothryx iris and Dixiphia pipra. The third category included those species whose occurrence included islands smaller than 100 hectares despite a low abundance (e.g., Conopophaga roberti, Phaethornis superciliosus). The fourth category included highly abundant species occurring across the forest island gradient surveyed (Lophotriccus galeatus, Pteroglossus aracari, Thamnophilus stictocephallus). The first three categories illustrate the potential decline of forest-dependent populations and, consequently, the increase in regional scale extinction risk due to the high degree of deforestation within and around the THR. The species in our extinction-prone candidate list were classified as Vulnerable, but they spanned the entire vulnerability gradient.

Avifaunal inventory effort and data limitations: general overview

Here, we present a comprehensive landscape-scale avifaunal inventory effort focused on the THR influence area after 21–23 years since the onset of maximum reservoir inundation. Our inventories draw upon different and complementary sampling methods, which tend to survey bird communities in heterogeneous habitat mosaics most efficiently. Our sampling protocol was based on quantitative techniques (mist-netting, point counts, and boat censuses) supplemented by qualitative efforts involving less systematic ad hoc field observations. This effort resulted in a total of 404 species (316 species on the left bank; 307 on the right bank), with over 100 species recorded for the first time for the avifauna of this region. This high number of new records occurred in part because previous studies had focused on more haphazard collections of voucher specimens using mist-nets and firearms. Both methods are highly selective and were used with little regard for establishing a general quantitative baseline from which future compositional changes in the avifauna could be best monitored and assessed. Even studies developed during the construction phase of the hydroelectric dam failed to provide such a quantitative baseline. For example, the book chapter on birds in the Inventory of Fauna Utilization Project (Plano de Inventário do Aproveitamento da Fauna—PIAF, in Portuguese; Eletronorte, 1985a) presented a qualitative checklist, without assessing the complete hydrological cycle and lacking information on sampling effort or number of replicates. The rather expensive and controversial wild vertebrate rescue operation named Operação Curupira (Eletronorte, 1985b), which cost approximately US$30 million to deploy over several years, and contributed little information on the avifauna of this region, since most individuals rescued were relatively large-bodied and often remained unidentified at the species level. Therefore, many relatively common aquatic or canopy-dwelling species had not been recorded in the THR region prior to our study, and we do not have abundance data for most of avian species before the Tocantins River was dammed.

Despite the lack of previous quantitative data, our large sampling effort, with additional recent records from online citizen science databases, indicates that remarkable changes occurred in the THR avifauna after a 36 years of post-inundation history with an increase in extinction risk as consequence of the synergistic effects of flooded, deforestation and forest degradation surrounding the reservoir. Even confirming the large number of bird species persisting in the long term, our forest island occupancy analysis indicates that several forest species classified as vulnerable are locally extinct on forest islands smaller than 1,800 ha or smaller than 100 ha. This scenario can be extended to the entire THR influence area, even though the open-water matrix is more inhospitable than newly created anthropogenic habitats surrounding the reservoir (Mendenhall et al., 2014). Conversely, we recorded open-habitat species that were unexpected in this region, such as Athene cunicularia, Columbina squammata, Zenaida auriculata, Amazona aestiva, Myiothlypis flaveola and Gnorimopsar chopi. Although at low abundances, these open-habitat species have been gradually expanding their geographic ranges following deforestation. This gradual turnover consisting of local extinctions of forest specialists and replacements by open-habitat species has been documented elsewhere along the ‘Arc of Deforestation’ (e.g., Moura et al., 2013).

Species extinction after 36 years of post-inundation history

In a presence-absence analysis of species extinction of the THR, Henriques & Dantas (2009) suggested that most of the missing species in our sampling effort and part of the species historically recorded only downstream or upstream of the reservoir have in fact been extirpated (71 species with 27 additional extinct species on the right bank alone). They emphasized that 70% of these species are forest specialists, including 30% of riverine habitat specialists that were completely eliminated by the rising floodwaters. However, this may have been overestimated because some of these species were recorded only upstream or downstream of the THR influence area or may not belong to the regional species pool. In fact, although species presence records are assumed to be reliable, the same could not be said for absence records as failing to detect any given species does not necessarily equate to absences (e.g., Gu & Swihart, 2004). We found evidence for the possible extinction of 53 consistently recorded species for the THR influence area, all of which remained undetected since 1984, including Cnemotriccus fuscatus, erroneously listed as still occurring in THR by Bueno et al. (2018). We can also confirm the vulnerability of forest specialist species, particularly insectivores and fluvial habitat specialists, including river species associated with sand banks, river islands, and rocky outcrops (Table 1), which are usually highly vulnerable habitats to hydroelectric development (Lees et al., 2016). Specialized forest insectivores frequently show high sensitivity to habitat fragmentation (Stouffer & Bierregaard, 1995; Sekercioglu et al., 2002; Powell, Cordeiro & Stratford, 2015), even though other life history traits, such as large body size, specialized feeding habits, low dispersal capacity and narrow geographic distribution, are good predictors of bird sensitivity to extinction (e.g Bregman, Sekercioglu & Tobias, 2014).

