Fish climbing in the upper Congo Basin (Central Africa), first report for the shellear Parakneria thysi on the Luvilombo Falls

Waterfalls, dominant geo-hydrographic obstacles present in most of the rivers of the upper Lualaba ecoregion, are the most difficult to ascend for fishes, even for those with (pre-)adaptations to resist strong water currents^1,2. However, examples of fish species climbing similar vertical obstacles have been found in several phylogenetically distant rheophilic species and their lineages, e.g. five orders and eight families, from South America, Asia and Australia, referred to here (Supporting Note S1 and Supporting Table S1). Besides these species having a greater fall-climbing ability, reports on other species, e.g. the European freshwater blenny, Salariopsis fluviatilis (Blenniiformes), have also shown this propensity, but with much less efficiency (Supporting Note S2).

However, this phenomenon is poorly documented and remains mostly anecdotal for freshwater fishes on the African continent. Specimens of the Mochokidae catfish (Siluriformes), genus Chiloglanis, for example, have been mentioned to climb wet vertical surfaces using mainly their suckermouth as an adhesive structure^3,4. Indeed, Chiloglanis paratus has been reported to be able to climb damp barriers. A similar climbing ability to climb damp surfaces of barrier rocks and weirs using its suckermouth and pectoral fins has also been mentioned for the cyprinid, Labeo cylindricus (Cypriniformes)⁴.

This is also true for some species of the African endemic family Kneriidae (Gonorynchiformes)^5,6. This family contains four genera^7,8: Cromeria, with only two species, one distributed in the Nile Basin and the other in West African river basins⁹; Grasseichthys, with currently a single species distributed in the Ogooué Basin and middle Congo (sub)Basin, in Central Africa⁷; and the genera Kneria and Parakneria, with 15 species each, which are both well-known from the upper Lualaba (upper Congo Basin) but also far beyond the borders of the Congo Basin itself^10,11. The distribution of both Kneria and Parakneria is essentially conditioned by their affinity to torrential habitats in high elevation portions of sub-tropical Africa^7,12. However, kneriids are absent from the central part of the Congo Basin, being the Cuvette Centrale, where its sister family Phractolaemidae occurs^5,13. Phractolaemus ansorgii, the only species in the family⁸, is sometimes included in the family Kneriidae^14,15,16,17. Of the latter family, only two genera, Kneria and Parakneria, occur in the upper Lualaba ecoregion¹⁸. Most of the currently known species inhabit the south-eastern part of the Congo Basin at high altitudes^10,11, mainly on the plateaux of the upper Lualaba ecoregion¹⁹, such as the Kundelungu Plateau (KP), at ~ 1200–1700 m altitude above sea level (a.s.l.)²⁰.

Kneria, shellears, is the only genus, of this family, reported to be able to climb steep waterfalls. This is true for Kneria auriculata, which have been reported climbing the wall of an unnamed dam in Mozambique²¹. As some of the habitats occupied by K. auriculata are located at high altitudes (~ 1170 m a.s.l.), it has even been implied that these can only be occupied by a fish that, in the past at least, was able to climb and surmount a series of waterfalls with a total height of 154 m^22,23. Species of the genera Kneria and Parakneria appear, by some of their paired-fin related morphological features, to be well-adapted to torrential environments¹⁸ and thus possibly (pre-)adapted to waterfall climbing. These features are, the horizontal position of the pectoral fins, which tend to form an adhesive surface on their ventral side²⁴, and the presence of thickened skin pads on them. These pads were observed in Kneria paucisquamata, but only along the ventral surface of the first two rays of their pectoral and the first ray of their pelvic fins²⁵. Such pads could have an important role as adhesive structures on rocky substrates¹⁸, while alternative functions, such as protection against abrasion by the substrate, remain unconfirmed to date²⁵. Nevertheless, a distribution study of K. auriculata in the upper Crocodile River (Mpumalanga; Transvaal Region; South Africa) revealed no specimens upstream of 76 m high waterfalls²⁶. Moreover, considering its rather limited overall distribution, it was considered doubtful this species could surmount even waterfalls of 5 m height^24,26. Thus, although anecdotal²¹ and even dubious^24,26, this was, before our recent observations on the Luvilombo Falls²⁷, the only documented case of waterfall climbing known within the family Kneriidae. Based on these new observations, the present study aims, beyond tales, to document, for the first time and with available cinematographic and photographic evidence, the effective occurrence of waterfall climbing in Kneriidae.

