**3.1. A synopsis of published global biodiversity mapping**

Generally, we found little congruence between global mapping of biodiversity and global mapping of freshwater habitats (**Table 4**). Only two studies in **Table 4** used spatial products from recent global habitat mapping efforts [19, 72]. Richman et al. [19] summarized crayfish range maps from IUCN and georeferenced occurrences (from experts) in Hydro1K basins [28] to examine factors responsible for their decline. All but one of the studies outlined in **Table 4** have been published within the last 15 years, and opposite as expected, species mapping efforts do not display a clear trend of increasing spatial granularity over time. In contrast, studies seem to summarize biogeographical information at the coarsest scales sufficient to achieving their purpose, which in most cases, was related to examining declines in species and threats to their existence. Spatial resolutions of freshwater species mapping ranged from biogeographic regions and range estimates (polygons) to 96-km2 gridded cells and small watersheds (e.g., Hydro1K).



methods of detection, and are not spatially comprehensive. Hence, extending species presences into unsampled areas requires various levels of inference ranging from summarization into regions or watersheds to sophistical statistical models predicting probability of presence using a suite of predictor variables characterizing habitat. Obviously, the first approach requires less resources and information, whereas the latter approach requires rich informa-

Generally, we found little congruence between global mapping of biodiversity and global mapping of freshwater habitats (**Table 4**). Only two studies in **Table 4** used spatial products from recent global habitat mapping efforts [19, 72]. Richman et al. [19] summarized crayfish range maps from IUCN and georeferenced occurrences (from experts) in Hydro1K basins [28] to examine factors responsible for their decline. All but one of the studies outlined in **Table 4** have been published within the last 15 years, and opposite as expected, species mapping efforts do not display a clear trend of increasing spatial granularity over time. In contrast, studies seem to summarize biogeographical information at the coarsest scales sufficient to achieving their purpose, which in most cases, was related to examining declines in species and threats to their existence. Spatial resolutions of freshwater species mapping ranged from

Major drainage basins

Major drainage basins

Freshwater ecoregions

(n = 325)

(n = 397)

Nature lake features Multiple published

(n = 292)

gridded cells and small

Multiple published sources

sources; International Lake Environment Committee Foundation (ILEC) global lake database [65]

Oberdorff et al. 1995 [63];

FishBase [67]

Multiple

tion on descriptions of habitat, not just the features themselves.

biogeographic regions and range estimates (polygons) to 96-km2

richness patterns across continents and show that species-area and speciesenergy relationships explain most of the

Developed models of fish species richness from natural lake characteristics

model to simulate scenarios of future fish species loss with losses in river discharge from climate change and withdrawal

global biogeographic regionalization of Earth's freshwater systems based on composition of freshwater fish species

variation

**Source Description Spatial resolution Source**

watersheds (e.g., Hydro1K).

68 Pure and Applied Biogeography

Amarasinghe & Welcome

Oberdorff et al. [63] Analyze fish species

Xenopoulos et al. [66] Use global hydrologic

Abell et al. [10] Developed first

**Fish**

[64]

**3.1. A synopsis of published global biodiversity mapping**



**Source Description Spatial resolution Source**

20

Range maps; point distributions

Biogeographic regions

Continents Multiple

Cell IUCN [73]

Cell IUCN [73]

grids Multiple

(n = 6)

0.50

0.250

96.3 km2

Cell Multiple

Geographic regions (n = 32) MUSSEL Project [84]

HydroIK river basins IUCN; expert georeference

collection efforts

Global Amphibian Assessment (IUCN) [75]; GBIF [81]; Check List Online Journal [82]

Multiple published sources

future interactions of climate change, land-use change, and spread of the pathogenic fungal disease chytridiomycosis on amphibian species declines

global range maps for

and global diversity of freshwater mussel species

diversity in crayfish

responsible for global declines in crayfish

protected areas in representing species diversity (includes amphibians, mammals, birds, turtles).

