Evaluating the process of ecological restoration

We developed a conceptual framework for evaluating the process of ecological restoration and applied it to 10 examples of restoration projects in the northern hemisphere. We identified three major phases, planning, implementation, and monitoring, in the restoration process. We found that evaluation occurred both within and between the three phases, that it included both formal and informal components, and that it often had an impact on the performance of the projects. Most evaluations were short-term and only some parts of them were properly documented. Poor or short-term evaluation of the restoration process creates a risk that inefficient methods will continue to be used, which reduces the efficiency and effectiveness of restoration. To improve the restoration process and to transfer the knowledge to future projects, we argue for more formal, sustained evaluation procedures, involving all relevant stakeholders, and increased and improved documentation and dissemination of the results.


INTRODUCTION
The accelerating degradation of the world's ecosystems has fostered a counter-movement to mitigate destructive impacts (Le Houerou 2000, Novacek and Cleland 2001, Lal 2004, Bernhardt and Palmer 2011).The topics of ecosystem restoration and ecological restoration have thus received increasing attention worldwide (e.g., Erwin 2009, Schmutz et al. 2014, Barral et al. 2015).For the last two decades the Web of Science (WoS) reports 2876 scientific papers on "ecological restoration" by 7 December 2015, but only 36 papers before that time.Papers addressing "ecological restoration" combined with "evaluation" are, however, much less common-only 177 according to the WoS, by 7 December 2015; the first one from 1995.Still, ecological restoration, defined as an "intentional activity that initiates or accelerates the recovery of an ecosystem with respect to its health, integrity and sustainability" (SER 2004), requires evaluation to document progress and inform adaptive management strategies (Williams 2011) in terms of the cost-efficiency of the restoration process and the positive effects on the recovery of degraded ecosystems.This is particularly important in complex systems (Gunderson and Light 2006).If many years pass before a restoration is evaluated, and if the restoration has failed, it follows that recovery will be delayed or failed.Ecosystems often require decades or centuries to recover after restoration has been initiated, especially in high latitude and high elevation ecosystems with short growing seasons (Forbes andMcKendrick 2002, Campbell andBergeron 2012).Under such circumstances, failures may be costly if repeated trials are required to restore the ecosystem (Aradóttir et al. 2013).To avoid problems arising from flawed design and implementation of restoration, the monitoring and evaluation of restoration should be given high priority.
A recent review of restoration in the Nordic countries indicates that ecological restoration projects in the region often completely lack formal evaluation (Halldórsson et al. 2012, Hagen et al. 2013).Other studies also show this to be the case in other parts of the world (e.g., Bernhardt et al. 2005, Suding 2011), although the number of empirical evaluations has increased during recent years (Wortley et al. 2013).If evaluation steps are properly described and justified, restoration processes can be improved in terms of cost-efficiency and ecosystem effects, and the lessons learned can be more easily transferred to other projects (Nilsson et al. 2015).Traditionally, evaluation has been equated to the monitoring of the postrestoration outcome, and such monitoring has often been restricted to a single or a few events (Kondolf and Micheli 1995, Zedler and Callaway 2000, Suding 2011).Such limited efforts are unlikely to provide a full picture of the restoration process and its outcomes.For more accurate and reliable results, restoration evaluation should be a continual activity that is an ongoing part of the entire restoration process (Allen et al. 2002).In other words, evaluation could consist of different subactions or steps during the entire restoration process from the beginning to the achievement of the restoration goal (Jungwirth et al. 2002, Hughes et al. 2011, Pander and Geist 2013).Another drawback is that the evaluations may often be too simple to allow reliable conclusions (Suding 2011, Morandi et al. 2014).These problems may backfire on the restoration itself in that project goals may not be reached and cost efficiency not secured, and future restorations not conducted.
