https://www.cs.princeton.edu/~schapire/papers/ecolmod.pdf

Maximum entropy modeling of species geographic distributions

Summary
1In recent years the use of species distribution models by ecologists and conservation managers has increased considerably, along with an awareness of the need to provide accuracy assessment for predictions of such models. The kappa statistic is the most widely used measure for the performance of models generating presence–absence predictions, but several studies have criticized it for being inherently dependent on prevalence, and argued that this dependency introduces statistical artefacts to estimates of predictive accuracy. This criticism has been supported recently by computer simulations showing that kappa responds to the prevalence of the modelled species in a unimodal fashion.
2In this paper we provide a theoretical explanation for the observed dependence of kappa on prevalence, and introduce into ecology an alternative measure of accuracy, the true skill statistic (TSS), which corrects for this dependence while still keeping all the advantages of kappa. We also compare the responses of kappa and TSS to prevalence using empirical data, by modelling distribution patterns of 128 species of woody plant in Israel.
3The theoretical analysis shows that kappa responds in a unimodal fashion to variation in prevalence and that the level of prevalence that maximizes kappa depends on the ratio between sensitivity (the proportion of correctly predicted presences) and specificity (the proportion of correctly predicted absences). In contrast, TSS is independent of prevalence.
4When the two measures of accuracy were compared using empirical data, kappa showed a unimodal response to prevalence, in agreement with the theoretical analysis. TSS showed a decreasing linear response to prevalence, a result we interpret as reflecting true ecological phenomena rather than a statistical artefact. This interpretation is supported by the fact that a similar pattern was found for the area under the ROC curve, a measure known to be independent of prevalence.
5Synthesis and applications. Our results provide theoretical and empirical evidence that kappa, one of the most widely used measures of model performance in ecology, has serious limitations that make it unsuitable for such applications. The alternative we suggest, TSS, compensates for the shortcomings of kappa while keeping all of its advantages. We therefore recommend the TSS as a simple and intuitive measure for the performance of species distribution models when predictions are expressed as presence–absence maps.

https://besjournals.onlinelibrary.wiley.com/doi/pdf/10.1111/j.1365-2664.2006.01214.x

Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS)

The area under the receiver operating characteristic (ROC) curve, known as the AUC, is currently considered to be the standard method to assess the accuracy of predictive distribution models. It avoids the supposed subjectivity in the threshold selection process, when continuous probability derived scores are converted to a binary presence‐absence variable, by summarizing overall model performance over all possible thresholds. In this manuscript we review some of the features of this measure and bring into question its reliability as a comparative measure of accuracy between model results. We do not recommend using AUC for five reasons: (1) it ignores the predicted probability values and the goodness-of-fit of the model; (2) it summarises the test performance over regions of the ROC space in which one would rarely operate; (3) it weights omission and commission errors equally; (4) it does not give information about the spatial distribution of model errors; and, most importantly, (5) the total extent to which models are carried out highly influences the rate of well-predicted absences and the AUC scores.

/pdf/auc-a-misleading-measure-of-the-performance-of-predictive-qqj6tbb550.pdf

AUC: a misleading measure of the performance of predictive distribution models

Most methods for modeling species distributions from occurrence records require additional data representing the range of environmental conditions in the modeled region. These data, called background or pseudo-absence data, are usually drawn at random from the entire region, whereas occurrence collection is often spatially biased toward easily accessed areas. Since the spatial bias generally results in environmental bias, the difference between occurrence collection and background sampling may lead to inaccurate models. To correct the estimation, we propose choosing background data with the same bias as occurrence data. We investigate theoretical and practical implications of this approach. Accurate information about spatial bias is usually lacking, so explicit biased sampling of background sites may not be possible. However, it is likely that an entire target group of species observed by similar methods will share similar bias. We therefore explore the use of all occurrences within a target group as biased background data. We compare model performance using target-group background and randomly sampled background on a comprehensive collection of data for 226 species from diverse regions of the world. We find that target-group background improves average performance for all the modeling methods we consider, with the choice of background data having as large an effect on predictive performance as the choice of modeling method. The performance improvement due to target-group background is greatest when there is strong bias in the target-group presence records. Our approach applies to regression-based modeling methods that have been adapted for use with occurrence data, such as generalized linear or additive models and boosted regression trees, and to Maxent, a probability density estimation method. We argue that increased awareness of the implications of spatial bias in surveys, and possible modeling remedies, will substantially improve predictions of species distributions.

/pdf/sample-selection-bias-and-presence-only-distribution-models-5al33pm4t5.pdf

Sample selection bias and presence-only distribution models: implications for background and pseudo-absence data

We study the problem of modeling species geographic distributions, a critical problem in conservation biology. We propose the use of maximum-entropy techniques for this problem, specifically, sequential-update algorithms that can handle a very large number of features. We describe experiments comparing maxent with a standard distribution-modeling tool, called GARP, on a dataset containing observation data for North American breeding birds. We also study how well maxent performs as a function of the number of training examples and training time, analyze the use of regularization to avoid overfitting when the number of examples is small, and explore the interpretability of models constructed using maxent.

