Delineating environmental envelopes to improve mapping of species distributions, via a hurdle model with CART &/or MaxEnT

Species distribution modelling (SDM) typically analyses species’ presence together with some form of absence information. Ideally absences comprise observations or are inferred from comprehensive sampling. When such information is not available, then pseudo-absences are often generated from the background locations within the study region of interest containing the presences, or else absence is implied through the comparison of presences to the whole study region, e.g. as is the case in Maximum Entropy (MaxEnt) or Poisson point process modelling. However, the choice of which absence information to include can be both challenging and highly influential on SDM predictions (e.g. Oksanen and Minchin, 2002). In practice, the use of pseudo- or implied absences often leads to an imbalance where absences far outnumber presences. This leaves analysis highly susceptible to ‘naughty-noughts’: absences that occur beyond the envelope of the species, which can exert strong influence on the model and its predictions (Austin and Meyers, 1996). Also known as ‘excess zeros’, naughty noughts can be estimated via an overall proportion in simple hurdle or mixture models (Martin et al., 2005). However, absences, especially those that occur beyond the species envelope, can often be more diverse than presences. Here we consider an extension to excess zero models. The two-staged approach first exploits the compartmentalisation provided by classification trees (CTs) (as in O’Leary, 2008) to identify multiple sources of naughty noughts and simultaneously delineate several species envelopes. Then SDMs can be fit separately within each envelope, and for this stage, we examine both CTs (as in Falk et al., 2014) and the popular MaxEnt (Elith et al., 2006). We introduce a wider range of model performance measures to improve treatment of naughty noughts in SDM. We retain an overall measure of model performance, the area under the curve (AUC) of the Receiver-Operating Curve (ROC), but focus on its constituent measures of false negative rate (FNR) and false positive rate (FPR), and how these relate to the threshold in the predicted probability of presence that delimits predicted presence from absence. We also propose error rates more relevant to users of predictions: false omission rate (FOR), the chance that a predicted absence corresponds to (and hence wastes) an observed presence, and the false discovery rate (FDR), reflecting those predicted (or potential) presences that correspond to absence. A high FDR may be desirable since it could help target future search efforts, whereas zero or low FOR is desirable since it indicates none of the (often valuable) presences have been ignored in the SDM. For illustration, we chose Bradypus variegatus, a species that has previously been published as an exemplar species for MaxEnt, proposed by Phillips et al. (2006). We used CTs to increasingly refine the species envelope, starting with the whole study region (E0), eliminating more and more potential naughty noughts (E1–E3). When combined with an SDM fit within the species envelope, the best CT SDM had similar AUC and FPR to the best MaxEnt SDM, but otherwise performed better. The FNR and FOR were greatly reduced, suggesting that CTs handle absences better. Interestingly, MaxEnt predictions showed low discriminatory performance, with the most common predicted probability of presence being in the same range (0.00-0.20) for both true absences and presences. In summary, this example shows that SDMs can be improved by introducing an initial hurdle to identify naughty noughts and partition the envelope before applying SDMs. This improvement was barely detectable via AUC and FPR yet visible in FOR, FNR, and the comparison of predicted probability of presence distribution for pres/absence.

Delineating Ecological Boundaries of Hanuman Langur Species Complex in Peninsular India Using MaxEnt Modeling Approach

Hanuman langur is one of the widely distributed and extensively studied non-human diurnal primates in India. Until recently it was believed to be a single species - Semnopithecus entellus. Recent molecular and morphological studies suggest that the Hanuman langurs consists of at least three species S. entellus, S. hypoleucos and S. priam. Furthermore, morphological studies suggested that both S. hypoleucos and S. priam have at least three subspecies in each. We explored the use of ecological niche modeling (ENM) to confirm the validity of these seven taxa and an additional taxon S. johnii belonging to the same genus. MaxEnt modeling tool was used with 19 bioclimatic, 12 vegetation and 6 hydrological environmental layers. We reduced total environmental variables to 14 layers after testing for collinearity and an independent test for model prediction was done using ENMTools. A total of 196 non-overlapping data points from primary and secondary sources were used as inputs for ENM. Results showed eight distinct ecological boundaries, corroborating the eight taxa mentioned above thereby confirming validity of these eight taxa. The study, for the first time provided ecological variables that determined the ecological requirements and distribution of members of the Hanuman langur species complex in the Indian peninsula.

