Analyzing riparian zone ecosystem services bundles to instruct river management

ABSTRACT The ecosystem service framework is now well accepted for focussing management strategies to preserve and restore ecosystems. Its implementation remains challenging, however, due to the environment’s complexity and dynamics that interfere with ecosystems’ ability to provide the services. Here, we question whether we can show where and how to intervene in riparian corridors to restore specific ecosystem services without endangering others. Specific hypotheses in this context are for the spatial aggregation of ecosystem services delivered by riparian corridors with respect to naturalness (1), to the existence of bundles of ecosystem services (2), and finally for the scale sensitivity of this congruence (3). Within a Geographical Information System framework, we analyse the capacity of riparian corridors to provide ecosystem services over three river basins in the Bresse region (France) based on high-resolution data of the riparian corridor hydromorphology and land use. Specifically, we compare the capacity to provide two services: in-stream water purification and riparian retention of nutrients that are critical goals for river management and rehabilitation strategies. We observe little spatial association and high spatial variability for the two emphasized ecosystem services. Surprisingly, no congruence of ecosystem services with riparian corridor naturalness is present. The absence of associations between ecosystem services and their spatial variability will oblige environmental managers to identify underpinning environmental processes and patterns at local scales. In conclusion, we plead for fine-grained multifunctional assessment of ecosystems’ capacity to deliver services, especially in environments such as river corridors that exhibit high environmental heterogeneity. EDITED BY Neville Crossman


