On the Edge: Identifying priority areas for conservation of Fishing Cat, a threatened wetland felid, amidst rapidly altering freshwater landscapes

Freshwater ecosystems have been most severely impacted in the Anthropocene with 27% of its species threatened with extinction. Fishing Cat is a globally threatened South and South-east Asian wetland felid that is also a highly rated Evolutionarily Distinct and Globally Endangered (EDGE) species, i.e., it is a global priority for conservation and research. Being an understudied species, knowledge gaps exist on its basic ecology, such as distribution and niche. To address this, ensemble species distribution modeling (ESDM) was used to clarify doubts on its potential distribution and niche. To provide a relatable current context, loss of suitable habitat to urbanization (2010-2020) was estimated by analyzing range-wide survey data with environmental and anthropogenic variables (night-time lights and land surface temperature as proxies for urbanization). Wetlands (18.36%) and elevation (17.15%) are the most important variables determining the ecological niche of Fishing Cat. It was predicted to be mainly restricted to low-elevation (<111 m) wetlands in river basins of South and South-east Asia. An estimated 23.74% suitable habitat was lost to urbanization. Incrementally building on the ESDM outputs, high priority movement corridors and landscape conservation units were identified. South Asia holds the core of the global Fishing Cat population with two very important regions - Ganges Brahmaputra Basin and Indus Basin - sharing transboundary areas with highly suitable habitat and many priority conservation units. The former is strategic to maintaining connectivity between South and South-east Asian Fishing Cat populations while isolation effects in the latter need investigation. Coastal wetlands of South-east Asia, though severely impacted, are crucial for the felid’s persistence. More than 90% of Fishing Cat’s potential range lies outside the protected area network. Here, the felid can be adopted as a flagship species to conserve rapidly degrading low- elevation wetlands within a socio-ecological framework by involving multiple stakeholders.

Thailand and Cambodia (Cutter, 2015;Thaung et al., 2018;Chutipong et al., 2019). Moreover, 111 recent records of its occurrence from relatively non-moist drier areas (Sadhu and Reddy, 2013; In addition, a circuit theory-based connectivity analysis was also performed using the tool 220 Circuitscape v4.0 (McRae et al., 2013), which implements the circuit and random walk theory 221 (McRae et al., 2009). This tool computes a resistance-based connectivity metric based on a series 222 of combinatorial and numerical analyses (Shah and McRae, 2008). The prime difference between 223 a circuit and an LCP is that the circuit corridor is identified after computation of multiple paths 224 between the population cores and the conductive probability of every raster cell is measured. In 225 contrast, the LCP is based on a single least-cost route (Bowman et al., 2020). The circuit theory 226 accommodates the assumption that animals may not have the knowledge of the perfect least-cost  Being narrow sections in the corridor, removal of these regions might hamper the connectivity of 241 populations in a landscape (Liu et al., 2020). Linkage paths were constructed using the 'Linkage 242 Pathways' tool, which were further used as inputs to the Pinch Point Mapper tool. Pairwise 243 calculation was set for the Circuitscape mode and 4.6 km was set as the minimum cost-weighted 244 corridor width. Since, all raster data in this study were processed at a spatial resolution of 2.5 245 minutes (~ 4.6 km), this was the lowest possible corridor width that could be parameterised. costs that would be incurred if a given 100 km 2 grid was to be selected for conservation. The raster 257 output from the ESDM for UrbT2 and the CIFOR wetlands were used as feature layers.
Optimization targets ranging from 10%-50% of the landscape, with increments of 5% were 259 implemented for each scenario resulting in nine solutions for each respective scenario.

263
Hereafter, the solutions from all the nine prioritized targets were added into a single solution and 264 units encompassing protected areas were selected.  Asia.

Discussion
To the best of our knowledge, this study was the first attempt to model the global distribution of 331 Fishing Cat, an under-studied, globally endangered freshwater mammalian carnivore of high 332 conservation and research value (Veron et al., 2008;Tensen, 2018;Zanin and Neves, 2019). RF 333 and SVM, selected as the best performing algorithms for the SDM of Fishing Cat, have been found 334 to be powerful classifiers in other studies too (Fukuda et al. 2013;Thahn Noi and Kappas, 2018).

335
The felid was predicted to be mainly restricted to low-elevation (<111 m above sea level) wetlands 336 of South and South East Asian river basins (Fig 3). Low elevation was also estimated to be an  Table 3). The situation in South-east Asia seems precarious and therefore worthy of  Cat was recorded from drier regions in Sri Lanka, also predicted to be unsuitable/sub-optimal by our model.  suggesting that the species is critically endangered in the country (Willianto pers. comm., 2021).

461
No suitable habitat for Fishing Cat was predicted to be present in Sumatra. This is backed by 462 absence of either historical or recent evidence of its presence in the country. How the Fishing Cat 463 came to occur in Java therefore is a mystery.

512
While spatial prioritization for conservation research and action has foregrounded biodiversity in 513 the global context and eased the process of systematic planning and fund acquisition (Myers,514 2003), it has also been substantially criticized for being bereft of 'situated knowledge' and its 515 failure to engage with locally obtained experience (Wyborn and Evans, 2021). In our study, we 516 attempted to interpret and discuss the findings of our model based on our collated local and suggesting that this region provides high permeability conducive for species dispersal and requires 533 lesser investments for conservation and restoration of the species' habitat (Table 3). The Lower

534
Indus Basin had the second highest number of LCUs (Table 3)    Red River Delta South-east Asia 1 -1 2