4.1 Has new research utilized the strengths of new technologies or are we doing the same old stuff with more expensive data?
The vast quantity of readily available spatial data sets over large areas, new methods for collecting spatial data with smaller and smaller grain sizes, and the improved computing power of spatial analytic software suggest that whole new approaches to landscape riverine ecology are now possible. By providing synoptic views of streams and their catchments at multiple resolutions, and by providing the spatial analytical framework to exploit the power of digital maps, these new technological tools provide the capability to examine interrelationships between streams and their catchments in ways that were either extremely limited or impossible in the past. Our first question is whether new research is truly harnessing the power of these new resources or whether we are simply doing the same old things with new data that are more resource-intensive to collect.
Habitat parameters particularly amenable to remote sensing include water temperature and stream depth; substrate type can also be collected under certain conditions. Analyses that apply these new types of data have made considerable progress in our understanding of how streams and landscapes interact. The work of Torgersen et al. (1999, 2001), for example, demonstrated how high resolution airborne FLIR systems could be used to assess the spatial distribution of thermal habitats relevant to stream fishes of the Pacific Northwest, USA. Techniques developed by Fonstad and Marcus (2005*) link spectral reflectance, stream discharge measured at local gages, and equations describing stream resistance into predictive models of depth termed “hydraulically assisted bathymetry”. These techniques and similar efforts by Bjerklie et al. (2003, 2005) hold promise for mapping stream depth over entire stream networks without the need for field data. Additionally, retrospective studies using archival photography have examined temporal trends in stream depth due to land use changes (Fonstad and Marcus, 2005). In a separate study, Lorang et al. (2005) linked field-derived estimates of water depth, flow velocity, shear stress, and stream power to multispectral imagery to spatially model geomorphic processes in gravel-bed rivers. While these techniques are promising, Legleiter and Roberts (2005) found that channel morphology could affect the accuracy of image-derived depth estimates.
Recent years have seen remarkable advances in application of high resolution remote sensing to the study of stream physical habitat structure. Newly available systems enable pixel resolutions of 1-meter or better, providing the fine spatial resolution needed to detect and map instream features of small rivers and streams. Wright et al. (2000) used 1-meter resolution multispectral (e.g., 4 wavelength band) digital imagery to map stream morphological units (eddies, glides, riffles, scour pools, etc.) of 3rd and 4th order streams with moderate to high levels of success. Then, in a series of pioneering studies (Legleiter et al., 2004*; Legleiter, 2003*; Marcus et al., 2003*), high spatial resolution hyperspectral imagery was shown to be effective at mapping instream habitat features including woody debris, depth, and substrate. The authors tested spectral band ratios and classification methods for extracting stream habitat information from spectral imagery (Legleiter, 2003; Marcus et al., 2003*) and developed the basis for physical models relating spectral reflectance to instream features (Legleiter et al., 2004). Their results demonstrate the feasibility of these technologies for mapping small streams, although higher accuracies were possible in larger streams (Marcus et al., 2003). Leckie et al. (2005*) exploited high spatial resolution (80 cm) multispectral data for mapping instream habitat and were able to reliably map substrate, woody debris, and depth classes. Feurer et al. (2008*) reviewed progress in mapping underwater topography using radiometric models and through-water photogrammetry. McKean et al. (2009a*,b*, 2008*); Madriñán (2008); Kinzel et al. (2007*) have had success mapping stream and riparian habitat using LiDAR.
In addition to physical habitat features, direct sensing of submerged aquatic vegetation and algae are now possible by combining high resolution imagery and GIS modeling. Lehmann and Lachavanne (1997) provided a comprehensive review of the topic, although at the time most applications were limited to larger rivers or estuaries, primarily because of a lack of data at the appropriate scale. The interaction of light with the water column introduces spectral classification and interpretation issues unique to aquatic environments, but the theoretical basis for remote sensing of aquatic vegetation has been established (Silva et al., 2008). Although applications to small streams are still rare, the potential of hyperspectral imagery for mapping aquatic vegetation over entire catchments is evident (Govender et al., 2007; Nelson et al., 2006).