Introduction
The overall frameworks to assess ecosystems' capacities to provide services are now well accepted and expected to deliver operational measures for management strategies and planning (Haines-Young and Potschin 2010;Lautenbach et al. 2012;Allan et al. 2013). Especially the supply of multiple functions and services in these frameworks is seen as a valuable asset for management strategies. Initially, it was strongly embraced as a framework to reconcile societal and ecological demands and visions, in an assumed harmony of services delivered, for planners and managers to 'cherry-pick'. Some limits to this harmonious picture have arisen from the observation that biodiversity is not always served by an ecosystem services-targeted approach, and vice versa (Adams 2014), giving way to a strong debate and the new discipline of Biodiversity Ecosystem Services research (Cardinale et al. 2012). The concept of a spatial and temporally consistent association between services, or ecosystem service bundles sensu Raudsepp-Hearne et al. (2010), is a very attractive idea for management.
Indeed, in practice, it is generally accepted and applied in that way. However, the dynamics of the ecosystem's ability to deliver the services in space and time still needs more attention. Most operational ecosystem service assessments undertaken (Burkhard et al. 2010;De Groot et al. 2010;Paetzold et al. 2010;Pinto et al. 2010) elaborated a comprehensive work to appraise a status at a specific point in time. Up to now, these assessments pay little attention to the spatial and temporal dynamics of the ecosystems. Riparian corridors provide a unique opportunity to explore such a focus because they are dynamic networks, influenced by strong directional connectivity that integrates processes across multiple scales and broad distances through time (McCluney et al. 2014). This flow context and network structuring of river basins has not yet been investigated with respect to the delivery of ecosystem services.
The question we try to answer in this paper is whether the presumed ES (ecosystem services) bundles of riparian zones can be detected over larger territories with land cover-based matrix model methods. This not only points at the scale sensitivity in the analysis of the cascade of ecosystem structure and functions to services (Burkhard et al. 2014), but also at the problem of riparian management options for improving ES delivery. We will address these issues via the following hypotheses: (1) Natural systems with undisturbed ecosystem functions offer maximal ES (within the specific geographic and societal context). Systems with maximum capacity to provide services are assumed to perform key ecological roles both for wildlife and for human well-being (Liquete et al. 2015). As such, we hypothesize the generally assumed strong association of service supply in so-called 'ES bundles'. (2) The ES approach is a bridge between societal and ecological demands, embracing social and natural sciences, and as such requires interand multidisciplinary methods; the more functions and services we will be able to detect and assess, the better the method is believed to reveal and reflect the whole picture of the ecosystem's services delivered (Schindler et al. 2014).
(3) Spatial aggregation of societal and natural functions within ES approaches is sufficiently clear with low-resolution spatial grids or entities (e.g. CORINE land cover units). The need for spatially detailed information to identify processes behind services does not hamper the lower-resolution ES assessments to reveal significant and accurate patterns. Riparian corridors are deemed to deliver an exceptional amount of ecosystem services (Capon et al. 2013;Thorp et al. 2010), thanks to their arterial position in the landscape, agglomerating solid and liquid fluxes above-and below ground. Riparian corridors and wetlands have been ranked the second best ecosystems globally for ecosystem services provision (Costanza 2008;Acreman et al. 2011). Riparian forests are generally appraised to deliver services for water quality control, especially for nutrient retention both by plant uptake and by denitrification (Hill 1979(Hill , 1996Haycock et al. 1993;Dodds and Oakes 2006;Curie et al. 2011;Van Looy et al. 2013). However, strong disparities in proposed strategies exist when ES frameworks are applied to their rehabilitation (Bark et al. 2016). This can be illustrated by different proposed strategies to solve the eutrophication problems in Chesapeake Bay in the USA. First, a catchment-scale analysis of riparian buffer zone effects on nitrogen retention suggested that restoration of 70% riparian forest cover over the basin would make an end to eutrophication problems (Weller et al. 2011). Consecutively, a denitrification-oriented analysis suggested in-stream flow restoration as the most effective solution (Filoso et al. 2015). Both studies investigated the social-ecological system and the ecosystem service of water quality improvement but in terms of biogeochemical processes and functions one focused on catchment and riparian buffer retention and the other on aquatic denitrification. Making recourse to ES frameworks to validate these approaches gives the false hope of an integrated and multidisciplinary vision to the question. This example illustrates clearly the problems of association, multidisciplinarity and the dynamics in the capacity to deliver services.
Here, we direct attention to specific ecosystem functions as the 'means' of ES provision (Wallace 2007) that we assume as delivery of services (Danley and Widmark 2016). For consistency in our arguments, we follow the ES nomenclature of Villamagna et al. (2013). Generally, due to inherent complexity in ecosystems, a single process or function intervening in the delivery of services is investigated within a wide array of intervening processes (Bennett et al. 2009), as illustrated above for the eutrophication problem of Chesapeake Bay. Even though some straightforward solutions might arise, most ecosystem processes involved are highly complex with many biotic and abiotic factors entering the analysis. As an example of this, in an attempt to model the different processes and pathways of a freshwater ES, Johnston et al. (2011) highlighted over 7000 variables. Moreover, the processes underpinning the capacity of an ecosystem to deliver services are often strongly spatially variable depending on local climatic, geomorphic and biotic factors (Feld et al. 2013;Grêt-Regamey et al. 2014). Here, using riparian corridors, we highlight some of these methodological and practical issues in the deployment of the ES approach. It should be noted that assessment of ES on rivers is still in its infancy (Gilvear et al. 2013). Exceptions include the incorporation of nitrogen retention (Grizzetti et al. 2008;Liquete et al. 2011;Lautenbach et al. 2012;Natho et al. 2013), water quality (Keeler et al. 2012;Brauman 2015), water provision (Notter et al. 2012) and flood regulation (Nedkov and Burkhard 2012) within the river network. Gilvear et al. (2013) have proposed a framework for assessing range of ecosystem services within river networks, and Large and Gilvear (2015) have identified potential river attributes and data sources for undertaking such assessment. In this work, within a Geographical Information System (GIS) framework, we examine ES delivered by riparian corridors over three river basins in one geographic regionnamely the Bresse region of France. More specifically, we compare two processes: in-stream and riparian retention processes of nutrient control by river management. To answer the hypotheses cited above, a scale-sensitive finegrained analysis to the continuity of riparian (forest) cover is needed to identify and to infer physical and biotic responses (Tormos et al. 2014a) and associated ecosystem functions (Tormos et al. 2014b). For this purpose, geospatial data within a GIS are analysed to characterize the physical nature of the riparian zones with a focus on riparian forests. Furthermore, a set of spatial indicators, available over the whole of the river network of France, is deployed in this regional study.
We can translate the stated hypotheses to our specific question for ecosystem service provision capacity of in-stream and riparian nutrient retention: Are these two ecosystem functions spatially associated? Are they congruent with the provision of other ES? Are they different at small scale or more 'regionally' or upstream-downstream organized?

