Habitat classification modeling with incomplete data: pushing the habitat envelope

Habitat suitability, live abundance and their link to road mortality of Tasmanian wildlife

ContextTasmania has been called the roadkill capital of Australia. However, little is known about the population-level impact of vehicle mortality on native mammals in the island state. AimsThe aims were to investigate the predictability of roadkill on a given route, based on models of species distribution and live animal abundance for three marsupial species in Tasmania – the Tasmanian pademelon (Thylogale billardierii), Bennett’s wallaby (Macropus rufogriseus) and the bare-nosed wombat (Vombatus ursinus) – and to assess the possibility of predicting the magnitude of state-wide road mortality based on live animal abundance. MethodsRoad mortality of the three species was measured on eight 15-km road segments in south-eastern Tasmania, during 16 weeks over the period 2016–17. Climate suitability was predicted using state-wide geographical location records, using species distribution models, and counts of these species from 190 spotlight survey roads. Key resultsThe Tasmanian pademelons were the most frequently killed animal encountered over the study period. Live abundance, predicted by fitting models to spotlight counts, did not correlate with this fatality rate for any species. However, the climate suitability index generated by the species distribution models was strongly predictive for wombat roadkill, and moderately so for pademelons. ConclusionsAlthough distributional and wildlife abundance records are commonly available and well described by models based on climate, vegetation and land-use predictors, this approach to climate suitability modelling has limited predictability for roadkill counts on specific routes. ImplicationsRoad-specific factors, such as characteristics of the road infrastructure, nearby habitats and behavioural traits, seem to be required to explain roadkill frequency. Determining their relative importance will require spatial analysis of roadkill locations.

Effects of Climate Change on Habitat Availability and Configuration for an Endemic Coastal Alpine Bird

North America's coastal mountains are particularly vulnerable to climate change, yet harbour a number of endemic species. With little room "at the top" to track shifting climate envelopes, alpine species may be especially negatively affected by climate-induced habitat fragmentation. We ask how climate change will affect the total amount, mean patch size, and number of patches of suitable habitat for Vancouver Island White-tailed Ptarmigan (Lagopus leucura saxatilis; VIWTP), a threatened, endemic alpine bird. Using a Random Forest model and a unique dataset consisting of citizen science observations combined with field surveys, we predict the distribution and configuration of potential suitable summer habitat for VIWTP under baseline and future (2020s, 2050s, and 2080s) climates using three general circulation models and two greenhouse gas scenarios. VIWTP summer habitat is predicted to decline by an average of 25%, 44%, and 56% by the 2020s, 2050s, and 2080s, respectively, under the low greenhouse gas scenario and 27%, 59%, and 74% under the high scenario. Habitat patches are predicted to become fragmented, with a 52-79% reduction in mean patch size. The average elevation of suitable habitat patches is expected to increase, reflecting a loss of patches at lower elevations. Thus ptarmigan are in danger of being "squeezed off the mountain", as their remaining suitable habitat will be increasingly confined to mountaintops in the center of the island. The extent to which ptarmigan will be able to persist in increasingly fragmented habitat is unclear. Much will depend on their ability to move throughout a more heterogeneous landscape, utilize smaller breeding areas, and survive increasingly variable climate extremes. Our results emphasize the importance of continued monitoring and protection for high elevation specialist species, and suggest that White-tailed Ptarmigan should be considered an indicator species for alpine ecosystems in the face of climate change.

Vulnerability of dynamic genetic conservation units of forest trees in Europe to climate change

A transnational network of genetic conservation units for forest trees was recently documented in Europe aiming at the conservation of evolutionary processes and the adaptive potential of natural or man-made tree populations. In this study, we quantified the vulnerability of individual conservation units and the whole network to climate change using climate favourability models and the estimated velocity of climate change. Compared to the overall climate niche of the analysed target species populations at the warm and dry end of the species niche are underrepresented in the network. However, by 2100, target species in 33-65 % of conservation units, mostly located in southern Europe, will be at the limit or outside the species' current climatic niche as demonstrated by favourabilities below required model sensitivities of 95%. The highest average decrease in favourabilities throughout the network can be expected for coniferous trees although they are mainly occurring within units in mountainous landscapes for which we estimated lower velocities of change. Generally, the species-specific estimates of favourabilities showed only low correlations to the velocity of climate change in individual units, indicating that both vulnerability measures should be considered for climate risk analysis. The variation in favourabilities among target species within the same conservation units is expected to increase with climate change and will likely require a prioritization among co-occurring species. The present results suggest that there is a strong need to intensify monitoring efforts and to develop additional conservation measures for populations in the most vulnerable units. Also, our results call for continued transnational actions for genetic conservation of European forest trees, including the establishment of dynamic conservation populations outside the current species distribution ranges within European assisted migration schemes.