Delineating Ecological Boundaries of Hanuman Langur Species Complex in Peninsular India Using MaxEnt Modeling Approach

Hanuman langur is one of the widely distributed and extensively studied non-human diurnal primates in India. Until recently it was believed to be a single species - Semnopithecus entellus. Recent molecular and morphological studies suggest that the Hanuman langurs consists of at least three species S. entellus, S. hypoleucos and S. priam. Furthermore, morphological studies suggested that both S. hypoleucos and S. priam have at least three subspecies in each. We explored the use of ecological niche modeling (ENM) to confirm the validity of these seven taxa and an additional taxon S. johnii belonging to the same genus. MaxEnt modeling tool was used with 19 bioclimatic, 12 vegetation and 6 hydrological environmental layers. We reduced total environmental variables to 14 layers after testing for collinearity and an independent test for model prediction was done using ENMTools. A total of 196 non-overlapping data points from primary and secondary sources were used as inputs for ENM. Results showed eight distinct ecological boundaries, corroborating the eight taxa mentioned above thereby confirming validity of these eight taxa. The study, for the first time provided ecological variables that determined the ecological requirements and distribution of members of the Hanuman langur species complex in the Indian peninsula.

Predicting loss and fragmentation of habitat of the vulnerable subtropical rainforest tree Macadamia integrifolia with models developed from compiled ecological data

Habitat models are widely used in ecology, however there are relatively few studies of rare species, primarily because of a paucity of survey records and lack of robust means of assessing accuracy of modelled spatial predictions. We investigated the potential of compiled ecological data in developing habitat models for Macadamia integrifolia, a vulnerable mid-stratum tree endemic to lowland subtropical rainforests of southeast Queensland, Australia. We compared performance of two binomial models—Classification and Regression Trees (CART) and Generalised Additive Models (GAM)—with Maximum Entropy (MAXENT) models developed from (i) presence records and available absence data and (ii) developed using presence records and background data. The GAM model was the best performer across the range of evaluation measures employed, however all models were assessed as potentially useful for informing in situ conservation of M. integrifolia, A significant loss in the amount of M. integrifolia habitat has occurred (p < 0.05), with only 37% of former habitat (pre-clearing) remaining in 2003. Remnant patches are significantly smaller, have larger edge-to-area ratios and are more isolated from each other compared to pre-clearing configurations (p < 0.05). Whilst the network of suitable habitat patches is still largely intact, there are numerous smaller patches that are more isolated in the contemporary landscape compared with their connectedness before clearing. These results suggest that in situ conservation of M. integrifolia may be best achieved through a landscape approach that considers the relative contribution of small remnant habitat fragments to the species as a whole, as facilitating connectivity among the entire network of habitat patches.