Study region and methods
Three stream catchments of the Chalaronne, Veyle and Reyssouze rivers, all tributaries of the Saône river, in the Bresse region in East-France were selected for study. These all have catchments with a mixed agriculturalforest landscape with scattered villages and small towns. The climate is moderately continental with an annual precipitation of between 700 and 900 mm. Summer temperatures are high with a July average of between 19.6 and 21.5°C. The three river systems are rain-fed, creating strong flow contrasts between autumn/winter flows and severe summer low flows. The combination of these hydrological characteristics and the strong land use pressures gives an overall strong risk of eutrophication in these catchments. Valley slopes are between 0% and 0.7% and the basins are between 300 and 700 km 2 in size.

Ecosystem function and service selection
Establishing the spatial congruence of services does not necessarily mean that they arise from the same process. Therefore, in our approach, we applied an ecosystem function-based ES definition that distinguishes for the specific processes. A range of ecosystem functions and services provided by riparian zone and floodplain ecosystems have been identified (Costanza et al. 1997;Atkins and Burdon 2006;Acreman et al. 2011). For our analysis, we determined, based on expert knowledge, the main riparian zone functions and processes that determine the level of ES provision. These functions were the effect of vegetation presence and structure in the riparian zone to (1) habitat availability and (2) connectivity allowing movement of organisms through the river networks, the buffering functions of (3) pollution retention and of (4) microclimate control, the (v) water purification in the river bed and (vi) carbon sequestration in riparian zones. The riparian ES can be quantified at relevant scale and precision, using the matrix method Grêt-Regamey et al. 2014).

River network segmentation
To be able to develop strategies focussed at nutrient control and riparian and in-stream rehabilitation, we link and weight the ecosystem functions relevant to the riparian zone and processes present (Stürck et al. 2014). Therefore, an important step in ES valuation is to define the appropriate spatial scale and the possible service variation in space, especially in land use analysis Luisetti et al. 2011;Gilvear et al. 2013). A river segmentation procedure applied to the river network undertaken in an earlier study (see Van Looy et al. 2015) produced a total of 292 river segments (homogenous hydromorphic units) between 0.8 and 8 km in length, with an average value of 2.3 km. For each of these segments, land use and hydromorphological data were identified to characterize the segments. Catchments' surfaces are between 300 and 700 km 2 , river lengths between 20 and 75 km and annual average discharges between 2 and 7 m 3 /s (Chalaronne: 333 km 2 , 52 km, 2 m 3 /s; Veyle :670 km 2 , 67 km, 6 m 3 /s; and Reyssouze: 495 km 2 , 75 km, 5 m 3 /s). For the riparian corridor specifically, elements of riparian forest cover and infrastructure/urbanization in riparian buffers of 10, 30 and 100 m from the river's edge were identified (valley floors range from 0.2 to 2 km in width) and mapped based on orthophotograph interpretation (0.5 × 0.5 m resolution). This results in a spatially explicit data set on riparian cover with calculation of continuity at river segment level and upstream corridor. To these segments, we applied the method of Large and Gilvear (2015) for ecosystem services evaluation of river reaches. This approach is based on expert-judged scores per ES, with emphasis on selecting relevant services and functions.

Scoring and quantification
We apply the currently used ES framework of the matrix method Crossman et al. 2013). Under this umbrella, the approaches still show panoply of specific methods and concepts to assess the delivery of ES (Schägner et al. 2013;Burkhard et al. 2014). Newly developed methods for ES assessment of riparian corridors are directed at gathering catchment-scale spatial information for the river network (Liquete et al. 2015). Even though wellknown examples of economic valuation of river corridor services exist (Dubgaard et al. 2005;Murray et al. 2009), we chose a non-economic expert-based approach Gret-Regamey et al. 2013;Jacobs et al. 2015;Stoll et al. 2015). The scoring (Table 1) Table 1. Rules relating to attributing river (corridor) features or land cover types to potential ecosystem service supply score (0 is very low/absent and 4 high). and Grabowski 2016), and for most factors, we had recourse to the evaluations developed for the hydromorphological conditions of the river segments (Van Looy et al. 2015). Where we make recourse to land use and other proxies for the functions, our approach is both proxy based and phenomenological, according to the description of Lavorel et al. (2017), and definitely to a higher resolution (0.5 m × 0.5 m) than the land cover proxy-based approaches mostly referred to.
All ES score assessments are normalized, i.e. converted into the same ordinal scale (see Liquete et al. 2015) to allow an integration and comparison at river segment level. With our objective of evaluating the spatial configuration of delivery of ecosystem services, we present the services equally valid with our relative scoring, an objective way to evaluate geographical distributions of ecosystem services (Thomas et al. 2013).
To evaluate specific functions with regard to riparian corridor land use and configuration, the elements contributing to riparian retention and in-stream purification are based on the river retention model GREEN (Grizzetti et al. 2008;Bouraoui and Grizzetti 2011;La Notte et al. 2017). Nutrient retention was estimated based on documented retention rates from flow volumes, flood duration and habitatspecific retention rates (Olde Venterink et al. 2003) and in-stream retention estimates from De Klein and Koelmans (2011). The relative scoring of in-stream retention is in the first place based on the 'risk of alteration' identified for the river bed that mainly addresses the risks for the natural bed substrate to be disturbed by clogging with fine sediments that disrupt the exchanges with the interstitial spaces of the river bed where the purification processes occur. Furthermore is the stream longitudinal and crossprofile alteration identified, since both the variety of stream facies, stream velocity variation and the length of the stream are all quantitative measures for the purification process.
Carbon sequestration annual accumulation is documented for floodplain forest and marshland as 0.1 and 2 ton C ha −1 year −1 for wetlands and woodlands, respectively (Nabuurs and Schelhaas 2002), allowing relative scoring of these land use types in the floodplain. For the biodiversity ES, we looked at the river corridor habitat ecosystem function. We used a combination of the undisturbed hydromorphological character of the river bed, and a proxy for habitat availability, the area of natural areas in the floodplain. Under these, natural land use classes are natural herbaceous vegetation (including wetlands) and riparian forests of the area that are mixed Alno-Padion alluvial forests with mainly ash, alder and oak trees. This alluvial (and aquatic) vegetation is the strongest biodiversity value (protected under the Natura2000 network) in the area. To distinguish for this biodiversity/ habitat classification in the riparian corridor, tree plantations (mainly poplar cultivars) are not in this land use category.
For the connectivity function, we combine the riparian corridor continuity with the aquatic environment's continuity with respect to weir presence. We do not incorporate provisioning services as they are either not relevant for the studied area (i.e. the riparian forests are not harvested and no hydropower is present) or not significant/measurably influenced by corridor management (e.g. commercial or recreational fisheries). Overall, the riparian forest continuity and cover is not the only element here under consideration, but a recent overview study qualifies it as the index that is able to assess the largest number of ecosystem services in the fluvial and riparian system (Vidal-Abarca et al. 2016).

Analysis of associations and bundles of ecosystem services
The first step following determination of scores is to identify spatial patterns, across all three river networks in the services per river segment. We follow the proposed method of Mouchet et al. (2014) for quantifying ecosystem service associations, using correlation testing over the river segment ES matrix. Since we apply relative scoring, the Kendall correlation test is used in order to give the least weight to the actual values.
In a second step, we define ES bundles. The analysis uses the approach proposed by Raudsepp-Hearne et al. (2010) and subsequently developed further (Mouchet et al. 2014). We identify ES bundles by hierarchical cluster analysis (using Ward's method). This identification of bundles confirms whether the identified associations are spatially consistent.