Most aquatic mapping applications to date have used passive optical remote sensing to relate stream properties to spectral reflectance patterns in imagery. Interpreting and extracting information from these data requires a high level of operator skill and data acquisition over large areas is expensive. Marcus and Fonstad (2007*) reviewed applications to date and discussed the potential and limitations of this technology for advancing river science. While they predicted that future developments of optical remote sensing would create continuous meter-scale maps of stream habitats across entire river systems, they rightly discussed obstacles and issues that must be addressed to insure future advancement. These problems include difficulty mapping areas with steep terrain (valleys and canyons), differential illumination of stream habitats because of shading and surface turbulence, the need for clear water conditions for assessing instream habitat features and vegetation, timing constraints in image acquisition because of satellite overpass or weather conditions, the cost of acquiring repeat imagery, difficulties assessing mapping accuracy of stream features, and ethical issues related to revealing habitat locations that could be easily exploited by anglers (Marcus and Fonstad, 2007*). As well, promising new data are often available for ecological research only in limited pilot areas and there is a disproportionate amount of data for large versus small rivers. Some parameters are still limited to clear waters or surface waters (e.g., temperature). It often remains prohibitively costly to assess dynamic stream habitat environments from repeated flights of static airborne imagery. The increased availability of remote sensing data, as well as field data collected from different large-scale monitoring programs, creates substantial data aggregation problems (Roper et al., 2010; Independent Multidisciplinary Science Team, 2009*) and the costs of obtaining, storing, manipulating, and interpreting such imagery are nontrivial.
In spite of these limitations, great potential exists for further refinement of optical remote sensing techniques, especially when combined with other sensing technologies to assess the complete aquatic environment (Marcus and Fonstad, 2007; Leckie et al., 2005). While just beginning, the application of high resolution active remote sensing methods such as LiDAR for mapping stream morphology and habitat features (McKean et al., 2009a,b, 2008; Feurer et al., 2008; Kinzel et al., 2007) holds promise for precise mapping of channel morphology and water surface elevation. When combined with optical methods (Hall et al., 2009) and thermal imaging, the possibility exists for mapping multiple stream parameters over large regions for a nearly complete assessment of fish habitats. Host et al. (2005), for example, published a complex spatial analysis that provided an efficient method for the identification of reference areas along the Great Lakes coast, USA.
All relevant technological advances will not be solely in the collection and processing of spatial data. In ichthyological and fisheries sciences, researchers have recently developed a technique to reconstruct migration history of individual fishes based on chemical analysis of bony tissue, otoliths. Otolith microchemistry, especially when combined with landscape-scale data, could provide new avenues to study fish migration across large areas and to identify effects of landscape disturbances such as dams on riverine fishes (e.g., Clarke et al., 2007; Hogan and Walbridge, 2007).
4.2 Have we incorporated key concepts from landscape ecology to improve our understanding of how landscapes affect rivers?
The field of landscape ecology has focused on the relationships between pattern and process as well as on planning for patterns of human land use (Forman and Godron, 1986). Within landscape ecology, rivers have traditionally been characterized as landscape elements (Wiens, 2002*) rather than as the target of large-scale processes or as the research focus. There have been many calls to apply principles of landscape ecology in river ecology (e.g., Wang et al., 2006a*; Wiens, 2002*) or to incorporate the dynamic elements of rivers into landscape thinking. Below, we briefly describe five elemental concepts from landscape ecology and how they have been applied to improve our understanding of how landscapes alter rivers.
A fundamental underpinning of the field of landscape ecology is the study of relationships between patterns and processes, usually over broad spatio-temporal scales (Turner, 2005). Specifically, landscape ecologists seek to understand how ecosystem processes act to form patterns on the landscape, and of equal interest, how landscape patterns can influence ecological processes. Aquatic ecologists have made great strides toward understanding ecosystems by including this important concept (Fausch et al., 2002*; Ward et al., 2002; Townsend, 1996). For example, researchers have investigated how ecosystem processes such as climate, hydrology, and geomorphology influence stream characteristics (e.g., Benda et al., 2004*; Swanson et al., 1988) and fish distribution patterns (e.g., Pess et al., 2002). More prevalent are examples of the possible effect of upland landscape composition on the functioning of instream processes. Researchers have related distributions of natural features in the surrounding landscape, such as topography and soils (e.g., Richards et al., 1996*), and human land use (e.g., Independent Multidisciplinary Science Team, 2010; Brown et al., 2005*; Kershner et al., 2004*; Van Sickle et al., 2004*; Snyder et al., 2003*; Paul and Meyer, 2001*; Roth et al., 1996*) to instream biodiversity and viability of fishes.