Species Favourability Shift in Europe due to Climate Change: A Case Study for Fagus sylvatica L. and Picea abies (L.) Karst. Based on an Ensemble of Climate Models

Climate is the main environmental driver determining the spatial distribution of most tree species at the continental scale. We investigated the distribution change of European beech and Norway spruce due to climate change. We applied a species distribution model (SDM), driven by an ensemble of 21 regional climate models in order to study the shift of the favourability distribution of these species. SDMs were parameterized for 1971–2000, as well as 2021–2050 and 2071–2100 using the SRES scenario A1B and three physiological meaningful climate variables. Growing degree sum and precipitation sum were calculated for the growing season on a basis of daily data. Results show a general north-eastern and altitudinal shift in climatological favourability for both species, although the shift is more marked for spruce. The gain of new favourable sites in the north or in the Alps is stronger for beech compared to spruce. Uncertainty is expressed as the variance of the averaged maps and with a density function. Uncertainty in species distribution increases over time. This study demonstrates the importance of data ensembles and shows how to deal with different outcomes in order to improve impact studies by showing uncertainty of the resulting maps.

The use of a predictive habitat model and a fuzzy logic approach for marine management and planning

Bottom trawl survey data are commonly used as a sampling technique to assess the spatial distribution of commercial species. However, this sampling technique does not always correctly detect a species even when it is present, and this can create significant limitations when fitting species distribution models. In this study, we aim to test the relevance of a mixed methodological approach that combines presence-only and presence-absence distribution models. We illustrate this approach using bottom trawl survey data to model the spatial distributions of 27 commercially targeted marine species. We use an environmentally- and geographically-weighted method to simulate pseudo-absence data. The species distributions are modelled using regression kriging, a technique that explicitly incorporates spatial dependence into predictions. Model outputs are then used to identify areas that met the conservation targets for the deployment of artificial anti-trawling reefs. To achieve this, we propose the use of a fuzzy logic framework that accounts for the uncertainty associated with different model predictions. For each species, the predictive accuracy of the model is classified as 'high'. A better result is observed when a large number of occurrences are used to develop the model. The map resulting from the fuzzy overlay shows that three main areas have a high level of agreement with the conservation criteria. These results align with expert opinion, confirming the relevance of the proposed methodology in this study.

The importance of distance to resources in the spatial modelling of bat foraging habitat

Many bats are threatened by habitat loss, but opportunities to manage their habitats are now increasing. Success of management depends greatly on the capacity to determine where and how interventions should take place, so models predicting how animals use landscapes are important to plan them. Bats are quite distinctive in the way they use space for foraging because (i) most are colonial central-place foragers and (ii) exploit scattered and distant resources, although this increases flying costs. To evaluate how important distances to resources are in modelling foraging bat habitat suitability, we radio-tracked two cave-dwelling species of conservation concern (Rhinolophus mehelyi and Miniopterus schreibersii) in a Mediterranean landscape. Habitat and distance variables were evaluated using logistic regression modelling. Distance variables greatly increased the performance of models, and distance to roost and to drinking water could alone explain 86 and 73% of the use of space by M. schreibersii and R. mehelyi, respectively. Land-cover and soil productivity also provided a significant contribution to the final models. Habitat suitability maps generated by models with and without distance variables differed substantially, confirming the shortcomings of maps generated without distance variables. Indeed, areas shown as highly suitable in maps generated without distance variables proved poorly suitable when distance variables were also considered. We concluded that distances to resources are determinant in the way bats forage across the landscape, and that using distance variables substantially improves the accuracy of suitability maps generated with spatially explicit models. Consequently, modelling with these variables is important to guide habitat management in bats and similarly mobile animals, particularly if they are central-place foragers or depend on spatially scarce resources.