Using Remote Sensing for Modeling and Monitoring Species Distributions
- 5.1k Downloads
Interpolated climate surfaces have been widely used to predict species distributions and develop environmental niche models. However, the spatial coverage and density of meteorological sites used to develop these surfaces vary among countries and regions, such that the most biodiverse regions often have the most sparsely sampled climatic data. We explore the potential of satellite remote sensing (S-RS) products—which have consistently high spatial and temporal resolution and nearly global coverage—to quantify species-environment relationships that predict species distributions. We propose several new environmental metrics that take advantage of high temporal resolution in S-RS data and compare these approaches to classic climate-only approaches using the live oaks (Quercus section Virentes) as a case study. We show that models perform similarly but for some species, particularly in understudied regions, show less precision in predicting spatial distribution. These results provide evidence supporting efforts to enhance environmental niche models and species distribution models (ENMs/SDMs) with S-RS data and, when combined with other approaches for species detection, will likely enhance our ability to monitor biodiversity globally.
KeywordsSpecies distribution models Environmental niche models Biotic niche Abiotic niche Remotely sensed environmental data Climate variables Phenology
What drives species distributions? This is one of the most fundamental questions in ecology, evolution, and biogeography, and it drew the attention of early naturalists (Gaston 2009; Guisan et al. 2017). Although the question is classic and its answers sometime seem obvious—for example, Alfred Russel Wallace recognized the effect of geographical and environmental features on species distributional ranges (Wallace 1860)—the answers are highly complex as a consequence of historical evolutionary and biogeographic processes and the spatial and temporal dynamics of abiotic and biotic factors (Soberón and Peterson 2005; Soberón 2007; Colwell and Rangel 2009).
Here we explore the potential of satellite remote sensing (S-RS) products to quantify species-environment relationships that predict species distributions. We propose several new metrics that take advantage of the high temporal resolution in Moderate Resolution Imaging Spectroradiometer (MODIS) leaf area index (LAI) and MODIS normalized difference vegetation index (NDVI) data products. Evaluating the potential of remotely sensed data in environmental niche modeling (ENM) and species distribution modeling (SDM) is an important step toward the long-term goal of improving our ability to monitor and predict changes in biodiversity globally. To achieve this, we first modeled the environmental/ecological niches for the American live oak species (Quercus section Virentes) using environmental variables derived from (1) interpolated climate surfaces data (i.e., WorldClim) and (2) S-RS products. Live oaks are a small lineage, descended from a common ancestor (for a discussion on phylogenetics, see Meireles et al., Chap. 7) that includes seven species that vary in geographic range size and climatic breadth and are distributed in both temperate and tropical climates from the southeastern United States, Mesoamerica, and the Caribbean (Cavender-Bares et al. 2015). Their variation in range size and climatic distributions, their distributions in both highly studied and understudied regions of the globe, and the second author’s expert knowledge of their distributions make them an interesting case study for comparing SDM/ENMs that rely on classic data sources to those that use remotely sensed data sources, which have more consistent data accuracy and resolution. We used the live oaks as a test clade to evaluate the relationships among the modeled niches estimated from both sources of environmental data.
Given that the interpolated climate surfaces from WorldClim (Hijmans et al. 2005) are the most widely used data set for the study of species-environment relationships, we compare the performance of SDMs based on S-RS products to those based on WorldClim data. If there is a tight relationship between models from the two sources, this would indicate that the resultant models from S-RS products have similar performance to the resultant models from the WorldClim climatic predictors. Remotely sensed data products may provide an advantage in predicting species distributions in regions where climatic data is sparsely sampled. Although WorldClim provides interpolated climate surfaces for land areas across the world at multiple spatial resolutions, from 30 arc seconds (~1 km) to 10 arcmin (~18.5 km) (Hijmans et al. 2005), the spatial distribution of the base information (i.e., weather or climatic stations) used for interpolations is unevenly distributed across the world (Fig. 9.2c). This is not a small issue given the uncertainty associated with interpolated climatic variables when modeling species-environment relationships, especially in many tropical countries, where weather stations are frequently few and far apart (Soria-Auza et al. 2010). Given that tropical regions are precisely the regions where most species occur (Fig. 9.2c), finding alternative means to predict species is important for efforts to monitor and manage biodiversity globally. S-RS products, which provide quasi-global coverage of land and sea surfaces at high temporal and spatial resolution, represent promising alternatives that may be particularly important in the world’s most biologically diverse regions. Our aim here is to provide an understanding of the potential of S-RS products to quantify species ecological niches and estimate species distributions rather than to develop a definitive ecological and geographical profile for the live oaks themselves. If the consistent accuracy and high spatial resolution of S-RS products can actually improve estimates of species distributions, they will represent an advance in our ability to predict where species are likely persist under changing environments. Ultimately, such predictions can be combined with other remote sensing (RS) means of detecting species and biodiversity (Meireles et al., Chap. 7; Bolch et al., Chap. 12; Record et al., Chap. 10) to enable global-scale biodiversity change detection.