The restoration process can be seen as consisting of three phases: planning, implementation, and monitoring (Hobbs andNorton 1996, Tischew et al. 2010).Actors responsible for each of these three phases should evaluate and improve their "within-phase" work, at least informally, to guide adaptive decision making, thus reducing risk of failure (Williams 2011, Loftin 2014).In addition, the interactions between restoration phases can also be evaluated.For example, the implementation of restoration requires that planners communicate with practitioners, i.e., the people responsible for the practical work ("This is how we want to restore").Similarly, the monitoring benefits from communication between practitioners and monitoring experts ("This is how we restored"), and the monitoring teams need to pass on their findings to the planners ("This is what happened").We propose Table 1.Summary of the information used for evaluation, how the evaluation was made, and by whom in the six evaluation steps (Figs. 1 and 2).

Evaluation step
What type of information was used as a basis for evaluation?
What was the information evaluated against?

Types of communication evaluated
Feedback to planning and implementation phases.Well documented evaluation could benefit new projects that all these three restoration phases can be improved by appropriate, within-phase as well as between-phase six-step evaluation (Fig. 1).In this paper, we discuss the relevance and usefulness of these 6 steps by analyzing 10 restoration projects in 8 northern countries.We ask the following questions: (1) Does the selected data set of 10 projects include examples of evaluation at each of the 6 steps?; (2) Are there examples, among the cases, of steps at which evaluation had an impact with respect to modification of current or future restoration projects?; (3) Based on the selected cases, what are the major limitations in the evaluation of restoration projects?

METHODS
We took inspiration from the paper by Hagen et al. (2013) who discussed ecosystem restoration in the Nordic countries.We searched for large, completed, or well-established, long-term restoration projects representing different ecosystem types, and added Greenland, Scotland, and Canada to cover a larger and more diverse area with such restoration projects.Among the preselected projects we chose 10 projects that fulfilled our criteria (Box 1).We included a variety of ecosystems typical of the northern hemisphere.We then analyzed the projects with respect to how evaluation had been made at the six steps, within-phase as well as between-phase as presented in Fig. 1.This analysis was made during an expert workshop where criteria for identifying the different evaluation steps were produced.We agreed that any type of value-laden information exchange within and between restoration phases could be categorized as evaluation.Because there were no previous frameworks available for such analyses, we constructed a new framework to assist in the data collection (Table 1).Our analysis then relied on information from scientific literature, reports, websites, oral communication with the restoration community, and our own knowledge about the specific projects.It should be noted that the selected case studies were chosen to illustrate the tool rather than to assess them per se.Given the diversity of restoration projects examined, we do not expect another choice of examples to produce very different results.1.Note that the evaluation process is not a single-loop, one-time practice; instead evaluation activity can go back and forth and be upheld during several years.
Birch woodland, Hekluskógar, Iceland.This project aims at restoring natural woodlands on about 900 km² of degraded land in the vicinity of the volcano Mt Hekla in South Iceland, to increase the resilience of the area against fall-out and secondary distribution of volcanic ash (Aradottir 2007(Aradottir , Óskarsson 2009a, b), b).Around 200 landowners participate in the project and the restoration takes place partly on their properties, but also to a large extent on public land.Actions involve establishment of plant cover and strategic establishment of seed sources of native birch and willows to facilitate natural distribution of woodlands.Restoration activities and results are regularly monitored and discussed with participating landowners (Óskarsson 2009a,b, 2011).
Rangeland, Farmers Heal the land (FHL), Iceland.This is a costshare revegetation project aimed at enhancing restoration and improving rangeland management and stewardship (Arnalds 2005).About 600 landowners participate in the project and the restoration takes place on their properties.FHL is organized by the Soil Conservation Service of Iceland (SCSI), which provides extension services, seed and funding to buy fertilizers, while the farmers provide land, machinery, labor, and in some cases additional fertilizers and mulch.SCSI officers make annual or biennial visits to all participating farms, during which restoration activities are planned, discussed, and subjectively assessed (Arnalds 2005, Berglund et al. 2013).Thus, the FHL operates as an "umbrella," but much of the planning and monitoring are done on an individual farm basis.Information about restoration activities is kept in the SCSI database and has been used in ad hoc studies that involve more in-depth evaluation on a subset of the FHL.Examples involve studies on stakeholder interactions and experiences (Schmidt 2000, Berglund et al. 2013, Petursdottir et al. 2013a), vegetation succession (Elmarsdottir et al. 2003, Petursdottir et al. 2013b), and carbon sequestration (Aradóttir et al. 2000).
