Mapping wetland loss and restoration potential in Flanders (Belgium): an ecosystem service perspective.

. With the case of Flanders (northern part of Belgium) we present an integrated approach to calculate accurate losses of wetlands, potentials for restoration, and their ecosystem services supplies and illustrate how these insights can be used to evaluate and support policy making. Flanders lost about 75% of its wetland habitats in the past 50–60 years, with currently only 68,000 ha remaining, often in a more or less degraded state. For five different wetland categories (excluding open waters) we calculated that restoration of lost wetland is still possible for an additional total area of about 147,000 ha, assuming that, with time and appropriate measures and techniques, the necessary biophysical and ecological conditions can more or less be restored or created. Wetland restoration opportunities were mapped according to an open and forested landscape scenario. Despite the fact that for 49,000 ha wetland restoration is justifiable by the actual presence of an appropriate spatial planning and/or protection status, the official Flemish nature policy only foresees 7,400 to 10,600 ha of additional wetland (open waters excluded) by 2050. The benefits of a more ambitious wetland restoration action program are underpinned by an explorative and quantified analysis of ecosystem service supply for each of the two scenarios, showing that the strongly increased supply of several important regulating and cultural ecosystem services might outweigh the decrease of food production, especially if extensive farming on temporary wet soils remains possible. Finally, we discuss the challenges of wetland restoration policies for biodiversity conservation and


INTRODUCTION
It is estimated that the world lost at least 50% of its wetlands during the 20th Century (UNWWAP 2003, Davidson 2014).Some two thirds of the European wetlands have been lost in the same period (European Commission 1995), leading to a substantial decrease in the number, size, and quality of bogs, marshes, wet grasslands, and shallow lakes.Over a timespan of multiple centuries wetland loss is much higher because draining, conversion, and infilling of coastal and inland wetlands in Europe has been ongoing since at least Roman times (Russi et al. 2013).During that period wetlands were converted from uninhabitable, remote areas, with often harsh and unhealthy living conditions, into more productive, accessible, and human-friendly rural landscapes.
Exact figures of contemporary wetland loss in Europe are hard to find.For the period 1950-1985 wetland loss in six European countries was roughly estimated to lie between 55% and 67% (European Commission 2007), however without providing any supporting data or references.Based on Corine (the EEA's "coordination of information on the environment") land cover data, it was estimated that between 1990 and 2006 another 5% (1267 km²) of Europe's marshes and bogs were lost (EEA 2010).On the other hand, coastal wetlands remained more or less stable and open waters had even increased by 4.4% (1581 km²) in the same period (EEA 2010), the latter probably mainly in the form of artificial water bodies such as new dam and water storage constructions (Acreman 2012).The European Habitats Directive now protects 47 different wetland habitat types and 290 species (plants, invertebrates, fish, amphibians, reptiles, mammals) linked to wetlands.In 2006 however, nearly two-thirds of the species and more than three-quarters of the habitats throughout the EU member states were in unfavorable conservation status (ETC/BD 2008).Even more worrying, for the Boreal and Atlantic regions where large areas of wetlands (used to) occur, none of the habitats was in a favorable status.The successive (2007)(2008)(2009)(2010)(2011)(2012) assessment indicates a further decrease (European Commission 2015).
This ongoing loss and deterioration of European wetlands contrasts sharply with their well-known values for society, recognized decades ago (e.g., Thibodeau and Ostro 1981, Batie and Shabman 1982, Farber 1987, Costanza et al. 1989, Folke 1991, Gren et al. 1994).The TEEB-review study ("The Economics of Ecosystems and Biodiversity") for water and wetlands (Russi et al. 2013) clearly mentions the major ecosystem services provided: flood protection, water supply, water purification, carbon sequestration, climate regulation, production of raw materials and food, tourism and recreation, aesthetic and cultural values.In 2012, United Nations Environment Programme (UNEP) called for urgent integration of the key role of wetlands into decision making, and the need for their future protection, restoration, and sustainable use as a vital component of the transition into a resource-efficient, sustainable world economy (http://www.unep.org/newscentre/Default.aspx?DocumentID=2697&ArticleID=9305&l= en).The importance of wetlands has, on various occasions, been recognized within the framework of the CBD (e.g., COP Decision X/28, UNEP/CBD/COP/DEC/X/28, 29 October 2010; COP Decision XI/23, UNEP/CBD/COP/DEC/XI/23, 5 December 2012; Message of the Executive Secretary of the Convention on Biological Diversity Braulio F. De Souza Dias of 2 February 2016, https://www.cbd.int/doc/speech/2016/sp-2016-01-28-wwd-en.pdf).
Because many benefits of wetlands are of nonmarket and public nature they are rarely represented nor defended in decisionmaking processes.Governments who want to develop an evidence-based policy on wetlands will rely heavily on the availability of scientific information.A first and essential step in this process is to make the consequences of different land-use scenarios as explicit as possible.In the present paper we follow a spatially explicit approach balancing past wetland losses with Yes potential gains from restoration and associated ecosystem service benefits and translating this into restoration opportunities.This approach enables us to identify synergies and trade-offs between alternative land-use planning policies and restoration scenarios.
We test this integrated approach in Flanders (northern part of Belgium), one of the most degraded wetland regions in Europe.
We successively describe and discuss the following: 1. the loss of different wetland types since the 1950s (as an important turning period in Flemish land-use development and with detailed soil maps available from that same period) ; 2. the mapping of two realistic wetland restoration (and creation) scenarios; 3. the potential ecosystem service benefits of both scenarios; 4. the evaluation of current wetland restoration policies with a brief discussion on future challenges, as an illustration how results from our integrated approach could be used in decision making processes.