Our extinction probability estimates for 20 species (for which the number of historical records was ≥ 5) indicate that (i) 15 species (75%) have gone extinct in the THR influence area, (ii) part of these extinctions is related to the impacts of land-use change prior to damming, but (iii) the rate of species loss increased from 0.25 species/year from 1916 to 1944 to 0.34 species/year from 1980 to 2003. To put this in perspective, in a long-term (>160 years) analysis of bird community erosion of the region around the eastern Amazonian city of Belém, the rate of species loss was estimated at 0.28 species/year (Moura et al., 2014). The extinction rate may be even higher in the THR influence area since we are confident that other species were also last detected in 1984, and several extinction-prone species are restricted to riverine habitats that were permanently flooded by the rising floodwaters in the aftermath of river damming. Most of these habitat specialist species are in fact visually or acoustically conspicuous, and very unlikely to be overlooked over three years of intensive sampling. For example, we recorded several of these species during less than 60 days of fieldwork as part of the Xingu River surveys using the same sampling methods (Table 1). However, our data also indicate that Myrmotherula multostriata, Xiphorhynchus obsoletus and Campylorhamphus multostriatus may still be extant, even though their habitat had been permanently flooded. In a study on methods for inferring species extinction based on sight records, Boakes, Rout & Collen (2015) show that Solow’s equation provides low predictive power if extinction time is near the end of the observation period, which suggests that these species may now be extinct according to Solow’s equation (Table 1).

Considering species for which we were unable to calculate extinctions probabilities (those with fewer than 5 historical records), we are confident that Hylophylax punctulatus and Hylopezus berlepschi, two alluvial forest species, Vireolanius leucotis, one canopy terra-firme forest, Pygochelidon melanoleuca and Atticora fasciata, two gregarious habitat-specialist species of rocky outcrop habitats, are also locally extinct given our field experience with these species at other Amazonian sites that indicate they are common and conspicuous. The alluvial forest and the rocky outcrop habitats were extirpated by permanent flooding and high deforestation rates of upland forest. A recent multitemporal analysis of deforestation in response to the construction of the Tucuruí Dam indicates the highest deforestation rate occurred in the first period of the analysis (1984–1988), due to the areas submerged by the reservoir and due to the anthropogenic disturbances, such as timber extraction, road construction, and the conversion of large tracts of forests into croplands (Velastegui-Montoya, De Lima & Adami, 2020). The other 25 species with fewer than five records include one globally threatened species and one near-threatened (IUCN, 2019), respectively: Lophornis gouldii and Zebrilus undulatus. These and several other extinction-prone species in our hypothetical list are conspicuous and very unlikely to go undetected over three years of intensive sampling. These missing species indicate that Solow’s equation as used here likely underestimates the full impact of the Tucuruí Hydroelectric Dam on the avifauna.

Our analysis also revealed the impacts of land-use change in the bird community of THR in the first half of the 20th century, because seven species became presumably extinct between 1916 and 1944. These species were extinct during the course of the construction of the Tocantins Railway, used to transport Brazil nuts and natural rubber. Between the inauguration of the first stretch of the railway in 1908 and the last, 36 years were required to build 117 km, which connected Alcobaça (present-day Tucuruí) to Jatobal, an occupation process that led to the near extinction of three indigenous groups of the lower Tocantins basin (Laraia & Da Matta, 1978). Although these findings are less severe than the impacts of land-use change on the bird community of the San Antonio cloud forest over 100 years (Kattan, Alvarez-Lopez & Giraldo, 1994; Palacio, Kattan & Pimm, 2020), it stresses our scant knowledge about the historical impacts of the modern colonization of Amazonia on biodiversity.

Implications for the long-term impact assessments of hydropower infrastructure

Our results show that (1) the lack of a pre-damming systematic inventory likely masks some of the temporal changes in extinction patterns; (2) it is critical to initiate conservation measures before the main impact, when vulnerable species can be more easily surveyed; and (3) further studies and the creation of protected areas beyond the immediate area impacted by hydropower infrastructure are critically required to protect habitats and species at the landscape scale.

Our results are consistent with the notion that avian assemblages in human-modified Amazonian forests likely pay a heavy extinction debt over long timescales. Much like the Belém case study (Moura et al., 2014), the conservation value of remaining forest patches at Tucuruí decays over time because disturbance-adapted species are recruited into novel assemblages at the expense of the gradual loss of forest-dependent species (a typical case of the ‘winner-loser’ paradigm: Filgueiras et al., 2021). We expect that insular forest patches created by the reservoir typically harbored depauperate avian assemblages comprised primarily of naturally abundant and habitat generalist species (see also Bueno et al., 2018). Biodiversity loss in Amazonia is largely irreversible and a consequence of deforestation, and the THR influence area is no exception because of the elevated number of threatened bird species (Bird et al., 2011). However, other threats have also been detected over the 36 years since the damming of the Tocantins River, such as illegal logging, forest wildfires, overhunting and the spread of domestic mesopredators (e.g., cats and dogs) (Barlow et al., 2006; Whiteman et al., 2007; Benítez-López et al., 2017). The synergistic effects of these threats coupled with forest habitat loss clearly aggravates avian population declines and local extinctions in the Tucuruí region.

The high levels of avian gamma-diversity hosted by the THR places it among the most diverse across the Amazon basin. Contributing to this high diversity is the fact that the THR influence area includes portions of both the Xingu and Belém lowland areas of endemism, with several parapatric species replacing one another on opposite riverbanks. This diversity can be higher if we consider infra-specific taxa and recent molecular studies, which confirm deep molecular divergence between opposite riverbanks (e.g., Ribas et al., 2012 for Psophia complex; Batista et al., 2013 for Dendrocolaptes certhia complex; Maldonado-Coelho et al., 2013 for Pyriglena leuconota complex). In addition to this high diversity, we recorded 44 threatened or near-threatened species (IUCN, 2019; ICMBio, 2018), 34 of which are still extant in the study region. These species have been further threatened by increasing habitat loss since the dam was built (see Fig. 1), highlighting the relevance of this region for biodiversity conservation.