To date, most landscape-scale research on lotic systems has attempted to evaluate relationships between pattern and process by quantifying the effect of the composition of surrounding upland landscapes (e.g., 40% of a watershed is in low-density residential land use) on some instream feature (e.g., fish abundance Paulsen and Fisher, 2001*). Inclusion of metrics of landscape structure in both upland landscapes (e.g., connectivity of non-impervious areas or average patch size of high quality riparian forest; Gergel et al., 2002) and within aquatic landscapes (e.g., longitudinal connectivity of suitable habitat patches; Isaak et al., 2007*; Torgersen et al., 2006; Benda et al., 2004*; Wiens, 2002) is an opportunity to clarify our understanding of the interaction between terrestrial and aquatic systems.
Integral to both landscape ecology and aquatic ecology is recognition that relationships between organisms and their habitats depend on the spatio-temporal scales at which they are observed (e.g., Talley, 2007; Durance et al., 2006*) (Figure 1*). Scale is clearly important in identifying and analyzing spatial pattern (Feist et al., 2010*; Wu, 2004; Wu et al., 2002; Turner et al., 1989) and has garnered much attention over the past 20 years or so (Levin, 1992; Wiens, 1989). Schneider (2001) dedicated a special section to addressing the many ways that scale can be defined and Jenerette and Wu (2000) wrote an entire essay on the multiple definitions of scale. The emphasis of landscape ecology on scale has led riverine landscape research to use multi-scale studies in order to gain a better understanding of processes acting in a stream network (Lowe et al., 2006; Leclerc and DesGranges, 2005; Fausch et al., 2002; Lammert and Allan, 1999; see Table 3 in Johnson and Host, 2010*, for further examples). Because of the strong migratory behaviour of many fishes, they are ideal study subjects for testing hypotheses about scale (Schmutz and Jungwirth, 2001). Processes affecting fish assemblages range from global to local scales (Durance et al., 2006*; Tonn, 1990).
Existing cross-scale studies have provided contradictory results giving either more weight to local (Walters et al., 2003) or catchment factors (Mugodo et al., 2006; Marsh-Matthews and Matthews, 2000). Studies that examine instream response to land use at multiple scales report, unsurprisingly, mixed influence (Feist et al., 2010*; Stewart, 2001; Fitzpatrick et al., 2001; Richards et al., 1996; Roth et al., 1996*; Johnson and Host, 2010*; see Table 3 in Johnson and Host, 2010, for further examples). For example, near-stream connected imperviousness had a stronger influence on fish assemblages than did comparable amounts of impervious surface located farther from the stream, apparently owing to increased severity and frequency of high-flow events and lowered baseflow (Wang et al., 2001); Yet, catchment-scale influence may be greatest when the primary mechanism is flow instability, nutrients, or some other factor related to the entire landscape (Allan, 2004a*). Wang et al. (2006a) found that fish assemblages were mainly influenced by local factors in undisturbed catchments whereas the relevance of catchment scale factors increased with increasing landscape disturbance.
Opportunities for advancing the study of scale in landscape riverine research are emerging from the advances in technology described in Section 2.3, which are providing fine-grain data over large extents even on small rivers. To explicitly measure the influence of scaling on fish populations, we must hold the grain constant as we vary the extent, and vice-versa. There are many good examples in the literature of varying the analysis extent while holding the grain constant (e.g., Moerke and Lamberti, 2006; Creque et al., 2005*; Santoul et al., 2005; Feist et al., 2003*). It is unusual to find riverine examples in which researchers varied the grain while holding the extent constant. There are examples from theoretical ecology, but these are usually in small systems under controlled conditions.
Landscape ecologists define connectivity as the degree to which the landscape facilitates or impedes the ability of organisms to move among resource patches (Taylor et al., 1993). Hitt and Angermeier (2008b*, 2006*) found that failure to consider spatial connectivity may bias measures of biotic integrity. However, classic connectivity metrics used in two-dimensional ecosystems such as terrestrial or oceanic landscapes (Calabrese and Fagan, 2004) are not easily applied in stream networks for several reasons. First, quantifying connectivity in stream networks is more challenging than quantifying connections in two-dimensional habitats (Fagan, 2002). Second, resource patches in temporally dynamic aquatic systems typically experience a high degree of patch turnover (Gresswell et al., 2006*). Third, organisms must contend with the force of current velocity, which may affect locomotive abilities and alter directional movement among habitats (e.g., Olden, 2007; Gresswell et al., 2006). Fourth, the apparent existence of different riverine zones that appear to support different fish assemblages suggests that fish species pools are filtered by various longitudinally varying river characteristics such that some species are limited to specific areas (Ibañez et al., 2009; McGarvey and Hughes, 2008; Vannote et al., 1980; Hawkes, 1975; Huet, 1949; Fritsch, 1872). Some recent progress has been made towards quantitatively measuring and studying connectivity in streams (Cote et al., 2009; Hitt and Angermeier, 2008a; Hughes, 2007; Isaak et al., 2007; Ganio et al., 2005; Fagan et al., 2002) but many opportunities to refine and explore the topic exist.
