Federico Montesino Pouzols

Global protected area expansion is compromised by projected land-use and parochialism

Protected areas are one of the main tools for halting the continuing global biodiversity crisis caused by habitat loss, fragmentation and other anthropogenic pressures. According to the Aichi Biodiversity Target 11 adopted by the Convention on Biological Diversity, the protected area network should be expanded to at least 17% of the terrestrial world by 2020 (http://www.cbd.int/sp/targets). To maximize conservation outcomes, it is crucial to identify the best expansion areas. Here we show that there is a very high potential to increase protection of ecoregions and vertebrate species by expanding the protected area network, but also identify considerable risk of ineffective outcomes due to land-use change and uncoordinated actions between countries. We use distribution data for 24,757 terrestrial vertebrates assessed under the International Union for the Conservation of Nature (IUCN) ‘red list of threatened species’, and terrestrial ecoregions (827), modified by land-use models for the present and 2040, and introduce techniques for global and balanced spatial conservation prioritization. First, we show that with a coordinated global protected area network expansion to 17% of terrestrial land, average protection of species ranges and ecoregions could triple. Second, if projected land-use change by 2040 (ref. 11) takes place, it becomes infeasible to reach the currently possible protection levels, and over 1,000 threatened species would lose more than 50% of their present effective ranges worldwide. Third, we demonstrate a major efficiency gap between national and global conservation priorities. Strong evidence is shown that further biodiversity loss is unavoidable unless international action is quickly taken to balance land-use and biodiversity conservation. The approach used here can serve as a framework for repeatable and quantitative assessment of efficiency, gaps and expansion of the global protected area network globally, regionally and nationally, considering current and projected land-use pressures.

Global priorities for national carnivore conservation under land use change

Mammalian carnivores have suffered the biggest range contraction among all biodiversity and are particularly vulnerable to habitat loss and fragmentation. Therefore, we identified priority areas for the conservation of mammalian carnivores, while accounting for species-specific requirements for connectivity and expected agricultural and urban expansion. While prioritizing for carnivores only, we were also able to test their effectiveness as surrogates for 23,110 species of amphibians, birds, mammals and reptiles and 867 terrestrial ecoregions. We then assessed the risks to carnivore conservation within each country that makes a contribution to global carnivore conservation. We found that land use change will potentially lead to important range losses, particularly amongst already threatened carnivore species. In addition, the 17% of land targeted for protection under the Aichi Target 11 was found to be inadequate to conserve carnivores under expected land use change. Our results also highlight that land use change will decrease the effectiveness of carnivores to protect other threatened species, especially threatened amphibians. In addition, the risk of human-carnivore conflict is potentially high in countries where we identified spatial priorities for their conservation. As meeting the global biodiversity target will be inadequate for carnivore protection, innovative interventions are needed to conserve carnivores outside protected areas to compliment any proposed expansion of the protected area network.

Evolving fuzzy optimally pruned extreme learning machine for regression problems

This paper proposes an approach to the identification of evolving fuzzy Takagi–Sugeno systems based on the optimally pruned extreme learning machine (OP-ELM) methodology. First, we describe ELM, a simple yet accurate learning algorithm for training single-hidden layer feed-forward artificial neural networks with random hidden neurons. We then describe the OP-ELM methodology for building ELM models in a robust and simplified manner suitable for evolving approaches. Based on the previously proposed ELM method, and the OP-ELM methodology, we propose an identification method for self-developing or evolving neuro-fuzzy systems applicable to regression problems. This method, evolving fuzzy optimally pruned extreme learning machine (eF-OP-ELM), follows a random projection based approach to extracting evolving fuzzy rulebases. In this approach systems are not only evolving but their structure is defined on the basis of randomly generated fuzzy basis functions. A comparative analysis of eF-OP-ELM is performed over a diverse collection of benchmark datasets against well known evolving neuro-fuzzy methods, namely eTS and DENFIS. Results show that the method proposed yields compact rulebases, is robust and competitive in terms of accuracy.

Autoregressive time series prediction by means of fuzzy inference systems using nonparametric residual variance estimation

We propose an automatic methodology framework for short- and long-term prediction of time series by means of fuzzy inference systems. In this methodology, fuzzy techniques and statistical techniques for nonparametric residual variance estimation are combined in order to build autoregressive predictive models implemented as fuzzy inference systems. Nonparametric residual variance estimation plays a key role in driving the identification and learning procedures. Concrete criteria and procedures within the proposed methodology framework are applied to a number of time series prediction problems. The learn from examples method introduced by Wang and Mendel (W&M) is used for identification. The Levenberg-Marquardt (L-M) optimization method is then applied for tuning. The W&M method produces compact and potentially accurate inference systems when applied after a proper variable selection stage. The L-M method yields the best compromise between accuracy and interpretability of results, among a set of alternatives. Delta test based residual variance estimations are used in order to select the best subset of inputs to the fuzzy inference systems as well as the number of linguistic labels for the inputs. Experiments on a diverse set of time series prediction benchmarks are compared against least-squares support vector machines (LS-SVM), optimally pruned extreme learning machine (OP-ELM), and k-NN based autoregressors. The advantages of the proposed methodology are shown in terms of linguistic interpretability, generalization capability and computational cost. Furthermore, fuzzy models are shown to be consistently more accurate for prediction in the case of time series coming from real-world applications.