9.2 Theoretical Background
9.2.1 The BAM Diagram
9.2.2 Where Are We Now?
Although the BAM framework was developed to understand and quantify species-environment relationships (Soberón 2007; see also Soberón and Peterson 2005), the concept and investigation of species-environment relationships are long-standing, dating back to Wallace (Wallace 1860) and early ecologists (Grinnell 1904, 1917; Elton 1927; Holdridge 1947; Hutchinson 1957). These early naturalists originally established the theoretical principles to analyze and describe biogeographical distributions in relation to environmental patterns (Colwell and Rangel 2009). Interestingly, despite the large body of theoretical advances and empirical applications, the quantification of ecological niches and estimation of species distributions is still a challenging task (but see Sanín and Anderson 2018; Smith et al. 2018) and one of the most active areas in macroecological and biodiversity research (Franklin 2010; Peterson et al. 2011; Anderson 2013; Guisan et al. 2017).
Parallel to the development and evolution of ENM and SDM theory and applications, we have witnessed the growth of technological tools and S-RS products (Pettorelli et al. 2014a; Turner 2014). Many of these are particularly applicable for describing, quantifying, and mapping the spatial and temporal patterns of vegetation structure and function, the impacts of human activities, and environmental change (Turner et al. 2003; Pinto-Ledezma and Rivero 2014; Jetz et al. 2016; Cord et al. 2017) and more recently are used as predictors of broad patterns of biodiversity, including the associations between species co-occurrence patterns and ecosystem energy availability (Phillips et al. 2008; Pigot et al. 2016; Hobi et al. 2017). In addition, an unprecedented number of S-RS data and data products (S-RS) have been made freely available (Turner et al. 2003; Hobi et al. 2017) with the potential to track the spatial variation in the chemical composition of vegetation (Wang et al. 2019; Serbin and Townsend, Chap. 3), physiology, structure, and function (Lausch et al., Chap. 13; Serbin and Townsend, Chap. 3; Myneni et al. 2002; Saatchi et al. 2008; Jetz et al. 2016).
Despite the potential of S-RS products for measuring and modeling biodiversity (Gillespie et al. 2008; Pettorelli et al. 2014a, b; Turner 2014; Cord et al. 2013; Zimmermann et al. 2007), attention has only recently turned to using these data in studies of species-environment relationships (Cord et al. 2013; West et al. 2016), and most studies use bioclimatic data such as WorldClim (but see Paz et al., Chap. 11; Record et al., Chap. 10). Although early attempts indicated that S-RS products do not seem to improve the accuracy in estimating species distributions (Pearson et al. 2004; Thuiller 2004; Zimmermann et al. 2007), more recent publications (Kissling et al. 2012; Cord et al. 2013) suggest that despite these apparent limitations, S-RS products provide better spatial resolution that allow the discrimination of habitat characteristics not captured when bioclimatic data are used (Saatchi et al. 2008; Cord et al. 2013), and they can be used as surrogates of biotic and/or functional predictors such as LAI that increase the performance of individual species models (Kissling et al. 2012; Cord et al. 2013).
9.3 Modeling Ecological Niches and Predicting Geographic Distributions
Although the terms ENM and SDM are often used synonymously in the literature, the two are not equivalent (Anderson 2012; Soberón et al. 2017). A comprehensive discussion of this topic is beyond the scope of this chapter but is provided elsewhere (see Peterson et al. 2011; Anderson 2012; Soberón et al. 2017). A crucial step in differentiating the two terms is to establish a distinction between environmental space and geographical space (Hutchinson’s duality; Colwell and Rangel 2009). On the one hand, environmental space corresponds to a suite of environmental conditions at a given time (e.g., climate, topography); on the other hand, geographical space is the extent of a particular region or study area (Soberón and Nakamura 2009; Peterson et al. 2011) and includes important historical context. Thus, when modeling species ecological niches, we are modeling the existing abiotically suitable conditions for the species or the biotically reduced niche (Peterson et al. 2011; see also Fig. 9.1). However, when modeling species distributions, the intent is to project objects into geographical space (Fig. 9.1), and, depending on the factors considered, it is possible to estimate the occupied distributional area or the invadable distributional area (Soberón and Nakamura 2009; Peterson et al. 2011; Anderson 2012; Soberón et al. 2017).