Study area
The region of Flanders is situated in the northern half of Belgium and covers 13,522 km².Bordering the North Sea and the Netherlands, the area is rather flat, partly reclaimed from the sea (polders), and large parts are dominated by wide river valleys and a dense network of slow-running watercourses.The highest point only reaches 156 m above sea level.It has a maritime climate with an annual precipitation of 800 mm and mild winters and summers (average of 3°C in January and 21°C in July).These conditions explain the large historical density of wetlands.Currently 45% of the region is used for intensive agriculture, heavily fertilized and drained or irrigated.Another 26% of the land is urbanized (470 inhabitants/km²) and 13% of the soils are sealed (De Meyer et al. 2011).This has resulted in a substantial and steady increase of the number of recorded floods since the 1970s and an average yearly economic damage of 50 million euro (VMM 2014a).
Remaining wetlands cover only 5% of the region and suffer from eutrophication, pollution, and disturbed hydrological regimes.All 25 wetland habitat types protected by the Habitats Directive are in an unfavorable conservation status (Louette et al. 2013).Most peat soils were extracted in medieval times and nearly all of the 6000 ha of remaining peat soils are heavily fragmented and assumed to be in a degraded, mineralized state.

Wetland classification and mapping
In this paper wetlands are defined as temporary or permanently wet, nonmarine areas where typical wetland biodiversity is (still) more or less present.Consequently, "lost wetlands" must be understood as areas that, apart from ditches and small rivers or ponds, can no longer be considered as "wet" and lack typical wetland communities, including the temporary residing of migrating waterfowl.In our case wetland loss is not to be confused with degraded, damaged, or polluted wetlands, as is sometimes discussed in other literature (Davidson 2014).
We distinguished seven wetland categories (Table 1) based on drainage class (open water, permanently or temporary wet soil, tidal marsh) and trophic state (meso-, eu-or oligotrophic).Open waters (artificial water bodies, lakes, and large ponds) were included in the mapping and calculation of historical wetland loss, but were not considered in the restoration scenarios because their restoration/creation preconditions are less stringent.
Spatial analysis considered the following main maps:  Ecosystem services considered in this study, the quantification units used for supply mapping (indicator), and the respective reference chapter of the Flanders regional ecosystem service assessment (Jacobs et al. 2014a, Stevens et al. 2015).References of individual chapters are listed in Appendix 9.For the biophysical and monetary estimates (wood production, climate regulation, food production, water quality regulation, and flood risk regulation) additional data from the ECOPLAN project was used.habitat types based on local abiotic and biotic conditions (Wouters et al. 2013);

5.
Additional maps on tidal marshes, historic forest cover, and current land use (De Keersmaeker et al. 2001).
For GIS analysis all maps were transformed into grid cells of 20 x 20 m.The area of water courses was considered constant over time and excluded from the analysis to avoid large errors in the calculation of area because of this grid transformation.All currently urbanized areas were considered to be not suitable for wetland restoration.For nonurbanized areas we assumed that in the long term the environmental conditions, as they were recorded in the 1950s, can be restored with appropriate measures.Further details about maps and spatial analysis are provided in supplementary material (Appendix 1 and 2).

Scenarios for potential wetland restoration
For the calculation of the area of potential wetlands we distinguished two management scenarios: (1) an open (not forested) landscape scenario, and (2) a closed (forested) landscape scenario.To obtain realistic scenarios, the legal protection ("standstill principle") for existing forests and open habitat types with nature value were taken into account.

Socioeconomic potential of wetland restoration
We followed two approaches.First, the ecosystem service (ES) supply potential of the restoration scenarios was estimated for a broad bundle of services (see Table 2), based on the results of the Flanders Regional Ecosystem Assessment (Jacobs et al. 2014a, 2015, Stevens et al. 2015).Additionally, the socioeconomic relevance of this change in supply is demonstrated by estimating a monetary value for a selection of five services for which reliable monetary data are available from the ECOPLAN project (http:// www.ecosysteemdiensten.be/cms/, https://www.uantwerpen.be/en/rg/ecoplan/).
The ES profile of the different wetland categories was obtained by a direct overlay of the wetland maps with ES supply maps from the Flanders regional ecosystem assessment (Stevens et al. 2015), and the consequent calculation of median supply per wetland category.Prior to the overlay, maps were normalized (0-100) to allow cross-service comparison and graphing along the same unitaxis.Scenario-changes in relative provision of the entire bundle of ESs were estimated based on the surface changes of 114 landuse classes for the whole of Flanders and their averaged and normalized ES-supply per ha and per year.This derived ES supply map based on 114 land-use classes allows direct translation of land-use scenarios to ecosystem service supply impact without redefining all biophysical and socioeconomic variables in the original quantification maps.To calculate the total impact on the level of the Flanders region, first, the total supply per ES of each habitat was multiplied with the surface area of this habitat, and normalized to obtain an ES supply profile for the whole of Flanders.Second, changes in these surfaces for each scenario provided alternative profiles.Finally, the relative difference between the scenario profile and the reference (current) profile provides the impact of the scenario in terms of increases and decreases in ES supply.
Monetary estimates were performed for wood production, climate regulation (as carbon storage in soils), food production, water quality regulation, and flood risk regulation (as water quantity regulation).The quantification and valuation methods have all been developed specifically for the Flemish Region (Broekx et al. 2013a,b, VITO 2014) and adapted to spatially explicit models at high resolution (ECOPLAN-project, details in Appendix 3).This monetization only aims at demonstrating the socioeconomic relevance of the multiple benefits from wetlands.Valuation for (societal) cost-benefit analysis, would have to include many other services, cost estimates, discount rates, assumptions on constant demand per service, additional quantification of nonmarket values, etc. (Dendoncker et al. 2014, Boeraeve et al. 2015).For all five services, the two scenarios were compared to the actual land use as baseline.
To translate the forest and open landscape scenarios into the datadriven models, the following assumptions were made: . In the forest scenario wetland restoration on current low biodiversity crop and grasslands is realized by spontaneous succession.Following the abandonment of agriculture the restored wetlands on permanently wet soils gradually become forested wetlands with alder, willow, and birch and on temporary wet soils other species such as oak and ash.This scenario assumes an active water retention where mean highest groundwater tables can be above surface.Also nutrient retention and carbon storage in belowground stocks is considered to be maximized because there is no drainage and commercial timber harvesting is absent (permanently wet soils) or reduced (temporary wet soils).
. The open landscape scenario assumes open landscapes are maintained actively through conservation and agricultural management.Current low biodiversity crop and grasslands on permanently wet soils are converted into botanically and/ or faunistically more biodiverse grassland.This implies minimal drainage to a level that still allows conservation management, e.g., mowing or grazing.Mean highest groundwater levels are close to the surface and fertilizer application is absent.In temporarily wet zones extensive agricultural management would be possible with limited maintenance fertilization.
. In both scenarios, rewetting of existing forest sites is assumed.Commercial wood production assumes conversion to native species.