A method for calculating minimum biodiversity offset multipliers accounting for time discounting, additionality and permanence

Biodiversity offsetting, which means compensation for ecological and environmental damage caused by development activity, has recently been gaining strong political support around the world. One common criticism levelled at offsets is that they exchange certain and almost immediate losses for uncertain future gains. In the case of restoration offsets, gains may be realized after a time delay of decades, and with considerable uncertainty. Here we focus on offset multipliers, which are ratios between damaged and compensated amounts (areas) of biodiversity. Multipliers have the attraction of being an easily understandable way of deciding the amount of offsetting needed. On the other hand, exact values of multipliers are very difficult to compute in practice if at all possible. We introduce a mathematical method for deriving minimum levels for offset multipliers under the assumption that offsetting gains must compensate for the losses (no net loss offsetting). We calculate absolute minimum multipliers that arise from time discounting and delayed emergence of offsetting gains for a one-dimensional measure of biodiversity. Despite the highly simplified model, we show that even the absolute minimum multipliers may easily be quite large, in the order of dozens, and theoretically arbitrarily large, contradicting the relatively low multipliers found in literature and in practice. While our results inform policy makers about realistic minimal offsetting requirements, they also challenge many current policies and show the importance of rigorous models for computing (minimum) offset multipliers. The strength of the presented method is that it requires minimal underlying information. We include a supplementary spreadsheet tool for calculating multipliers to facilitate application.

RobOff: software for analysis of alternative land‐use options and conservation actions

Summary 1. Habitat restoration is increasing in importance as a conservation action, compared with more traditional establishment of conservation areas. It is applied, for example, in the context of biodiversity offsetting, in which environmental impacts of economic activity are offset by additional compensating conservation efforts. 2. We present a publicly available decision support software tool for the comparison of ecological impacts of alternative land-use options. 3. Methods implemented account for uncertain consequences of alternative land-use options, including conservation actions. These actions have different costs and effects on different biodiversity features, including species, guilds or habitat types, in different environments. Consequences of actions are uncertain through time, and time discounting is allowed in the investigation of temporal preferences. 4. This tool facilitates analyses relevant for planning of habitat restoration or management, environmental impact avoidance, biodiversity offsetting and scenario development for systematic conservation planning. RobOff derives its name from Robust Offsetting.

A method for building corridors in spatial conservation prioritization

We introduce a novel approach to building corridors in spatial conservation prioritization. The underlying working principle is the use of a penalty structure in an iterative algorithm used for producing a spatial priority ranking. The penalty term aims to prevent loss or degradation of structural connections, or, equivalently, to promote to a higher rank landscape elements that are required to keep networks connected. The proposed method shows several convenient properties: (1) it does not require a priori specification of habitat patches, end points or related thresholds, (2) it does not rely on resistance coefficients for different habitats, (3) it does not require species targets, and (4) the cost of additional connectivity via corridors can be quantified in terms of habitat quality lost across species. Corridor strength and width parameters control the trade-off between increased structural connectivity via corridors and other considerations relevant to conservation planning. Habitat suitability or dispersal suitability layers used in the analysis can be species specific, thus allowing analysis both in terms of structural and functional connectivity. The proposed method can also be used for targeting habitat restoration, by identifying areas of low habitat quality included in corridors. These methods have been implemented in the Zonation software, and can be applied to large scale and high resolution spatial prioritization, effectively integrating corridor design and spatial conservation prioritization. Since the method operates on novel principles and combines with a large number of features already operational in Zonation, we expect it to be of utility in spatial conservation planning.

Learning deep belief networks from non-stationary streams

Deep learning has proven to be beneficial for complex tasks such as classifying images. However, this approach has been mostly applied to static datasets. The analysis of non-stationary (e.g., concept drift) streams of data involves specific issues connected with the temporal and changing nature of the data. In this paper, we propose a proof-of-concept method, called Adaptive Deep Belief Networks, of how deep learning can be generalized to learn online from changing streams of data. We do so by exploiting the generative properties of the model to incrementally re-train the Deep Belief Network whenever new data are collected. This approach eliminates the need to store past observations and, therefore, requires only constant memory consumption. Hence, our approach can be valuable for life-long learning from non-stationary data streams.

Automatic clustering-based identification of autoregressive fuzzy inference models for time series

We analyze the use of clustering methods for the automatic identification of fuzzy inference models for autoregressive prediction of time series. A methodology that combines fuzzy methods and residual variance estimation techniques is followed. A nonparametric residual variance estimator is used for a priori input and model selection. A simple scheme for initializing the widths of the input membership functions of fuzzy inference systems is proposed for the Improved Clustering for Function Approximation algorithm (ICFA), previously introduced for initializing RBF networks. This extension to the ICFA algorithm is shown to provide the most accurate predictions among a wide set of clustering algorithms. The method is applied to a diverse set of time series benchmarks. Its advantages in terms of accuracy and computational requirements are shown as compared to least-squares support vector machines (LS-SVM), the multilayer perceptron (MLP) and two variants of the extreme learning machine (ELM).

Fuzzy inference based autoregressors for time series prediction using nonparametric residual variance estimation

We apply fuzzy techniques for system identification and supervised learning in order to develop fuzzy inference based autoregressors for time series prediction. An automatic methodology framework that combines fuzzy techniques and statistical techniques for nonparametric residual variance estimation is proposed. Identification is performed through the learn from examples method introduced by Wang and Mendel, while the Marquard-Levenberg supervised learning algorithm is then applied for tuning. Delta test residual noise estimation is used in order to select the best subset of inputs as well as the number of linguistic labels for the inputs. Experimental results for three time series prediction benchmarks are compared against LS-SVM based autoregressors and show the advantages of the proposed methodology in terms of approximation accuracy, generalization capability and linguistic interpretability.