Forest and peatland in the Green Belt of Finland.Green Belt LIFE project encompassed restoration of 600 ha of forest, 362 ha of peatland, forest roads, and quarries in 13 Natura 2000 areas in the eastern Finnish region of the Fennoscandian Green Belt zone during 2004-2008.Established restoration methods developed for protected areas were principally used, and some alternative methods were experimentally tested.Intensive scientific monitoring was an essential part of the project, requiring continuous discussion between managers, planners, and scientists.The project is representative of LIFE projects targeted to forest and peatland restoration in Finland.Information on the project was compiled from scientific (Laine et al. 2011, Similä and Junninen 2012, Tarvainen et al. 2013, Hekkala et al. 2014a,b, Similä et al. 2014, Hägglund et al. 2015, Tarvainen and Tolvanen 2015) and public papers (Similä andJunninen 2012, Similä et al. 2014), project proposals, restoration plans, maps, and practical experiences.
Grasslands, northern Great Plains, Canada.Millions of ha of central North America have been planted to low-diversity, highproductivity introduced grasses, on both private lands (farms, ranches) and public areas (parks, common pastures, roadsides).Restoring diversity involves removing or decreasing these species and introducing native species.Ongoing annual evaluation revealed the surprising persistence of the introduced grasses (Bakker et al. 2003, Wilson andPinno 2013).This challenge was addressed by a change in attitude: instead of completely removing introduced grasses, grazing was used to decrease their cover and incorporate them at a low abundance into the diverse community.
Montane grassland, Trotternish, Skye, Scotland.In response to Scottish Government concerns that excessive numbers of sheep were causing slope erosion on montane grassland, and were overgrazing grasslands within the Trotternish Ridge Special Area of Conservation, a vegetation and erosion monitoring program was set up in 1998.Small experimental plots were established on the steep slopes to enable monitoring of the effects of different grazing treatments on vegetation structure and erosion.The response to grazing removal was found to be very slow, with monitoring extended after 11 years for a further 6 years.During the time of this project other factors were also influencing the results: decline in number of farmers due to aging population and decline in sheep numbers due to changes in EU subsidies (Brown andBirnie 2012, Hewison et al. 2016).
Peatland, Caithness and Sutherland, Scotland.Restoration of peatlands and blanket bogs is occurring in many areas in Caithness and Sutherland.The restoration usually involves some form of drainage blocking to restore hydrological regimes, and where the area has been planted with commercial forestry, removal of the trees.Changes in hydrological regimes need to be carefully planned and the impacts on neighboring land taken into account.Some of this work has been funded by EU-LIFE projects with the evaluation of the project occurring in part during the reporting process of the projects (Lunt et al. 2010).
River, Skjern River, Denmark.In the 1960s the Skjern River was channelized and nearby meadows were ditched to increase agricultural production.After a few decades pollution due to N and P leaching and mobilization of ochre became obvious and it was decided to restore the river with a focus on pollutant removal.Different methods were much debated in the mid-1980s.The parliament decided in 1987 on a large restoration project focusing not only on cleaning processes but also on habitat improvements and recreation.The physical work, based on a construction law and environmental impact assessment from 1998, was implemented 1999-2002.Afterward stakeholders discussed the use of the area and scientists evaluated the outcome.The entire project is so far the largest restoration project in Denmark (Pedersen et al. 2007a,b, Pedersen 2010).
