Rediscovery of the critically endangered Hill's horseshoe bat (Rhinolophushilli) and other new records of bat species in Rwanda

Abstract Background For forty years, there has been growing uncertainty about whether Hill's horseshoe bat (Rhinolophushilli) still persists in Nyungwe National Park, Rwanda. Only known from one small area within the National Park, R.hilli is listed as Critically Endangered by the International Union for the Conservation of Nature (IUCN), based on its extremely small geographic range and presumed low number of mature individuals. Here, we present and describe bat species occurrence data contributed to the Global Biodiversity Information Facility (GBIF) that we collected as part of a long-term collaborative project to rediscover this lost species. This data paper describes the survey methods and findings resulting from cave roost surveys, capture surveys, and acoustic sampling of bat echolocation activity in Nyungwe National Park and surrounding areas in south-western Rwanda from 2013-2020 and their conservation relevance. New information We report the discovery of an extant population of Hill's horseshoe bat (Rhinolophushilli) in Nyungwe National Park, Rwanda, 40 years since the last reported observation of the species in 1981. We also report the first record of Lander's horseshoe bat (Rhinolophuslanderi) in Nyungwe National Park and the first record of the Damara woolly bat (Kerivoulaargentata) in Rwanda. The dataset contributed to GBIF and described in this paper includes 278 occurrence records from 10 bat species of five families detected at 71 locations in or near Nyungwe National Park, Rwanda. We include a description of the morphological descriptions of R.hilli and present the first acoustic echolocation signatures and phylogenetic information for this species.


Introduction
Nyungwe National Park in south-western Rwanda is one of the most biologically important montane rainforests in central Africa. The Park protects 1,019 km of montane rainforest, which is home to a diversity of wildlife species, including many species endemic to the Albertine Rift region of Africa (Rwanda Environment Management Authority 2015). The International Union for Conservation of Nature (IUCN) Red List of Threatened Species (Version 2021-3; www.iucnredlist.org) lists fifty-four species of bats as occurring in Rwanda, as well as additional, previously undocumented, species likely occur in Rwanda's forest habitats. Habitat loss and fragmentation from activities, such as mining, logging, hunting, agriculture, and fires caused by wild honey collection, pose a significant threat to the region's biodiversity (Crawford 2012). The amount of intact forest in Nyungwe National Park has steadily reduced due to human encroachment and the forests surrounding the Park have disappeared almost entirely (Rwanda Environment Management Authority 2015). Due to the level of habitat change, the continued existence of rare and endangered species, such as the IUCN Critically Endangered Hill's horseshoe bat (Rhinolophus hilli), depends on accurate and up-to-date information being made available to resource managers who can utilise the data to implement effective conservation actions to protect species' at greatest risk of extinction.

Hill's horseshoe bat (Rhinolophus hilli)
First described by Aellen (1973), Rhinolophus hilli was considered a synonym of Rhinolophus ruwenzorii by Smith and Hood (1980), which they regarded as a subspecies of R. maclaudi. Csorba et al. (2003) considered R. hilli a subspecies of R. ruwenzorii, but Fahr et al. (2002) showed clear differentiation between the two species, re-instating R. hilli as a distinct species. Despite the activity determining its taxonomic status, R. hilli has only been observed on two occasions, once in 1964 and again in 1981, at two locations less than 8 km apart in the Uwinka region of Nyungwe National Park (Aellen 1973, Baeten et al. 1984. Basic knowledge about the species, such as where it roosts, population status, foraging habitat and behaviour and whether it can persist in degraded forests, is lacking. Based on what is known about closely-related species, R. hilli is presumed to be a cavedwelling species that likely roosts in small colonies in forest caves (Fahr et al. 2002). Rhinolophus hilli is believed to be severely threatened with extinction due to its limited geographic distribution, small population size, disturbance at caves and the degradation and loss of forest habitat in the Albertine Rift region (Webala et al. 2021). Caves within the boundaries of Nyungwe National Park were presumed to contain the last remaining populations of R. hilli (Fahr et al. 2002). Here, we present the results collected by our team as part of a multi-phased effort to conduct targeted, cost-effective surveys to prioritise species detection.