Extinction risk in cloud forest fragments under climate change and habitat loss

Aim: To quantify the consequences of major threats to biodiversity, such as climate and land-use change, it is important to use explicit measures of species persistence, such as extinction risk. The extinction risk of metapopulations can be approximated through simple models, providing a regional snapshot of the extinction probability of a species. We evaluated the extinction risk of three species under different climate change scenarios in three different regions of the Mexican cloud forest, a highly fragmented habitat that is particularly vulnerable to climate change. Location: Cloud forests in Mexico. Methods: Using Maxent, we estimated the potential distribution of cloud forest for three different time horizons (2030, 2050 and 2080) and their overlap with protected areas. Then, we calculated the extinction risk of three contrasting vertebrate species for two scenarios: (1) climate change only (all suitable areas of cloud forest through time) and (2) climate and land-use change (only suitable areas within a currently protected area), using an explicit patch-occupancy approximation model and calculating the joint probability of all populations becoming extinct when the number of remaining patches was less than five. Results: Our results show that the extent of environmentally suitable areas for cloud forest in Mexico will sharply decline in the next 70 years. We discovered that if all habitat outside protected areas is transformed, then only species with small area requirements are likely to persist. With habitat loss through climate change only, high dispersal rates are sufficient for persistence, but this requires protection of all remaining cloud forest areas. Main conclusions: Even if high dispersal rates mitigate the extinction risk of species due to climate change, the synergistic impacts of changing climate and land use further threaten the persistence of species with higher area requirements. Our approach for assessing the impacts of threats on biodiversity is particularly useful when there is little time or data for detailed population viability analyses. © 2013 John Wiley & Sons Ltd.

Priority threat management of invasive animals to protect biodiversity under climate change

Climate change is a major threat to global biodiversity, and its impacts can act synergistically to heighten the severity of other threats. Most research on projecting species range shifts under climate change has not been translated to informing priority management strategies on the ground. We develop a prioritization framework to assess strategies for managing threats to biodiversity under climate change and apply it to the management of invasive animal species across one-sixth of the Australian continent, the Lake Eyre Basin. We collected information from key stakeholders and experts on the impacts of invasive animals on 148 of the region's most threatened species and 11 potential strategies. Assisted by models of current distributions of threatened species and their projected distributions, experts estimated the cost, feasibility, and potential benefits of each strategy for improving the persistence of threatened species with and without climate change. We discover that the relative cost-effectiveness of invasive animal control strategies is robust to climate change, with the management of feral pigs being the highest priority for conserving threatened species overall. Complementary sets of strategies to protect as many threatened species as possible under limited budgets change when climate change is considered, with additional strategies required to avoid impending extinctions from the region. Overall, we find that the ranking of strategies by cost-effectiveness was relatively unaffected by including climate change into decision-making, even though the benefits of the strategies were lower. Future climate conditions and impacts on range shifts become most important to consider when designing comprehensive management plans for the control of invasive animals under limited budgets to maximize the number of threatened species that can be protected.

Potential Distribution and Risk Assessment of an Invasive Plant Species: A Case Study of Hymenachne amplexicaulis in Australia

Given the limited resources available for weed management, a strategic approach is required to give the best bang for your buck. The current study incorporates: (1) a model ensemble approach to identify areas of uncertainty and commonality regarding a species invasive potential, (2) current distribution of the invaded species, and (3) connectivity of systems to identify target regions and focus efforts for more effective management. Uncertainty in the prediction of suitable habitat for H. amplexicaulis (study species) in Australia was addressed in an ensemble-forecasting approach to compare distributional scenarios from four models (CLIMATCH; CLIMEX; boosted regression trees [BRT]; maximum entropy [Maxent]). Models were built using subsets of occurrence and environmental data. Catchment risk was determined through incorporating habitat suitability, the current abundance and distribution of H. amplexicaulis, and catchment connectivity. Our results indicate geographic differences between predictions of different approaches. Despite these differences a number of catchments in northern, central, and southern Australia were identified as high risk of invasion or further spread by all models suggesting they should be given priority for the management of H. amplexicaulis. The study also highlighted the utility of ensemble approaches in indentifying areas of uncertainty and commonality regarding the species invasive potential.

Predictive distribution modeling for rare Himalayan medicinal plant Berberis aristata DC

Predictive distribution modelling of Berberis aristata DC, a rare threatened plant with high medicinal values has been done with an aim to understand its potential distribution zones in Indian Himalayan region. Bioclimatic and topographic variables were used to develop the distribution model with the help of three different algorithms viz. GeneticAlgorithm for Rule-set Production (GARP), Bioclim and Maximum entroys(MaxEnt). Maximum entropy has predicted wider potential distribution (10.36%) compared to GARP (4.63%) and Bioclim (2.44%). Validation confirms that these outputs are comparable to the present distribution pattern of the B. atistata. This exercise highlights that this species favours Western Himalaya. However, GARP and MaxEnt's prediction of Eastern Himalayan states (i.e. Arunachal Pradesh, Nagaland and Manipur) are also identified as potential occurrence places require further exploration.