188.8.131.52 Oak Species Data Sets
Occurrence data were downloaded from iDigBio between 20 and 24 July 2018, including localities collected by the authors, and cleaned for accuracy. Any botanical garden localities were discarded. All points were visually examined, and any localities that were outside the known range of the species, or in unrealistic locations (e.g., water bodies), were discarded.
184.108.40.206 Environmental Data Sets
Environmental variables from S-RS products were obtained from MODIS over a 15-year period (2001–2015) from NASA using the interface EOSDIS Earthdata (https://earthdata.nasa.gov). Data include two MODIS Collection 5 land products: LAI (8-day temporal resolution) and NDVI (16-day temporal resolution). LAI and NDVI products (Fig. 9.3b, c) are derived from Terra/Aqua MOD15A2 and Terra MOD13A2, respectively (see Myneni et al. 2002 for a detailed explanation of MODIS products). We also obtained precipitation data from Climate Hazards group Infrared Precipitation with Stations (CHIRPS, Fig. 9.3a), an S-RS product designed for monitoring drought and global environmental land change (Funk et al. 2015). Notice that the original MODIS products present a spatial resolution of 1 km and CHIRPS, a spatial resolution of 3 arcmin or ~5.5 km at the equator. To standardize the spatial resolution of both MODIS products and CHIRPS, we upscaled the spatial resolution of MODIS products to that of CHIRPS.
220.127.116.11 Modeling Procedure
Combinations of environmental variables used for modeling live oak species-environment relationship under an ensemble framework
Climate Hazards group Infrared Precipitation with Stations
MODIS maximum leaf area index calculated over a year
MODIS mean leaf area index calculated over a year
MODIS seasonality of leaf area index calculated over a year
MODIS minimum leaf area index calculated over a year
Mean elevation from Shuttle Radar Topography Mission
MODIS maximum normalized difference vegetation index calculated over a year
MODIS mean normalized difference vegetation index calculated over a year
MODIS seasonality of normalized difference vegetation index calculated over a year
MODIS minimum normalized difference vegetation index calculated over a year
Mean annual temperature
Minimum temperature of coldest month
Mean temperature of warmest quarter
Mean annual precipitation
Within our ensemble framework, species’ ecological niches are modeled using the six algorithms by fitting the occurrences of a single species and the predictors. The resulting six species models (one for each algorithm) are stacked into a single species model by averaging all models (Araújo and New 2007). We chose this approach because a major source of uncertainty in ENM/SDM arises from the algorithm used for modeling (Diniz-Filho et al. 2009; Qiao et al. 2015) and because the choice of the “best” modeling algorithm depends on the aims of the modeling applications (Peterson et al. 2011). Finally, using the stacked species models, we estimated macroecological patterns of species richness and the uncertainty associated with model parametrization. These patterns are less interesting in their own right for a small clade with only seven species, but they demonstrate an effective approach that can be applied to much larger groups of species.
We estimated live oak species richness by summing the projected potential species distributions; uncertainty was estimated as the variance attributable to the source of uncertainty (i.e., algorithms and their interactions) by performing a one-way analysis of variance (ANOVA) without replicates (Sokal and Rohlf 1995). The resulting uncertainty map shows regions with low and high uncertainty associated with the source of uncertainty (i.e., algorithm).
To explore the performance of environmental data derived from RS for ENM/SDM compared to traditionally used environmental data from climatic variables (e.g., WorldClim), we evaluated the relationship between the modeled ecological niches from: (1) S-RS products; and (2) environmental variables from WorldClim. In doing so, we used correlation analyses corrected according Clifford’s method to obtain the effective degrees of freedom for Pearson’s coefficients while controlling for spatial autocorrelation (Clifford et al. 1989). Statistical analyses were performed in R using the package SpatialPack (Vallejos and Osorio 2014).