Benthic life of Parakneria thysi and its ability to climb waterfalls

Specimens of Parakneria thysi (Fig. 1a) were observed in typical riffle habitats of the Luvilombo River, a left bank tributary of the lower Lufira River, located in the Upemba National Park (UNP)²⁸ (Fig. 1b). They were apparently clinging to the river bedrock with their paired fins. Their local name in the Sanga (Bantu) language of the Kyubo region is “Tulumbu” (plural) and “Kalumbu” (singular), derived from the verb “kulumba”, which literally means ‘to stick’, and thus refers to their clinging behaviour. Additionally, numerous P. thysi specimens were observed (2009, 2018 and 2020) climbing the Luvilombo Falls (~ 15 m high), located on the border between the UNP and the Kundelungu National Park (KNP) (Fig. 1b–d). Fish performing this climbing do so at the end of the rainy and the beginning of the dry season (April vs. May: see Supporting Note S3). They move along the splash zone, i.e. the part of the vertical wall of the falls with a limited water flow, but still moistened by regular water spills (see Supporting Video S1).

However, when the flood level is at its maximum and the water flow and spill together still cover the maximum width of the waterfall itself, fish only climb along the rocky lateral edges of the river. This was observed on the right bank of the river (see Supporting Video S2). To climb these falls, small- to medium-sized P. thysi specimens [~ 37–48 mm standard length (SL), vs. ~ 96 mm maximum SL] follow one another in large numbers (estimated in thousands of individuals) (Fig. 2a; see Supporting Videos S1 & S3). They cling to the vertical rock surface of the falls first using their pectoral fins, and then using their pelvic fins to reinforce their resistance to the current (Fig. 2b). During vertical movement and rest, these two pairs of fins are fully extended (Fig. 2b). Fish that have successfully clung to the base of the vertical rock surface of the falls will propel themselves upwards using lateral undulatory movements, mostly of the body’s posterior half, and orientating movements of the well-deployed paired fins, being mostly the pectoral supported by the pelvic fins (Fig. 2a,b, see Supporting Video S4).

External ventral structure of paired fins and their weight carrying capacity

The ventral surface of these paired fins possesses skin thickened, cushion-like areas, called pads²⁵ on their first rays, i.e. the first 7–9 and 2–3 rays in pectoral and pelvic fins, respectively. These distinct pads (see Supporting Note S4) are composed of tiny unicellular hook-like projections, called unculi. These microstructures are, however, longer and more densely implanted on the pectoral (Fig. 2c) compared to the pelvic fins (Fig. 2d) (see Supporting Note S5). Unlike the presence of pads on the ventral surface of the paired fins, the abdominal surface of their body is devoid of scales but also of unculi, or any similar adhesive structures (see Supporting Fig. S1).

The surface area of the pads for both pairs of fins, especially their proportion (in %) in relation to the total surface area of the fin (see Supporting Note S6), shows (almost) no allometric trend as a function of size (SL: mm). This result is accompanied by a very low correlation coefficient: 0.006 and 0.008, for the pectoral and pelvic fins, respectively (Fig. 3a1,b1). However, although there is no proportional increase in the pad surface (in %) with size, the specimen weight tends to increase exponentially, especially for those measuring over 50 mm SL (Fig. 3a2,b2). Therefore, larger specimens (> 50 mm SL) seem to lose the weight carrying capacity needed to effectively climb the falls.