Barriers and dams are a particularly pressing ecological problem (Graf, 1999). They fragment aquatic systems both by impeding fish migrations and by disconnecting physical processes such as the transport of wood, water, and sediment. In Europe, for example, dams and water pollution in the Rhine and Danube Rivers have reduced the native fish fauna, particularly salmonids and sturgeons (Bacalbasa-Dobrovici, 1989; Lelek, 1989). Likewise, dams have eliminated several diadromous fish species from the Seine River, France (Oberdorff and Hughes, 1992). And, a vast network of dams and flow alterations on the Colorado River and Rio Grande, USA, have resulted in deteriorated water quality, flows frequently failing to reach the sea, and replacement of endemic fish faunas by invasive non-indigenous species (Hughes et al., 2005b*). Among major USA rivers draining to the Pacific Ocean, those with mainstem dams without fish passage or with inadequate fish passage (e.g., Snake, Columbia, Klamath, Sacramento–San Joaquin Rivers) have all experienced substantial flow and channel alterations, water pollution, species extirpations, and reaches dominated by non-indigenous species (Brown et al., 2005; Hughes et al., 2005b; Ebel et al., 1989). Even small low-head dams can eliminate native species and alter the distributions of non-indigenous species (LaVigne et al., 2008a*; Holden et al., 2005; Meffe, 1984). The fish species most affected by instream anthropogenic barriers are species with a limited range of habitat types and migratory species. Understanding the magnitude of fragmentation caused by dams requires investigations over large extents.
One area where riverine landscape research has made strong progress is in quantifying the impacts of landscape fragmentation by dams on aquatic systems, and, in particular, on fish assemblages. Fukushima et al. (2007) were able to quantify impacts of watershed fragmentation on fish assemblages through use of spatially explicit models applied to the occurrence data of multiple freshwater fish species inhabiting Hokkaido Island, Japan (ca. 80,000 km2). Sheer and Steel (2006*) quantified a relationship between the amount and distribution of lost aquatic habitat as a result of dams or barriers and salmon population performance. Letcher et al. (2007) found that habitat fragmentation by barriers increased the likelihood of local and system-wide extinction. We did not identify research examples in which other human impacts, such as pollution, altered thermal regimes, or flow alterations, were explicitly considered through the lens of landscape fragmentation or reduced connectivity between habitats.
There are global inequities in our understanding of the impacts of river fragmentation on fish assemblages and communities. The scarcity of quantitative data in developing countries presents a tremendous challenge to assessing effects of human disturbances on riverine ecosystems. The lack of scientific data may be contributing to aggressive human development in these same regions. For example, the Mekong River, the 11th-longest river in the world, is being fragmented by a series of large dams (Baran et al., 2007*). Dam construction within the Mekong watershed is on the rise with increasing demand for hydroelectricity. The Yangtze River, the third longest river in the world, now has the world’s largest dam, the Three-Gorges Dam, fragmenting an area of about 58,000 km2 (Xie, 2003*). These enormous hydrologic alterations are predicted to significantly reduce both terrestrial (Wu et al., 2003) and aquatic (Park et al., 2003; Xie, 2003) biodiversity, eventually diminishing fisheries resources and food security (Baran et al., 2007). A “geowiki” is being developed by Mark Mulligan at King’s College in London to help coordinate data-sharing and improve spatial coverage of data on dams (http://www.kcl.ac.uk/schools/sspp/geography/research/emm/geodata/geowikis.html). The database combines the visualization capabilities of Google Earth with the power of multiple Internet users to develop a comprehensive global database of dams.