/pdf/a-maximum-entropy-approach-to-species-distribution-modeling-2av58hym9x.pdf

A maximum entropy approach to species distribution modeling

(Vieglais et al., 2000). Museum data are high quality because voucher specimens can be examined if identification is questionable. However, like all point data, museum specimens provide only a limited view of the actual species' range (Krohn, 1996), hence the need for predictive approaches. A limited number of ecological data sets are also available, worldwide, including physio-chemical parameters (NOAA, 1999) and bathymetry (Smith and Sandwell, 1997). Ecological niche modeling uses the primary point data and the ecological data to build a partial niche The niche model is defined in ecological space by ecological parameters. It can be projected into geographic space by dividing the area of interest into rows and columns to create a grid, and then identifying the grid cells where the ecological parameters match those of the niche model. The landscape for this study is the Central Western Atlantic, roughly bounded by the definitions of the Food and Agricultural Organizations definition of Fishing Area 31, approximately the area of the Atlantic Ocean, Caribbean Sea, and the Gulf of Mexico between 35°N and 5°N Latitude and west of 40°W Longitude. Many tools have been used to develop models of ecological niches. Among the simplest is BIOCLIM (Nix, 1986), which involves intersecting the ranges (slightly trimmed) inhabited by the species along each environmental axis (e.g. 0–50 m depth x annual surface temperature average of 20–22°C, etc.). Other approaches include general linear models, distance-based algorithms , and regression-tree analyses (Austin et al. These relatively straightforward algorithms, however, suffer from their focus on a search for a single decision rule, or a small set of decision rules. The reality of species' ranges is that many factors affect them, and indeed different decision rules may govern distributional limits in different sectors of a species' distribution Biological communities are changing drastically in response to global climate change (Walther et al., 2002), changes in use by human populations (Krishtalka et al., 2002), and introduction of exotic species (Carlton, 1996; Enserink, 1999). To study the impact of such changes in the marine environment, biologists require a detailed understanding of the diversity and distributions of marine organisms on macroscopic scales, such as across entire ocean basins, in order to improve understanding of the actual distributions of species, and gain an overall impression of the potential community structures that exist in particular habitats. A major obstacle to such an improved understanding is the fact that existing biodiversity records …

Niche Modeling and Geographic Range Predictions in the Marine Environment Using a Machine-learning Algorithm

/pdf/niche-modeling-perspective-on-geographic-range-predictions-1tnaqev7ut.pdf

Niche Modeling Perspective on Geographic Range Predictions in the Marine Environment Using a Machine-learning Algorithm

The Spotted Owl (  Strix occidentalis ) is the focus of intense concern as a species threatened by the destruction and fragmentation of primary forest in the Pacific Northwest ( U.S.A. ). Aside from habitat concerns, an additional peril exists for the species: the larger and more aggressive Barred Owl (  S. varia ) is invading the Pacific Northwest and has the potential to overrun much of the range of the endangered species. We evaluated the dimensions of this potential invasion by using ecological niche models based on point-occurrence data. With these tools, we developed models with significant predictions for the native distributions of both Spotted and Barred owls ( tested via independent occurrence data ) and for the invasive range of Barred Owls. Overlap between the models for the two species suggests that most of the northern portion of the Spotted Owl's distribution ( south to about lat. 38°N ) is vulnerable to Barred Owl invasion. We present an example of the potential effects of species invasions on endangered species conservation.

Using ecological-niche modeling to predict Barred Owl invasions with implications for Spotted Owl conservation

ISHNET 1 is a distributed information system that seeks to provide data towards answering two fundamental questions regarding marine biodiversity: how many vouchered records exist for each species of fishes, and what are their distributions in time and space? Recent advances in information technology make possible efficient access to one of the major resources for such knowledge, and the only resource in which data are backed up by actual specimens of the organisms: natural history museums. Scientific collections housed at world natural history museums constitute the most complete and authoritative record of global biodiversity (PCAST, 1998; Krishtalka and Humphrey, 1998, in press). The value of scientific collections lies in the fact that records are directly tied to actual voucher specimens. Changes in taxonomic names and discovery of new species can always be related to actual specimens which are available to all qualified investigators. Efficient access to such data can furnish investigators with baseline data regarding where and when collections have been made. Clearly, this information provides an important tool in planning future collecting efforts. Combining these data with environmental data and sophisticated artificial-intelligence algorithms (see Stockwell and Noble, 1991; Stockwell and Peters, 1993) makes it possible to predict species' distributions (Peterson et al., 1999), study the possible impact of the introduction of exotic species, and to model effects of environmental changes such as global climate change. The purpose of FISHNET is seamless integration of diverse data sets available to all scientists. FISHNET is a consortium of natural history museum fish collections that have agreed to share specimen data openly among all users. As a distributed information system, the data reside on computer servers at each participating institution who thus maintain ownership and stewardship of the data. Because data are served on FISHNET is a consortium of natural history museum fish collections that have agreed to share specimen data openly among all users. request to remote users, FISHNET is not an entity but a cooperative community. The concept of sharing fish specimen data is an old idea that has grown out of the earlier initiatives such as FISHGOPHER 2, MUSE 3 and NEODAT 4 projects. The fish community is almost unique among communities in the extent to which museum holdings have been captured in electronic form and a willingness to share data over the Internet. This particular initiative had its origins when Vieglais began exploring the application of the ANSI/NISO Z39.50 …

https://kuscholarworks.ku.edu/bitstream/handle/1808/16626/PetersonT_Oceanography_13(3)10.pdf;sequence=1

C. Richard Robins

Papers

Niche Modeling and Geographic Range Predictions in the Marine Environment Using a Machine-learning Algorithm

Niche Modeling Perspective on Geographic Range Predictions in the Marine Environment Using a Machine-learning Algorithm

Using ecological-niche modeling to predict Barred Owl invasions with implications for Spotted Owl conservation

Harnessing Museum Resources for the Census of Marine Life: The FISHNET Project