Evaluation of the current policy for restoration of wetland biodiversity
An important indicator for the ambition level of any European government to restore part of the lost biodiversity are its (legally defined) conservation objectives for the implementation of the Habitats and Birds Directives.We translated the habitat types and habitats of protected species into our seven wetland categories and compared the Flemish objectives to increase wetland habitat area with the calculated "restoration opportunities" (defined as potential area for wetland restoration in the two scenarios, reduced with already existing wetland area).

Change in wetland area over time
In the 1950s 244,000 ha (19% of Flanders) could still be considered wetland (Table 3).Currently only 68,000 ha (5% of Flanders) remain, implying a substantial loss of almost 75% of wetland habitats over 50-60 years' time.Thirty-seven thousand ha (15%) has been urbanized; the rest was mainly lost to intensification of agriculture and to a lesser extent also to an increase in forest production.The proportion of wetland loss differs between categories.Moist to wet heathlands and nutrient-poor grasslands decreased by 95% (-24,000 ha), with an identical rate of loss for the forested parts on these soils (-7,000 ha).Wet floodplain grasslands and polders decreased by 75% (-95,000 ha) and floodplain forests by 55% (-8,000 ha).Historical rich fens and Ecosystem service profiles for different wetland-type categories in Flanders.Scores are derived as the median of normalized ecosystem service supply from the Flanders Regional Ecosystem Assessment supply maps (Jacobs et al. 2014a, 2016b, Stevens et al. 2015).For instance, in permanent wet, oligotrophic, nonforested habitats (upper left panel), the pixels of this habitat on the climate regulation supply map of Flanders (R6, normalized from 0-100) have a median score of 80%.Legend: C1 -Cultural Services; R1 -Flood Risk Regulation; R3 -Water Quality Regulation; R4 -Pollination; R5 -Air Quality Regulation; R6 -Climate Regulation ; R7 -Erosion Risk Regulation; R8 -Sound buffer; P1 -Energy Crop Production; P2 -Wood Production; P3 -Wild Meat Production; P4 -Ground Water production; P5 -Food Production.Details on quantitative units are depicted in Table 2.
marshes decreased by 95% (-41,400ha) and swamp forests by 60% (-4,500 ha).Permanently wet heathlands and open bogs showed a loss of 95% (-4,000 ha), while the forested version of this habitat decreased by 50% (-500 ha).Tidal marshes showed a reduction of 80% in area (-2,400 ha), mainly for land acquisition in the neighborhood of the port of Antwerp.In general, 20,000 ha (10%) of open wetland habitats disappeared because of active or spontaneous afforestation, with those on permanently wet soils proportionally being most affected.In contrast with the dramatic numbers above, deep waters tripled and shallow waters doubled in surface area over those years.At present, 100% of the deep waters and 90% of the shallow waters are eutrophic.Their trophic state could not reliably be reconstructed for the 1950s, but it is fair to assume that many meso-and oligotrophic waters have shifted to a eutrophic or even hypertrophic state.

Potential for wetland restoration
According to our calculations (Table 3) there is still a potential to restore 147,000 ha of wetland in Flanders (deep and shallow waters excluded).In the long term this could bring the total amount of wetland to 215,000 ha or 17% of the territory.With appropriate measures to restore the conditions of the 1950s, floodplain grasslands and forests and wet polder areas can theoretically triple in surface area to a significant 120,000 ha.Oligotrophic wetland habitats on temporary wet soils could increase 14-fold to 26,500 ha.Restoration of wetlands on permanently wet soils would lead to a 6-fold increase of open and forested wetland habitats: 36,500 ha on meso-eutrophic soils and 4500 ha on oligotrophic soils, or 72% and 88%, respectively, of the original surface area of the 1950s.There is a huge potential for the restoration of tidal marsh along the river Schelde if embankments are moved inland.With many of these embankments already in place in the 1950s, this implies a 3-fold increase in area compared to the reference period and a 15-fold increase in area compared to the current situation.The potential for restoration of shallow waters was not calculated: in principle they can be artificially created in many sites.Maps with the modeled distribution of historical, current, and potential wetland categories in Flanders are provided in Appendix 4-7.