Conservation relevance
Results of this work contribute to the overall knowledge of the biodiversity of the region and provide specific information to guide protective measures to save a critically-endangered bat from global extinction. Nyungwe National Park is one of the most biologically important montane rainforests in Central Africa. Documentation of the diversity of bat species occurring in the Park aids national Rwandan conservation efforts, which must balance competing interests. Since decades passed without any reported observation of R. hilli, uncertainty over its continued persistence impeded implementation of species-specific management actions, including safeguarding critical habitat for the species against encroachment or disturbance. Rediscovery of R. hilli and documenting new records of Rhinolophus landeri and Kerivoula argentata reinforces the universal value of Rwanda's committed stewardship of Nyungwe National Park as a global biodiversity hotspot.
By recording the first echolocation signature for R. hilli, we provide a beneficial tool for conservation managers to conduct cost-effective monitoring that provides information for conservation planning. The on-going acoustic monitoring conducted by Nyungwe National Park staff continues to identify the core range of the species within the Park, which greatly improves knowledge about critical habitat needs of the species. Furthermore, by contributing the echolocation signature to the ChiroVox global acoustic call database (Görföl et al. 2022), we aid any future acoustic monitoring projects conducted by researchers and conservation managers working in the Albertine Rift region of Central Africa.

Project description
Study area description: Bat surveys were conducted in Nyungwe National Park in southwestern Rwanda and in a few locations near the park boundary. Sampling locations within Nyungwe National Park were primarily caves and forest trails. Surveys for bats were conducted within the Uwinka region of Nyungwe and in similar habitats in the Park to determine the presence of an extant population of R. hilli and document all bat species encountered.
Design description: Surveys were conducted in four phases: Initial reconnaissance surveys (2013)(2014)(2015); reconnaissance for cave suitability by Nyungwe National Park Rangers (2018); a survey expedition with trapping efforts in forest habitats, cave surveys to assess bat use, and acoustic sampling of bat echolocation activity (2019); and on-going acoustic monitoring conducted by Nyungwe National Park Rangers (2019-present). Initial reconnaissance surveys were carried out by a small team (P. Webala and J. Nziza) with the intention of determining if the presence of R. hilli in Nyungwe National Park was readily detectable with minimal survey effort. Surveys were targeted in the Uwinka region of Nyungwe National Park and surrounding areas where bat roosts had been reported. Over the course of two years, 10 sites were surveyed with eight species detected (Fig. 1A). However, R. hilli was not detected during these rapid assessment surveys.
In the cave suitability reconnaissance phase, Nyungwe National Park Rangers identified caves within Nyungwe National Park with features suitable for bat occupancy. We provided Nyungwe National Park Rangers with a pictorial cave survey form to describe the size, type, and location of caves and abandoned mines and to report on any sign of bats using subterranean features. Rangers reported caves encountered during patrols and queried local communities to identify sites. Rangers identified and located a total of ten caves, one abandoned mine, and one building as potential bat roost habitats prior to the planned survey expedition in early 2019 (Fig. 1B).
We conducted a 10-day intensive field survey from 13-23 January 2019 that focused on: (1) surveys for bat use at caves identified by Nyungwe Rangers as suitable and likely to be occupied by bats, (2) capture surveys in forested habitats in the Uwinka region and similar surrounding habitats in Nyungwe National Park, and (3) acoustic sampling of bat echolocation activity using SongMeter 4BAT-FS recorders (Wildlife Acoustics, Inc). In total, 17 locations were surveyed, ten within the Nyungwe National Park boundary (Fig. 1C), with 55 bats from five families caught, including Hill's horseshoe bat (Rhinolophus hilli), Lander's horseshoe bat (Rhinolophus landeri) and the Damara woolly bat (Kerivoula argentata) (Fig. 2, Table 1, Suppl. material 1). This survey effort was scheduled to occur during the short dry season as the first of several planned survey trips intended to sample in the dry and wet seasons to determine the seasonality of bat occurrences. Future survey expeditions have been delayed until further notice due to safety precautions and travel restrictions during the COVID-19 pandemic. Since 2019, Nyungwe National Park Rangers have conducted acoustic sampling within the Park as part of a long-term bat acoustic monitoring project using two SongMeter 4BAT-FS recorders. With this dataset, we report the acoustic detections of R. hilli, R. landeri, and R. clivosus resulting from sampling effort at 35 locations within the Park over a total of 166 nights from July 2019 through November 2020 (Fig. 3). Capture surveys in forest habitats: Capture surveys were conducted with harp traps (a 2-bank 4.2 m harp trap by Ausbat and the 'cave-catcher' 2-bank 0.9 m harp trap by Bat Conservation and Management) and use of three to five mist-nets of 2 m, 6 m and 12 m lengths (Avinet). We placed harp traps and mist-nets parallel or perpendicular to forest trails in locations selected to maximise capture probability. Harp traps were deployed from sunset until sunrise. We opened mist-nets at sunset and monitored for approximately 4 hours and then re-opened 1-2 hours before sunrise. We monitored mist-nets continuously while open every 10-15 mins. We held bats individually in clean, cloth bags until processed and then released bats at the location of capture. See 'Step Description' for the description of data collected from captured bats. Acoustic sampling: Nyungwe Park Rangers deployed SongMeter 4BAT-FS acoustic recorders (Wildlife Acoustics, Inc) at locations along forest trails or near cave entrances during multi-day patrols and collected recorders when returning from patrol. The SM4BAT-FS recorders with SMM-U2 microphones were programmed to record in full-spectrum at 384 kilohertz sampling frequency with 12 dB gain and 16 k high filter. All other settings were set to default. The SM4BATs were set to record 30 minutes before sunset to 30 minutes after sunrise and were typically deployed for 3-5 nights at each location. We embedded geo-location coordinates on all files using the GPS attachment available from Wildlife Acoustics, Inc. Data were transferred to external hard drives and sent to Bat Conservation International in the USA for processing. See 'Step Description' for the description of the processing of acoustic data for species identification of R. hilli.
Quality control: For a subset of tissue samples, we compared species identification determined from morphological measurements with genetic data using BLASTN. As we were unable to obtain viable DNA from the holotype R. hilli specimen collected in 1981, we inspected both museum samples and compared morphological features with measurements of the two R. hilli caught during our survey. In addition, we compared the sequence data from the two suspected R. hilli samples with sequence data from closelyrelated species (Demos et al. 2019) to confirm that our classification was accurate.
Step description: Species Identification and Morphometrics: We assessed captured bats for age (juvenile/sub-adult/adult), sex, and reproductive condition (females: nonbreeding/pregnant/lactating/post-lactating; males: reproductively active/non-reproductively active as determined by enlarged testes) (Racey 2009). We measured standard morphometrics, including forearm length, tibia length, hind-foot length, tail length, ear length, tragus length, body length, and mass. We used the Mammals of Africa Volume IV (Hedgehogs, Shrews and Bats) (Kingdon 2013) as the primary key for species identification. We sampled skin tissue using a 3-mm biopsy punch from the wing membrane and stored skin tissue in desiccant until the DNA was extracted.

