**2. Global freshwater habitat mapping efforts**

Global estimates of freshwater ecosystem coverages have been developed through both theoretical [20] or empirical means [21], or a combination of both [11]. Theoretical constructs, for example, might assume relationships between the size, distribution, and bifurcation of rivers (i.e., network theory) to quantify size and distribution of rivers within a region [20]. Likewise, theoretical relationships of size versus distribution are commonly used to estimate the frequency and size of unobserved waterbodies [22]. In contrast, empirical estimates typically rely on spatial observations from remote sensing data. Because the geospatial representation of waterbodies is limited to the spatial fidelity of mapping efforts, the number and areas of waterbodies provided through empirical observation is consistently smaller than that estimated theoretically. This comparison is important, however, in that it yields insights into the current state (i.e., comprehensiveness and granularity) of global freshwater mapping efforts. In the following sections, we review and compare approaches to obtaining global scale estimates of three different freshwater ecosystem types: rivers and streams, lakes and reservoirs, and wetlands. Estimation methods and datasets vary for each of these aquatic ecosystem types and influence their respective global estimates. We also devote particular attention to trends in freshwater mapping efforts within the United States.

#### **2.1. River and streams**

**1. Introduction**

58 Pure and Applied Biogeography

habitat mapping.

Our knowledge of Earth's ecosystems and biodiversity is growing at rates that exceed our ability to accurately predict regional species pools [1]. Recent estimates of Earth's biodiversity suggest that the planet boasts a total of 8.7 million species, 87% of which are yet to be described [2]. Yet while our comprehension of the magnitude and appreciation of species diversity grows, many have suggested we are currently within the Earth's six mass extinction event [3, 4], in which rates of species loss are unprecedented compared to past extinction events. Indeed, cataloguing biodiversity is a catalyst for global conservation efforts. The International Union for the Conservation of Nature (IUCN) has assessed over 77,300 species, of which 29,530 (38%) are classified as threatened, endangered, or critically endangered, and >10,000 more (13%) species listed as vulnerable [5]. While only 0.01% of Earth's surface water occurs in rivers, lakes, and swamps, >126,000 (7%) of the Earth's described species are found in freshwaters [6, 7]. Therefore, freshwater species especially are in serious jeopardy of extinction. Dudgeon et al.'s [6] review of threats and conservation challenges to global freshwater biodiversity came at a much-needed time and addressed information gaps limiting our knowledge of these systems. The authors suggested (correctly) that there was no global comprehensive analysis of freshwater biodiversity comparable to those conducted for terrestrial systems [8]. Additionally, there was no comprehensive mapping of inland waters. The lack of this information prohibited our collective ability to inform large-scale conservation and prioritizing species and habitat protection. Since that time, many have answered the call to map global freshwater habitats and biodiversity to inform large-scale conservation. Just 2 years later, in 2008, the first seamless high-resolution map of global river hydrography was developed [9], and the first

global biogeographical regionalization of freshwater biodiversity was completed [10].

In more recent years, significant advances in mapping aquatic habitats—specifically rivers, lakes, and wetlands—have been made at the global scale (e.g., [11–13]). Much of the progress in spatially depicting freshwater ecosystems has been the result of new globally comprehensive remote sensing technologies [13], but also significant efforts by scientists to collate disparate data sources [14]. As new datasets and geospatial products emerge with increasing spatial resolution, estimates of the spatial extent and importance of freshwater ecosystems in global biogeochemical cycles have also increased [15–17]. While efforts to develop comprehensive inventories and maps of the distribution of the world's freshwater fauna have dramatically increased [18, 19], these efforts have remained separate from those of freshwater