Model accuracy assessment for the three combinations of environmental predictors
Spatial correlations between live oak ENMs estimated under three combinations of environmental variables
Spatial correlation between estimations of live oak species richness and uncertainty quantified under three combinations of environmental variables
The field of ecological niche and species distribution modeling contributes significantly to our capacity to evaluate and describe the effect of geographical and environmental features on species distributions and has become in one of the most widely applied tools for the assessment of the impact of climate change and human activities on species and communities, biological invasions, epidemiology, and conservation biology (Peterson et al. 2011; Guisan et al. 2017). However, despite important advances in theory (Soberón 2007; Colwell and Rangel 2009; Soberón and Nakamura 2009; Peterson et al. 2011), methods, and algorithms (reviewed in Duarte et al. 2019; see also Warren et al. 2018) and practical applications (Guillera-Arroita et al. 2015; Cord et al. 2017; Sanín and Anderson 2018), most studies still rely on the use of interpolated climate data as environmental predictors (Saatchi et al. 2008; Waltari et al. 2014). In this article we compare the performance of environmental data derived from interpolated climate surfaces data (i.e., WorldClim) and S-RS products data (i.e., LAI and NDVI). Specifically, using live oaks as a case study, we show the advances and potential caveats in using S-RS data in describing and predicting species-environment relationships. Overall, our analyses show that S-RS products perform, as well as products from interpolated climate surfaces as environmental predictors (Tables 9.2, 9.3, and 9.4), and indeed present quite similar results for both species environmental suitability and macroecological patterns (Figs. 9.5 and 9.6), similar to Paz et al. (Chap. 11). However, they have the potential to provide more precise estimates of species distributions at higher spatial resolution.
In our example, we used different grain sizes for both data sets: WorldClim (10 and 2.5 arcmin or ~18.5 and ~ 4.5 km at the equator, respectively) and S-RS products (3 arcmin or ~5.5 km at the equator). Although changing grain size in the predictors might be anticipated to affect model performance, different lines of evidence indicate that model performance is not affected by grain resolution, but rather by species response to the environmental conditions in the study region (Guisan et al. 2007, see also Fig. 9.5). Our results show that enhancing spatial resolution of interpolated climatic data does not improve the spatial resolution at which species distributions can be accurately predicted. The quality of interpolated climate surfaces such as WorldClim, which depends on climatic stations as data sources, has been ignored as a source of uncertainty in studies of species-environment relationships—for example, Hijmans et al. (2005) used a variable number of weather stations for their interpolations, 47.554, 24.542, and 14.930 for precipitation, mean temperature, and maximum and minimum temperature, respectively—especially in the tropics (Fig. 9.2c), where weather stations are sparse (Hijmans et al. 2005; Soria-Auza et al. 2010). This source of uncertainty can be avoided using S-RS products, which have continuous (from daily to monthly) and quasi-global environmental information, including precipitation, temperature, and biophysical variables that represent different components of vegetation and ecosystems (Funk et al. 2015; Cord et al. 2017; Radeloff et al. 2019).
Fine and broad spatial and temporal scale data derived from S-RS, which have only been available in the last ~20 years (Turner 2014), can be used to improve the evaluation of species-environment relationships. A number of research avenues remain to be pursued to better understand the potential of S-RS data and their products in quantifying species ecological niches and estimating species distributions. For example, applying the same framework presented here to other species or clades (including vertebrates and invertebrates) or applying more complex frameworks (e.g., Peterson and Nakazawa 2008; Waltari et al. 2014) may shed light on the potential of S-RS products as predictors for the analysis of species-environment relationships. This is important because ENMs/SDMs are used as predictive models that can be extrapolated across space and time to forecast and monitor biodiversity under a changing global climate (Peterson and Nakazawa 2008; Warren 2012).
9.4.1 Should We Use S-RS Data for ENM/SDM?
Whether S-RS data should replace other environmental data in modeling niches and projecting species distribution depends on the modeling purposes (Peterson et al. 2011). In fact, modeling species niches and projecting distributions involves relating a set of species occurrences to relevant environmental predictors. In essence, ENM/SDM based only on climatic variables would tend to return broad predictions (Coudun et al. 2006, see also right panel in Fig. 9.5), particularly because climatic data are useful in describing macroecological patterns of species distributions and communities (Lin and Wiens 2017; Manzoor et al. 2018), while ENM/SDM based on S-RS data alone allows the discrimination of local features not captured by climatic information (Coudun et al. 2006; Saatchi et al. 2008; Radeloff et al. 2019). Nevertheless, it seems that overall statistical model accuracy in this example is not improved (Pearson et al. 2004; Thuiller 2004; see also Table 9.2). Given that climatic and S-RS data provide information at different spatial and temporal scales, a promising option would be to use both sources of environmental predictors to model species distributions to achieve “the best of both worlds” (Saatchi et al. 2008; Pradervand et al. 2013).