Ecosystem service supply by wetlands in Flanders
Wetlands in Flanders provide a broad bundle of services (Fig. 1).Ecosystems on permanently wet soils provide most ecosystem services, especially forested habitats.Provision of water quality regulation, pollination, and climate regulation are the most prominent.Cultural services and flood risk regulation are also important, as are air quality regulation, sound buffer, and wood production in forested permanently wet habitats.Systems on temporary wet soils have a very similar profile, but perform poorer on water quality regulation.The meso-and eutrophic habitats include seminatural grasslands that can be combined with food production (haymaking, grazing).Tidal marshes differ from the former wetlands by a high cultural value and hunting potential, but deliver a lower supply of water and air quality regulation and of sound buffer.Shallow waters provide a remarkably high supply of flood risk regulation but lower supplies of water and air quality and sound buffer.
The different restoration scenarios have a significant impact on total ecosystem services supply (Fig. 2).Both the forested and the open landscape scenario lead to a decrease in food production (-16% to -19%), an increase in both flood risk regulation and climate regulation (5% to 10%), and a strong increase in water quality regulation (31% to 46%).The forested scenario leads to an additional increase in sound buffer and wood production (9%), while these services slightly decrease in the open landscape scenario (respectively, -9% and -2.5%).Slight decreases occur also in the supply of coastal protection, air quality regulation, and production of energy crops.Benefits for water quality regulation (total nitrate release to surface water) are comparable for both scenarios, but depend on different aspects.While nitrate leaching is reduced most in the forested scenario, denitrification decreases significantly.In the open landscape scenario, nitrate leaching decreases less dramatically (11.3% instead of 16% for forested), but denitrification is relatively more performant because the decrease is only 4.2% (instead of 12% for forested).The monetary benefit ranges from €15 to €225 million per year.The high estimate (€74/ kg N-NO3) is based on shadow prices of effectively implemented policy measures for nitrate in surface water (marginal cost method).Implementing a large scale restoration scenario that decreases nitrate release up to 20% would make a range of current technical measures dispensable.For correct valuation one should be able to derive a mean value for the dispensable measures.On the other hand, most water bodies do not meet the water quality standards despite the current measures.

Monetary valuation of selected ES
Carbon sequestration in soils is relatively insensitive compared to the total stock in the Flemish Region, but highest under the forested scenario.The nature conservation management and agricultural management imply harvest of aboveground biomass, which results in less input to the soil compartment.This, however, does not mean that local changes cannot be important.Especially for the permanently wet ecosystems, active peat formation could be restored.Unfortunately, the quantification methods for soil organic carbon do not incorporate carbon stocks from potential peat formation.
Water quantity regulation is an ecosystem service that is likely to become more important in the next decades.Rewetting former wetland ecosystems allows increasing water retention by 7.6% under the forested scenario and by 4.8 % under the open landscape scenario.This volume of additionally retained water compares to a river with a steady flow of 2.8 and 1.8 m³/s, respectively.Whether the retained water could all be used for consumption can be disputed, but on the other hand this would be a service that is of strategic and crucial importance in terms of climate adaptation.

Evaluation of the current policy for wetland restoration
Comparing the restoration opportunity (potential area for wetland restoration in the two scenarios, reduced with already existing wetland area) and the objectives for wetland expansion in the Flemish Natura 2000 policy (Table 5), a significant discrepancy between the two figures appears.Present policy foresees a total wetland expansion of 8900-13,000 ha (or 7400-10,600 ha with open waters excluded) in 2050, including 1800-3000 ha forested wetland and 2500 ha tidal marsh.All figures are much lower than what could be reached with a more ambitious policy.The ambitions for oligotrophic and mesoeutrophic wetlands on temporary wet soils and meso-eutrophic wetlands on permanently wet soils appear to be especially modest with an increase of only 1-8% of the restoration opportunity.With a projected increase of 19-26% of the potential restorable surface, ambition levels are significantly higher for tidal marsh and wetlands on oligotrophic permanently wet soils.

Area estimations for wetland loss and restoration potential
Reliable estimations for (sub)national wetland loss, subdivided into different wetland subtypes and for an identical time period are very rare in literature (see Davidson 2014).The combination with an accurate and spatially explicit analysis of the wetlands that can potentially still be restored makes our study rather unique.Obviously, possible errors in area estimations and their mapping largely depend on the accuracy and scale of the map layers that are available for GIS analysis (Joao 1998).In the case of Flanders we were fortunate with the availability of detailed maps on a scale of 1:25,000 describing the soil conditions in the 1950s and the recent distribution of 180 habitat types, including 40 types of wetland habitat.Such accurate maps may not be available in other regions of the world and this poses a challenge to the exact replication of our methods (see also Clare and Creed 2014).Another error source may be the use of discrete values derived from the basic map layers to define the different classes of abiotic and biotic conditions.Within the limitations of the used basic data we believe our approach provides the best possible proxy for estimating wetland loss and the present potential for wetland restoration in Flanders.However, the maps that were generated (see Appendix 4-7 in Supplementary materials) should http://www.ecologyandsociety.org/vol21/iss4/art46/Table 5. Ambition level for the expansion of wetland habitats within the Flemish Natura 2000 policy framework in perspective with the available restoration opportunities (defined as potential area for wetland restoration in the two scenarios, reduced with already existing wetland area).† All increase of alluvial forests (habitat type 91E0) was assigned to meso-eutrophic wetlands on temporary wet soil.‡ No distinction could be made between forested and nonforested tidal marsh.be interpreted with caution when zooming in on the individual site level.The transformation in to grid cells of 20 x 20 m, in combination with possible errors in the used basic map layers, may inevitably generate inaccuracies when maps are scaled down.
Concerning application of this approach in other areas, one should be aware of the impact of accuracy on the final area estimates, as was also observed by Davidson (2014).Especially in areas with low data availability, applying a min-max fork estimate could provide confident and transparent estimates.