Herein, we briefly review the status and recent history of global mapping of freshwater habitats, their biodiversity, and human disturbances. First, we provide an overview of the efforts and datasets to empirically map rivers, lakes, reservoirs, and wetlands at the global scale, and compare these to theoretical estimates of the spatial coverage of unobserved features. This provides an assessment of the accuracy and comprehensiveness of global freshwater habitat mapping. Secondly, we discuss the current state of global freshwater biodiversity mapping and provide sources of information and various approaches used. We compare the spatial scales and resolution of biodiversity and freshwater habitat mapping to identify potential Global estimates of river and stream mileage and area range widely, with aerial estimates provided more frequently than distances. The latest and largest estimates of river length and area are over 88.3 million km and 662,100 km2 , respectively [20]. To provide these estimates, Downing et al. [20] used two approaches, one reliant on stream network theory and empirical data on stream widths and the other estimating the fraction of continental area occupied by streams while correcting for the unresolved small stream portion. The authors first estimated global river number, length, and area according to stream order by relying on relying on river geometry and scaling laws [23, 24] and known bifurcation ratios and stream length-order equations [25]. Stream widths among different order streams were obtained from literature or aerial imagery and applied to the number and lengths of streams. In the second method, estimates of the fraction of river area per land for well-studied landscapes were extrapolated to the global land area, which led to a very close second approximation, 640,400 km2 .

Empirical estimates of global river length and area from mapping efforts are far less than the maximum theoretical estimates [20]. The Digital Chart of the World (DCW) estimates global stream length at 16.6 million km [26, 27]. HydroSheds (basins and stream networks) were developed from global digital elevation models (DEMs) which increased the estimate to 27.3 million km (derived from 15 arc-second resolution) (**Figure 1**) [9]. The Hydro1K database is currently the highest resolution empirical estimates of global stream length [28], which constitutes 53% of the highest theoretical estimates [20]. Previous estimates of global river area range from 360,000 to 510,000 km2 (**Table 1**). The Global Lakes and Wetlands Database (GLWD) is a compilation of at least 17 different datasets of regional to global registers, inventories, and digital maps according to different spatial extents [21]. Their estimate of 360,000 km2 of global river area was dependent upon aerial and satellite imagery of >5th order rivers and streams [20].

The spatial distribution and quantification of global river and stream mileage is limited to the resolution of widespread DEMs and, in turn, derived stream networks [31, 32]. Increased spatial resolution [33] and new algorithms for deriving stream networks [31] have continually increased the accuracy of spatial representations of global rivers (**Figures 1** and **2**). The finest resolution of consistent global-extent elevation grids is >90 m [9, 28], which will grossly underrepresent small stream systems. According to the DCW, the length of streams and rivers within the conterminous-US (CONUS) totals 727,326 km (almost 29,000 reaches) whereas the HydroSheds database (15 arc-second) estimates the same distance as almost 1.9 million km (238,405 reaches) (**Figure 3**). In contrast, the total mileage is 5.7 million km (2.98 million reaches) according to the NHD plus medium resolution dataset (1:100k scale) [34], which was constructed on the basis of 30-m DEM resolution [35]. The NHD High-Resolution Dataset (1:24k scale), however, estimates stream length for the CONUS at 1.2 million km (**Figure 3**) [36]. While mapping perennial systems seems straightforward, accurately mapping ephemeral systems from flow accumulation thresholds is difficult. Even the NHDplus dataset under-represents the small headwater systems apparent in the high-resolution National Hydrography Dataset (1:24k scale), which also under-represents potential ephemeral systems (**Figure 2**).

**Figure 1.** HydroSHED 15s basin boundaries (left). Example of improved accuracy of rivers mapped in HydroSHEDs 15s versus the Digital Chart of the World in the Congo River Basin, Africa.


stream length at 16.6 million km [26, 27]. HydroSheds (basins and stream networks) were developed from global digital elevation models (DEMs) which increased the estimate to 27.3 million km (derived from 15 arc-second resolution) (**Figure 1**) [9]. The Hydro1K database is currently the highest resolution empirical estimates of global stream length [28], which constitutes 53% of the highest theoretical estimates [20]. Previous estimates of global river

(GLWD) is a compilation of at least 17 different datasets of regional to global registers, inventories, and digital maps according to different spatial extents [21]. Their estimate of 360,000 km2 of global river area was dependent upon aerial and satellite imagery of >5th order rivers and

