Modeling Forest Species Distributions in a Human ... - Semantic Scholar
International Journal of Applied Geospatial Research, 4(3), 39-57, July-September 2013 39
Modeling Forest Species Distributions in a HumanDominated Landscape in Northeastern, USA Stephen McCauley, George Perkins Marsh Institute, Clark University, Worcester, MA, USA John Rogan, Graduate School of Geography, Clark University, Worcester, MA, USA Jennifer Miller, Department of Geography and the Environment, University of Texas, Austin, USA
ABSTRACT Mapping forest species distributions is notoriously difficult in human-dominated mixed temperate forest environments. The goal of this study is to evaluate the effectiveness of species distribution modeling techniques for estimating the distribution of forest canopy species on Cape Cod, Massachusetts, USA. Binary maps of estimated presence/absence were produced for four canopy species using a classification tree approach and a generalized linear modeling approach, both including and excluding a spatial dependence term, and the results are evaluated using several assessment measures. Secondary goals of the study are to examine the influence of past land use on species distributions at the landscape scale and to consider the effect of explicitly including information on spatial dependence. Findings suggest that these techniques are broadly applicable in such human-dominated landscapes, but that complex disturbance histories introduce significant challenges. Keywords:
Forest Composition, Human-Dominated Landscape, Massachusetts, Species Distribution Models, Temperate Forests
INTRODUCTION Mixed temperate forests are the dominant ecosystem in eastern North America, Europe, and East Asia (Defries & Townshend, 1999; Friedl et al., 2002; Hansen et al., 2003) and they represent one of the most altered biomes on the planet (Wade et al., 2003). Temperate
forests are expected to face significant further changes in composition related to global environmental change, including climate warming (Barbler et al., 2006; Iverson & Prassad, 1998), human-induced land change (Sala et al., 1986), and altered disturbance regimes (Dale et al., 2001). While these disturbance dynamics occur at landscape to global scales (Turner et
DOI: 10.4018/jagr.2013070103 Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
40 International Journal of Applied Geospatial Research, 4(3), 39-57, July-September 2013
al., 2001), the biophysical response to change is primarily a function of the physiological characteristics of individual tree species (Aber et al., 2001; Hansen et al., 2001). Monitoring and analysis of vegetative dynamics, therefore, requires characterization of mixed temperate forest composition at fine categorical detail over landscape and regional scales. Species distributions in temperate forest ecosystems are notoriously complex and difficult to map due to both ecological and instrumental reasons. Gradients in dominant tree species are often not significant at the scale of stands (Masaki et al., 1992) or even ecoregions (McDonald et al., 2005), as ubiquitous generalist species fill numerous niches (Brokaw & Busing, 2000). Meanwhile, the ability to distinguish individual forest species using medium or coarse resolution satellite data is severely constrained by the spectral similarity of species and the mixed species composition within individual grid cells (Metzler & Sader, 2005; Woodcock & Franklin, 1989; Woodcock et al., 1994). High-spatial and high-spectral resolution sensors allow for individual tree crown identification (Lamar et al., 2005; Leckie et al., 2005) and leaf chemical differentiation (Martin et al., 1997; Pontius et al., 2005), respectively, but the spatial extent of these efforts is severely limited by resource and time costs (Fuller, 2006; Wulder & Nelson, 2003). Species distribution modeling (SDM) is an alternative approach for estimating species distributions which relies on quantifying relationships between observed species distributions and environmental variables based on landscape ecological theory (Franklin, 1995). Advances in modeling techniques over the last decade, particularly through the integration of GIS platforms with semi-parametric and nonparametric modeling techniques, have expanded the range of applications using SDM from those involving individual charismatic species to more complex, landscape mapping exercises (Elith & Leathwick, 2009). Increasingly, the approach has been used for mapping species composition in complex temperate forest landscapes characterized by more shallow environmental
gradients and greater human impacts (Pearson et al., 2004; Tasser et al., 2007; Taverna et al., 2005). Species distribution models generally assume species are in equilibrium with their environments, or at least at a quasi-equilibrium where change is slow relative to the lifespan of the biota (Austin, 2007). However, land use conversion and other human alternations of ecosystems significantly alter species-environment relationships (Bellemere & Foster, 2002; Foster, 2002). The role of human activities in shaping the distribution of forest species in the northeastern US is well-established (Hall et al., 2002), though studies have generally been limited to plot level analyses and have not been used to generate landscape scale distribution estimates. European deforestation goes back much further in time with several phases of abandonment and reforestation (Hermy & Verheyen, 2007), and numerous studies have demonstrated the effects of human land use in shaping the complex ecological landscape in this region (Koerner et al., 1997; Thuiller et al., 2004). Other studies have demonstrated similar patterns in temperate forest regions around the world (Ito et al., 2004; Vila & Pujadas, 2001). Recruitment and dispersal lead to a new forest composition, and species establish a new quasi-equilibrium with the human-altered landscape (Benjamin et al., 2005; Hermy & Verheyen, 2007). Including past land use as a predictor variable in species distribution models in temperate regions therefore is critical. The Cape Cod region in Eastern Massachusetts presents many characteristic challenges associated with forest composition mapping in human-dominated mixed temperate forest ecosystems. The region’s vegetation is dominated by a pitch pine-oak community which is distributed widely across the region, with only small areas within the region not supporting this community. The limited range of climatic and landform variables across the study area can confound efforts to predict species distributions based on these variables (Fischer, 1990; Leathwick & Austin, 2001). Furthermore, the distribution of several canopy species owes
Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
17 more pages are available in the full version of this document, which may be purchased using the "Add to Cart" button on the product's webpage: www.igi-global.com/article/modeling-forest-speciesdistributions-human/77924?camid=4v1
This title is available in InfoSci-Journals, InfoSci-Journal Disciplines Engineering, Natural, and Physical Science. Recommend this product to your librarian: www.igi-global.com/e-resources/libraryrecommendation/?id=2
Related Content Similarities between Competitors and the Implications for Location Strategies Lawrence Joseph (2010). International Journal of Applied Geospatial Research (pp. 45-62).