Economic and population growth dramatically influence the functioning of ecosystems. For example, economic prosperity has been associated with declines in biodiversity worldwide (Clausen and York, 2008; Leprieur et al., 2008b*; Naidoo and Adamowicz, 2001), and may be especially problematic for freshwater fishes (Miller Reed and Czech, 2005; Rose, 2005). Urbanization has significant impacts on freshwater ecosystems. The consistent hydrological effect of urbanization, including flashy flows (more frequent, short-lived, more intense), is a result of the increased amount of impervious surfaces in urban streams (e.g., Booth et al., 2004*). Researchers have also been able to detect more diffuse effects of urbanization and land use (e.g., Randhir and Ekness, 2009; Bilby and Mollot, 2008; Alberti et al., 2007; Walsh et al., 2005; Kershner et al., 2004; Paulsen and Fisher, 2001; Paul and Meyer, 2001; Roth et al., 1996*). Human threats to imperiled fish populations also include the spread of non-native species (Leprieur et al., 2008a*) and global climate change (Crozier et al., 2008; Rieman et al., 2007*; Flebbe et al., 2006*). A concept in landscape ecology that is becoming more evident in recent years is the notion that humans should be considered as part of the ecosystem being managed, rather than as an outside factor exerting negative impacts on natural systems (Otte et al., 2007; Wu and Hobbs, 2002). Few riverine studies or approaches have incorporated this mindset.
Results of landscape analyses have significant implications for management of streams and catchments as well as for developing informed policies on land management. The genesis of this field lies in great part with policy directives, the increasing need for catchment-scale management, and wide-ranging environmental problems as discussed in Section 2. Therefore, landscape analyses are often intended to have on-the-ground impacts. However, evaluating the degree to which landscape analyses have actually been used for making management decisions is difficult because many peer-reviewed publications define the possible or intended application rather than the actual application. Allan (2004a*) noted that there are limitations in the degree to which these types of analyses can inform management prescriptions and fisheries management. He and others have indicated the need for a more experimental approach in which systematic variations in land use are considered. In this section, we identify the most common potential applications of landscape analyses for riverine management and those papers that explicitly focus on these applications.
Because landscape ecology has traditionally involved evaluation of spatial patterns of land use and species distributions, it has followed naturally that newfound knowledge was applied to ecosystem management such as the creation of conservation reserve networks (Thieme et al., 2007; Margules and Pressey, 2000). Landscape-scale riverine analyses often apply high spatial resolution satellite and aerial imagery to identify conservation and restoration needs of large and complex aquatic ecosystems. The Missouri Aquatic GAP Project, for example, used landscape-scale data to identify types of aquatic habitats not adequately represented within the existing conservation network (Sowa et al., 2007). Randhir and Tsvetkova (2009) used landscape-scale data to explore conservation implications of relationships between water conservation and management strategies. And, Ballinger and Mac Nally (2006) used a landscape perspective to explore the impacts of spatial and temporal flooding variability on wildlife habitats in the Murray–Darling Basin, Australia. Statistical models were built and tested using species occurrence data for multiple fish species in the Iberian Peninsula (Filipe et al., 2004). These models used landscape characteristics to quantify the multi-species conservation value of particular river reaches and, eventually, to select a series of reserve reaches. Ekness and Randhir (2007) identified spatial criteria for watershed-scale policy development. They concluded that basing conservation networks on stream order, riparian condition, and land use could lead to increased riparian areas, protection of headwaters, and minimized disturbance in headwater areas. However Li et al. (1996) and Dunham et al. (2008) recommended protection of mainstem reaches and entire channel networks, and Osborne and Wiley (1992*) and Hitt and Angermeier (2008b*) have reported on the importance of confluences to fish species occurrences in tributaries.
Species-based examples of applying landscape-scale data to conservation management include restoration of the Middle Rio Grande in New Mexico, USA for the endangered silvery minnow (Cowley, 2006), estimation of potential steelhead habitat above barriers in the Willamette River basin, Oregon, USA (Steel et al., 2004*), or probability of occurrence of two diadromous fish species in New Zealand (Eikaas et al., 2005). Landscape condition has also been used to estimate the frequency and severity of ecosystem threats, such as the frequency of human-induced stressors, to provide a more informed basis for conservation planning (Mattson and Angermeier, 2007; Kracker, 2006). An example of the use of large-scale analyses to identify specific landscape stressors and the impact of those stressors on fish assemblages is Scott (2006*), who quantified the impacts of urbanization on endemic versus broad-ranging fishes.
