skip to main content


Title: Matching expert range maps with species distribution model predictions
Abstract

Species’ range maps based on expert opinion are a critical resource for conservation planning. Expert maps are usually accompanied by species descriptions that specify sources of internal range heterogeneity, such as habitat associations, but these are rarely considered when using expert maps for analyses. We developed a quantitative metric (expert score) to evaluate the agreement between an expert map and a habitat probability surface obtained from a species distribution model. This method rewards both the avoidance of unsuitable sites and the inclusion of suitable sites in the expert map. We obtained expert maps of 330 butterfly species from each of 2 widely used North American sources (Glassberg [1999, 2001] and Scott [1986]) and computed species‐wise expert scores for each. Overall, the Glassberg maps secured higher expert scores than Scott (0.61 and 0.41, respectively) due to the specific rules (e.g., Glassberg only included regions where the species was known to reproduce whereas Scott included all areas a species expanded to each year) they used to include or exclude areas from ranges. The predictive performance of expert maps was almost always hampered by the inclusion of unsuitable sites, rather than by exclusion of suitable sites (deviance outside of expert maps was extremely low). Map topology was the primary predictor of expert performance rather than any factor related to species characteristics such as mobility. Given the heterogeneity and discontinuity of suitable landscapes, expert maps drawn with more detail are more likely to agree with species distribution models and thus minimize both commission and omission errors.

 
more » « less
NSF-PAR ID:
10375931
Author(s) / Creator(s):
 ;  ;  ;  
Publisher / Repository:
Wiley-Blackwell
Date Published:
Journal Name:
Conservation Biology
Volume:
34
Issue:
5
ISSN:
0888-8892
Page Range / eLocation ID:
p. 1292-1304
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Hamer, Gabriel (Ed.)
    Abstract Many species distribution maps indicate the ranges of Aedes aegypti (Linnaeus) and Aedes albopictus (Skuse) overlap in Florida despite the well-documented range reduction of Ae. aegypti. Within the last 30 yr, competitive displacement of Ae. aegypti by Ae. albopictus has resulted in partial spatial segregation of the two species, with Ae. aegypti persisting primarily in urban refugia. We modeled fine-scale distributions of both species, with the goal of capturing the outcome of interspecific competition across space by building habitat suitability maps. We empirically parameterized models by sampling 59 sites in south and central Florida over time and incorporated climatic, landscape, and human population data to identify predictors of habitat suitability for both species. Our results show human density, precipitation, and urban land cover drive Ae. aegypti habitat suitability, compared with exclusively climatic variables driving Ae. albopictus habitat suitability. Remotely sensed variables (macrohabitat) were more predictive than locally collected metrics (microhabitat), although recorded minimum daily temperature showed significant, inverse relationships with both species. We detected minor Aedes habitat segregation; some periurban areas that were highly suitable for Ae. albopictus were unsuitable for Ae. aegypti. Fine-scale empirical models like those presented here have the potential for precise risk assessment and the improvement of operational applications to control container-breeding Aedes mosquitoes. 
    more » « less
  2. Machine learning (ML) methods, such as artificial neural networks (ANN), k-nearest neighbors (kNN), random forests (RF), support vector machines (SVM), and boosted decision trees (DTs), may offer stronger predictive performance than more traditional, parametric methods, such as linear regression, multiple linear regression, and logistic regression (LR), for specific mapping and modeling tasks. However, this increased performance is often accompanied by increased model complexity and decreased interpretability, resulting in critiques of their “black box” nature, which highlights the need for algorithms that can offer both strong predictive performance and interpretability. This is especially true when the global model and predictions for specific data points need to be explainable in order for the model to be of use. Explainable boosting machines (EBM), an augmentation and refinement of generalize additive models (GAMs), has been proposed as an empirical modeling method that offers both interpretable results and strong predictive performance. The trained model can be graphically summarized as a set of functions relating each predictor variable to the dependent variable along with heat maps representing interactions between selected pairs of predictor variables. In this study, we assess EBMs for predicting the likelihood or probability of slope failure occurrence based on digital terrain characteristics in four separate Major Land Resource Areas (MLRAs) in the state of West Virginia, USA and compare the results to those obtained with LR, kNN, RF, and SVM. EBM provided predictive accuracies comparable to RF and SVM and better than LR and kNN. The generated functions and visualizations for each predictor variable and included interactions between pairs of predictor variables, estimation of variable importance based on average mean absolute scores, and provided scores for each predictor variable for new predictions add interpretability, but additional work is needed to quantify how these outputs may be impacted by variable correlation, inclusion of interaction terms, and large feature spaces. Further exploration of EBM is merited for geohazard mapping and modeling in particular and spatial predictive mapping and modeling in general, especially when the value or use of the resulting predictions would be greatly enhanced by improved interpretability globally and availability of prediction explanations at each cell or aggregating unit within the mapped or modeled extent. 
    more » « less
  3. Abstract