www.igi-global.com/article/similarities-between-competitors-implicationslocation/46935?camid=4v1a Adopting BIM Standards for Managing Vision 2030 Infrastructure Development in Qatar Fatima Al Mohannadi, Mohammed Arif, Zeeshan Aziz and Phillip A. Richardson (2013). International Journal of 3-D Information Modeling (pp. 64-73).
www.igi-global.com/article/adopting-bim-standards-for-managing-vision2030-infrastructure-development-in-qatar/99618?camid=4v1a Describing and Selecting Collections of Georeferenced Media Items in Peerto-Peer Information Retrieval Systems Daniel Blank and Andreas Henrich (2013). Geographic Information Systems: Concepts, Methodologies, Tools, and Applications (pp. 666-684).
www.igi-global.com/chapter/describing-selecting-collections-georeferencedmedia/70469?camid=4v1a
Overview, Classification and Selection of Map Projections for Geospatial Applications Eric Delmelle and Raymond Dezzani (2009). Handbook of Research on Geoinformatics (pp. 89-98).
www.igi-global.com/chapter/overview-classification-selection-mapprojections/20391?camid=4v1a
Modeling Forest Species Distributions in a HumanDominated Landscape in Northeastern, USA Stephen McCauley, George Perkins Marsh Institute, Clark University, Worcester, MA, USA John Rogan, Graduate School of Geography, Clark University, Worcester, MA, USA Jennifer Miller, Department of Geography and the Environment, University of Texas, Austin, USA
ABSTRACT Mapping forest species distributions is notoriously difficult in human-dominated mixed temperate forest environments. The goal of this study is to evaluate the effectiveness of species distribution modeling techniques for estimating the distribution of forest canopy species on Cape Cod, Massachusetts, USA. Binary maps of estimated presence/absence were produced for four canopy species using a classification tree approach and a generalized linear modeling approach, both including and excluding a spatial dependence term, and the results are evaluated using several assessment measures. Secondary goals of the study are to examine the influence of past land use on species distributions at the landscape scale and to consider the effect of explicitly including information on spatial dependence. Findings suggest that these techniques are broadly applicable in such human-dominated landscapes, but that complex disturbance histories introduce significant challenges. Keywords:
Forest Composition, Human-Dominated Landscape, Massachusetts, Species Distribution Models, Temperate Forests
INTRODUCTION Mixed temperate forests are the dominant ecosystem in eastern North America, Europe, and East Asia (Defries & Townshend, 1999; Friedl et al., 2002; Hansen et al., 2003) and they represent one of the most altered biomes on the planet (Wade et al., 2003). Temperate
forests are expected to face significant further changes in composition related to global environmental change, including climate warming (Barbler et al., 2006; Iverson & Prassad, 1998), human-induced land change (Sala et al., 1986), and altered disturbance regimes (Dale et al., 2001). While these disturbance dynamics occur at landscape to global scales (Turner et
DOI: 10.4018/jagr.2013070103 Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
40 International Journal of Applied Geospatial Research, 4(3), 39-57, July-September 2013
al., 2001), the biophysical response to change is primarily a function of the physiological characteristics of individual tree species (Aber et al., 2001; Hansen et al., 2001). Monitoring and analysis of vegetative dynamics, therefore, requires characterization of mixed temperate forest composition at fine categorical detail over landscape and regional scales. Species distributions in temperate forest ecosystems are notoriously complex and difficult to map due to both ecological and instrumental reasons. Gradients in dominant tree species are often not significant at the scale of stands (Masaki et al., 1992) or even ecoregions (McDonald et al., 2005), as ubiquitous generalist species fill numerous niches (Brokaw & Busing, 2000). Meanwhile, the ability to distinguish individual forest species using medium or coarse resolution satellite data is severely constrained by the spectral similarity of species and the mixed species composition within individual grid cells (Metzler & Sader, 2005; Woodcock & Franklin, 1989; Woodcock et al., 1994). High-spatial and high-spectral resolution sensors allow for individual tree crown identification (Lamar et al., 2005; Leckie et al., 2005) and leaf chemical differentiation (Martin et al., 1997; Pontius et al., 2005), respectively, but the spatial extent of these efforts is severely limited by resource and time costs (Fuller, 2006; Wulder & Nelson, 2003). Species distribution modeling (SDM) is an alternative approach for estimating species distributions which relies on quantifying relationships between observed species distributions and environmental variables based on landscape ecological theory (Franklin, 1995). Advances in modeling techniques over the last decade, particularly through the integration of GIS platforms with semi-parametric and nonparametric modeling techniques, have expanded the range of applications using SDM from those involving individual charismatic species to more complex, landscape mapping exercises (Elith & Leathwick, 2009). Increasingly, the approach has been used for mapping species composition in complex temperate forest landscapes characterized by more shallow environmental
gradients and greater human impacts (Pearson et al., 2004; Tasser et al., 2007; Taverna et al., 2005). Species distribution models generally assume species are in equilibrium with their environments, or at least at a quasi-equilibrium where change is slow relative to the lifespan of the biota (Austin, 2007). However, land use conversion and other human alternations of ecosystems significantly alter species-environment relationships (Bellemere & Foster, 2002; Foster, 2002). The role of human activities in shaping the distribution of forest species in the northeastern US is well-established (Hall et al., 2002), though studies have generally been limited to plot level analyses and have not been used to generate landscape scale distribution estimates. European deforestation goes back much further in time with several phases of abandonment and reforestation (Hermy & Verheyen, 2007), and numerous studies have demonstrated the effects of human land use in shaping the complex ecological landscape in this region (Koerner et al., 1997; Thuiller et al., 2004). Other studies have demonstrated similar patterns in temperate forest regions around the world (Ito et al., 2004; Vila & Pujadas, 2001). Recruitment and dispersal lead to a new forest composition, and species establish a new quasi-equilibrium with the human-altered landscape (Benjamin et al., 2005; Hermy & Verheyen, 2007). Including past land use as a predictor variable in species distribution models in temperate regions therefore is critical. The Cape Cod region in Eastern Massachusetts presents many characteristic challenges associated with forest composition mapping in human-dominated mixed temperate forest ecosystems. The region’s vegetation is dominated by a pitch pine-oak community which is distributed widely across the region, with only small areas within the region not supporting this community. The limited range of climatic and landform variables across the study area can confound efforts to predict species distributions based on these variables (Fischer, 1990; Leathwick & Austin, 2001). Furthermore, the distribution of several canopy species owes
Copyright © 2013, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
17 more pages are available in the full version of this document, which may be purchased using the "Add to Cart" button on the product's webpage: www.igi-global.com/article/modeling-forest-speciesdistributions-human/77924?camid=4v1
This title is available in InfoSci-Journals, InfoSci-Journal Disciplines Engineering, Natural, and Physical Science. Recommend this product to your librarian: www.igi-global.com/e-resources/libraryrecommendation/?id=2
Related Content Similarities between Competitors and the Implications for Location Strategies Lawrence Joseph (2010). International Journal of Applied Geospatial Research (pp. 45-62).
www.igi-global.com/article/similarities-between-competitors-implicationslocation/46935?camid=4v1a Adopting BIM Standards for Managing Vision 2030 Infrastructure Development in Qatar Fatima Al Mohannadi, Mohammed Arif, Zeeshan Aziz and Phillip A. Richardson (2013). International Journal of 3-D Information Modeling (pp. 64-73).
www.igi-global.com/article/adopting-bim-standards-for-managing-vision2030-infrastructure-development-in-qatar/99618?camid=4v1a Describing and Selecting Collections of Georeferenced Media Items in Peerto-Peer Information Retrieval Systems Daniel Blank and Andreas Henrich (2013). Geographic Information Systems: Concepts, Methodologies, Tools, and Applications (pp. 666-684).
www.igi-global.com/chapter/describing-selecting-collections-georeferencedmedia/70469?camid=4v1a
Overview, Classification and Selection of Map Projections for Geospatial Applications Eric Delmelle and Raymond Dezzani (2009). Handbook of Research on Geoinformatics (pp. 89-98).
www.igi-global.com/chapter/overview-classification-selection-mapprojections/20391?camid=4v1a