One of the most obvious and far-reaching impacts of landscape analyses for land use management has been the promotion of the riparian zone as a critical transition area between streams and their catchments. In rehabilitating lateral connectivity, studies have emphasized establishing or enlarging riparian wooded buffer zones (Miltner et al., 2004; Freeman et al., 2003*; Snyder et al., 2003*; Wichert and Rapport, 1998). Riparian management is particularly attractive because of the riparian zone‘s immediate influence on stream condition and because it promises benefits that are highly disproportionate to the land area required (Allan, 2004a*). The landscape perspective can improve management of riparian lands by identifying the best strategies for particular areas. For example, the landscape perspective enabled fine-tuning of forest buffer restoration priorities in agricultural areas of Maryland, USA by quantifying those underlying geologies for which forested buffers had the greatest impact on indices of biotic integrity (Barker et al., 2006*). FEMAT (1993) also took a landscape approach in its recommendation of a riparian buffer equivalent to 2 site-potential tree heights along fish bearing streams, because the size of trees varies by ecoregion. Multi-scale analyses have been used to demonstrate that riparian management shows greater effectiveness in protecting streams from the negative impacts of land use at the local or reach scale than at the catchment scale. Intensive land use at the scale of entire catchments may have impacts too great for a riparian zone to moderate (Wang et al., 2006b*; Allan, 2004a*; Snyder et al., 2003; Morley and Karr, 2002; Nerbonne and Vondracek, 2001; Roth et al., 1996).
Use of landscape analyses to improve land management across entire catchments has been more challenging. Reversal of land use to a less-developed state over vast extents is usually infeasible; however, improvement of stream conditions can be accomplished by promoting best management practices (BMPs) and improvements in landscape management, e.g., reduced fertilizer applications in the catchment (Wang et al., 2006c*) or conservation tillage (Yoder et al., 2005). However, where native fish species have been extirpated, improvements in land use alone may not be enough to restore species distributions to their historic states (Wang et al., 2006c).
The landscape perspective has improved fish-based bio-assessments through validation of biotic indices and identification of appropriate sampling scales. Correlations between indices of biotic integrity and landscape conditions have validated expected relationships between degraded landscapes and stream condition (e.g., Carlisle et al., 2008*; Pinto et al., 2006*; Bramblett et al., 2005; Dauwalter and Jackson, 2004*; Joy and Death, 2004*). However linking instream biological responses to landscape (land use, economic growth, population growth; Schleiger, 2000; Leprieur et al., 2008a) and riverscape (dams, diversions; Stanford and Ward, 2001; Ward and Stanford, 1995) stressors requires different analytical approaches than the site-specific studies that comprise most biological assessments. For example, Hawkins et al. (2000) found that while landscape data can explain more variation in aquatic biota than what might be expected by chance alone, the amount of variation explained is not large, and thus will have limited value in making reach scale predictions. Herlihy et al. (2006) reported that landscape classifications accounted for approximately half the variability in fish assemblage clusters of the conterminous USA; however Pinto et al. (2009) reported that ecoregion and fish species clusters had similar classification strengths at a river basin scale.
Many analyses correlating land use with indices of biotic integrity have compared models at varying spatial scales such as riparian buffer scales and catchment scales (e.g., Barker et al., 2006; Van Sickle et al., 2004*). Such analyses have begun to identify the specific scales at which land use drives various types of biotic indices. Fish-based and macrophyte-based biological indices are often more closely correlated with ecological quality at the river basin scale whereas indices based on macroinvertebrates and benthic diatoms are more closely correlated with environmental metrics at the reach and stream scale (Springe et al., 2006). Brazner et al. (2007*) and Allen et al. (1999*) reported similar differences among biological indicators of disturbance in northeastern USA lakes and Great Lakes wetlands, respectively. In addition, recent focus on stream network topology has demonstrated the importance of considering position in the stream network and dispersal rates when using bioindicators to assess fish assemblage responses to the environment (Hitt and Angermeier, 2008b). The landscape perspective inherent in these studies has led to improved use of biological indicators of stream health; however, the lack of specific and tested mechanistic relationships between remote landscape conditions and instream biological response continues to plague on-the-ground applications and to limit the applicability of results beyond the study basins.
Another common application of the landscape perspective is in determining where, within a catchment, rehabilitation activities should be undertaken. Some approaches have used riparian condition to identify those sub-basins most likely to respond to rehabilitation activities (e.g., Fullerton et al., 2006*) others have identified locations within basins that are most likely to increase aquatic connectivity or provide cumulative positive impacts (Jansson et al., 2007). Landscape characteristics have been used to estimate the quantity of lost habitat behind migration barriers (Sheer and Steel, 2006) and to predict the quality of habitat where fish currently do not exist, such as behind migration barriers, and prioritize barrier removal projects (Steel et al., 2004).
A large-scale perspective is necessary for rehabilitation prioritization for three reasons. First, the watershed is the unit of observation. Abiotic conditions and biological assemblages within a watershed are interdependent and cannot be effectively examined in isolation. Second, funding is often distributed for political equity and/or by watershed and so decisions about what restoration actions to fund must prioritize across all possible actions within a political jurisdiction or watershed. And, third, rehabilitation actions are often aimed at reducing impacts associated with large scale degrading processes such as wood, sediment, nutrient, and water delivery (Lake et al., 2007). Understanding these large-scale habitat-forming processes (Beechie and Bolton, 1999) across entire catchments aids us in identifying and prioritizing aquatic rehabilitation activities at the appropriate temporal and spatial scale. Weber et al. (2007) credit the failure of many rehabilitation and restoration projects to a failure to consider degrading factors over large enough spatio-temporal scales.
Rehabilitation of entire catchments would be ideal but is rarely practical. However, the condition of the catchment can still inform decisions about which types of rehabilitation to initiate first. As pressures increase in the catchment, the importance of local factors may decline. In catchments experiencing intensive pressure from human development, instream projects should be initiated only after substantial progress in removing major causative factors is made (Schmutz et al., 2007*; Booth et al., 2004*). Large-scale analyses suggest that rehabilitation prioritization in degraded catchments should be geared toward rehabilitating upslope processes first (e.g., rehabilitating the sediment, wood and water flow regimes; Bohn and Kershner, 2008) or should emphasize rehabilitation of upslope processes in combination with local rehabilitation measures (Schmutz et al., 2007*; Wang et al., 2006b*; Booth et al., 2004; Soulsby et al., 2001). In relatively undisturbed catchments, local instream habitat and riparian improvement will be most effective (Wang et al., 2006b*).
Many kinds of decision tools are now developed to set priorities in catchment-scale rehabilitation planning (e.g., Roni et al., 2008; Steel et al., 2008*; Schmutz et al., 2007*; Fullerton et al., 2006; Shriver and Randhir, 2006; Reynolds and Peets, 2001). Landscape-scale analyses can inform predictions of future conditions, often through scenario-based tools that integrate landscape evaluations and restoration or rehabilitation planning (e.g., Randhir and Hawes, 2009; Fullerton et al., 2009; Steel et al., 2008*; Reynolds and Hessburg, 2005*). Decision support frameworks such as the Ecosystem Management Decision Support (EMDS) have been used in salmon recovery planning efforts in the Pacific Northwest USA (Reynolds and Hessburg, 2005). Rieman et al. (2001) evaluated the effect of a variety of land management schemes on salmonid viability, and Burnett et al. (2007*) evaluated how potential habitat available to salmonids is distributed among different land use classes. Using aerial photos, Freeman et al. (2003) not only evaluated possible future states in a floodplain, but also evaluated how land use has changed through time using aerial photos. The landscape perspective enables scenario planning to consider far-ranging impacts of particular actions and to provide spatially explicit predictions of trade-offs in future condition that would result from alternative action schemes. The maps of potential future conditions that often result from landscape-based scenarios can be particularly helpful in soliciting citizen input and engaging local landowners (e.g., Baker et al., 2004; Hulse et al., 2004).
Landscape-scale analyses will be essential in predicting impacts of climate change (e.g., Battin et al., 2007; Rieman et al., 2007*; Flebbe et al., 2006) and in identifying management alternatives that are robust to the suite of likely climate change scenarios. Matulla et al. (2007*) modeled the impact of IPCC (Intergovernmental Panel on Climate Change) emission scenarios on the fish assemblages of the Mur River, Austria. They used downscaled temperature and precipitation predictions to model instream river temperatures over the entire basin in order to identify native species at risk of extirpation and non-native species with increased potential for invasion. Rieman et al. (2007) examined potential effects of climate change on bull trout (Salvelinus confluentus) over the entire interior Columbia Basin. Working over such large extents, they were able to identify populations of bull trout that faced particularly high potential habitat losses.