The spatial distribution and quantification of global river and stream mileage is limited to the resolution of widespread DEMs and, in turn, derived stream networks [31, 32]. Increased spatial resolution [33] and new algorithms for deriving stream networks [31] have continually increased the accuracy of spatial representations of global rivers (**Figures 1** and **2**). The finest resolution of consistent global-extent elevation grids is >90 m [9, 28], which will grossly underrepresent small stream systems. According to the DCW, the length of streams and rivers within the conterminous-US (CONUS) totals 727,326 km (almost 29,000 reaches) whereas the HydroSheds database (15 arc-second) estimates the same distance as almost 1.9 million km (238,405 reaches) (**Figure 3**). In contrast, the total mileage is 5.7 million km (2.98 million reaches) according to the NHD plus medium resolution dataset (1:100k scale) [34], which was constructed on the basis of 30-m DEM resolution [35]. The NHD High-Resolution Dataset (1:24k scale), however, estimates stream length for the CONUS at 1.2 million km (**Figure 3**) [36]. While mapping perennial systems seems straightforward, accurately mapping ephemeral systems from flow accumulation thresholds is difficult. Even the NHDplus dataset under-represents the small headwater systems apparent in the high-resolution National Hydrography Dataset (1:24k scale), which also under-represents potential ephemeral systems (**Figure 2**).

**Figure 1.** HydroSHED 15s basin boundaries (left). Example of improved accuracy of rivers mapped in HydroSHEDs 15s

versus the Digital Chart of the World in the Congo River Basin, Africa.

(**Table 1**). The Global Lakes and Wetlands Database

area range from 360,000 to 510,000 km2

streams [20].

60 Pure and Applied Biogeography

Downing et al. [20] use three different approaches to estimating stream and river area as denoted by A, B, and C (see text).

**Table 1** Theoretical and empirical estimates of global stream and river length and area provided by different studies and datasets**.**

**Figure 2.** Comparison of HydroSHEDs to NHDPlus (1:100k) flowlines in the Ohio and Tennessee River Basins of the US (left). Example of the increased spatial resolution provided by the National Hydrography Dataset (High-resolution, 1:24k) over that of NHDPlus in Bear Creek, near Oak Ridge, Tennessee, USA. However, ephemeral channels are likely even underestimated by the NHD High-resolution dataset.

**Figure 3.** Total continental US stream distance represented by four spatial datasets depicting river networks.

Interestingly, global length-stream order relationships do not follow global area-stream order relationships. For example, the number and length of 1st order systems in the world are, by far, numerically dominant constituting 52% of global river length (28.5 million and 45.7 million km2 , respectively) [20]. However, global river area is dominated by larger order systems (≥6th order), which represent 65% of total river area. Size-specific stream distribution estimates are extremely important for accurately portraying or modeling the distribution of aquatic organisms.

#### **2.2. Lakes, reservoirs, and farm ponds**

Studies estimating the global extent of lakes and reservoirs were more numerous than those estimating river and stream distributions. Global numbers of lakes range from 800,000 to 304 million whereas cumulative area of world lakes ranges from 2.3 to 5 million km2 (**Table 2**, **Figure 4**). Human construction of reservoirs has been extensive, the most current estimate at 16.7 million waterbodies with a cumulative surface of 305,723 km2 , an area equivalent to increasing the world's naturally occurring terrestrial water surface by 7.3% [11]. Other estimates of global reservoir surface area range from 150,000 to 600,000 km2 , depending on the source and whether regulated natural lakes are included (**Table 2**). Only one study provided an estimate of global farm pond coverage (77,000 km2 ) using relationships between the fraction of farm pond area within farm land and annual precipitation [22].

Similar to rivers and streams, lakes and reservoirs have been estimated using both empirical observation of available geospatial datasets or via extrapolation of observed data to unobserved features. Until recently, theoretical estimates of lakes exceeded that of empirically derived estimates. New high-resolution satellite imagery provided means to observe lakes A Synopsis of Global Mapping of Freshwater Habitats and Biodiversity: Implications... http://dx.doi.org/10.5772/intechopen.70296 63


**Figure 3.** Total continental US stream distance represented by four spatial datasets depicting river networks.

million km2

aquatic organisms.