River, Vindel River LIFE, Sweden.This project, located in northern Sweden, restores the river network after the impact of timber-floating between the mid-1800s and 1976.Restoration measures include removal of structures like piers and dams, recreation of fish spawning beds, and diversification of channel morphology by putting back coarse sediment and tree trunks in channels.In some areas, experimental restoration has introduced large boulders from adjacent uplands into the channels (Gardeström et al. 2013).The results of the restoration are monitored, especially with respect to riparian vegetation and fish (Helfield et al. 2007, Palm et al. 2007, Polvi et al. 2014, Hasselquist et al. 2015, Nilsson et al. 2015).

RESULTS
We summarize our major findings on how the evaluation was made in the 10 chosen restoration projects (Box 1, Appendix 1), following the 6 steps in the conceptual framework model (Fig. 1, Table 1).For consistency, we standardized the names of the actors identified in each of the three phases of restoration, although we recognize that their actual roles and denotations varied between projects and countries.Thus, in short, planners planned, practitioners implemented, and monitoring teams monitored.Our framework for the evaluation of restoration (Table 1) was designed to maximize the information gained from the restoration projects that can be used to maximize the effectiveness of future restorations.We are aware that some evaluations encompass two or more of the six steps.For the sake of simplicity, however, we assigned them to the step where they had their main focus.It was not possible to consistently collect information on the types of evaluation metrics or qualitative information types used in the different projects.For this reason, the assignment of scores or ratings for the quality of evaluations was not possible.On the other hand, a scoring framework could add substantial value to the evaluation process.To make such a framework meaningful, however, restoration actors would be required to document their evaluation steps very carefully, and this was not the case in the chosen examples.Irrespective of the type of evaluation data, however, oral interaction between actors was the most common and effective way of sharing information.It should also be stated that our goal was not to provide detailed descriptions of the evaluations that had been done.Our main goal was to describe when and how evaluation should be done, with examples of how it influenced restoration.
Step 1: Evaluation within planning All projects included some form of evaluation and adjustment of original ideas during planning.However, the choice of partners in this process differed between projects.In most cases evaluation involved interaction between restoration practitioners and landowners.In a few cases evaluation and adjustment of plans were made by restoration practitioners and experts together (e.g., Dovre Mountains in Norway and Green Belt LIFE in eastern Finland; LIFE is the EU's financial instrument supporting environmental, nature conservation and climate action projects throughout the EU).There were examples when restoration practitioners involved more partners in addition to landowners, such as governmental bodies, funding agencies, NGOs, scientists, and conservationists (e.g., Skjern River in Denmark).There were several examples of changes made to the plan based on the evaluation.In the Vindel River LIFE restoration project in Sweden, a dialogue with landowners resulted in (temporary) withdrawal of some landowners from the project.In the Hekluskógar project in Iceland, some areas were excluded from the project because of farmers being concerned about continued use of grazing commons.In the Skjern River restoration project, negotiations with landowners and NGOs resulted in modifications of plans to protect migrating salmon and trout.In the Northern Great Plains restoration project in Canada, scientists were involved to deal with questions about seed origin, seed types, seed nativeness, and sowing practices.In the Green Belt forest restoration project, a dialogue with scientists resulted in establishment of control sites and specific studies on the impact of reindeer grazing on plant regeneration.
Step 2: Evaluation between planning and implementation Many projects used meetings, workshops, and media to share information about the practical implementation and to receive feedback on the techniques and location of restoration and its anticipated outcome.For example, in the peatland restoration projects in eastern Finland (Green Belt LIFE) and Scotland (Caithness and Sutherland), feedback from landowners influenced the planning if restoration jeopardized the economic benefits from forestry or grazing on neighboring, private land.The restoration plans were then modified or restoration activities moved elsewhere.The example from the Vindel River LIFE restoration project shows that information and visits to restored sites stimulated landowners, who earlier had rejected restoration actions on their land, to change their decision.Restoration could thereafter be implemented on their land.Interestingly, even more radical practices than those originally suggested were then used because implementation practices had developed over years based on previous evaluations.On the other hand, restoration projects initiated at high governmental levels, such as parliamentary decisions, may restrict the flexibility of restoration plans as seen in the example from the Skjern River.Only minor adjustments were made following the mandatory environmental impact assessment after the initial project plan.Although the Dovre Mountains project has clear obligations to the Norwegian parliament, and the Norwegian Defense Estate Agency had been fully responsible for the project, both expert knowledge and stakeholder opinions were incorporated into the final implementation.In the Northern Great Plains grassland http://www.ecologyandsociety.org/vol21/iss1/art41/restoration project, scientific experts were partners with the National Park staff and collaborated on the technique and location of restoration, as well as evaluation of its effectiveness.Technical evaluation of whether the project was implemented as planned varied among projects.The Green Belt project applied similar procedures to the Canadian project to make sure that the size and number of restoration sites and the applied practices followed the restoration plan.