Analysis of river size, network and geographical structuring
To identify the spatial variation of services in relation to river type, network position and geographical context in general, we perform a correlation test to the major structuring elements of river and geographical context with respect to the hypotheses formulated earlier: (1) the river size determined by the Strahler system of stream ordering as a basic proxy of flow quantity; (2) the hydrological alteration risk determined by aspects of water abstraction (pumping and irrigation) and flow regulation (presence of ponds, impoundments and lakes); and (3) the upstream basin area as a measure of the landscape geographic setting.

Associations of ES
A number of associations of ES were found to be present with varying levels of correlation present from near zero to nearly 0.7 (Table 2). In terms of level of correlation and numbers of correlations, habitat, microclimate, pollution retention and carbon sequestration score highly. In-stream purification and connectivity score poorly.

Ecosystem service bundles
Two bundles are identified by hierarchical cluster analysis (55% of variance grouped): a first group with Corridor habitat provision-Microclimate control-Carbon sequestration-Pollution retention and a non-associated group of two separate functions: Connectivity and In-stream purification. Even though there is a high evenness in capacity to deliver services over the river network (averaged overall ES for river segments 2.5 with small standard deviation of 0.27), the associations are not that strong. Comparison of in-stream (self-purification) with out-of-stream pollution retention shows only for the Reyssouze basin some congruence (Figure 1). Although this basin also shows a strong difference for the downstream tributaries, where the in-stream quality allows purification functioning, there is no contact with banks and valley for retention function. For Veyle and Chalaronne rivers, in-stream and riparian retention capacity are strongly differing spatially, especially for downstream sections.
The scores for ES delivery, even averaged over all services (Figure 2), show strong spatial heterogeneity and little continuity, with no geographic/environmental gradients (elevation, distance to source, tributary-river main stem differentiation). For the riparian corridor functioning to pollution retention, only minor significant correlation to river size (Strahler order Kendall's tau coefficient 0.153) or discharge is present (Table 3). In contrast, in-stream functioning has a minor negative correlation, due to higher rates of river hydromorphological alteration for the larger systems, especially for the rivers' main stem. Forest cover-related services are generally stronger present downstream, since there is a higher agricultural pressure to small streams on the upstream plateau, whereas connectivity and in-stream purification (Kendall's tau −0.261 and −0.138, respectively) are lower downstream, since flow regulation and alteration are generally greater in downstream reaches.
The sum of the services (Figure 2) delivered by the corridor functions is very weakly but significantly correlated to river size (Strahler order Kendall's tau coefficient and upstream basin surface Kendall's tau coefficient), meaning a slight accumulation of services downstream.