Ecosystem service supply of wetland restoration
Estimations of ecosystem service (ES) supply per wetland category as well as the impact estimation of the scenarios on total ES supply should be handled with caution.Here, we want to point out three caveats for ES supply estimates, which apply for any case study engaging in ES quantification.First, the data used for this exercise are the best available data on ES supply at this moment.These indicators are often combinations of several data layers and combined models involving a number of reasonable (and checked) assumptions.Indicators and maps should therefore be interpreted alongside their confidence and used within the boundaries of their specific purpose.Although the indicators used in this study robustly support our conclusions, they cannot be applied to answer just any question, especially not questions that require much higher accuracy and confidence, e.g., accounting, trend analysis, development of payment schemes, etc. (see also Jacobs et al., in press).Second, the maps are made and reviewed for the regional scale.Zooming in to local levels will bring to light biases caused by local physical, ecological, or social conditions that are not captured by the models.This has little repercussions for regional-scale analyses, but it is clear that the local ES supply of wetland types will differ strongly from one location to another.Third, and following from this local scale, there might be ecosystem services relevant on the Flanders scale that are not important at all at some locations (because either supply or demand is lacking).In fact, there might be important services missing from this analysis when scaling down to the implementation level, while the basic valuation at this level does not include a differential societal importance of the services.Conclusions drawn on this exercise are strictly general and may not be used to guide a local planning process.
Our monetary estimations indicate that benefits derived from the regulating services (water quantity regulation, water quantity regulation and carbon storage in soils) range from 20 to €268 million/yr.The decrease in production services (agriculture and timber production) ranges from 137 to €186 million/yr.Much can be debated about the quantification and valuation methods, including the validity of the scenario.Nevertheless these estimates demonstrate that for at least three services, substantial benefits could be obtained.A more sophisticated scenario would probably allow decreasing the impact on agricultural production and timber production, while maintaining these regulating services.Moreover, including health benefits, tourism, and recreation could tip the balance to positive numbers (e.g., Broekx et al. 2013b).Also the current mean cost of €50 million/yr to compensate for economic damage due to flood hazards (VMM 2014a) needs to be taken into account.
Despite these caveats, this analysis clearly shows the overall importance of specific wetland habitats for supply of mainly http://www.ecologyandsociety.org/vol21/iss4/art46/regulating and cultural services on the scale of the entire region.
Restoring or creating wetland habitat in Flanders can result in a strongly increased supply of several important services, and in a decrease of food production.Basic economic valuation demonstrates the high societal importance of these services.Without being conclusive, this simple valuation opens a rational debate on whether the benefits and costs involved in food production might be outweighed by the broader benefits supplied by restored areas.
Our results broadly concur with earlier valuation studies (e.g., Thibodeau and Ostro 1981, Batie and Shabman 1982, Farber 1987, Costanza et al. 1989, Folke 1991, Gren et al. 1994, Russi et al. 2013) but especially highlight that valuation exercises should be broadened to include more than monetizeable benefits.First, not all ES are increasing, and societal trade-offs have to be made between benefits and losses of Flemish wetlands.Second, benefits and losses for different users should be disentangled to account for the governance issues involved in actual realization of a certain scenario.Third, a broader value typology to integrate intrinsic values, instrumental values, and relational values should be applied to go beyond an eye-opening study toward actual decision support (Jacobs et al. 2016a).
Projected losses in food production also consider the current production model, which involves substantial financial support from public budgets, as well as issues concerning food waste and caloric efficiency of meat production.Even a slightly different production model might easily compensate projected losses in wetland areas, or provide ways of farming that can be combined with the multiple services provided in these landscapes (Jacobs et al. 2014b, Van Gossum et al. 2014).Rather than retreating to the typical historical struggle for monofunctional land-service allocation and grinding on trade-offs between services and stakeholder groups, the many existing synergies on a practical and local level could offer concrete solutions for a multifunctional, biodiversity-rich wetland use.Such an approach could evoke a more transparent and rational debate on restoration of natural habitats in intensively used areas and might be more effective in obtaining biodiversity goals.

Past and current policy context
To understand the current state of ecosystems in any country or region and develop new policies, insights into past and current policies are essential.With almost 75% of its wetlands lost since the 1950s, Flanders ranks highest amongst the European regions (see data in European Commission 2007).The high population density and inappropriate spatial planning and urbanization policy were important drivers, as well as the lack of coordination of water management, which is traditionally very complex with many actors on different government and administrative levels.
The European Water Framework Directive (2000), the Floods Directive ( 2007), and increasing socioeconomic costs of flood events in urbanized areas (VMM 2014a) were important turning points in the mind setting of the Flemish water policy makers.Nowadays some of the most prestigious nature restoration projects in Flanders go hand in hand with flood protection, e.g., for the large rivers Schelde, Grensmaas, and IJzer).The once common practice of widening and straightening of rivers and urbanization of flood-prone areas has virtually stopped.A more detailed overview on Flanders-specific water management practice can be found in Appendix 8.

Wetland restoration and biodiversity conservation
Our integrated and spatially explicit approach delivers data that can be useful in the societal debate on more and better wetland restoration and is helpful to develop guidance for future decision making.In the case of Flanders we found that 35% of the remaining wetlands have no spatial planning or protection status, while 49,000 ha (33%) of potential wetlands lack investments for restoration despite their appropriate status (see Appendix 8).As was demonstrated in many other countries (e.g., Birol et al. 2009, Buijs 2009, Scholte et al. 2016) flood protection is more widely accepted as a motivation for wetland restoration than biodiversity conservation.In the Flemish floodplains this is demonstrated by the still dominant, more or less intensive agricultural use with fertilization and active drainage of wet grasslands.Outside the floodplains, restoration projects of nutrient-poor wet grasslands and heaths on temporary or permanently wet soils remain rare and small in scale.They are mainly restricted to nature reserves in the upstream and interfluvial areas.Conflicts with the surrounding land use in terms of water levels and water quality often hamper these projects.It is the public perception that such wetlands would not contribute to flood prevention and therefore they stay beyond the reach of the (traditionally much bigger) budgets of water management administrations.In general, the lack of interest in the restoration of wetland biodiversity is also reflected in the rather low ambition level for expansion of Natura 2000 wetland habitat types and habitats for Natura 2000 wetland species, particularly those of open landscapes.
For wetland restoration in general, and for the Flanders case specifically, we conclude that more awareness raising beyond direct biodiversity values will be essential to implement a more effective long-term restoration policy for the different types of biodiversityrich wetlands.Fostering public support is not only essential, different stakeholder groups will need different kinds of information and opportunities for participation (see also Johansson and Henningsson 2011, Tolvanen et al. 2013, Aggestam 2014, Scholte et al. 2016).Studies like ours are essential to identify priority areas for restoration and create a more robust ecological network of wetlands (see also Gibbs 2000, Vos et al. 2010).