    Recent studies have used occupancy models (OM) and ecological niche models (ENM) to provide a better understanding of species’ distributions at different scales. One of the main ideas underlying the theoretical foundations of both OM and ENM is that they are positively related to abundance: higher occupancy implies higher density and more suitable areas are likely to have more abundant populations. Here, we analyze the relationship between habitat use measured in terms of occupancy probabilities from OM and environmental suitability derived from ENM in three different Neotropical mammal species: Leopardus wiedii, Cuniculus paca, and Dasypus novemcinctus. For ENM, we used climatic and vegetation cover variables and implemented a model calibration and selection protocol to select the most competitive models. For OM, we used a single-species, single-season model with site covariates for camera-trap data from six different sites throughout the Neotropical realm. Covariates included vegetation percentage, normalized difference vegetation index, normalized difference water index, and elevation. For each site, we fit OM using all possible combinations of variables and selected the most competitive (ΔAICc < 2) to build an average OM. We explored relationships between estimated suitability and occupancy values using Spearman correlation analysis. Relationships between ENM and OM tended to be positive for the three Neotropical mammals, but the strength varied among sites, which could be explained by local factors such as site characteristics and conservation status of areas. We conjecture that ENM are suitable to understand spatial patterns at coarser geographic scales because the concept of the niche is about the species as a whole, whereas OM are more relevant to explain the distribution locally, likely reflecting transient dynamics of populations resulting from many local factors such as community composition and biotic processes.

     
    more » « less
  4. Abstract Aim

    We sought to illuminate the history of the arachnid orders Schizomida and Uropygi, neither of which have previously been subjected to global molecular phylogenetic and biogeographical analyses.

    Location

    Specimens used in this study were collected in all major tropical and subtropical areas where they are presently found, including the Americas, Africa, Australia and the Indo‐Pacific region.

    Methods

    From field‐collected specimens, we sequenced two nuclear and two mitochondrial markers, combined these with publicly available data, and conducted multi‐gene phylogenetic analyses on 240 Schizomida, 24 Uropygi and 12 other arachnid outgroups. Schizomid specimens included one specimen from the small family Protoschizomidae; other schizomid specimens were in Hubbardiidae, subfamily Hubbardiinae, which holds 289 of the order's 305 named species. We inferred ancestral areas using the Dispersal‐Extinction‐Cladogenesis model of range evolution, and we used fossil calibrations to estimate divergence times.

    Results

    We recovered monophyletic Schizomida and Uropygi as each other's sister group, forming the clade Thelyphonida, and terminals from the New World were usually positioned as the earliest diverging lineages. The ancestral area for schizomids reconstructed unambiguously to the region comprised of Mexico, Southern California and Florida (the xeric New World subtropics). Optimal trees suggested a single colonization of the Indo‐Pacific in both orders, although this did not receive bootstrap support. Molecular dating gave an Upper Carboniferous origin for each order, and a mid‐Cretaceous expansion of Schizomida, including the origin and initial diversification of those in the Indo‐Pacific.

    Main conclusions

    Ancestral area reconstructions, molecular dating and fossil evidence all support an Upper Carboniferous, tropical Pangean origin for Thelyphonida, Schizomida and perhaps Uropygi. Much of this region became unsuitable habitat for these arachnids during the breakup of Pangea, but they persisted in the area that is now Meso‐ and South America. From there they then expanded to the Indo‐Pacific, where schizomids today display an idiosyncratic combination of microendemism and long‐range dispersal.

     
    more » « less
  5. Ice-rich permafrost in the circum-Arctic and sub-Arctic (hereafter pan-Arctic), such as late Pleistocene Yedoma, are especially prone to degradation due to climate change or human activity. When Yedoma deposits thaw, large amounts of frozen organic matter and biogeochemically relevant elements return into current biogeochemical cycles. This mobilization of elements has local and global implications: increased thaw in thermokarst or thermal erosion settings enhances greenhouse gas fluxes from permafrost regions. In addition, this ice-rich ground is of special concern for infrastructure stability as the terrain surface settles along with thawing. Finally, understanding the distribution of the Yedoma domain area provides a window into the Pleistocene past and allows reconstruction of Ice Age environmental conditions and past mammoth-steppe landscapes. Therefore, a detailed assessment of the current pan-Arctic Yedoma coverage is of importance to estimate its potential contribution to permafrost-climate feedbacks, assess infrastructure vulnerabilities, and understand past environmental and permafrost dynamics. Building on previous mapping efforts, the objective of this paper is to compile the first digital pan-Arctic Yedoma map and spatial database of Yedoma coverage. Therefore, we 1) synthesized, analyzed, and digitized geological and stratigraphical maps allowing identification of Yedoma occurrence at all available scales, and 2) compiled field data and expert knowledge for creating Yedoma map confidence classes. We used GIS-techniques to vectorize maps and harmonize site information based on expert knowledge. We included a range of attributes for Yedoma areas based on lithological and stratigraphic information from the source maps and assigned three different confidence levels of the presence of Yedoma (confirmed, likely, or uncertain). Using a spatial buffer of 20 km around mapped Yedoma occurrences, we derived an extent of the Yedoma domain. Our result is a vector-based map of the current pan-Arctic Yedoma domain that covers approximately 2,587,000 km 2 , whereas Yedoma deposits are found within 480,000 km 2 of this region. We estimate that 35% of the total Yedoma area today is located in the tundra zone, and 65% in the taiga zone. With this Yedoma mapping, we outlined the substantial spatial extent of late Pleistocene Yedoma deposits and created a unique pan-Arctic dataset including confidence estimates. 
    more » « less