Step 3: Evaluation within implementation Most projects used already established ("best-practice") restoration practices.However, because of evaluation during the implementation step ("learning by doing"), adjustments were also common.In the Northern Great Plains grassland restoration example, observations during implementation suggested a need to allow soil-seed contact by removing extant vegetation, which led to higher success of reintroduced plants.Improved scientific methods were also tested: in the Green Belt peatland restoration project, a new and more expensive method of wood removal was applied by the implementer (a researcher) without extra cost to the project owner.In cases of insufficient experience, possibilities for adjustments were incorporated into the implementation plan.For example, in the long-term Skjern River restoration project, tenders were made stepwise to incorporate initial experiences in later phases of implementation.Evaluations took place in the field or as regular meetings and included project owners, planners, and contractors.This also occurred in the Dovre Mountains where the project owner and experts regularly met with the machine operators in the field.This was also the case when public officers and farmers in the project Farmers Heal the Land discussed field methods.Such evaluations were often informal and produced limited documentation.The outcome of evaluation was seen as modifications on a practical level, e.g., in several projects where distribution of plants, turf size, selected type of gravel, and technical equipment used, led to changes in implementation.Large-scale modifications were also made, as in the Skjern River example where modifications in the planned movement of soil led to an enlargement of an already planned lake.
Step 4: Evaluation between implementation and monitoring In this step practitioners passed information to the monitoring teams about the accomplished work and its location.Deviations from the original plan were highlighted so that monitoring could be conducted in the areas where it made most sense.For example, in the restoration of montane grassland in Scotland, maps showing which type of fence (sheep only, or sheep and rabbit) and their locations were passed on to the monitoring team.In the Green Belt forest restoration project, monitored sites had to be moved because of mistakes made in the placement of restoration.In the Vindel River, practitioners updated the monitoring team on changes in methods to facilitate monitoring.The bulk of information flow in this step went from the practitioners to the monitoring teams.However, the monitoring could also feed back to the practitioners, potentially resulting in direct modification of ongoing work on the site.This was the case at the alpine heathland restoration project in Dovre Mountains, and in the rangeland restoration project in Iceland, where there were clear avenues for informal dialogue between the practitioners and the people carrying out the monitoring.When the restoration work was conducted as part of an experiment, such as the montane grassland restoration in Scotland, the monitoring team had to communicate with the practitioners to ensure that the layout of the plots provided sufficient replication.Evaluation in this step was often ongoing throughout the project.There were also examples when monitoring started before implementation of the restoration to gather baseline data, e.g., in the Green Belt peatland and forest restoration projects.In these cases the implementation activities such as ditch-closing and logging were planned to avoid damaging the groundwater sampling wells and monitoring gear.
Step 5: Evaluation within monitoring Formal monitoring usually started after the practical work was finished.Some exceptions were found, however, as in the Dovre Mountains alpine heathland restoration project where monitoring started four years before the full-scale restoration was implemented in order to support the project with relevant background information and allow "before and after" comparisons.In some cases, the chosen variables did not show much change.For example, very little biotic change was found in the Vindel River during the first two decades after restoration.To ensure that monitoring succeeded in detecting an extremely slow response, monitoring was continued using the same variables.In the montane grassland restoration project in Scotland, a similar situation was solved by adding 6 more monitoring years to the original 11.Long-term monitoring also allowed the documentation of restoration failure, e.g., the grassland restoration project in Canada in which native species were replaced by successional waves of exotic invaders over 18 years.To ensure that a lack of response was not caused by monitoring the wrong variables, additional variables were sometimes chosen for monitoring, but monitoring of the original variables was maintained.In the Canadian grassland restoration project, the unexpected flowering of prominent target species was quantified.In other cases, the restoration led to emergence of new microhabitat types, such as those in ditches following restoration of drained peatlands in Finland.In these cases, new plots were established in addition to the original ones in order not to miss any change in the restored site.