Discussion
It is well acknowledged that continuous forested corridors lead to an improvement of physical and biotic conditions of streams and rivers (Hill 1979(Hill , 1996Haycock et al. 1993). Nevertheless, questions as to whether the configuration and specific rate of riparian canopy gaps is crucial in the pollution retention processes (Weller et al. 2011), and whether the upstream or downstream basin context prevails for biotic corridor functioning (Brown et al. 2011), remained unanswered. We present an ES framework that deals with these aspects over multiple catchments. To the presumed congruence of different services and the necessary scale of analysis, we show that when we look in more detail to the different services provided by riparian corridors, we find striking spatial separations and little congruence. Where some studies on ecological restoration indicate synergies between multiple ecosystem services (Jiang et al. 2016), here we find no real support to the strong spatial aggregation presumed in the ES bundles concept (Raudsepp-Hearne et al. 2010). In line with the findings of Bai et al. (2011), the services can be divided into two groups that should be managed and conserved independently. Also supporting the findings of Bai et al. (2011), we find that the corridor habitat biodiversity service was positively correlated with carbon sequestration, but contrary to their observation, it is also positively correlated with nutrient retention. This difference can be attributed to the difference in scale and in landscape contexts of the analyses. The same is true when we compare our findings to national scale study in Great Britain by Thomas et al. (2013) who found that biodiversity and carbon storage ES were negatively correlated. Clearly a different scale of analysis can lead to different observations, yet these authors also conclude for a 'combined' strategy of conservation.
For the delivery of the two specific services of water purification in-stream and nutrient retention in the corridor, there clearly is a strong spatial incongruence through the different catchments, without a clear pattern linked to geography or position along the river continuum. So, strategies oriented in-stream and at the river bank can be complementary. But emphasizing only one of the two processes lacks efficiency. Managing for in-stream self-purification capacity does not enhance the other riparian ecosystem functions and services. Prioritizing riparian management at the river banks and floodplain has the benefit of improving a series of services. The same observation goes for the service of connectivity that is disconnected from the delivery of other riparian corridor services. Combining it with corridor habitat provision measures will highly improve the delivery of ES. So, we refute the first and third hypotheses that represent the premises of many actually used ES approaches and for the use of ES bundles.
For the second hypothesis, although the multidisciplinary analysis clearly adds information for planning and preserving ecosystem functions, the congruence of services is not increased by adding more functions and services to the analysis. Yet, we can identify specific strategies and operational bases for managers to improve the capacity to deliver services. The presented approach highlights features that are less obvious, unexpected from just the mapping of land use and hydromorphology elements. For our three basins, we observe generally highest capacity to produce services for the Chalaronne, followed by the Veyle and the Reyssouze lowest. The mapping of this capacity allows to identify significant spatial 'gaps' in ecosystem functioning, for which solutions can be proposed.
The identified incongruence confirms the reported risk that traditional conservation strategies oriented towards biodiversity may not be effective at protecting the economic benefits of an ecosystem and vice versa (Adams 2014). The pressure on ecosystems to provide various different and often conflicting services is immense and likely to increase (Moilanen et al. 2011). The spatial variability of the capacity to produce services stresses the importance of looking at the dynamics of ES. Ecosystem services exist at the point of interaction between ecosystem function and human activity (de Groot et al. 2010). Therefore, even with a constant biophysical supply of an ecosystem service, changes in human activity can alter service delivery . Furthermore is the human influence on the ES delivery a crucial aspect in the assessment; here, the nutrient status of these alluvial plain rivers lies close to adverse thresholds for the ecosystem. The heterogeneity in ES supply is also reported in other contexts of analyses of demand and provision of ES (Verhagen et al. 2016). But even in an absence of ES bundles, strategies to ES delivery are possible.
Riparian buffers are highly sensitive in their efficiency to filter pollution according to practices of drainage (Petersen and Petersen 1991;Petersen et al. 1992) and run-off (Weller et al. 2011). So, a finegrained analysis of processes and functions, managing the relationships among ecosystem services, can enhance the provision of multiple services and help avoid catastrophic shifts in ecosystem's capacity to provide services (Bennett et al. 2009). The influence that spatial scale has on these relationships was recently illustrated comparing a national scale with a river basin assessment (Holland et al. 2011). Here, we go up one scale level in spatial resolution to the functions at the level of individual riparian zones. Preserving and restoring riparian corridor green infrastructure will gain in importance and effectiveness when land use practices are more at risk. Filtering and purification services will be larger in intensive agricultural areas, and thus, restoration of green infrastructures in these areas has the highest efficiency and priority.
With the application of the matrix method Grêt-Regamey et al. 2014), we advocate ES delivery by land use as a proxy in the absence of detailed process models and system understanding. Nevertheless, we suggest high-resolution information on riparian corridor features (canopy cover) and hydromorphological characteristics (weir presence, bed alteration) up to the phenomenological identificationaccording to Lavorel et al. (2017) of the ecosystem functions assessed. Obviously, even relatively good proxies are likely to be unsuitable for identifying hotspots (Eigenbrod et al. 2010). Still, Lavorel et al. (2017) do admit that perhaps the greatest obstacle to substantial progress in assessing ecosystem services is a lack of datathere is simply none available for most services in most of the worldand that it remains a crucial first step in global efforts to conserve key ecosystem services by mapping their  spatial distributionseven if assessment precision is inaccurate. The services we deal with in this paper are highly variablein time especially, for instance, nutrient retention, making measurements and real estimations of the service very challenging scientifically and logistically. Therefore, we rely mostly on the process understanding derived from modelling approaches (like SWAT model at the scale here applied). These models are mostly developed at the catchment scale and obviously also the relevant scale for management and ecosystem service identification (Doody et al. 2016). The choice for land cover-based ES assessment is in this study nevertheless brought to a detailed level, thanks to high-resolution image analysis for the riparian forest continuityidentified in the scientific literature as most relevant factor (see Weller et al. 2011). Vermaat et al. (2016) quantified ES for riparian corridors per reach and summed the ES to annual economic value normalized per reach area. The resulting value ratio's differed more than 10-fold between restored sites. Here, our relative scoring is a simplification compared to the more economic approaches, yet it clearly highlights managerial contexts and priorities. Especially for comparison between catchments, with consistent scoring, this approach has strong merits for management and planning approaches.