62 Pure and Applied Biogeography

**2.2. Lakes, reservoirs, and farm ponds**

Interestingly, global length-stream order relationships do not follow global area-stream order relationships. For example, the number and length of 1st order systems in the world are, by far, numerically dominant constituting 52% of global river length (28.5 million and 45.7

tems (≥6th order), which represent 65% of total river area. Size-specific stream distribution estimates are extremely important for accurately portraying or modeling the distribution of

Studies estimating the global extent of lakes and reservoirs were more numerous than those estimating river and stream distributions. Global numbers of lakes range from 800,000 to 304

**Figure 4**). Human construction of reservoirs has been extensive, the most current estimate

increasing the world's naturally occurring terrestrial water surface by 7.3% [11]. Other esti-

source and whether regulated natural lakes are included (**Table 2**). Only one study provided

Similar to rivers and streams, lakes and reservoirs have been estimated using both empirical observation of available geospatial datasets or via extrapolation of observed data to unobserved features. Until recently, theoretical estimates of lakes exceeded that of empirically derived estimates. New high-resolution satellite imagery provided means to observe lakes

million whereas cumulative area of world lakes ranges from 2.3 to 5 million km2

at 16.7 million waterbodies with a cumulative surface of 305,723 km2

tion of farm pond area within farm land and annual precipitation [22].

an estimate of global farm pond coverage (77,000 km2

mates of global reservoir surface area range from 150,000 to 600,000 km2

, respectively) [20]. However, global river area is dominated by larger order sys-

(**Table 2**,

, an area equivalent to

) using relationships between the frac-

, depending on the

**Table 2** Global estimates of the area and number of lakes, reservoirs, and farm ponds according to different studies**.**

>0.002 km2 [13]. Using this technology, the GLObal WAter BOdies database (GLOWABO) was developed for 117 million lakes with a total surface area of 5 million km2 [13]. This surface area estimate exceeds that of the highest theoretical estimate [20], but is still smaller in total lake abundance (**Figure 4**).

The development of reservoir mapping datasets has provided valuable spatial representations of waterbodies in recent years. For example, the GLWD dataset consists of polygon shapefiles of approximately 250,000 lakes and reservoirs >0.1 km2 and raster datasets of other lakes, reservoirs, and wetland coverages [21]. The GLWD included only information for the world's largest reservoirs (storage >0.5 km<sup>3</sup> ) either because spatial information was limiting or existing lake datasets did not explicitly clarify whether a given waterbody was manmade. Because of the incomplete nature of global datasets on impoundments, the Global Reservoir and Dam database (GranD) was developed as a compilation of spatial coverages of 6862 reservoir polygons and associated dams and attributes [11]. More recently, a new geospatial coverage of

**Figure 4.** Global lake abundance estimated by several different studies.

global lakes and reservoirs, HydroLakes, was developed and includes hydrologic attributes, such as volume and residence time, using a geo-statistical model [42] (**Figure 5**). Within the US, the NHDplus (1:100k) dataset provides coverage of lakes and areas as polygons, an area estimated at almost 250,000 km2 ; however, this dataset misses small waterbodies, especially farm ponds. The NHD high-resolution (1:24k) dataset estimates lake and reservoir area coverage as approximately 890,000 km2 , almost 3.5 times higher than that of NHDplus.

**Figure 5.** HydroLakes database depiction of global lakes and reservoirs.

The most numerous lake and reservoir waterbodies are very small (<0.1 km2 ) (**Figure 4**), yet these are typically omitted from most maps (with recent exceptions, [13]). To estimate the size and distribution of these smaller waterbodies, Pareto distributions of log-abundance versus log-size are fit to observed larger lakes and then those coefficients are used to extrapolate the abundance of smaller, unobserved lakes [43] or reservoirs [11]. Obviously, these estimates do not come without error, with some suggesting that numbers of small lakes and any related scaling estimates (e.g., carbon fluxes) are unreliable [44].