Internal morphological (pre-)adaptations of paired fins

The pectoral fins of Parakneria as their main clinging support are also osteologically different from those of the morphologically similar genus Kneria, which seems to present limited adaptations to a torrential lifestyle and, conversely, adaptations to a more pelagic lifestyle. For example, the proximal portion of the pectoral-fin rays in Parakneria extends well beyond the adjacent proximal radials, and each ray has a scythe-like shaped head that partially overlaps that of the subsequent ray (vs proximal portion in Kneria not extending beyond the adjacent radials, but has a plate-like extension on the anterior side of the mid-proximal half of the most anterior rays) (Fig. 4a1 vs. b1 and a2 vs. b2). Furthermore, in Parakneria, the pectoral fins seem particularly well attached to the pectoral girdle by a broad scapula bridge, with the three most posterior radials (n°2–4) of a filiform shape [vs a less strong scapula bridge in Kneria with the most posterior radial (n°4) of a more or less triangular shape and distinctly larger (Fig. 4a3 vs. b3 and 4a2 vs. b2)].

The pelvic girdle is not directly connected with the rest of the skeleton, particularly with the axial skeleton, in neither P. thysi nor K. stappersii (see Fig. 5). Nevertheless, morphological differences of the pelvic girdle have been noticed between both species. In P. thysi, the basipterygium is broader with a more developed medial wing (vs narrower in K. stappersii) (see Fig. 5a1,b1). Furthermore, in P. thysi, the medial wings of right and left girdles have an extended contact zone, about half the basipterygium length (vs contact zone restricted to anterior-most third or quarter of the basipterygium) (see Fig. 5a2,b2). Moreover, the posterior portion of the basipterygium is more robust in P. thysi than in K. stappersii. The former has a broad, deep medial arm connected to a broad, triangular posterior process, and a pointed and upwardly directed process on the outer margin of the basipterygium (vs narrower medial arm, without a well-developed posterior process, and without a distinct process on its outer margin). Furthermore, in P. thysi, the anterior portion of the first pelvic-fin ray is triangular and about three times broader than the anterior portion of the adjacent ray (vs rounded and only about twice as broad as the anterior portion of the adjacent ray). Finally, the pelvic splint is better developed and curved in P. thysi. An analysis of the ventral surface muscles shows the presence of paired infracarinalis anterior muscles^29,30, also better developed in P. thysi (see Fig. 6a1,a2) than in K. stappersii (see Fig. 6b1,b2) (see Supporting Note S7). Interestingly, the pelvic splint is well developed in Phractolaemus ansorgii (Phractolaemidae³¹), a species belonging to the sister family of the Kneriidae. However, the paired infracarinalis anterior muscles are almost absent in this species (for more details: see Supporting Fig. S2 and Note S8).

Waterfall-climbing abilities of Parakneria thysi

During the migration period, the density of migratory fish seems to vary with the water level. Observations made in 2020 showed that their density is low in early April, when the water level is at its highest, which seems to correspond to the onset of the migration period. From then on, their density increases and reaches a peak by mid-April, when the water level drops to its average rainy season level. Their density then declines further by the end of April and early May, when the water level drops even further, corresponding to the end of the rainy season and thus the end of the migration period.

Furthermore, the number of migratory fish seems to be rather limited during the day. However, it increases towards the end of the day, at sunset (around 4–6 p.m.), then decreases again at sunrise (around 6–8 a.m.).

Nevertheless, during an ascent the upward movement of individual fish is not continuous. Two types of pauses can intermittently be observed. First, short, more or less regular pauses, usually about ~ 15–60 s long, are observed in-between different bouts of vertical movement. During these pauses, the individual grip to the vertical wall of the falls (see Supporting Video S5). Second, longer pauses, often lasting 2–15 min, or even one hour or more, are observed when the fish reach one of the horizontal ledges during their ascent. Consequently, large numbers of fish congregate on those ledges before engaging the next climbing phase (Fig. 7a,b). This seems to indicate that these upward movements require a lot of energy and as observed, the specimens therefore need longer resting periods to recover from the effort.