Step 6: Evaluation between monitoring and planning Monitoring teams often had double roles in that they reported back to both restoration practitioners and planners.In both these cases, information could be evaluated (steps 4 and 6, respectively).An important part of the reports consisted of more or less standardized documentation of the project results.Reports to planners suggested modifications of the plans and project designs that could either be ignored, left for consideration in future projects, or assimilated in current projects.We found several examples of monitoring results being adopted in current and future projects.For instance, in the Vindel River, studies showed that water was not slowed down during high flows because of the lack of large elements such as big boulders and tree trunks.The large elements were consequently incorporated in the project and suggested to planners for inclusion in future projects.A similar example comes from the removal of military roads in the Dovre Mountains where monitoring provided hands-on advice on how to modify the plans, a modification that was also implemented.In the grassland restoration in Canada, monitoring showed that the original plan to eliminate nonnative plants was unrealistic, and the agreed compromise was to keep them but manage the land in such a way that they were kept at a lower abundance than http://www.ecologyandsociety.org/vol21/iss1/art41/before.In the Skjern River, monitoring led the project owners to adjust the boundaries of the restored site and to compensate the landowners economically.Monitoring also led to changed grazing strategies in the restored area.In the Finnish forest restoration and other similar forest restoration projects, monitoring showed that tree cutting was an inefficient method to initiate succession or bring back the desired threatened species and this information was used to make subsequent plans.
Our results indicate that evaluation could have an impact in all of the steps, but that the importance varied among steps (Fig. 2).Although the first four evaluation steps were most important for ongoing restoration projects, the last two evaluation steps were most likely to affect future projects.Fig. 2. Potential of evaluation to cause changes in project planning, implementation, monitoring, and future projects.Black = high potential; Dark gray = moderate potential (modifies some but not all projects); Light gray = possible changes on rare occasions but lack of documentation (at least in the projects analyzed in the present study); White = changes are unlikely to happen.Evaluation steps are defined in the Introduction and in Table 1.

DISCUSSION
Our analysis clearly demonstrates that ecological restoration projects can include evaluation throughout the restoration process.In our 10 case studies, the three basic restoration phases, planning, implementation, and monitoring, all involved components of evaluation and reflection, within as well as between phases.Thus, we got a clear "yes" answer to our first question on whether the selected data set includes examples of evaluation at each of the six steps.Our case studies also demonstrated that evaluations can be formal as well as informal, and in many projects both kinds occurred.Formal evaluations were usually linked to mandatory processes such as environmental impact assessments, hearings or land-use planning, research and scientific publication, or strict protocols.Examples of informal evaluation included discussions and other exchange of information between actors involved in restoration, but also critical thinking by individual actors.No actor was formally excluded from the evaluation process, but the combinations of actors involved varied among phases and projects.In general, informal evaluations were poorly documented, if at all, but still built up important experiences and knowledge among the actors involved.
