Species distribution modeling in regions of high need and limited data: waterfowl of China

Abstract

Background: A number of conservation and societal issues require understanding how species are distributed on the landscape, yet ecologists are often faced with a lack of data to develop models at the resolution and extent desired, resulting in inefficient use of conservation resources. Such a situation presented itself in our attempt to develop waterfowl distribution models as part of a multidisciplinary team targeting the control of the highly pathogenic H5N1 avian influenza virus in China. Methods: Faced with limited data, we built species distribution models using a habitat suitability approach for China's breeding and non-breeding (hereafter, wintering) waterfowl. An extensive review of the literature was used to determine model parameters for habitat modeling. Habitat relationships were implemented in GIS using land cover covariates. Wintering models were validated using waterfowl census data, while breeding models, though developed for many species, were only validated for the one species with sufficient telemetry data available. Results: We developed suitability models for 42 waterfowl species (30 breeding and 39 wintering) at 1 km resolution for the extent of China, along with cumulative and genus level species richness maps. Breeding season models showed highest waterfowl suitability in wetlands of the high-elevation west-central plateau and northeastern China. Wintering waterfowl suitability was highest in the lowland regions of southeastern China. Validation measures indicated strong performance in predicting species presence. Comparing our model outputs to China's protected areas indicated that breeding habitat was generally better covered than wintering habitat, and identified locations for which additional research and protection should be prioritized. Conclusions: These suitability models are the first available for many of China's waterfowl species, and have direct utility to conservation and habitat planning and prioritizing management of critically important areas, providing an example of how this approach may aid others faced with the challenge of addressing conservation issues with little data to inform decision making.

Waterfowl Spring Migratory Behavior and Avian Influenza Transmission Risk in the Changing Landscape of the East Asian-Australasian Flyway

“…Conversely, since many species such as northern pintail commonly fail to breed every season (Miller et al, 2005;Clark et al, 2014), intermediate-distance migrants may be exhibiting a short-stopping strategy that individuals use in seasons where they will not breed. Range maps of reported breeding ranges for waterfowl species in Asia are generally inferred based on expert knowledge and anecdotal evidence rather than documented with tools such as satellite transmitters, and it has been noted that such breeding ranges and timelines are not always indicative of species behavior (Prosser et al, 2018). Thus, it also is feasible that the intermediate-distance migrants are following a longstanding migration that has not been incorporated into species' range maps.…”

Seasonal occurrence and abundance of dabbling ducks across the continental United States: Joint spatio‐temporal modelling for the GenusAnas

“…Species distribution models are valuable tools in waterfowl ecology and have been broadly applied in assessing assessing waterfowl vulnerability to climate change (Reese & Skagen, ; Sofaer et al, ; Steen, Skagen, & Noon, ), delimiting endangered waterfowl ranges (Pickett et al, ; Si et al, ; Zeng et al, ) and in waterfowl habitat conservation prioritization (Barker, Cumming, & Darveau, ; Johnston et al, ; Runge, Martin, Possingham, Willis, & Fuller, ). Species distribution models have also been incorporated into a host of epidemiological studies, including those aimed at appraising the role of migratory waterfowl in global transmission of highly pathogenic avian influenza among wild bird populations and to domestic poultry (Belkhiria, Hijmans, Boyce, Crossley, & Martínez‐López, ; Martin et al, ; Prosser et al, , , ; Russell, ; Takekawa et al, ). Despite the demonstrated ability of species distribution models to correlate species presence or wildlife disease to underlying environmental factors (Elith & Leathwick, ; Guisan & Theurillat, ), several challenges remain in accurately forecasting species occurrence and abundance within a spatio‐temporal context (Franklin, ; Martnez‐Minaya, Cameletti, Conesa, & Pennino, ; Paradinas, Conesa, López‐Quílez, & Bellido, ).…”

Confronting models with data: the challenges of estimating disease spillover

“…Input data (c, m) for the donor hosts (wild waterfowl) were less available than for poultry, forcing a different iterative modelling approach. First steps defined suitable habitats for each species and across subannual seasons, as migratory behaviour results in very different seasonal distributions [27]. To create geospatial layers of d 1 , abundance estimates [35,36] were distributed across the predicted habitat ranges [28].…”