As such, to overcome a vertical surface one meter in height, for example, fish need an average of 30–60 s of movement, at an average speed of about 1.5–3.0 cm/sec (0.9–1.8 m/min). However, they will also need eight to nine short pauses [~ 120–480 (2.0–8.0) or ~ 135–540 (2.3–9.0) seconds (minutes), respectively]. Therefore, to reach the top of the Luvilombo Falls, a fish climbing at a maximum speed of about 3.0 cm/sec would need 900 s (15 min) of movement and 1800s (30 min) of short pauses, for a total of 2700 s (45 min). Moreover, nine main horizontal ledges where fish seem to rest, for example, for about an hour each, have been identified. It would therefore take an individual fish approximately 9 h and 45 min to cover the entire height of the falls. This suggests that the fish could need almost a whole day or night to fully overcome the falls.

Moreover, this arduous phenomenon is not without major difficulties. Indeed, some specimens fall abruptly, for example when they are suddenly hit by a jet of water. Such a fall is also more likely when the fish move upside down (Fig. 7b), when they try to circumvent an overhanging cliff (see Supporting Videos S6).

Waterfall-climbing abilities in Kneria and Parakneria

In their natural environment, Kneria generally swim in open water and have a rather pelagic lifestyle (see Supporting Videos S7 & S8). In contrast, Parakneria, due to their benthic lifestyle, tend to spend more time clinging horizontally to the rocky substrate with their two pectoral fins, supported by their pelvic fins (Fig. 2a; see Supporting Videos S9 & S10). From these field observations, it is apparent that, compared to Kneria, Parakneria seems to be better adapted to live in more extreme torrential environments. This is further supported by osteological details showing a broad scapula bridge, the proximal portion of the pectoral fin rays extending over the adjacent radials, and filiform radials (n°2–4) in Parakneria. These are probably related to the unique climbing ability of Parakneria. Its broad scapula bridge indicates the likely presence of hypertrophied musculature and the filiform radials and the scythe-like head of the pectoral fin rays suggesting a refined and more coordinated movement of the rays. The proximal filiform radials in Parakneria, known to result in a less flexible skeleton, are also observed in some Blenniidae³², a family referred to have at least a limited climbing ability³³. Conversely, there are plate-like anterior extensions towards the middle of the proximal half of at least the most anterior pectoral-fin rays in Kneria. These appear, in some respects, to be similar to the plate-like extensions found in Pantodon buchholzi (Osteoglossiformes: Pantodontidae), although this species prefers slow-moving waters, where they seem to enable its air-gliding behaviour³⁴. Thus, in Kneria, these extensions could also help the pectoral fin to function as a single wing/palm-like entity, which is here considered an adaptation to its more pelagic lifestyle. Nonetheless, the functionality of the pectoral fin and girdle, in Parakneria and Kneria, requires further analyses that will be explored in a subsequent paper.

To confirm a greater adaptation degree of Parakneria to torrential lifestyle, the present study shows, for the first time, that some P. thysi specimens climb falls. They propel themselves by lateral movements of the posterior part of their body and clinch to the rocky substrate with their paired fins. Both the pectoral and pelvic fins bear pads composed of numerous tiny, spike/hook-like unculi. These pads are more numerous in Parakneria than in Kneria (see Supporting Note S9). The lack of connection of the pelvic girdle with the axial skeleton suggests the more limited pelvic role for waterfall climbing in P. thysi. This contrasts with the case of some stream loaches (Balitoridae)³⁵, such as Cryptotora thamicola, in which the pelvic fins are implicated in their salamander-like lateral diagonal-couplet walking behaviour³⁶.

Nevertheless, there are clear differences between the pelvic fin structures of P. thysi and K. stappersii that are likely related to their different climbing abilities. These include: (i) a broader basipterygium with a more continuous contact zone between the basipterygia medial wings, (ii) a more robust posterior portion with a deep medial arm connected to a broad, triangular shaped posterior process, and (iii) a pointed and upward-directed process on its outer margin. All indicate a larger surface area for connecting ligaments and muscles, and thus the presence of hypertrophied muscles. For example, a broader basipterygium likely indicates the presence of a hypertrophied abductor and adductor muscles’ group. Their insertion points also include the processes of the posterior portion of the basipterygium and the infracarinalis adductor and abductor muscles’ group. In addition, the pelvic splint is well-developed in more rheophilic species^37,38, such as in P. thysi (vs less in K. stappersii)³¹. Furthermore, there is evidence of greater development of the infracarinalis anterior paired muscles in Parakneria than in Kneria. For these muscles, two main functions have also been identified in king salmon [Oncorhynchus tshawytscha (Salmoniformes: Salmonidae)], pulling the pelvic girdle forward and strong ventral flexion of the body³⁹. Considering that the infracarinalis anterior muscles are connected to the more developed pelvic fin muscles (adductor and abductor muscles’ group), we speculate that in P. thysi, the former might be the most important for waterfall climbing. This muscle, when contracted, pulls the pelvic fin forward and causes a strong ventral flexion of the body. Thus, it probably increases the attrition between the pectoral and pelvic pad (unculi) areas with the wall.

Compared to the two general locomotion modes identified so far in climbing of vertical obstacles by Hawaiian Oxudercidae (Gobiiformes)⁴⁰, that of P. thysi seems to be the most similar to the powerburst, as documented for two species, Awaous guamensis^41,42 and Lentipes concolor^40,41,42. This type of locomotion is achieved by axial body undulations and adduction, mostly by the pectoral fins⁴¹. It thus differs from the inching, observed in the species Sicyopterus stimpsoni, which is performed by using alternatively a pelvic and a buccal sucker^40,41,42.

Among the references reporting the ability of K. auriculata to climb waterfalls, the only concrete evidence is the written testimony of Jackson in 1961²¹. He reported observing these fish attempting to migrate up a few inches of an artificial dam in the Gorongosa Mountain area (Mozambique)²¹, currently the type region of Parakneria mossambica⁴³. This was more than a decade before the original description of the latter species, and even of the genus Parakneria, in 1965¹⁸. Therefore, the species now named P. mossambica could have been mistaken for K. auriculata. Similar identification problems still persist today, as underwater, at first glance, Kneria and Parakneria specimens can be quite easily confused with each other and thus also with some of their behaviours.

However, despite the climbing abilities documented here for P. thysi and the lack of morphological adaptations in Kneria to overcome steep waterfalls, some specimens of the latter genus, unlike Parakneria, occupy higher altitude environments in terms of distribution, e.g. ~ 1170 m a.s.l. for K. auriculata^22,23. This also seems to be the case for the middle Lufira, (upper Lualaba Basin)¹⁹, and the lower Luapula (Luapula-Mweru Basin)⁴⁴, draining the western and eastern flanks of the KP, respectively, in and around the KNP (Fig. 1a). In these rivers, Kneria are abundant on the KP, on the flanks, and also in the nearby plains below the falls, on both sides of the plateau²⁰. Instead, Parakneria have so far only been found in the plains, downstream the falls, on both sides of the plateau (PKM pers. obs. 2015–2017). This pattern contrasts with the so-called Parakneria zone in the Luansa (Luanza) River (eastern flank of the KP), located at altitudes between 1060 and 1310 m a.s.l. on the KP, as identified by Malaisse^45,46. According to the RMCA collection records, all available Parakneria specimens collected from this flank of the KP, being both historical (Malaisse collections: 1966–1967) as well as recent (new collections: 2015–2017) ones, were collected from below the identified Parakneria zone, i.e. between 1004 and 1024 m a.s.l. (PKM, pers. obs. 2015–2025). This is likely another case of ambiguous generic identification, with Kneria specimens being misidentified as Parakneria malaissei. The occurrence of Kneria specimens at the high altitudes of the KP and the absence of the highly modified and climbing specialist Parakneria, still needs research attention.

