An accurate representation of the geographical distribution of species is central to ecological research and conservation science and practice. Species’ distributions can be represented using a variety of approaches: geographical ranges, which represent the geographical limits of distributions; point locality data, which represent species’ known occurrences; or inductive or deductive models, which usually represent species’ habitat within geographic ranges. Representations of distributions may contain false presences (commission errors) and/or false absences (omission errors). Recently, Area of Habitat (AOH) maps, a type of deductive model, have gained traction as a tool to represent global distribution of species, reducing the often high rate of commission errors in range maps. AOH models map the distribution of suitable habitat for a species inside its distributional limits. One of the key challenges in producing AOH maps is to translate knowledge of a species’ habitat (a complex and species-specific concept) into specific land-cover classes in existing land use/cover layers. Three different methods (expert-based crosswalk, translation table and global maps of terrestrial habitat types) have been developed to date to overcome this challenge to produce the AOH maps. (‘Crosswalk’ is a table translating habitat types in a Habitat Classification Scheme to land-cover classes in a land-cover layer.) However, the performance of these methods has not yet been tested. One of the key parts of modeling is validation of the model outputs. This is done by comparing the model output with real world observations, to quantify omission and commission errors in the models. The aim of this thesis is to produce and compare AOH models for terrestrial mammals and birds using different habitat mapping and validation methods. In the second chapter, I developed a map of global terrestrial habitat types based on the IUCN Red List Habitat Classification Scheme, and a novel method to estimate the omission and commission error of the global map of terrestrial habitat types using presence-only data of habitat specialist species downloaded from open repositories like GBIF (Global Biodiversity Information Facility), eBird (www.ebird.com), PREDICTS (Projecting Responses of Ecological Diversity In Changing Terrestrial Systems) and the IBA (Important Bird and Biodiversity Areas) dataset. To date, AOH maps have been validated using presence-only data for small subsets of species for different taxonomic groups, but no standard validation method exists for cases where absence data are not available. In Chapter 3, I developed a novel two-step validation protocol for AOH maps which includes first a model-based evaluation of model prevalence (i.e, the proportion of a species’ range that contains suitable habitat), and second a validation using species point localities (point prevalence) using presence-only data. I used 48,336,141 point localities for 4,889 bird species and 107,061 point localities for 420 mammal species. Where point prevalence exceeded model prevalence, the AOH was taken to be a better reflection of species’ distribution than random. In Chapter 4, I used the global map of terrestrial habitat types to produce AOH maps for 10,651 terrestrial birds and 4,581 terrestrial mammals. I then applied the validation protocol developed in Chapter 3 to AOH maps of terrestrial birds and mammals produced using translation table and global maps of terrestrial habitat types. I found that the average model prevalence for AOH maps produced using the global map of terrestrial habitat type was lower (0.55±0.28 for birds and 0.51±0.29 for mammals) than those produced using the translation table (0.64±27 for birds and 0.65±0.28 for mammals). This led to higher omission errors in the AOH maps produced using the global map of terrestrial habitat types. Also, the number of AOH maps which were better than random was higher in the AOH mapset produced using the translation table. I also found a high similarity between these two sets of maps (53.44% mapped as suitable and 23.22% mapped as unsuitable in both datasets for birds and 58% mapped as suitable and 19% mapped as unsuitable in both datasets for mammals). Each AOH map produced using the global map of terrestrial habitat types was effectively a subset of the equivalent AOH map produced using the translation table, because the former was based on a single map for each habitat type, whereas the latter was based on one-to-many relationships between habitat types and land-cover classes. I conclude that, overall, AOH maps based on the translation table are more robust than AOH maps based on the global map of terrestrial habitat types in terms of reducing commission errors of the geographic range maps without introducing large omission errors. However, for species occurring primarily in human- modified habitats, the AOH maps based on the global map of terrestrial habitat types are more robust as few human-modified habitats are not mapped by the translation table but are mapped in the global map of terrestrial habitat types. The AOH modeling and validation methods developed in this thesis can help update the AOH maps in the future with latest data on land-cover, habitat and elevation. Furthermore, the validation metrics can be used as a guideline by the users to select the most appropriate AOH map for their use.