assessing the extent of protected land coverage for representation of biodiversity including amphibians, mammals, freshwater turtles and tortoises, and globally threatened birds

and commonalities in biodiversity and rare and threatened species among amphibians, mammals,

and birds

biodiversity loss in mussels including threats and

amphibians

solutions

Hof et al. [79] Assess the current and

70 Pure and Applied Biogeography

Ficetola et al. [80] Assessment of error in

Graf and Cummings [83] Review of systematics

Nobles and Zhang [85] Assessment of global

Crandall and Buhay [86] Description of global

Richman et al. [19] Evaluation of factors

Rodrigues et al. [87] Examination of global

Rodrigues et al. [88] Global gap analysis

Grenyer et al. [89] Examine congruence

**Mussels**

**Crayfish**

**Multiple taxa**

**Table 4.** Examples of studies developing or utilizing global freshwater biogeography databases.

In most cases, global mapping of biodiversity has been achieved by summarizing occurrence or estimated range information into spatial units as opposed to developing predictive species distribution models (SDMs) (**Table 4**). There are, however, several global-scale species modeling efforts, many of which are provided as interactive online resources (see following sections). Of freshwater taxa, amphibians and fish mapping efforts have been documented more than crayfish and mussels (**Table 4**), possibly because more vertebrate species have been described and more is known about the details of their life histories, habitat requirements, and conservation status. Additionally, global mapping efforts for amphibians are more common because of the wealth of data for that taxa. In particular, the Global Amphibian Assessment conducted by the International Union for the Conservation of Nature (IUCN) produced polygon range maps for >6000 known amphibian species [75] (**Figure 8**) and was used in six different studies (**Table 4**). The IUCN provides similar spatial data for mammals, reptiles, and marine and freshwater taxa [73]. The range maps are many times converted to gridded raster datasets [74] (**Figure 8**) or overlapped with region polygons to provide summaries of species within those areas (e.g., [76]).

The IUCN recently produced a set of higher-resolution global maps of ranges of freshwater taxa (IUCN) within HydroBasins (240,000 basins globally) [12] (**Figure 9**). One study relied on this resource to examine spatial relationships between fish biodiversity and planned hydropower dam construction in three large basins of the world [72]. The authors suggested that site selection for dams not be conducted purely on the grounds of energy, but should be conducted strategically through tradeoff analyses to conserve the most biodiversity while financing new dams. The IUCN data is currently the best openly available global information on freshwater species occurrences, but has many gaps in spatial coverage (e.g., **Figure 9**). While the Congo and Mekong River (China) basins had sufficient information at the resolution of HydroBasins, the Amazon Basin did not have comprehensive biodiversity mapping at that resolution; hence, reference [72] relied on biodiversity estimates in Freshwater Ecoregions [10], a far coarser alternative. The Amazon basin is over 7 million km2 yet only contains 13 Freshwater Ecoregions. Obviously, for conservation purposes, higher-resolution granularity is required to inform dam site selection in many areas of the globe. To compensate for lack of knowledge in many areas of the world, other mapping efforts have relied on published resources to compile freshwater species lists within regions or basins [63, 70]. While these resources can fill in important knowledge gaps, they are coarse (presented at the resolution of large basins) and leave large regions of the globe vacant of information (**Figure 9**).

**Figure 8.** Global amphibian richness from the International Union for the Conservation of Nature (IUCN) Global Amphibian Assessment.

A Synopsis of Global Mapping of Freshwater Habitats and Biodiversity: Implications... http://dx.doi.org/10.5772/intechopen.70296 73

**Figure 9.** Global maps of fish richness provided by the IUCN [73] and Bross [70].

**Figure 8.** Global amphibian richness from the International Union for the Conservation of Nature (IUCN) Global

wealth of data for that taxa. In particular, the Global Amphibian Assessment conducted by the International Union for the Conservation of Nature (IUCN) produced polygon range maps for >6000 known amphibian species [75] (**Figure 8**) and was used in six different studies (**Table 4**). The IUCN provides similar spatial data for mammals, reptiles, and marine and freshwater taxa [73]. The range maps are many times converted to gridded raster datasets [74] (**Figure 8**) or overlapped with region polygons to provide summaries of species within those areas (e.g., [76]). The IUCN recently produced a set of higher-resolution global maps of ranges of freshwater taxa (IUCN) within HydroBasins (240,000 basins globally) [12] (**Figure 9**). One study relied on this resource to examine spatial relationships between fish biodiversity and planned hydropower dam construction in three large basins of the world [72]. The authors suggested that site selection for dams not be conducted purely on the grounds of energy, but should be conducted strategically through tradeoff analyses to conserve the most biodiversity while financing new dams. The IUCN data is currently the best openly available global information on freshwater species occurrences, but has many gaps in spatial coverage (e.g., **Figure 9**). While the Congo and Mekong River (China) basins had sufficient information at the resolution of HydroBasins, the Amazon Basin did not have comprehensive biodiversity mapping at that resolution; hence, reference [72] relied on biodiversity estimates in Freshwater Ecoregions

Freshwater Ecoregions. Obviously, for conservation purposes, higher-resolution granularity is required to inform dam site selection in many areas of the globe. To compensate for lack of knowledge in many areas of the world, other mapping efforts have relied on published resources to compile freshwater species lists within regions or basins [63, 70]. While these resources can fill in important knowledge gaps, they are coarse (presented at the resolution of

large basins) and leave large regions of the globe vacant of information (**Figure 9**).

yet only contains 13

[10], a far coarser alternative. The Amazon basin is over 7 million km2

Amphibian Assessment.

72 Pure and Applied Biogeography

#### **3.2. What is limiting global high-resolution freshwater species distribution models (SDMs)?**

Although many of the world's freshwater species lack formal description, are prone to misidentification, and have few georeferenced occurrences, databases of species observations and species characteristics are growing rapidly. For example, the Global Biodiversity Information Facility (GBIF) currently has over 730 million occurrences for over 1.64 million species and harnesses global community participation [81]. GBIF operates through more formal data publishing, whereas other databases, such as iSPOT [96] provides a platform for crowd-sourced species observations. Additionally, rich databases on species ecology and conservation status have emerged to assist with linking biodiversity with their global freshwater habitat requirements [67, 93]. The wealth of information from georeferenced occurrence databases and descriptive databases suggests that global freshwater biodiversity SDM efforts are not limited by observations, but the inability to extrapolate occurrences to fine-grain freshwater habitats via distribution modeling. This is not to suggest that global freshwater biodiversity SDM efforts are completely absent. Indeed, novel web tools are available to enable users to perform their own SDM projections, both current and future. The Life Mapper project is an online resource that utilizes GBIF observations and global climate, terrain and land cover information to model the current and future distributions of species (including freshwater) [97]. Models of current ranges of species and habitat specifications are calibrated based on existing observations and climate information and used to model future potential ranges based on four climate scenarios spanning 2050 and 2070, according to the International Panel on Climate Change (**Figure 10**). As another example, AquaMaps uses a simplistic "environmental-envelope" method to develop large-scale predictions of marine and freshwater species occurrences [98, 99]. Occurrence data are obtained from GBIF and literature available through FishBase and summarized within bounding basins to constrain subsequent projections of distribution to only natural ranges. Occurrence data are overlain with eight environmental parameters to create an envelope of environmental suitability, which is essentially using the percent of observations (percentiles) in conjunction with local habitat conditions to estimate probability of occurrence [98]. Environmental envelopes are then used to model probabilities of species occurrence based on local conditions. Both the Life Mapper project and Aquamaps are freely available and are a quick approach to developing distribution maps; however, they are still relatively coarse projections, currently set at 10 arc-minutes and 0.5° (30 arc-second) cells, respectively, and do not approximate freshwater habitat features.

We suggest that the current leading limitation of achieving high-resolution global freshwater biodiversity mapping efforts has been a matter of limiting global habitat characteristic data, as opposed to limitations in occurrence data. Even if occurrences for a species are limited, current modeling approaches (e.g., Maxent) are capable of developing SDMs with low sample sizes [100]. By high-resolution, we are referring to the spatial granularity that approximates that of global freshwater habitat features. Recent developments have produced high-resolution depictions of freshwater features in the landscape, but much of these features have little accompanying information on habitat requirements for species (e.g., temperature, hydrology, depth, etc). One exception is a database on world lakes (n = 217) provided by the International Lake Environment Committee Foundation (ILEC), which includes location, morphometric features, climate, water quality, and edaphic variables [65]. This provided an opportunity to model fish species richness in selected natural lakes across the globe [64].

In comparison to terrestrial ecosystems, habitats within freshwater systems are shaped by upstream hydrologic processes, which require sophisticated geospatial summarization methods for appropriate characterization. For example, suppose air temperature is being used as a surrogate of water temperature in a fish species distribution model at the resolution of stream reaches or small watersheds. In this case, air temperature summarized at the location of the individual stream reach is unlikely to be representative of actual water temperature conditions. In contrast, using stream network routing to accumulate air temperature values for the entire upstream drainage network of each reach would be more representative [35]. Until recently, this type of habitat characterization was globally unavailable to support high-resolution freshwater species distributions. A near-global dataset summarizing 324 layers describing climate, land cover, topography, geology, and soils was recently developed for upstream drainage network of HydroSHEDs river reaches [101]. For the US, a comparable dataset is the NHD plus system (1:24K scale), which provides climate, hydrology, and land-use information summarized within the entire upstream network above each stream reach. Many freshwater species distribution modeling efforts have utilized the NHDplus data (1:24k) and architecture A Synopsis of Global Mapping of Freshwater Habitats and Biodiversity: Implications... http://dx.doi.org/10.5772/intechopen.70296 75

information to model the current and future distributions of species (including freshwater) [97]. Models of current ranges of species and habitat specifications are calibrated based on existing observations and climate information and used to model future potential ranges based on four climate scenarios spanning 2050 and 2070, according to the International Panel on Climate Change (**Figure 10**). As another example, AquaMaps uses a simplistic "environmental-envelope" method to develop large-scale predictions of marine and freshwater species occurrences [98, 99]. Occurrence data are obtained from GBIF and literature available through FishBase and summarized within bounding basins to constrain subsequent projections of distribution to only natural ranges. Occurrence data are overlain with eight environmental parameters to create an envelope of environmental suitability, which is essentially using the percent of observations (percentiles) in conjunction with local habitat conditions to estimate probability of occurrence [98]. Environmental envelopes are then used to model probabilities of species occurrence based on local conditions. Both the Life Mapper project and Aquamaps are freely available and are a quick approach to developing distribution maps; however, they are still relatively coarse projections, currently set at 10 arc-minutes and 0.5° (30 arc-second)

We suggest that the current leading limitation of achieving high-resolution global freshwater biodiversity mapping efforts has been a matter of limiting global habitat characteristic data, as opposed to limitations in occurrence data. Even if occurrences for a species are limited, current modeling approaches (e.g., Maxent) are capable of developing SDMs with low sample sizes [100]. By high-resolution, we are referring to the spatial granularity that approximates that of global freshwater habitat features. Recent developments have produced high-resolution depictions of freshwater features in the landscape, but much of these features have little accompanying information on habitat requirements for species (e.g., temperature, hydrology, depth, etc). One exception is a database on world lakes (n = 217) provided by the International Lake Environment Committee Foundation (ILEC), which includes location, morphometric features, climate, water quality, and edaphic variables [65]. This provided an opportunity to

In comparison to terrestrial ecosystems, habitats within freshwater systems are shaped by upstream hydrologic processes, which require sophisticated geospatial summarization methods for appropriate characterization. For example, suppose air temperature is being used as a surrogate of water temperature in a fish species distribution model at the resolution of stream reaches or small watersheds. In this case, air temperature summarized at the location of the individual stream reach is unlikely to be representative of actual water temperature conditions. In contrast, using stream network routing to accumulate air temperature values for the entire upstream drainage network of each reach would be more representative [35]. Until recently, this type of habitat characterization was globally unavailable to support high-resolution freshwater species distributions. A near-global dataset summarizing 324 layers describing climate, land cover, topography, geology, and soils was recently developed for upstream drainage network of HydroSHEDs river reaches [101]. For the US, a comparable dataset is the NHD plus system (1:24K scale), which provides climate, hydrology, and land-use information summarized within the entire upstream network above each stream reach. Many freshwater species distribution modeling efforts have utilized the NHDplus data (1:24k) and architecture

cells, respectively, and do not approximate freshwater habitat features.

74 Pure and Applied Biogeography

model fish species richness in selected natural lakes across the globe [64].

**Figure 10.** Life map projections of Brook Trout (*Salvelinus fontinalis*) and Appalachian Brook Crayfish (*Cambarus bartonii*) (f-j) distributions for current conditions and future climate projections for 2050 and 2070 under low (4.5 W/m2 ) and high (8.5 W/m2 ) IPCC representative concentration pathways (RCPs) for radiative forcing levels related to projected greenhouse gas emissions scenarios. Green points represent GBIF occurrences.

because of topological connectivity and habitat predictors offered by the resource [102–107] (**Figure 11**). Although NHDplus is a convenient database to support freshwater species distribution modeling, it does not adequately represent 1st order streams, the majority of which provide habitat for freshwater taxa (**Figure 11**). The NHD High resolution database (1:100k) represents smaller stream systems, but does not provide pre-summarized habitat information.

**Figure 11.** Species distribution model (SDM) developed for Largescale stoneroller (*Campostoma oligolepis*) in the Ridge and Valley and the Southern Appalachian Plateau Ecoregions of the Tennessee River Basin, USA. SDMs are generated for NHDPlus (1:100k) stream reaches and do not account for occurrences in NHD High-resolution stream reaches (smaller gray lines).

For this reason, other studies have developed their own reach datasets with accumulated habitat variables to support freshwater SDMs at resolution comparable to the NHD high-resolution dataset [108].