In many of the case studies, the informal evaluation processes also led to important modification of the restoration work.This means that also the second question on whether there are steps at which evaluation led to modification of projects can be answered "yes."In general, the large EU projects Green Belt LIFE in Finland and Vindel River LIFE in Sweden had more mandatory evaluation processes than the other projects, which is reasonable given that large projects are expensive and may affect many peoples' lives.This does not necessarily mean, however, that the evaluation is always satisfactory.For example, Morsing et al. (2013) analyzed 13 completed LIFE projects in Denmark and found that their evaluation was focused on ecosystem structures, which was not considered sufficient to assess the recovery of ecosystem processes.Measuring structures is often the key to evaluate processes because direct quantification of processes can be rather difficult, although not impossible (Muotka and Laasonen 2002, Ruiz-Jaén and Aide 2005).Ecosystem processes related to the nutrient cycling, such as decomposition and mineralization rates were in fact monitored in the Green Belt LIFE project (Tarvainen et al. 2013), but these evaluations were made using other sources than the EU funding.As the basic knowledge on the restoration impacts on the ecosystem structure is continuously increasing, restoration projects can direct more funding to the evaluation of ecosystem processes in the future.
Evaluations in the analyzed projects had three aims: (1) to assist the restoration process, (2) to judge the outcome of the restoration works, and (3) to gather information that could serve as a basis for deciding whether experiences from recent projects could be used in future projects.With respect to the outcome, evaluations could either identify which restoration projects had achieved their goals, or identify weaknesses that could lead to changes in one or more of the planning, implementation, and monitoring phases, for future projects.To fully evaluate the outcomes, all relevant variables, including the social ones, should be included.Barthélémy and Armani (2015) noted that social processes and local experiences are often ignored in restoration projects and Aronson et al. (2010) and Blignaut et al. (2013Blignaut et al. ( , 2014) ) found that the benefits of restoration for society were not given due attention.In most of our case studies, however, the social component was fairly well included in the projects.Information used in the process of planning evaluation is especially interesting to the public, because people tend to resist change, including changes wrought by restoration (Oreg 2003).However, although most projects dealt with social processes, these processes were seldom quantified.Instead, the data available for judging the success of projects were biotic in most of the case studies.This is a potential weakness because many studies have found that biotic variables exhibit a poor recovery and that an early judgement can lead to biased conclusions (Palmer et al. 2010, Nilsson et al. 2015).In such cases, it is likely that restoration actors will have to wait longer before they can draw any conclusions from the evaluation.Irrespective of the results of evaluations, their quality may vary considerably.For example, in an analysis of 62 European evaluation studies, Kleijn and Sutherland (2003) showed that many evaluations were too poorly designed to allow any conclusions about whether projects had reached their goals.To avoid such failures, welldesigned, standardized protocols are needed (Palmer et al. 2005, Kurth andSchirmer 2014).
Our final question was about limitations to the evaluation of restoration.A major challenge with regard to evaluation is how it is documented and reported, if at all.It is true that very few restoration projects undergo formal evaluation (e.g., Kondolf and Micheli 1995, Bernhardt et al. 2005, Brooks and Lake 2007), but if they do, it usually results in some kind of written documentation or photographs.Our analysis corroborated this view.Most of the investigated projects included components of informal evaluation and these were rarely documented or reported, which means that http://www.ecologyandsociety.org/vol21/iss1/art41/lessons learned cannot fully benefit future projects unless they are properly communicated between the respective groups of restoration actors.Such communication between for example scientists who have generated knowledge and practitioners who are expected to apply it is an intricate task (Hulme 2014).In our analysis, however, we were able to uncover much informal evaluation simply because we had personal knowledge about the projects and expanded our knowledge by collecting more information from restoration actors involved.Even if restoration actors do not document or report their findings, results can still be preserved if there are interested end-users.This means that the results of some kinds of evaluation can be gleaned from end-users and the public by way of field visits, websites, media articles, teacher education, school visits, roadside interpretive displays, and museum exhibits.Such evaluations are also important to share with actors in future restoration projects.
We were also able to get access to hidden examples of documented evaluation.The fact that evaluations are documented does not necessarily mean that they are made public.If published, this may have been done in reports or other gray literature that is poorly accessible to the wider scientific community (Aradottir and Hagen 2013).To make dissemination of all documented information possible, it should ideally be archived in open, searchable databases.