Wetland restoration and climate change adaptation
According to the most plausible scenarios described in the report of the Flanders Environment Agency (Brouwers et al. 2015) the mean temperature in Flanders may rise by up to 7.2% by 2100, which will lead to more extreme hot days and heat stress.Summers will get drier with more concentrated heavy rain events.Precipitation will be higher during winters.The sea level may rise up to 1 m.Combined with the predicted increase in population size and further urbanization of open space, the flood risk will further increase.Flood plain areas will hence become less valuable for agriculture and inhabitation, which is potentially facilitating their transition to (semi)natural wetlands.
Apart from reducing economic damage caused by floods (see e.g., Bullock and Acreman 2003, Acreman 2012, Acreman and Holden 2013, Walters and Babbar-Sebens 2016), both natural and artificial wetlands could produce additional adaptation services as water buffer areas to ensure sufficient water supply for the production http://www.ecologyandsociety.org/vol21/iss4/art46/ of food crops in hot and dry periods (e.g., Chester andRobson 2013, Downard andEndter-Wada 2013).Artificial wetlands in cities and urban areas will become more important to reduce heat island effects and to buffer heavy rain (e.g., Persson et al. 1999, Sun et al. 2012).More wetlands will also help to remove increased nutrient runoff from cultivated catchments in regions with increased rainfall or more intensive agriculture (e.g., Gilliam 1994, Gren 1995, Woltemade 2000, Verhoeven et al. 2006, Thiere et al. 2009, Jeppesen et al. 2011, Hefting et al. 2013, Ockenden et al. 2014).To avoid depletion of ground water acquifers the creation of more temporary and permanent wetlands can increase the infiltration rate of rain water (e.g., Winter 1999).On the other hand it is possible that suitable areas for wetland restoration or even existing wetlands get lost or suffer from increased pressures in regions with increased droughts, with agricultural expansion as a possible secondary effect (e.g., Hartig et al. 1997).In general, climate change is expected to increase demand for wetland types of eutrophic, temporary wet and tidal soils (e.g.Nicholls 2004, Temmerman et al. 2013).The combination of increased evapotranspiration and more extreme weather events will probably challenge the restoration of specific wetland types that depend on more or less stable, high water levels, especially those of oligo-mesotrophic conditions (e.g., Cusell et al. 2013).Of course, restoration success will also depend on how the local geographical location is impacted by climate change (Čížková et al. 2013).
All in all, the urgency of climate change and the obvious role wetlands can play to increase resilience in multifunctional landscapes, provide arguments for their further protection and restoration.The application of accurate maps and ecosystem service assessments like this study are needed to underpin these arguments scientifically and help them to get implemented into spatial planning.

CONCLUSION
We show that despite dramatic historic wetland loss and the unfavorable status of remaining wetlands, Flanders still has a large biophysical and ecological potential for wetland restoration with the proper spatial planning or protection status already in place to justify more action in the field.Based on the best available data, we demonstrated that restoring or creating wetland habitat will result in a strongly increased supply of several important regulating and cultural ecosystem services, and in a slight decrease of food production.Benefits supplied by restored or created wetlands and avoided costs of economic damage due to flood hazards might outweigh the costs involved and the loss in food production.Different policies, specific designs and local implementation examples could offer opportunities for multifunctional use, even with producing services, of restored wetlands.An exhaustive area-wide approach, supported by innovative GIS modeling and ecosystem service valuation techniques, provide a robust tool for assisting evidence-based policy decision making that could be applied for other ecosystem types or areas.
Responses to this article can be read online at: http://www.ecologyandsociety.org/issues/responses.php/8964 Appendix 1.Details on the GIS layer sources used for the modeling of wetland loss and their restoration potential in this study

Flemish Flood Hazard Map (VMM 2014b)
 Period of survey: 2004-2014  Scale: 1:10,000  Units: areas with actual high risk of flooding (i.e. more than one flood in 10 yrs), based on field observations and hydrodynamic modeling. Use in this study: provides information on regularly flooded areas (7,5% of Flanders), which can be natural (river valleys) or antropogenic (flooded areas due to changes in local urbanization and soil sealing).By excluding urban areas and arable land, the flood hazard map gives a picture where some biological value may still be present such as semi-natural vegetation relicts or at least temporary presence of (wintering) waterfowl.The map provides additional information to the Biological Valuation map for delineation of current wetlands.

Historic Forest Map (De Keersmaeker et al. 2001)
 Period of survey: topographic maps surveyed in the period 1910-1940  Scale: 1:20,000  Units: forested areas ( as detected by semi-automatic image recognition)  Use in this study: reconstruction of land use of historic wetlands

Tidal Marsh Maps
The actual distribution of tidal marshes was derived from the Biological Valuation Map.The historic distribution of tidal marshes along the river Schelde was based on the situation around 1960 as described by Van Braeckel et al. (2012).The potential for restoration of tidal marshes along the river Schelde was based on Van den Bergh et al. (2003).Actual and historic distribution of tidal marshes outside the area under influence of the river Schelde were the same.

POTNAT (Wouters et al. 2013)
 Integrated maps derived from different information sources by GIS modeling and grid transformation  Scale: 20x20m grid cells  Units: maps with the potentials for restoration of 18 terrestrial wetland habitat types under current (mainly Biological Valuation Map) and past (mainly Flemish Soil Map) abiotic conditions  Use in this study: distribution of wetlands that can be restored or created, assuming that on the long term Appendix 2. Consequences of adopting the legally obliged 'stand still principle' for the calculation of the total area (in hectares) for two scenarios of potential wetland.Depending on management and ownership structure (private, public) a harvest factor is applied that estimates the proportion of the annual maximal mean growth that is harvested annually.The harvested volumes are available from recent data (2009)(2010)(2011)(2012) on timber selling from public (state owned) forests and from forest owner cooperatives (privately owned, but the management is state coordinated).This data base has about 80.000 records of sold volumes per tree species and circumferences.For state owned forests, the harvest factor is 0.54.Privately owned forests are often unmanaged and have a lower (0.15) harvest factor.For private forests, there is an unknown fraction of harvest for private use and informal markets (especially for fire wood).
Valuation: Valuation of wood production has been done on the basis of annual m³ harvest per year and per tree species.The value for each species was based on the database of actual selling prices in the state-owned forests for the years 2009-2012.Although the records refer to tree species, volumes and associated circumferences, the selling prices often refer to a combination of several records sold as one single lot (in average 18 records/lot).Statistical analysis (SPSS 20.0) was used to reveal a selling price (€/m³) per species and circumference class (Table A .2).The average weighted selling price for all species and circumferences was estimated at 32.43 €/m³.Trees are sold as standing timber and prices are therefore considered as net added value of timber production.

CARBON SEQUESTRATION IN SOILS
Quantification: Soils under unmanaged, natural vegetation types typically have larger carbon stocks than managed vegetation types.Also soil hydrology plays a crucial role in the creation of soil organic carbon (SOC) stocks.Soil organic carbon is especially high for forests and/or hydric soils.The potential equilibrium state for soil organic carbon stocks can be calculated using the regression formula by Meersmans et. al. (2008), which includes parameters like water retention, soil texture and vegetation type.Changes in land-use typically affect both vegetation and/or drainage (ES water retention), which leads to a new potential equilibrium state for SOC stocks.Recent research by Dr. De Vos (2009) has revealed that this function systematically underestimated SOC-stocks in forest soils with 32 %.This correction factor to the regression formula of Meersmans is applied to all forests.Peatlands, wetlands and freshwater ecosystems can sequester higher carbon stocks than terrestrial ecosystems.Potential (maximal) stocks are very much dependent on hydrological regimes and how mature these ecosystems are.Depending on the hydrological regime, newly created wetlands sequester 2.5-3.5 ton C/ha*yr in the first 100 years.Older wetland systems often do not sequester much additional carbon, especially when they are not under permanent hydric conditions.On the other hand, pulsed hydrological conditions emit less methane.
Valuation: Stocks are difficult to consider in valuation exercises.Here we calculated a virtual scenario of changes in carbon stocks due to changes in land use (habitat types) and associated changes in water retention.The difference in SOC stocks would be gradually built or released at a rate of 2.5 % loss per year.The valuation method is identical to the valuation of carbon sequestration in biomass.

AVOIDED NITRATE LEACHING
Quantification: It can be debated if the cessation of fertilizer use can be categorized as an ES.It is imperative to include this since landscape level nutrient leaching is an important parameter for the ES "nutrient removal by denitrification".Infiltration on fertilized agricultural land results in nitrate leaching to groundwater and eventually surface water.Important variables are the specific combinations of soil texture, crop type, agricultural fertilizer application (kg N/ha) and atmospheric N-deposition (kg N/ha).Long-term data on autumn and spring nitrate residues in agricultural soils were available from the Flemish Land Agency (Geypens et al. 2005).The difference between fall and spring residue is assumed to be leached out by winter precipitation.Atmospheric nitrogen deposition data were provided by the Flemish Environment Agency (FEA 2011).We assumed that nutrient leaching also occurs on non-agricultural land with high deposition rates.Although declining, these values are still relatively high (Staelens et al. 2012).From the data on nutrient leaching from agricultural land we know these values range between 7 % and 33 % of the nitrate application.We applied the same range of values (7 -33 %) on non-agricultural land with N-deposition and varied the range of values accordingly to the natural sensitivity for nitrate leaching (soil texture).Nitrogen has many and complex pathways by which it is dispersed in the environment.For this study, we focus on the issue of excess nitrogen in groundwater and surface water.Nutrient leaching from agricultural land is one of the major pathways.Reduction of nitrate leaching has already been described in previous sections, but is an important variable for denitrification.For the current situation, the avoided nitrate leaching is zero.But the NCO's include both cessation of fertilizer application and cessation of drainage.Cessation of nitrate leaching implies a decrease of nitrate supply to the denitrification zones, which in their turn may have increased nitrate removal efficiency due to rewetting.The supply of nitrogen occurs through patterns of (local) infiltration (nitrate leaching) and seepage.Infiltration and seepage patterns are the result of processes that occur on a range of spatial scales.A topographic position index (TPI) is used to identify these patterns at multiple scales (Jenness 2006).This method has also been used in other studies for the Flemish Region and has proven its applicability (De Reu et al. 2013).We calculated the TPI at a range of spatial scales (radius: 250m -2000m) to indicate these local infiltration-seepage patterns.The multiscale TPI is then corrected for soil permeability to result in a seepage intensity map, indicating the water supply to a particular pixel (mm/day).The nitrate concentration of the supplied seepage water is calculated at the landscape level (2 km radius) by multiplying the annual nitrate leaching (kg N/ha) with the annual infiltration (m³/ha).This allows us to calculate denitrification by multiplying the removal efficiency with the annual nitrate load for each pixel.