Production and validation of Area of Habitat maps for terrestrial birds and mammals

DAHAL, PRABHAT RAJ
2022-05-26

Abstract

An accurate representation of the geographical distribution of species is central to ecological research and conservation science and practice. Species’ distributions can be represented using a variety of approaches: geographical ranges, which represent the geographical limits of distributions; point locality data, which represent species’ known occurrences; or inductive or deductive models, which usually represent species’ habitat within geographic ranges. Representations of distributions may contain false presences (commission errors) and/or false absences (omission errors). Recently, Area of Habitat (AOH) maps, a type of deductive model, have gained traction as a tool to represent global distribution of species, reducing the often high rate of commission errors in range maps. AOH models map the distribution of suitable habitat for a species inside its distributional limits. One of the key challenges in producing AOH maps is to translate knowledge of a species’ habitat (a complex and species-specific concept) into specific land-cover classes in existing land use/cover layers. Three different methods (expert-based crosswalk, translation table and global maps of terrestrial habitat types) have been developed to date to overcome this challenge to produce the AOH maps. (‘Crosswalk’ is a table translating habitat types in a Habitat Classification Scheme to land-cover classes in a land-cover layer.) However, the performance of these methods has not yet been tested. One of the key parts of modeling is validation of the model outputs. This is done by comparing the model output with real world observations, to quantify omission and commission errors in the models. The aim of this thesis is to produce and compare AOH models for terrestrial mammals and birds using different habitat mapping and validation methods. In the second chapter, I developed a map of global terrestrial habitat types based on the IUCN Red List Habitat Classification Scheme, and a novel method to estimate the omission and commission error of the global map of terrestrial habitat types using presence-only data of habitat specialist species downloaded from open repositories like GBIF (Global Biodiversity Information Facility), eBird (www.ebird.com), PREDICTS (Projecting Responses of Ecological Diversity In Changing Terrestrial Systems) and the IBA (Important Bird and Biodiversity Areas) dataset. To date, AOH maps have been validated using presence-only data for small subsets of species for different taxonomic groups, but no standard validation method exists for cases where absence data are not available. In Chapter 3, I developed a novel two-step validation protocol for AOH maps which includes first a model-based evaluation of model prevalence (i.e, the proportion of a species’ range that contains suitable habitat), and second a validation using species point localities (point prevalence) using presence-only data. I used 48,336,141 point localities for 4,889 bird species and 107,061 point localities for 420 mammal species. Where point prevalence exceeded model prevalence, the AOH was taken to be a better reflection of species’ distribution than random. In Chapter 4, I used the global map of terrestrial habitat types to produce AOH maps for 10,651 terrestrial birds and 4,581 terrestrial mammals. I then applied the validation protocol developed in Chapter 3 to AOH maps of terrestrial birds and mammals produced using translation table and global maps of terrestrial habitat types. I found that the average model prevalence for AOH maps produced using the global map of terrestrial habitat type was lower (0.55±0.28 for birds and 0.51±0.29 for mammals) than those produced using the translation table (0.64±27 for birds and 0.65±0.28 for mammals). This led to higher omission errors in the AOH maps produced using the global map of terrestrial habitat types. Also, the number of AOH maps which were better than random was higher in the AOH mapset produced using the translation table. I also found a high similarity between these two sets of maps (53.44% mapped as suitable and 23.22% mapped as unsuitable in both datasets for birds and 58% mapped as suitable and 19% mapped as unsuitable in both datasets for mammals). Each AOH map produced using the global map of terrestrial habitat types was effectively a subset of the equivalent AOH map produced using the translation table, because the former was based on a single map for each habitat type, whereas the latter was based on one-to-many relationships between habitat types and land-cover classes. I conclude that, overall, AOH maps based on the translation table are more robust than AOH maps based on the global map of terrestrial habitat types in terms of reducing commission errors of the geographic range maps without introducing large omission errors. However, for species occurring primarily in human- modified habitats, the AOH maps based on the global map of terrestrial habitat types are more robust as few human-modified habitats are not mapped by the translation table but are mapped in the global map of terrestrial habitat types. The AOH modeling and validation methods developed in this thesis can help update the AOH maps in the future with latest data on land-cover, habitat and elevation. Furthermore, the validation metrics can be used as a guideline by the users to select the most appropriate AOH map for their use.
File allegati a questo prodotto
File Dimensione Formato  
Tesi_dottorato_Dahal.pdf

accesso aperto

Tipologia: Tesi di dottorato
Licenza: Creative commons
Dimensione 3.84 MB
Formato Adobe PDF
3.84 MB Adobe PDF Visualizza/Apri PDF

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11573/1652626
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact