Although surface water and groundwater are increasingly referred to as one resource, there remain environmental and ecosystem needs to study the 10 m to 1 km reach scale as one hydrologic system. Streams gain and lose water over a range of spatial and temporal scales. Large spatial scales (kilometers) have traditionally been recognized and studied as river-aquifer connections. Over the last 25 years hyporheic exchange flows (1–10 m) have been studied extensively. Often a transient storage model has been used to quantify the physical solute transport setting in which biogeochemical processes occur. At the longer 10 m to 1 km scale of stream reaches it is now clear that streams which gain water overall can coincidentally lose water to the subsurface. At this scale, the amounts of water transferred are not necessarily significant but the exchanges can, however, influence solute transport. The interpretation of seemingly straightforward questions about water, contaminant, and nutrient fluxes into and along a stream can be confounded by flow losses which are too small to be apparent in stream gauging and along flow paths too long to be detected in tracer experiments. We suggest basic hydrologic approaches, e.g., measurement of flow along the channel, surface and subsurface solute sampling, and routine measurements of the water table that, in our opinion, can be used to extend simple exchange concepts from the hyporheic exchange scale to a scale of stream-catchment connection.
 Generally, surface water and subsurface water systems are understood to be connected in a manner in which a stream receives water and solutes from the subsurface; the water and solute continue to be transported without return to the subsurface (Figure 1a). However, the stream is not a pipe. In some hydrologic systems there can be continuing exchange between the stream and the connecting subsurface systems (Figure 1b). The volume of water in this continuing exchange is not necessarily significant; rather, it is the exchange that can further have influence on solute transport and aquatic ecosystems.
 The purpose of this article is to present a set of challenges in working to understand the connections of streams to their catchment at scales longer (in time and space) than hyporheic exchange and shorter than that of regional river-aquifer interactions. To build the context, we introduce an understanding of hyporheic exchange and the now common transient storage interpretation of solute transport. We close with our opinions on basic hydrologic approaches that may be useful in interpreting solute transport influenced by stream-catchment connections.
1.2. Hyporheic Flows: A Stream-Catchment Continuum
 Hyporheic exchange flow is stream water exchanging in the shallow subsurface. With these flows an apparent loss of water from the streams returns as a gain downstream (or, conversely, an apparent gain of water in the stream is the return of an upstream loss). The volumes of water involved in the exchanges are typically insignificant as a fraction of the stream flow. The significance of hyporheic exchange is in the movement of solutes from the stream into the subsurface (Figure 2). (See current, extensive reviews of hyporheic processes and significance from the viewpoints of hydrogeomorphology, ecology, and nutrient dynamics in Poole , Boulton et al. , and Mulholland and Webster , respectively. In the subsurface, solutes from the stream may be in an environment of different pH and dissolved oxygen and in which contact with solid (abiotic or biotic) surfaces is enhanced. The stream water, temporarily in the subsurface, may undergo reactions altering the biogeochemical signature of that water as it goes back into the stream. With hyporheic flow the stream and the catchment ecosystem can be viewed and studied as a continuum in which each is a boundary condition of the other rather than separate components of a system.
1.3. Transient Storage: Simple Conceptual View of Hyporheic Flows
 The transient storage model (TSM) is a simple conceptual model of the stream-catchment continuum that builds upon the fundamental 1-D model of solute advection and dispersion by providing a transient storage zone connected to the channel. This transient storage zone is conceptually a well-mixed reservoir of finite size that temporarily holds solute, delaying downstream transport for time scales longer than advection and dispersion (see Figure 3).
 The TSM has several variations in representing solute exchange between the main channel and the transient storage locations, though the most common is a single-rate mass transfer approach, as is implemented in the one-dimensional transport with inflow and storage (OTIS) model [Runkel, 1998]:
where C is the main channel conservative solute concentration (M L−3), CS is the storage zone conservative solute concentration (M L−3), A is the cross-sectional area of stream (L2), AS is the cross-sectional area of storage zone (L2), α is the storage zone exchange coefficient (T−1), qL is the lateral inflow rate (L3T−1L−1 of stream length), CL is the lateral inflow concentration of conservative solute (M L−3), D is the dispersion coefficient (L2T−1), x is the distance downstream (L), t is time (T), and Q is streamflow rate (L3T−1).
1.4. Limitations of the Transient Storage Concept
 The simple conceptual basis of the TSM is both its greatest strength and its greatest weakness. The transient storage zones in the conceptual and numerical models are lumped representations of a diverse array of real-world locations of temporary solute storage, such as eddies behind boulders, channel margins, pool margins, and hyporheic exchange. Whereas this is a great advantage from the perspective of simplifying the representation of the real world, not all storage zones have the same spatial and temporal scales nor provide the same biogeochemical function. Surface storage zones (e.g., eddies) host very different conditions than subsurface storage zones (hyporheic flow paths). Some storage zones may be “nested” within others; for example, water may have to pass through a surface transient storage zone to enter a subsurface transient storage zone. The physical processes that move mass and energy between the main channel and surface versus subsurface storage zones are likely to be very different. Further, the heterogeneity within a single storage zone type (e.g., all pool margins) may be great enough that the residence times and biogeochemical efficacy of each zone is substantially different. The TSM, as typically implemented, applies reach-representative characteristics to these processes without discrimination. Thus, in characterizing the “representative” storage zone influence on storage, the single-storage zone TSM lumps many zone types and exchange processes. In addition to these challenges, the most commonly used TSM (as implemented in OTIS) uses a finite difference solution scheme so that all mass that leaves the channel to move into transient storage must return in the same part of the channel. Hence, there is no solute transfer or “communication” from upstream to downstream along subsurface flow paths, as they are not explicitly represented in the solute transport model.
 Alternative to the TSM, several mathematical models have been produced with similar conceptual models, but different mathematical exchange characterizations. Several of these have focused on exchange that is not of the single rate mass transfer-type characterized in the TSM by an exponential storage residence time distribution. For example, Wörman et al.  developed a transient storage model with lognormal residence time distribution, and Haggerty et al.  and Gooseff et al.  have applied models with power law residence time distributions in their storage zones. A two-storage zone version of the TSM with differing storage zone parameters has been evaluated theoretically [Choi et al., 2000] and implemented practically [Harvey et al., 2005; Marion et al., 2008; Briggs et al., 2009, 2010] to separate surface transient storage from subsurface transient storage. Whereas all of these approaches yield excellent simulated fits to observed stream tracer data, each of these conceptual and numerical models has different interpretations of the influence of transient storage on solute transport. For example, mean storage residence times would be calculated differently, and therefore, implications for storage zone biogeochemical cycling would be interpreted differently as well [Gooseff et al., 2003].
 Stream solute transport studies (and therefore subsequent TSM modeling and/or mass balance analyses) are inherently limited by practical issues. To characterize return flows (i.e., hyporheic flows), the detection limit of the solute tracer is perhaps the most important restriction. Small fluxes of highly concentrated hyporheic water returning to the stream channel after much of the stream tracer has passed may not provide enough signal to be observed when concentration detection limits are high. Furthermore, without additional information (such as the discharge at the upstream and downstream ends of a reach of interest [see Payn et al., 2009], the stream tracer approach is generally insensitive to streamflow losses. When tracer-labeled water leaves the channel (as illustrated in Figure 4), it does not change the concentration of the remaining water in the channel. Dilution of the tracer by lateral inflowing water is, however, readily observed. The result of both gains and losses occurring along a stream reach may confound both the estimates of lateral inflows and therefore the quantification of stream-catchment connections. The transient storage concept and the associated implicit interpretation of tracer transport have been highly useful in developing an understanding of hyporheic processes. The limitations of the concept and its interpretation, however, call for caution as we seek to understand the connection of streams to their catchment at larger 10 m to 1 km scales.
2. Stream-Catchment Connections
 In discussing stream-catchment connections, we refer to a loosely specified portion of the continuum of surface water to groundwater interactions. These connections have one or more of the following characteristics.
 1. Flows are dispersed in the sense that an individual, well-delineated flow is insignificant.
 2. There may be exchange between the surface and subsurface occurring in both directions.
 3. The primary importance of the connection is on the flux of solutes, while over the longitudinal distance of the stream the dispersed exchanges may or may not change the stream discharge.
 4. The exchanges occur over larger (10 m to 1 km) spatial scales than typically associated with hyporheic flows (1–10 m).
 These interactions of surface water and groundwater are, again loosely, distinguished from the dominantly unidirectional flows at the 1–10 km scale of discharge of an alluvial aquifer feeding a river network or, conversely, a mountain range drainage system recharging a semiarid area.
 The distinction from the spatial scale of hyporheic flow is significant in that at the hyporheic scale flow (carrying solutes along) leaving the stream returns to the stream over a distance on the order of 10 m. At the larger spatial scale, water leaving the stream has flowed back into the catchment subsurface, seemingly a loss from the stream. If this water returns to the stream far down valley seemingly as now yet again a gain to the stream, it likely will have mixed extensively with other waters en route. At the scale of hyporheic flow, the exchange of solutes between the surface and the subsurface is commonly referred to as “transient storage”; at larger scales this terminology is not descriptive of the long-term storage that may occur. In this long-term storage, water and solute that leave the channel may return over time and space scales longer than can be quantified in the typical tracer study.
 We believe it is reasonable to assert that stream ecosystems with hyporheic exchange are now routinely understood and studied as a fully integrated ecosystem with surface and subsurface zones. Also at the larger river-aquifer scale, water is viewed as a single hydrologic resource [e.g., see Winter et al., 1998]. For the intermediate scale of stream-catchment connections we are proposing that studies need to be designed to recognize a complex hydrologic system with fundamentally disparate scales of surface and subsurface transport processes [e.g., Woessner, 2000]. This hydrologic system is distinct both from an integrated ecosystem in which detailed coupled processes can be identified and from a single resource in which the readily defined components of the system are linked. Stream-catchment connections at the intermediate 10 m to 1 km spatial scale have influences on public policy issues such as environmental flow management for aquatic ecosystem health and land use management, such as restoration of streams affected by mine drainage. A time frame in which these connections influence the system is most clearly the time of transport of water and solute along the stream. The spatially cumulative influence of hyporheic exchange flows acts in this time frame. The connections are also evident in a steady, or static, time frame of a study to identify the location and quantity of nutrient or contaminant load from the catchment. Alternatively, on an annual time scale, the summation of apparent catchment solute inputs to the stream may be different than the stream solute exports from the catchment, the differences being caused by losses and then apparent gains in the stream-catchment connections.
 Stream-catchment connections are the locations where the stream water column is in hydraulic contact with adjacent water tables. These connections may drive exchanges of water between the aquifer and the channel, where potential for exchange is a function of the spatiotemporal head distribution and flux is a function of both head gradients and hydraulic conductivity. That these are dynamic in both time and space is a general statement that most hydrologists can agree upon. However, the true temporal and spatial heterogeneity of connections between streams and aquifers is rarely quantified well and incorporated into conceptual and numerical models. As such, our understanding of exchange dynamics between streams and aquifers remains incomplete.
 Streamflow gains generally are easier to identify than losses of stream water. Some are obvious to the naked eye, such as stream-side seeps and springs. Stream tracer experiments will also generally identify locations of lateral inflows between tracer sampling sites as tracer concentrations at downstream locations are diluted (however, amounts of lateral inflows can be difficult to determine if streamflow losses have occurred upstream of the gain; see more detail about this in section 3.1). Inflows to channels may be hyporheic return flows (a mix of stream water and groundwater) or they may be contributions of pure groundwater from the larger aquifer. Streamflow losses are more difficult to identify because they are not visible and they do not influence stream tracer concentrations. Losses of water from streams to aquifers have one of two fates: they either mix with groundwater and reenter the channel at a downstream location or they follow deeper, more distal flow paths that mesh into a larger valley bottom aquifer. From the scales of grains to watersheds, the geologic structure of the stream and catchment is a primary control on the fate of streamflow gains and losses. Whereas we can generally identify short (spatially and temporally) exchanges or flow paths, our ability to observe and document temporally and spatially long exchange flow paths is poor. We often rely on models because direct observations of tracer signatures, etc., do not provide the resolution necessary to support evidence of stream-groundwater exchanges in long temporal or spatial contexts.
 In section 3 we set out three basic challenges to understanding the flow and solute transport pathways that may be significant in connection of streams and their catchment at scales longer than that of hyporheic exchanges. While describing these challenges, we offer suggestions of basic hydrologic approaches to address these challenges with field measurements.
3. Basic Hydrological Measurements Needed for Understanding at the Scale of Stream-Catchment Connections
3.1. Challenge 1: Measurement of In-Channel Flow
 During the past 2 decades, many mass-loading studies have been done in streams affected by acid mine drainage [Kimball et al., 2004, 2007]. The majority of these streams have been in upland catchments that can be considered gaining reaches, resulting in tracer dilution profiles that are straightforward and provide the data needed to calculate spatially detailed estimates of stream discharge and streamflow gain. And yet, in some catchments, a seemingly normal tracer dilution profile has resulted in a substantial overestimate of stream discharge at the end of the study reach. For example, in Red Butte Creek, a research catchment near Salt Lake City, Utah, a dilution profile for a bromide tracer is illustrated in Figure 5. This bromide profile resembles many that come from upland catchments, with clear gaining segments indicated by dilution of the tracer. Using the mass injection rate and injectate concentration, a tracer dilution discharge was calculated (see equations in the work by Kimball et al. ) and is illustrated along with the bromide concentration. Stream discharge increased at the tributary inflows but also along parts of the reach where no visible inflows were observed, suggesting dispersed inflow of water along the study reach from the subsurface. Area-velocity discharge measurements also were obtained in the stream the day of the synoptic sampling. Although not all of these were rated as good measurements because of the difficulties of measuring discharge in mountain streams by this method [Jarrett, 1992], taken together they clearly indicate a lower stream discharge for most of the study reach. If both the tracer dilution discharge and the area-velocity discharge represent sound data, then the simplest explanation for the discrepancy would be that this stream gains substantial quantities of water (Figure 4, flow paths A and B) but subsequently loses much of that water to flow paths on scales longer than hyporheic flows (Figure 4, flow paths C, D, and E).
 Faced with this dilemma in Red Butte Creek, various means were used to estimate an accurate discharge profile. For example, using area-velocity measurements that were rated good (five of the measurements at different locations along the stream), the bromide tracer data were used to estimate discharge by dilution both upstream and downstream from each of these points, generating a whole series of possible discharge profiles (Figure 6). Each of these results in a much smaller discharge than the tracer dilution discharge, but it is not clear which should be selected.
 This one example in Red Butte Creek illustrates the need for (1) recognizing the utility of measuring stream discharge not only at the catchment outlet but also along the stream reach and (2) increasing appreciation that tracer and physical measurements of discharge may be measuring different aspects of flow. A number of flow measurements by complementary methods, spread out along the stream, may be needed to fully understand the dynamics of exchange between the stream and its catchment. Differences between tracer and physical measurements of discharge have been previously noted [e.g., Zellweger et al., 1989] and ascribed to the observation that tracer methods will measure some portion of flow within the subsurface in addition to the flow strictly in the surface channel.