Valuation:
The valuation is based on the marginal reduction cost for nitrate removal.The Environmental Costing Model for Flanders compares different (technical) measures on cost-efficiency (€/kg reduction) and the applicability of those measures.The cost of the most expensive measure, considered in policy approved measure programs, can be seen as the cost the society is willing to pay for a further reduction of nitrate levels in ground and surface water.For nitrate, the marginal reduction cost is 74 €/kg N. As a low estimate we apply 5 €/kg N, based on a literature review (Cools et al. 2011, Broekx et al. 2013a, Broekx et al. 2013b).a mixed picture here.On one hand we see continued pressure from agriculture to have the water levels as favorable as possible for agricultural exploitation, resulting in for instance an intensive river and ditch management.On the other hand small scaled flood control areas have been constructed in many places, recognizing the need for stocking excess water during peak discharges.Unfortunately their design is often not very beneficial for biodiversity: they are generally constructed in the lowest places where permanent grassland persisted and many flood control areas function 'off line', meaning they are kept dry as long as possible for agriculture, excluding any possibility for natural riparian dynamics and spontaneous succession.Allowing natural buffer zones along the smaller rivers are still not a wide spread practice in Flanders.According to our calculations the area of valuable floodplain grasslands and forests can be increased with 78,000 ha, with 15,000 ha already protected for nature by spatial planning or Natura 2000 designation.
Restoration projects of nutrient poor wet grasslands and heaths on temporary or permanently wet soils are much rarer and more small scaled.Societal benefits such as flood protection are of no importance here.Hence they are mainly restricted to nature reserves where young forest encroachment is removed, often in combination with removal of the rich top layer of the soil to activate the seed bank.There are also examples of successful restoration starting from former intensively used agricultural grasslands.Fine-tuning of the local hydrological conditions is in all cases crucial.Fen meadows are mainly restored where modern agriculture has left the area.After the traditional mowing practice without fertilization is reinstalled biodiversity values can recover.New reed marshes are mostly found in the margins of newly created water bodies and on artificially raised land with heavy soils and poor drainage.Sedge marshes are most of the times a result of spontaneous succession of abandoned fen meadows.In all cases cessation of management will on the long term lead to a forested version of the habitat.The area suitable for restoration of all these wetland types combined is estimated at 59,000 ha of which 29,000 ha is already protected for nature by spatial planning or Natura 2000 designation.We conclude that for a large proportion of suitable sites for wetland restoration the legal protection status is already in place to get started.In this perspective, the rather low ambition level for expansion of Natura 2000 wetland habitat types and habitats for Natura 2000 wetland species is striking, particularly those of open landscapes.

Fig
Fig. 1.Ecosystem service profiles for different wetland-type categories in Flanders.Scores are derived as the median of normalized ecosystem service supply from the Flanders Regional Ecosystem Assessment supply maps(Jacobs et al. 2014a, 2016b, Stevens et al. 2015).For instance, in permanent wet, oligotrophic, nonforested habitats (upper left panel), the pixels of this habitat on the climate regulation supply map of Flanders (R6, normalized from 0-100) have a median score of 80%.Legend: C1 -Cultural Services; R1 -Flood Risk Regulation; R3 -Water Quality Regulation; R4 -Pollination; R5 -Air Quality Regulation; R6 -Climate Regulation ; R7 -Erosion Risk Regulation; R8 -Sound buffer; P1 -Energy Crop Production; P2 -Wood Production; P3 -Wild Meat Production; P4 -Ground Water production; P5 -Food Production.Details on quantitative units are depicted in Table2.

Flemish
Period of survey: 1947Period of survey:  -1970Period of survey:   (mainly 1950's)  's)    Scale: 1:5,000 (published: 1:20,000). Units: soil types according to 11 texture classes, 9 drainage classes, 13 profile development classes, 15 substrate classes  Use in this study: Abiotic profiles of the different wetland type categories allow the reconstruction of their historical presence based on drainage and texture class.Biological Valuation Map (INBO 2015, Vriens et al. 2011) Period of survey: 1998-2007  Scale: 1:10,000  Units: 180 habitat types including 40 types of wetland habitat, all based on vegetation. Use in this study: provides detailed information on the current distribution of wetlands.Abiotic profiles of the different habitat types allow the reconstruction of current drainage class and trophic state.

Table 3 .
Estimations of the historical(± 1950s), actual (± 2005)and potential presence of seven wetland categories in Flanders (in ha).For the theoretically restorable wetlands, distinction is made between a forested and open landscape scenario.Potentials for deep and open waters were not calculated.

Table 4 .
Absolute losses and gains in biophysical and monetary terms for the current situation compared to a forested and open landscape wetland restoration scenario.

Table A1 : Overview of the relationships between soil suitability and the maximal mean growth of stemwood (m³/ha*yr).
The classes 1-2 are associated with the P75 values of the crop revenue values, the classes 3-4 are associated with the P50 values and class 5 is associated with the P25 values.Wood production depends on soil characteristics and applied harvest regime.Species specific potential produced wood volumes can be found in table A.1, where differentiation is made according to the soil suitability.

Table A3 : parameter values for maximal fertilizer application, fall nitrate residues in soils and winter nitrate leaching in function of cultivation and soil texture.
transform combinations of the mean highest (MHG) and mean lowest groundwater (MLG) levels to an estimated nitrate removal efficiency (% of available nitrate removed).