Towards a better understanding of the waterfall-climbing behaviour in P. thysi

Based on in situ observations, the massive upstream migration of P. thysi specimens along the Luvilombo River, could be (I) triggered by the flooding, and the resulting formation of a lacustrine habitat, in the triangle-shaped area formed by the lower Lufira, the lower Luvilombo, and their confluence downstream of both the Kyubo and Luvilombo falls (see Supporting Fig. S3). This seems all the more plausible since this phenomenon apparently only occurs in years of abundant rainfall. Indeed, observations in April 2019, a year of low rainfall and no significant inundation downstream of both falls, suggest that no upstream migration of P. thysi specimens along the Luvilombo Falls did occur (P.K.M., pers. obs. 2019). In contrast, observations from April and May 2020, a year of abundant rainfall, showed that intensive upstream migration occurred throughout April and, to a lesser degree, also during the first half of May.

Furthermore, (II) given the low size variation of migratory specimens (~ 37–48 mm SL vs a maximum documented size of ~ 96 mm SL), the observed phenomenon can also be identified as a partial migration, characterised by the existence of two groups of individuals within the same population, migrants vs residents⁴⁷. In the present case, migrants may include fish washed downstream during previous major rains and those born downstream in the same year, which migrate upstream together to (re)occupy suitable riffle/rapid habitats upstream of the falls. This hypothesis shares at least some similarities with the upstream migration as reported for juvenile Hawaiian climbing Oxudercidae^41,48 and for Characidium cf. timbuiense (Crenuchidae) from the Crubixá-Mirim River in Brazil². The latter fishes migrate upstream, after being washed down by the currents, to reoccupy their habitats. According to the categorisation of partial migrations in fishes, three different types are distinguished based on the shared occurrence of migration and/or reproduction in all specimens of the same population⁴⁷. These are: (i) non-breeding partial migration, during which migrants and residents breed together but are separated from each other during the non-breeding period; (ii) breeding partial migration, when migrants and residents are together during the non-breeding period while separated during the breeding period; and (iii) skipped breeding partial migration, with some individuals migrating mainly for breeding, but not in all years⁴⁷. Referring to the observations made for P. thysi, in which only small- to medium-sized specimens migrate, it thus appears to be a form of non-breeding partial migration. Furthermore, the existing literature on partial migration reports body size (SL) as one of its major limiting factors. This is especially due to physiological reasons [e.g. thermal (cold) tolerance]⁴⁷. However, for P. thysi, although there is also an obvious restriction in body size, it is not physiological, but instead seems due to the size-related limitations in the physical carrying-capacity of paired fins (Fig. 3). This is also the case for Hawaiian Oxudercidae, for which climbing is only observed for specimens of 10–25 mm SL⁴¹, although species can attain a maximum size of 198 mm SL for S. stimpsoni⁴⁹ and 245 mm SL for A. guamensis⁵⁰.

Considering the large number of fish, not only of P. thysi, present in the flooded area downstream of the falls during the migration period, it is also conceivable that this upward migration is carried out to (III) avoid food competition. It has been pointed out that food availability is lower at the base of waterfalls than upstream⁵¹. Finally, this upward migration could also aim to (IV) avoid predation downstream of the falls, an environment that, during the flood period, supports quite dense populations of predatory fish. These include, for example, Schilbe intermedius (Siluriformes: Schilbeidae), which also seem to actively migrate to the base of the waterfalls for this purpose (P.K.M., pers. obs. 2020). Several ecological studies confirm that predation is generally greater down- than upstream of waterfalls^{41,48,49,50,51,52}.