Generality approach
Can we generalize our findings? For the functions and services we regarded, the main drivers are the riparian corridor features and human alterations present; the geographical context only plays a minor role. For other regions, distinction in the presented ES delivery might be for the presence-absence of functional floodplains, for which in this region there is no natural limitation. The generalities of the described results are obviously limited to the landscape context of our selected catchments.
The observed correlations can be explained by the interaction with the specific landscape context and ecosystem functioning. Carbon sequestration is correlated to pollution retention and habitat provision since these ES encompass the natural floodplain functioning. As these three are correlated, the overall capacity of service delivery is also strongest correlated to these three individual measures. But, even though correlated and thus congruent with several individual services, general corridor habitat ES not everywhere coincides with overall capacity. This implies that habitat enhancement and biodiversity-oriented management do not always or univocally means overall ES enhancement.
Strength of the approach presented is that it shows for groups as well as individual ecosystem services where improvement is possible. This can be compared with societal demand within the catchment in determining what improvements need to be made. Options for restoration of riparian zones in watershed contexts oriented to restore general ecosystem functions are moreover presented (Zedler and Kercher 2005;Capon et al. 2013), also stressing the need for differentiated measures (Fullerton et al. 2009). There still needs, however, to be a good understanding of the key components of ecosystem functioning that are a prerequisite for a good description of ES delivery (De Groot et al. 2010). The paradox in the strategy selection for nutrient control in Chesapeake Bay, stated in the Introduction, is linked to the identification of the trigger alteration of functioning; is the strongest potential for restoration in the in-stream nutrient control; or is it the retention in the floodplain and at the riverbanks. If we bring the analysis to this distinction, immediately the appropriate strategy for restoration will evolve (Thorp et al. 2010). The choice for in-stream (dam removal, remeandering, bed restoration and profiling) or corridor (bank replantation, floodplain/flood contact restoration) retention measures can be guided by the provided maps. Overall priorities need to be evaluated prior to this by evaluating the summed and other ecosystem functions and services supply potential.
With this analysis, we point at some caveats for using the ES framework in the design of restoration strategies. Even though some straightforward aspects in spatial and geographic context might arise, most ecosystem processes involved are highly complex and need fine-grained analysis and many biotic and abiotic factors entering the analysis. This need for finegrained analysis should never be left aside with the excuse of the multidisciplinarity and the larger-scale societal demand side of the ES approach.

Conclusion
To evaluate the potential for delivery of ecosystem services, the environmental and ecological processes behind the services need to be assessed at a relevant scale. Here, we looked at small rivers and the services they provide in relation to the riparian corridor functioning. The hydromorphological processes responsible for the delivery of services were evaluated at the reach scale based on a specific evaluation scheme.
We oriented our analysis to the central questions for the restoration manager: where, why and how to intervene in riparian corridors? We can highlight the strength of this fine-grained analysis to identify the local potential of ES delivery and subsequently the high resolution needed to consider and identify specific targets and functions to restore in riparian management. Ecosystem services appear as highly variable in space and associations or bundles of services are less evident than generally assumed.

Disclosure statement
No potential conflict of interest was reported by the authors.