More accurate predictions of species distributions are critical for the development of conservation and management actions if we are to meet the challenges posed by global change (Coudun et al. 2006; Zimmermann et al. 2007; Cord et al. 2013). Our point here is to facilitate and demonstrate the potential for the use of S-RS data for predicting species distributions and modeling environmental niches. The results we show here and those of others (e.g., Saatchi et al. 2008; Waltari et al. 2014) indicate that S-RS data provide a valuable complement to other environmental variables for ENM/SDM.
Another potential and important research direction is the use of S-RS products that have high temporal resolution, such as LAI (Fig. 9.3b), as biophysical variables that represent ecosystem functions (Cord et al. 2013, 2017). These products allow the exploration of dynamics of vegetation growth and seasonality in vegetation function, fundamental features that characterize vegetation form and function (Myneni et al. 2002; Hobi et al. 2017). Here, using metrics derived from MODIS LAI in combination with other S-RS products (Table 9.1), we show that, using relevant biophysical variables, it is possible to predict distributions similar to those predicted from climate data alone (Fig. 9.5, Tables 9.2 and 9.3). In fact, a recent study (Simões and Peterson 2018) found that including biotic predictors can improve ENMs even while increasing model complexity, such that the combination of abiotic and biotic predictors improves model performance (Simões and Peterson 2018). To confirm this conclusion, substantial effort would be needed, including new methodological and conceptual approaches, to disentangle the real contribution of S-RS products—spatial and temporal features of S-RS products that improve statistical model performance—as predictors of species distributions. Nonetheless, our results highlight an advance on the use of relevant predictors for modeling species-environment relationships.
In addition, recent macroecological studies have used these products to relate annual vegetation productivity to continental and global patterns of species richness (Pigot et al. 2016; Hobi et al. 2017; Coops et al. 2018), providing spatially explicit support for the use of satellite data products in predicting biodiversity. These advances point to an exciting avenue for the study of the distribution and assembly of biological communities (Ferrier and Guisan 2006). For example, S-RS products can be used for the development of stacked species distribution models (S-SDM, see Fig. 9.6) that can be integrated into novel biodiversity modeling frameworks, such as Spatially Explicit Species Assemblage Modelling (SESAM, Guisan and Rahbek 2011) or the Hierarchical Modelling of Species Communities (HMSC, Ovaskainen et al. 2017), aimed at predicting composition and distribution of species and communities (Mateo et al. 2017).
Finally, the example presented here is meant to spur further theoretical, methodological, and empirical research aimed at developing a Global Biodiversity Observatory (Geller et al., Chap. 20). Explicit incorporation of biotic information into species-environment modeling may turn our focus away from the use only of climatic information toward the “complete” evaluation of the drivers that determine the species distributions (Fig. 9.1).
9.4.2 Enabling Large-Scale Biodiversity Change Detection
Since the last millennium, rising human population and activity have been major drivers of environmental change on Earth, with consequences for the distribution and abundance of biodiversity and associated ecosystem functioning (Tilman 1997; Tylianakis et al. 2008). Thus, improving large-scale biodiversity change detection is crucial to the development of effective policies that advance conservation and management of species and communities.
In addition, enhancing predictive models of the species expected to be present in a given geographic region can be coupled with other means of detecting which species are present based on spectroscopic imaging (Serbin et al. 2015; Bolch et al., Chap. 12; Meireles et al., Chap. 7), reducing the complexity of species identification algorithms. For example, imaging spectroscopy allows mapping of functional traits, by estimating vegetation traits for each pixel in an image (Wang et al. 2019; Asner et al. 2017; Martin, Chap. 5). Plant spectra obtained from imaging spectroscopy at different spatial resolutions can in turn be used to detect different aspects and traits—within- and between-species differences in morphology, foliar chemistry, life history strategies—of plant species (Ustin and Gamon 2010; Cavender-Bares et al. 2017; Schweiger et al. 2018) and the correct identification of different taxonomic levels from populations to species to clades (Cavender-Bares et al. 2016). Thus, the integration of spectral approaches with techniques for modeling species ecological niches has the potential to produce reliable information of species distributions and co-occurrence, filling current gaps about species-environment relationships at a range of spatial scales and levels of organization—from species to communities—increasing the accuracy of direct detection assignments, and enabling monitoring of changes in biodiversity, one of the premises for the sustainable management of the biosphere (Pinto-Ledezma and Rivero 2014; Fernández et al., Chap. 18).
