Evaluating expert-based habitat suitability information of terrestrial mammals with GPS-tracking data

AIM

Macroecological studies that require habitat suitability data for many species often derive this information from expert opinion. However, expert-based information is inherently subjective and thus prone to errors. The increasing availability of GPS tracking data offers opportunities to evaluate and supplement expert-based information with detailed empirical evidence. Here, we compared expert-based habitat suitability information from the International Union for Conservation of Nature (IUCN) with habitat suitability information derived from GPS-tracking data of 1,498 individuals from 49 mammal species.

LOCATION

Worldwide.

TIME PERIOD

1998-2021.

MAJOR TAXA STUDIED

Forty-nine terrestrial mammal species.

METHODS

Using GPS data, we estimated two measures of habitat suitability for each individual animal: proportional habitat use (proportion of GPS locations within a habitat type), and selection ratio (habitat use relative to its availability). For each individual we then evaluated whether the GPS-based habitat suitability measures were in agreement with the IUCN data. To that end, we calculated the probability that the ranking of empirical habitat suitability measures was in agreement with IUCN's classification into suitable, marginal and unsuitable habitat types.

RESULTS

IUCN habitat suitability data were in accordance with the GPS data (> 95% probability of agreement) for 33 out of 49 species based on proportional habitat use estimates and for 25 out of 49 species based on selection ratios. In addition, 37 and 34 species had a > 50% probability of agreement based on proportional habitat use and selection ratios, respectively.

MAIN CONCLUSIONS

We show how GPS-tracking data can be used to evaluate IUCN habitat suitability data. Our findings indicate that for the majority of species included in this study, it is appropriate to use IUCN habitat suitability data in macroecological studies. Furthermore, we show that GPS-tracking data can be used to identify and prioritize species and habitat types for re-evaluation of IUCN habitat suitability data.

Camera settings and biome influence the accuracy of citizen science approaches to camera trap image classification

Scientists are increasingly using volunteer efforts of citizen scientists to classify images captured by motion-activated trail cameras. The rising popularity of citizen science reflects its potential to engage the public in conservation science and accelerate processing of the large volume of images generated by trail cameras. While image classification accuracy by citizen scientists can vary across species, the influence of other factors on accuracy is poorly understood. Inaccuracy diminishes the value of citizen science derived data and prompts the need for specific best-practice protocols to decrease error. We compare the accuracy between three programs that use crowdsourced citizen scientists to process images online: Snapshot Serengeti, Wildwatch Kenya, and AmazonCam Tambopata. We hypothesized that habitat type and camera settings would influence accuracy. To evaluate these factors, each photograph was circulated to multiple volunteers. All volunteer classifications were aggregated to a single best answer for each photograph using a plurality algorithm. Subsequently, a subset of these images underwent expert review and were compared to the citizen scientist results. Classification errors were categorized by the nature of the error (e.g., false species or false empty), and reason for the false classification (e.g., misidentification). Our results show that Snapshot Serengeti had the highest accuracy (97.9%), followed by AmazonCam Tambopata (93.5%), then Wildwatch Kenya (83.4%). Error type was influenced by habitat, with false empty images more prevalent in open-grassy habitat (27%) compared to woodlands (10%). For medium to large animal surveys across all habitat types, our results suggest that to significantly improve accuracy in crowdsourced projects, researchers should use a trail camera set up protocol with a burst of three consecutive photographs, a short field of view, and determine camera sensitivity settings based on in situ testing. Accuracy level comparisons such as this study can improve reliability of future citizen science projects, and subsequently encourage the increased use of such data.

Effects of body size on estimation of mammalian area requirements

Accurately quantifying species' area requirements is a prerequisite for effective area-based conservation. This typically involves collecting tracking data on species of interest and then conducting home-range analyses. Problematically, autocorrelation in tracking data can result in space needs being severely underestimated. Based on the previous work, we hypothesized the magnitude of underestimation varies with body mass, a relationship that could have serious conservation implications. To evaluate this hypothesis for terrestrial mammals, we estimated home-range areas with global positioning system (GPS) locations from 757 individuals across 61 globally distributed mammalian species with body masses ranging from 0.4 to 4000 kg. We then applied block cross-validation to quantify bias in empirical home-range estimates. Area requirements of mammals <10 kg were underestimated by a mean approximately15%, and species weighing approximately100 kg were underestimated by approximately50% on average. Thus, we found area estimation was subject to autocorrelation-induced bias that was worse for large species. Combined with the fact that extinction risk increases as body mass increases, the allometric scaling of bias we observed suggests the most threatened species are also likely to be those with the least accurate home-range estimates. As a correction, we tested whether data thinning or autocorrelation-informed home-range estimation minimized the scaling effect of autocorrelation on area estimates. Data thinning required an approximately93% data loss to achieve statistical independence with 95% confidence and was, therefore, not a viable solution. In contrast, autocorrelation-informed home-range estimation resulted in consistently accurate estimates irrespective of mass. When relating body mass to home range size, we detected that correcting for autocorrelation resulted in a scaling exponent significantly >1, meaning the scaling of the relationship changed substantially at the upper end of the mass spectrum.