3.2. Challenge 2: Estimating the Distribution of Stream-Catchment Exchange Flux
 The issues of discharge measurement in a given stream reach having subsurface flow translate into challenges for calculating an accurate catchment mass loading of solutes to the stream. In extending studies to multiple, successive stream reaches, judgments are needed to interpret the influence on mass load of inflows individually too small to be measured and of subsurface flow paths longer than the scale of hyporheic exchange observed in the standard tracer experiment. The longitudinal profile of sulfate concentrations along Red Butte Creek (Figure 7) indicates areas of the stream where the concentration increases from less than 80 mg/L to almost 160 mg/L. Dilution of sulfate concentration also is indicated where Parley's Fork enters the stream. The zone of greatest increase from 1902 to 2953 m included only a few visible inflows, suggesting that the increases in sulfate had to result from dispersed subsurface inflow of water with higher sulfate concentration than the stream concentration. To calculate load from this sulfate concentration profile requires the choice of just one of the possible discharge profiles (Figure 6). Two possible versions of sulfate mass load are indicated in Figure 7, representing the maximum and minimum discharge profiles in Figure 6. The first uses the tracer dilution discharge and the second uses the minimum adjusted-discharge profile based on the area-velocity measurement at 4600 m. Comparing the cumulative in-stream loads (Figure 7, solid squares and triangles), the tracer dilution discharge results in almost 5 times greater load for tracer dilution load than the adjusted-discharge load (Figure 7). By summing the load from each of the sampled inflows (see Kimball et al.  for equations), a cumulative inflow load is calculated. Comparing the cumulative in-stream and inflow loads using the tracer dilution discharge (Figure 7, solid and open squares), the cumulative in-stream load is almost 5 times greater than the inflow load at the end of the study reach. For the adjusted-discharge loads (Figure 7, solid and open triangles), the in-stream load is over 3 times the inflow load. The difference between cumulative in-stream and inflow loads has often been considered “unsampled load” in mass loading studies [Kimball et al., 2006]. However, some of this difference also could be a result of water returning to the stream from the catchment after traveling along flow paths on the scale of 10 m to 1 km. This circumstance could result in water and solute load being “double counted.” A closer examination of two sections of the discharge and sulfate-loading profiles (Figures 5 and 7) can illustrate these possibilities.
 Focusing on three stream segments from 2183 to 2369 m (shaded area A in Figures 5 and 7), we see no dilution of bromide in the first segment and then two slight decreases in bromide concentration in the next two segments. However, because the sulfate concentration increases in that first segment, a small increase in load is indicated. This could imply the inflow of an amount of sulfate-rich water too small to be observed by tracer dilution. Larger increases in sulfate load over the next two segments result from higher sulfate concentration and water inflow indicated by tracer dilution. Using the change in mass load of sulfate and the change in discharge, it is possible to calculate an effective inflow concentration, which is the concentration that would be necessary to cause the measured change in load for a given stream segment [Kimball et al., 2002]. The calculated inflow concentrations for the second and third segments, using either of the in-stream loading curves, compare very well with the range of the sampled inflows downstream at the “left bank springs.” Thus, despite the great difference between cumulative in-stream and inflow loads through these segments, the calculated increases result from plausible sources given the range of sulfate concentration among the sampled inflows.
 In contrast to this, looking at two segments downstream from Parley's Fork, from 4133 to 4804 m (shaded area B), the discharge increases in both segments but the sulfate concentration remains constant in the stream. Calculating the effective sulfate concentration as above results in a value of approximately 139 mg/L for both segments, which is very similar to the in-stream concentration of 142 mg/L along this reach. This could be a condition indicating that the increase in discharge represents inflow of water that has traveled a flow path on the scale of more than 10 m and then reentered the stream (similar to flow path C illustrated in Figure 4). The loading would be “double counted.” On this same scale, if part of the solute mass in a stream segment were lost to a longer flow path so that it was not accounted for at the end of the study reach, it may be a “lost” or “overcounted source” (similar to flow path D illustrated in Figure 4). This is significant because, in terms of loading profiles, that loss would have the same effect on a loading profile as would the removal of load through biogeochemical processes, complicating the interpretation of mass transfer processes. These two possibilities point to the need for an understanding of basic hydrology to quantify solute loading at this scale.
 Sampling of stream and subsurface chemistry has been widely used in identifying the temporal variation in the flow components of run-off at the catchment scale to streams. Here we are suggesting that sampling in the subsurface along 10 m to 1 km reaches of stream can be combined with spatially discrete stream discharge measurements to enhance our interpretation of the significance of local flow losses in a stream that is gaining water overall, as in this discussion.