Semi-supervised SVMs for classification with unknown class proportions and a small labeled dataset

In the design of practical web page classification systems one often encounters a situation in which the labeled training set is created by choosing some examples from each class; but, the class proportions in this set are not the same as those in the test distribution to which the classifier will be actually applied. The problem is made worse when the amount of training data is also small. In this paper we explore and adapt binary SVM methods that make use of unlabeled data from the test distribution, viz., Transductive SVMs (TSVMs) and expectation regularization/constraint (ER/EC) methods to deal with this situation. We empirically show that when the labeled training data is small, TSVM designed using the class ratio tuned by minimizing the loss on the labeled set yields the best performance; its performance is good even when the deviation between the class ratios of the labeled training set and the test set is quite large. When the labeled training data is sufficiently large, an unsupervised Gaussian mixture model can be used to get a very good estimate of the class ratio in the test set; also, when this estimate is used, both TSVM and EC/ER give their best possible performance, with TSVM coming out superior. The ideas in the paper can be easily extended to multi-class SVMs and MaxEnt models.

Extension of TSVM to multi-class and hierarchical text classification problems With general losses

Transductive SVM (TSVM) is a well known semi-supervised large margin learning method for binary text classification. In this paper we extend this method to multi-class and hierarchical classification problems. We point out that the determination of labels of unlabeled examples with fixed classifier weights is a linear programming problem. We devise an efficient technique for solving it. The method is applicable to general loss functions. We demonstrate the value of the new method using large margin loss on a number of multi-class and hierarchical classification datasets. For maxent loss we show empirically that our method is better than expectation regularization/constraint and posterior regularization methods, and competitive with the version of entropy regularization method which uses label constraints.

Predicting the effects of climate change on sea turtle nesting habitat in Florida

Rising global temperatures threaten the survival of many plant and animal species. Having already risen at an unprecedented rate in the past century, temperatures are predicted to rise between 0.3 and 7.5C in North America over the next 100 years (Hawkes et al. 2007). Studies have documented the effects of climate warming on phenology (timing of seasonal activities), with observations of early arrival at breeding grounds, earlier ends to the reproductive season, and delayed autumnal migrations (Pike et al. 2006). In addition, for species not suited to the physiological demands of cold winter temperatures, increasing temperatures could shift tolerable habitats to higher latitudes (Hawkes et al. 2007). More directly, climate warming will impact thermally sensitive species like sea turtles, who exhibit temperature-dependent sexual determination. Temperatures in the middle third of the incubation period determine the sex of sea turtle offspring, with higher temperatures resulting in a greater abundance of female offspring. Consequently, increasing temperatures from climate warming would drastically change the offspring sex ratio (Hawkes et al. 2007). Of the seven extant species of sea turtles, three (leatherback, Kemp’s ridley, and hawksbill) are critically endangered, two (olive ridley and green) are endangered, and one (loggerhead) is threatened. Considering the predicted scenarios of climate warming and the already tenuous status of sea turtle populations, it is essential that efforts are made to understand how increasing temperatures may affect sea turtle populations and how these species might adapt in the face of such changes. In this analysis, I seek to identify the impact of changing climate conditions over the next 50 years on the availability of sea turtle nesting habitat in Florida given predicted changes in temperature and precipitation. I predict that future conditions in Florida will be less suitable for sea turtle nesting during the historic nesting season. This may imply that sea turtles will nest at a different time of year, in more northern latitudes, to a lesser extent, or possibly not at all. It seems likely that changes in temperature and precipitation patterns will alter the distribution of sea turtle nesting locations worldwide, provided that beaches where the conditions are suitable for nesting still exist. Hijmans and Graham (2006) evaluate a range of climate envelope models in terms of their ability to predict species distributions under climate change scenarios. Their results suggested that the choice of species distribution model is dependent on the specifics of each individual study. Fuller et al. (2008) used a maximum entropy approach to model the potential distribution of 11 species in the Arctic Coastal Plain of Alaska under a series of projected climate scenarios. Recently, Pike (in press) developed Maxent models to investigate the impacts of climate change on green sea turtle nest distribution and timing. In each of these studies, a set of environmental predictor variables (including climate variables), for which ‘current’ conditions are available and ‘future’ conditions have been projected, is used in conjunction with species occurrence data to map potential species distribution under the projected conditions. In this study, I will take a similar approach in mapping the potential sea turtle nesting habitat in Florida by developing a Maxent model based on environmental and climate data and projecting the model for future climate data. (PDF contains 5 pages)