Even if restoration projects undergo evaluation and result in published reports, we found it difficult to get a grasp of entire projects by reading the disseminated work from the 10 studied cases.Our analysis suggests that only the most "interesting" and "successful" outcomes of the different evaluation steps are widely disseminated to the public.This is a general issue that leads to underrepresentation of "failed" restoration projects in the literature (cf.Zedler 2007).The low acceptance rates of scientific journals, and the time required to prepare and submit a paper, discourage publication, especially of local and statistically nonsignificant results or of seemingly failed projects.Therefore, making project evaluations public is a major challenge.
Our case studies also provided examples of the importance of public education along-side evaluation.In the Vindel River restoration project landowners who were reluctant to restore their streams changed their minds after having seen the result of other restoration projects (Gardeström et al. 2013).In addition, it also suggests that confidence between restoration actors and landowners can be built if restoration projects are not rushed or imposed on people (cf. Bunn et al. 2010).Other examples of education include information about the role of ecological restoration and the importance of native biodiversity (Hulme 2006) and stricter controls on introductions of nonnative species (van Wilgen and Richardson 2014).
Another important challenge is the persistence of an evaluation result.Formal evaluations, if carried out, usually apply to monitoring results gathered in step 5. Generally, such evaluations are based on short-term monitoring datasets because most monitoring programs do not supply the resources necessary to await the long-term effects of restoration (Suding 2011, Nilsson et al. 2015).Evaluations based on restricted time periods may miss considerable amounts of information that, properly used, could have led to modification of the entire restoration process and hopefully better restoration practices in the future, i.e., adaptive management (Williams 2011).Instead, cookbook solutions (Hilderbrand et al. 2005, Fryirs andBrierley 2009), often nonoptimal, are likely to become overused.To solve this problem, a revised approach to funding and monitoring is required.An example of such an approach can be found in northern Sweden, where a court verdict over a disputed railway construction through a Natura 2000 area led to the funding of a 100-year monitoring program (Länsstyrelsen Västerbotten 2015).The objectives of this program are to evaluate the long-term effects of the railway and a number of compensation measures on the wildlife in the area, but also to manage the area so it can maintain its wildlife values.A special foundation board with representatives from landowners, NGOs, public authorities, and scientists is responsible for the management of the monitoring and evaluation and also makes sure that the collected information is stored and made available to interested users (Länsstyrelsen Västerbotten 2015).
We conclude that ecosystem restoration practices are developing, although slowly, because the actors involved evaluate and modify their practices throughout the restoration process.In the majority of cases, however, the evaluation is informal and not documented, meaning that lessons learned can only be forwarded if communicated among actors involved or if restoration is implemented by the same actor who developed an evaluation step.To speed up the effectiveness of ecological restoration, we recommend that actors reflect about their practices, document their experiences and spread the word about their findings, both successes and failures.In addition to more traditional ways of interaction, the modern digital world offers numerous possibilities for sharing the lessons learned.Another important development would be an economic analysis of the cost-efficiency of monitoring and evaluation.Increased knowledge in this respect has the potential to foster a better understanding of the significance of budgeting for evaluation in every restoration project.
Responses to this article can be read online at: http://www.ecologyandsociety.org/issues/responses.php/8289 Appendix 1. Identified evaluation steps in major ecological restoration projects in the northern hemisphere.

Identity of restoration project
Step 1 Within planning Step 2 Between planning and implementation Step 3 Within implementation Step 4 Between implementation and monitoring Step 5 Within monitoring Step proposed by scientists to practitioners were also communicated to planners Helfield et al. 2007, Palm et al. 2007, Gardeström et al. 2013, Polvi et al. 2014, Hasselquist et al. 2015, Nilsson et al. 2015, Vindel River Life 2015 LITERATURE CITED

Fig. 1 .
Fig. 1.Conceptual diagram showing the three major restoration phases and the six evaluation steps, three within-phase (1, 3,and 5) and three between-phase (2, 4, and 6), discussed in this paper.Evaluation steps are further described in the Introduction and in Table1.Note that the evaluation process is not a single-loop, one-time practice; instead evaluation activity can go back and forth and be upheld during several years.