Anthropogenic threats to the ichthyofauna of the Luvilombo River

In the Luvilombo River, P. thysi and the rest of this river’s ichthyofauna are unfortunately subject to significant anthropogenic impacts. These are observable at the Luvilombo Falls by the fact that during the flood period (April–May), fishing intensity increases downstream of the waterfalls (P.K.M., pers. obs. 2018–2020). Consequently, Parakneria specimens, which congregate there to prepare for their ascent, are easily caught by fishermen using mosquito net seining. This fishing technique, however, is forbidden by both the Democratic Republic of the Congo law (see Provincial Arrest n°2007/0108/Katanga of 09 November 2007) and the Congo Safari Kyubo Lodge concession holders, which prohibits, in all its forms, the use of fishing nets with mesh sizes below 5 cm. Furthermore, just after this migratory period, i.e. at the beginning of the dry season (May–September)⁵³, especially in years of low rainfall, the river is diverted upstream of the falls, at the Sangala Village (~ 26°50′8.7″ E–9°47′26.5″S) (Fig. 1b). This water is used to irrigate land for cultivating off-season crops, mainly beans and peanuts. This results in the downstream section of the river to dry out completely (Fig. 8).

The latter practice confirms that agricultural development can be identified as one of the main anthropogenic threats affecting aquatic biodiversity in general and ichthyofauna diversity in particular⁵⁴. Especially, agriculture, through irrigation, can lead to changes in the regime of a watercourse^55,56 and, consequently, to changes in habitat configurations. This reality requires increased protection of the aquatic fauna of the (lower) Lufira Basin in general, and of the Luvilombo River and its waterfall in particular. Furthermore, due to the presence of this unique and well-observable fish migratory behaviour, the Luvilombo Falls constitute a potentially propitious site for the development of alternative and seasonal (eco-)touristic activities. Therefore, it is hoped that the present paper might generate increased interest in this enigmatic migratory phenomenon and highlight the need for its comprehensive protection. This could be made possible by paying more focused attention to these falls and identifying them as a natural monument and/or ecosystem of national interest, for which the legal framework in the DRC is available under the law of February 2014 on nature conservation.

The climbing ability of P. thysi was observed at the Luvilombo Falls, Luvilombo River, a left bank tributary of the lower Lufira River (Fig. 1). For this study, successive field monitoring campaigns were organised in 2018, 2019 and 2020 between April and May. The effective climbing phenomenon was observed once in 2018 (27 April) and three more times in 2020 (05–12 April, 24–29 April, and 06 May). The precise description of this waterfall-climbing process is mainly based on the two field campaigns in April 2020, during which 10 h of observations were devoted each day, including six hours in the morning (6–12 a.m.) and four hours in the afternoon (2–6 p.m.). For more details on field observations and photographic and cinematographic recordings, see Supporting Methods S1.

A few dip-net fishing sessions were conducted at the base, middle, and top of the falls. The Parakneria specimens collected from there, were directly anaesthetised and/or euthanised, in accordance with European Directive 2010/63/EU on the protection of animals used for scientific purposes (see Annex IV), by adding a few drops (n = 10–20) of clove oil to the water, i.e. about ~ 0.5–1.0 millilitre per litre. Furthermore, they were photographed in a field aquarium, fixed in a 10% formaldehyde solution, and stored in the RMCA fish collections in 70% alcohol (RMCA 2021–019-P-0017–0019) for long-term preservation. This allowed, among other things, an in-depth examination of the ventral surface of the pectoral and pelvic fins, as well as the body. To this end, SEM images were taken of the ventral surface of both paired fins and the body. This allowed to document their detailed structural composition. For more details on specimen preparation for SEM, see Supporting Methods S2.

To further discuss the climbing ability in relation to internal bony structures, especially the pectoral and pelvic girdles, selected specimens of Kneria and Parakneria previously stored at the RMCA were digitised using microcomputed tomography (µCT) using a RX EasyTom 150 (8.9–10 W, 80–100 kV, 10–25 µm) (RBINS). Segmentation was performed using Dragonfly software version 2020.1.0.797 for Windows (Object ResearchSystems Inc., Montreal, Canada, 2020; https://www.theobjects.com/dragonfly/index.html).

The author(s) and date of the original description of all genera and species cited in the text are listed in the Supporting Table S1. All references not cited in the main text have been added in the Supporting Information files and numbered from number 56 onwards according to their appearance in the Supporting material.