We thank Julián Velasco for his advice with ENM/SDM analyses. J.N.P-L. was supported by the University of Minnesota College of Biological Sciences’ Grand Challenges in Biology Postdoctoral Program. We acknowledge the bioDISCOVERY community at the University of Zürich for important discussions and the NSF Research Coordination Network “Biodiversity Across Scales.”
- Bivand R, Keitt T, Rowlingson B (2018) rgdal: bindings for the ‘geospatial’ data abstraction library. R package version 1.3-4. https://CRAN.R-project.org/package=rgdal.
- Duarte A, Whitlock SL, Peterson JT (2019) Species distribution modeling. In: Encyclopedia of Ecology, pp 189–198. https://doi.org/10.1016/b978-0-12-409548-9.10572-xCrossRefGoogle Scholar
- Elith J, Franklin J (2013) Species distribution modeling. In: Encyclopedia of biodiversity, pp 692–705. https://doi.org/10.1016/b978-0-12-384719-5.00318-xCrossRefGoogle Scholar
- Elton C (1927) Animal Ecology. Sedgwick and Jackson, LondonGoogle Scholar
- Greenberg JA, Mattiuzzi M (2018) gdalUtils: wrappers for the geospatial data abstraction library (GDAL) utilities. R package version 18.104.22.168. https://CRAN.R-project.org/package=gdalUtils.
- Grinnell J (1904) The Origin and Distribution of the Chest-Nut-Backed Chickadee. The Auk, 21(3), 364–382. https://doi.org/10.2307/4070199
- Grinnell J (1917) The Niche-Relationships of the California Thrasher. The Auk, 34(4), 427–433. https://doi.org/10.2307/4072271
- Guisan A, Rahbek C (2011) SESAM - a new framework integrating macroecological and species distribution models for predicting spatio-temporal patterns of species assemblages. Journal of Biogeography, 38(8), 1433–1444. https://doi.org/10.1111/j.1365-2699.2011.02550.x
- Hijmans RJ, Phillips S, Leathwick J et al (2017) dismo: species distribution modeling. R package version 1.1-4. https://CRAN.R-project.org/package=dismo.
- Hijmans RJ (2018) raster: Geographic Data Analysis and Modeling. R package version 3.0-7. https://CRAN.R-project.org/package=raster
- Hutchinson GE (1957) Concluding Remarks. Cold Spring Harbor Symposia on Quantitative Biology, 22(0), 415–427. https://doi.org/10.1101/sqb.1957.022.01.03
- Kissling WD, Dormann CF, Groeneveld J et al (2012) Towards novel approaches to modelling biotic interactions in multispecies assemblages at large spatial extents. J Biogeogr 39(12):2163–2178. https://doi.org/10.1111/j.1365-2699.2011.02663.x
- Liaw A, Wiener M (2002) Classification and regression by randomForest. R News 2:18–22Google Scholar
- Milborrow S (2016) Earth: multivariate adaptive regression splines. R package version 4.4.4. Retrieved from https://CRAN.R-project.org/package=earth.
- Nix HA (1986) A biogeographic analysis of Australian elapid snakes. In: Longmore R (ed) Atlas of elapid snakes of Australia: Australian Flora and Fauna Series, vol 7. Bureau of Flora and Fauna, Canberra, pp 4–15Google Scholar
- R Core Team (2018) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/
- Randin CF, Ashcroft MB, Bolliger J et al. (2020) Monitoring biodiversity in the Anthropocene using remote sensing in species distribution models. Remote Sens Environ 239:111626 https://doi.org/10.1016/j.rse.2019.111626
- Soberón J, Peterson AT (2005) Interpretation of models of fundamental ecological niches and species’ distributional areas. Biodivers Inform 2. https://doi.org/10.17161/bi.v2i0.4
- Sokal RR, Rohlf FJ (1995) Biometry: the principles and practice of statistics in biological research, 3rd edn. W. H. Freeman and Co., New YorkGoogle Scholar
- Wallace AR (1860) On the zoological geography of the Malay archipelago. Zool J Linnean Soc 4(16):172–184. https://doi.org/10.1111/j.1096-3642.1860.tb00090.xCrossRefGoogle Scholar
- West AM, Evangelista PH, Jarnevich CS et al (2016) Integrating remote sensing with species distribution models; mapping tamarisk invasions using the software for assisted habitat modeling (SAHM). J Vis Exp (116). https://doi.org/10.3791/54578
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.