How many predictors in species distribution models at the landscape scale? Land use versus LiDAR-derived canopy height

2014 ◽  
Vol 28 (8) ◽  
pp. 1723-1739 ◽  
Author(s):  
Gentile Francesco Ficetola ◽  
Anna Bonardi ◽  
Caspar A. Mücher ◽  
Niels L.M. Gilissen ◽  
Emilio Padoa-Schioppa

<em>Abstract</em>.—Increasingly, fisheries managers must make important decisions in complex environments where rapidly changing landscape and climate conditions interact with historical impacts to influence resource sustainability. Successful fisheries management in this setting will require that we adapt traditional management approaches to incorporate information on these complex interacting factors—a process referred to as resilient fisheries management. Large-scale species distribution data and predictive models have the potential to enhance the management of freshwater fishes through improved understanding of how past, present, and future natural and anthropogenic factors combine to determine species vulnerability and resiliency. Here we describe a resilient fisheries management framework that provides guidance on how and when these models can be incorporated into traditional approaches to meet specific goals and objectives for resource sustainability. In addition to elucidating complex drivers of distributional patterns and change, species distribution models can inform the prioritization, application, and implementation of management activities such as restoration (e.g., instream habitat and riparian), protection (e.g., areas where additional land use would result in a change in species distribution), and regulations (e.g., harvest restriction) in a way that informs resiliency to land use and climate change. Although considerable progress has been made with respect to applying species distribution models to the management of Brook Trout <em>Salvelinus fontinalis </em>and other aquatic species, there are several areas where a more unified research and management effort could increase the ability of distribution models to inform resilient management. Future efforts should aim to improve (1) data availability, consistency (sampling methodology), and quality (accounting for detection); (2) partnerships among researchers, agencies, and managers; and (3) model accessibility and understanding of limitations and potential benefits to managers (e.g., incorporation into publicly available decision support systems). The information and recommendations provided herein can be used to promote and advance the use of models in resilient fisheries management in the face of continued large-scale land use and climate change.


2016 ◽  
Vol 26 (1) ◽  
pp. 65-77 ◽  
Author(s):  
Jean-Sauveur Ay ◽  
Joannès Guillemot ◽  
Nicolas Martin-StPaul ◽  
Luc Doyen ◽  
Paul Leadley

2018 ◽  
Vol 28 (7) ◽  
pp. 1867-1883 ◽  
Author(s):  
Adrián Regos ◽  
Louis Imbeau ◽  
Mélanie Desrochers ◽  
Alain Leduc ◽  
Michel Robert ◽  
...  

2017 ◽  
Vol 28 (4) ◽  
pp. 581-592 ◽  
Author(s):  
LAURA CARDADOR ◽  
JOSÉ A. DÍAZ-LUQUE ◽  
FERNANDO HIRALDO ◽  
JAMES D. GILARDI ◽  
JOSÉ L. TELLA

SummaryKnowledge of a species’ potential distribution and the suitability of available habitat are fundamental for effective conservation planning and management. However, the quality of information on the distribution of species and their required habitats is highly variable in terms of accuracy and availability across taxa and regions, particularly in tropical landscapes where accessibility is especially challenging. Species distribution models (SDMs) provide predictive tools for addressing gaps for poorly surveyed species, but they rarely consider biases in geographical distribution of records and their consequences. We applied SDMs and variation partitioning analyses to investigate the relative importance of habitat characteristics, human accessibility, and their joint effects in the global distribution of the Critically Endangered Blue-throated MacawAra glaucogularis, a species endemic to the Amazonian flooded savannas of Bolivia. The probability of occurrence was skewed towards more accessible areas, mostly secondary roads. Variability in observed occurrence patterns was mostly accounted for by the pure effect of habitat characteristics (76.2%), indicating that bias in the geographical distribution of occurrences does not invalidate species-habitat relationships derived from niche models. However, observed spatial covariation between land-use at a landscape scale and accessibility (joint contribution: 22.3%) may confound the independent role of land-use in the species distribution. New surveys should prioritise collecting data in more remote (less accessible) areas better distributed with respect to land-use composition at a landscape scale. Our results encourage wider application of partitioning methods to quantify the extent of sampling bias in datasets used in habitat modelling for a better understanding of species-habitat relationships, and add insights into the potential distribution of our study species and opportunities for its conservation.


2019 ◽  
Vol 30 (2) ◽  
pp. 386-396 ◽  
Author(s):  
Rubén G. Mateo ◽  
Aitor Gastón ◽  
María José Aroca‐Fernández ◽  
Olivier Broennimann ◽  
Antoine Guisan ◽  
...  

2021 ◽  
Vol 13 (8) ◽  
pp. 1495
Author(s):  
Jehyeok Rew ◽  
Yongjang Cho ◽  
Eenjun Hwang

Species distribution models have been used for various purposes, such as conserving species, discovering potential habitats, and obtaining evolutionary insights by predicting species occurrence. Many statistical and machine-learning-based approaches have been proposed to construct effective species distribution models, but with limited success due to spatial biases in presences and imbalanced presence-absences. We propose a novel species distribution model to address these problems based on bootstrap aggregating (bagging) ensembles of deep neural networks (DNNs). We first generate bootstraps considering presence-absence data on spatial balance to alleviate the bias problem. Then we construct DNNs using environmental data from presence and absence locations, and finally combine these into an ensemble model using three voting methods to improve prediction accuracy. Extensive experiments verified the proposed model’s effectiveness for species in South Korea using crowdsourced observations that have spatial biases. The proposed model achieved more accurate and robust prediction results than the current best practice models.


Sign in / Sign up

Export Citation Format

Share Document