Prediction of mesophotic coral distributions in the Au‘au Channel, Hawaii

The primary objective of this study was to predict the distribution of mesophotic hard corals in the Au‘au Channel in the Main Hawaiian Islands (MHI). Mesophotic hard corals are light-dependent corals adapted to the low light conditions at approximately 30 to 150 m in depth. Several physical factors potentially influence their spatial distribution, including aragonite saturation, alkalinity, pH, currents, water temperature, hard substrate availability and the availability of light at depth. Mesophotic corals and mesophotic coral ecosystems (MCEs) have increasingly been the subject of scientific study because they are being threatened by a growing number of anthropogenic stressors. They are the focus of this spatial modeling effort because the Hawaiian Islands Humpback Whale National Marine Sanctuary (HIHWNMS) is exploring the expansion of its scope—beyond the protection of the North Pacific Humpback Whale (Megaptera novaeangliae)—to include the conservation and management of these ecosystem components. The present study helps to address this need by examining the distribution of mesophotic corals in the Au‘au Channel region. This area is located between the islands of Maui, Lanai, Molokai and Kahoolawe, and includes parts of the Kealaikahiki, Alalākeiki and Kalohi Channels. It is unique, not only in terms of its geology, but also in terms of its physical oceanography and local weather patterns. Several physical conditions make it an ideal place for mesophotic hard corals, including consistently good water quality and clarity because it is flushed by tidal currents semi-diurnally; it has low amounts of rainfall and sediment run-off from the nearby land; and it is largely protected from seasonally strong wind and wave energy. Combined, these oceanographic and weather conditions create patches of comparatively warm, calm, clear waters that remain relatively stable through time. Freely available Maximum Entropy modeling software (MaxEnt 3.3.3e) was used to create four separate maps of predicted habitat suitability for: (1) all mesophotic hard corals combined, (2) Leptoseris, (3) Montipora and (4) Porites genera. MaxEnt works by analyzing the distribution of environmental variables where species are present, so it can find other areas that meet all of the same environmental constraints. Several steps (Figure 0.1) were required to produce and validate four ensemble predictive models (i.e., models with 10 replicates each). Approximately 2,000 georeferenced records containing information about mesophotic coral occurrence and 34 environmental predictors describing the seafloor’s depth, vertical structure, available light, surface temperature, currents and distance from shoreline at three spatial scales were used to train MaxEnt. Fifty percent of the 1,989 records were randomly chosen and set aside to assess each model replicate’s performance using Receiver Operating Characteristic (ROC), Area Under the Curve (AUC) values. An additional 1,646 records were also randomly chosen and set aside to independently assess the predictive accuracy of the four ensemble models. Suitability thresholds for these models (denoting where corals were predicted to be present/absent) were chosen by finding where the maximum number of correctly predicted presence and absence records intersected on each ROC curve. Permutation importance and jackknife analysis were used to quantify the contribution of each environmental variable to the four ensemble models.

Wybiórczość siedliskowa dzięcioła średniego Dendrocopos medius w zróżnicowanych ekosystemach leśnych

Wydział Biologii