Echolocation Voucher Calls:
We recorded voucher echolocation calls upon release for each echolocating bat species using an M500 full-spectrum bat detector (Pettersson Electronics) at a sampling rate of 500 kHz. For constant-frequency (CF) bats (e.g. Rhinolophus spp.), we recorded resting echolocation calls while the bat was in hand. For species using frequency-modulated (FM) echolocation, we recorded echolocation activity in flight immediately upon release while visually following the bat with a light. Handrecorded bat echolocations were analysed using BatSound v.4.1 (Pettersson Electronics) to determine the following parameters for each pulse: duration (D), maximum frequency (FMAX), minimum frequency (FMIN), peak frequency (PF), and interpulse interval (IPI). We measured these parameters (D, FMAX, FMIN, and IPI) from spectrograms and the peak frequency (PF) from the power spectrum. We removed noise files and filtered the remaining files for constant frequency acoustic signatures (>15 ms call duration) using Kaleidoscope Pro (version 5.4.2, Wildlife Acoustics). Echolocation calls matching those of voucher calls collected from R. hilli (Fig. 4), R. landeri, and R. clivosus were identified during call analysis. All data are preserved to allow for future analysis once other call signatures are identified. Voucher calls will be contributed to the ChiroVox database (www. chirovox.org; Görföl et al. 2022).  Due to the age and preservation method, we were unable to obtain viable DNA from the two museum specimens of R. hilli . Instead, to verify species identification, we inspected both specimens and compared morphological features with measurements of the two R. hilli caught during our survey (Table 2, Fig. 5). In addition, we compared the sequence data from the two suspected R. hilli samples with sequence data from closely-related species to confirm that our classification was accurate (Fig. 6). For the remaining species, we compared species identification determined from morphological measurements with genetic data, using BLASTN for a subset of tissue samples.  Figure 5.

DNA Extraction for Species Confirmation
Sketches of noseleaf morphology of Rhinolophus hilli from photographs of the two individuals captured during this study. Drawings by Fiona Reid.