3.3. Challenge 3: Observing the Distribution of Where and When the Stream Is Connected to the Catchment
 Perhaps the most significant challenge to characterizing stream-catchment connections arising from the use of stream solute transport approaches is the simplicity of the conceptual model that oversimplifies a finite zone with which streams directly interact. The connections between streams and catchments are extensive in both space and time, with exchanges occurring along a continuum of flow path lengths, transport times, and flow capacities. Hence, assessing stream-catchment connections from the stream perspective alone will always be of limited utility.
 Several different types of stream-catchment connections exist and occur over different temporal and spatial scales. Each likely has a different influence on stream water conditions. One example is a temporally dynamic connection between the stream and adjacent riparian aquifers. Figure 8 presents a 10 day time series of stream and water table elevations from three wells adjacent to the stream, each 2–3 m apart, arranged normal to the stream. During a 3 day storm event, the water table elevations and local gradients shift substantially, particularly in relation to the far riparian well. These data suggest that flow direction in the riparian aquifer and connection to the channel change throughout this storm event. Likely, there is a coincident down-valley change in water table gradients that is not expressed here. At the scale of a 23 km2 mountain catchment, Jencso et al.  have documented the temporal dynamics of hillslope-riparian-stream hydraulic connections (i.e., when the hillslope is contributing water to the riparian-stream system) as a function of lateral contributing area. They found that larger lateral contributing areas had a more consistent aquifer-stream connection.
 We propose that a stream is a dynamic expression of local groundwater conditions, where exchanges of water between the catchment and the channel are consistently changing in response to heterogeneous temporal and spatial water table dynamics. These dynamic changes to boundary conditions of the channel occur at several spatial and temporal scales. Whereas we can generally acknowledge that streambeds are made of porous media, our development of understanding hydraulic process is largely confined to the boundaries of the channel. To improve our understanding of how streams are dynamically connected to catchments, we must at least (1) develop our conceptual and numerical models of streams in their catchment settings and (2) develop and apply improved techniques to detect the water table dynamics at space and time scales that influence stream-catchment hydrologic connections.
 One severe limitation in our approach to characterizing stream-catchment connections is that we have relied heavily upon observations of solute transport in the stream alone. The transport of solute downstream is affected by a host of processes that we ultimately attempt to interpret from solute breakthrough curves observed downstream, including advection, dispersion, and exchange between fast and slow water (and also reaction in the case of nonconservative tracers). Whereas our physical and mathematical understanding of advection and dispersion is quite mature, our conceptual and numerical models of exchange continue to evolve in an effort to better fit a simulation of transport to an observed solute breakthrough curve. The solution to this challenge is not extensive collection of solute concentrations from the subsurface, as demonstrated by Harvey and Bencala  and Wondzell . However, emerging new techniques, such as using electrical resistivity during stream tracer experiments to image solute distribution in the subsurface in two and three dimensions [Ward et al., 2010a], are developing with promise of advancing our conceptual and numerical models of solute transport [e.g., Ward et al., 2010b] and stream-catchment connections.
4. Opinion Summary
 The 10 m to 1 km scale of stream-catchment connections we have discussed is inherently “messy.” The physical process characteristics are heterogeneous throughout the catchment. This will make detailed process modeling unreliable; while treating the catchment as a box and the stream as a pipe will not account for the transitions and transformations we know occur. We propose here that through the previous 25 years of detailed study of smaller spatial scale hyporheic exchange flow we now have considerable understanding of the degree to which streams are continuously connected to their catchment. Flow and solute exchange with a stream and the connectivity of a stream to its catchment are seemingly straightforward characteristics of a stream. Basic hydrologic measurements at the stream scale done over the catchment scale are no doubt time and labor intensive to collect. However, such approaches, as we suggest (measurement of flow along the channel, surface and subsurface solute sampling, and routine measurements of the water table) are needed to establish and quantify the extent to which stream-catchment connections are determining solute transport and transformation.