In a recently published report of the European Parliament on climate change-induced water stress, land use management was identified as a key issue in adaptation strategies (Anderson et al., 2008). Land use measures that may support adaptation to climate change include, for instance, forestation, conservation agriculture, floodplain rehabilitation, the conversion or rehabilitation of natural land cover, and wetlands rehabilitation. In the Netherlands, projects are being implemented that limit development along rivers to reduce vulnerability to climate change-induced increases in flood risk. The Room for the River Program recognizes the need to widen the river floodplain, rather than increase the height of the dikes (http://www.ruimtevoorderivier.nl/).
Large-scale riverine monitoring programs have been driven by the needs of managers and scientists to understand how natural conditions and human impacts vary across watersheds and landscapes (USEPA, 2009*; Paulsen et al., 2008*; Hughes et al., 2006a; Pont et al., 2006; Yates and Bailey, 2006*; Stoddard et al., 2005*). Spatially extensive perspectives and landscape analyses have been essential in designing monitoring programs that capture watershed-scale habitat conditions and track performance of entire populations. For example, the USEPA’s Environmental Monitoring and Assessment Program (EMAP) began in 1989 and now samples roughly 900 probabilistically selected aquatic sites each year using a rotating panel design (i.e., lakes in 2007, rivers in 2008, streams in 2009, coastal waters in 2010, wetlands in 2011, lakes in 2012, etc.) (Shapiro et al., 2008; Hughes et al., 2000). Biological data, environmental data, and watershed parameters are measured at each site. The probabilistic design and nationally consistent methods allow rigorous statistical inference to all water bodies of each type. Based on the 2005 national wadeable stream survey, Paulsen et al. (2008*) reported that 42% of all wadeable stream length in the conterminous USA was in poor condition, and the major stressors were nutrients and excess fine sediments. USEPA (2009*) reported that 22% of USA lakes were in poor biological condition based on changes in their diatom assemblages, 36% had poor riparian vegetation cover, and 20% were considered to have poor levels of nutrients. A key component in these assessments was the use of large-scale monitoring data.
In 1990, the USGS began its National Water Quality Assessment Program (NAWQA), which focused monitoring and research on 42 study units representing different hydrologic regions and pressures (agriculture, urbanization). The NAWQA program rotates its intensive sampling on 14 of the study units every 3–4 years and proposes to develop models to predict conditions in unmonitored areas (Gilliom et al., 2001). In a 5-year NAWQA study of 9 urban areas, Brown et al. (2009*) found that the effects of urbanization on fish assemblages differed among those areas. They reported that urbanization affected stream habitats and fish assemblages differently because of natural landscape differences as well as the legacy effects of agriculture.
A major monitoring program (Sustainable Rivers Audit, SRA) for the Murray–Darling Basin in Australia was initiated in 2000, using subbasins as reporting units and systematically rotating annual sampling of 180–341 sites among those basins every 2–3 years (Davies et al., 2006*). The SRA stratifies sampling and data analyses by altitudinal zones (lowlands, slopes, uplands, montane) and sites are chosen randomly within each zone to ensure that they are representative.
Large-scale analyses are often applied to monitor the cumulative effectiveness of multiple instream rehabilitation (restoration) projects. For several decades a wide variety of agencies within the United States have implemented stream rehabilitation (restoration) projects, but rarely has the effectiveness of those projects been monitored (Alexander and Allan, 2006; Palmer and Bernhardt, 2006; Thompson, 2006*; Bernhardt et al., 2005; Roni et al., 2002).
The situation is somewhat better in central Europe where 58% of 50 projects were monitored (Kail et al., 2007). Statistically and ecologically rigorous monitoring of the effectiveness of individual rehabilitation projects is a substantial undertaking and scaling the inference up to evaluate effectiveness at larger scales has rarely been done effectively (Thompson, 2006; Ioannidis, 2005). A key issue in all monitoring programs is the question of sufficient and standardized sampling effort to ensure that site-scale noise or variance among field crews is low relative to the variance or signal from the population of sites being assessed. Standard sampling methods facilitate data compilation and comparison (Bonar et al., 2009; Independent Multidisciplinary Science Team, 2009). Sufficient site extents reduce the local variability that confounds comparison with landscape-scale variables (Flotemersch et al., 2010; Hughes et al., 2008). And high signal-noise ratios increase the potential R-squared values of regression or other statistical analyses when relating assemblage response variables to predictor variables (Stoddard et al., 2008). The 2009 IMST report found that lack of comparability in survey designs, indicators, and sampling methods hindered data aggregation despite millions of dollars dedicated to disparate research projects in the same area.