Assessment of Green Infrastructure in Riparian Zones Using Copernicus Programme

: This article presents an approach to identify Green Infrastructure (GI), its beneﬁts and condition. This information enables environmental agencies to prioritise conservation, management and restoration strategies accordingly. The study focuses on riparian areas due to their potential to supply Ecosystem Services (ES), such as water quality, biodiversity, soil protection and ﬂood or drought risk reduction. Natural Water Retention Measures (NWRM) related to agriculture and forestry are the type of GI considered speciﬁcally within these riparian areas. The approach is based on ES condition indicators, deﬁned by the European Environment Agency (EEA) to support the policy targets of the 2020 Biodiversity Strategy. Indicators that can be assessed through remote sensing techniques are used, namely: capacity to provide ecosystem services, proximity to protected areas, greening response and water stress. Speciﬁcally, the approach uses and evaluates the potential of freely available products from the Copernicus Land Monitoring Service (CLMS) to monitor GI. Moreover, vegetation and water indices are calculated using data from the Sentinel-2 MSI Level-2A scenes and integrated in the analysis. The approach has been tested in the Italian Po river basin in 2018. Firstly, agriculture and forest NWRM were identiﬁed in the riparian areas of the river network. Secondly, the Riparian Zones products from the CLMS local component and the satellite-based indices were linked to the aforementioned ES condition indicators. This led to the development of a pixel-based model that evaluates the identiﬁed GI according to: (i) its disposition to provide riparian regulative ES and (ii) its condition in the analysed year. Finally, the model was used to prioritise GI for conservation or restoration initiatives, based on its potential to deliver ES and current condition.


Introduction
In the view of human-induced climate change, ecosystem-based measures for disaster risk reduction (Eco-DRR) and climate change adaptation (CCA) have gained increasing attention [1]. Eco-DRR has been defined as "the sustainable management, conservation and restoration of ecosystems to reduce disaster risk, with the aim to achieve sustainable and resilient development" [2]. Eco-DRR is based on the concept that healthy, diverse and well-managed ecosystems increase the resilience of human societies and the environment to climate change impacts [3].
its surface is protected under the Birds Directive [26] (Figure 1). Therefore, it was selected as the study area to analyse GI condition using data from the Sentinel-2 satellite platform [46].

Copernicus Land Monitoring Service: Riparian Zones and Corine Land Cover
CLMS local component focuses on different hotspots, i.e., areas that are prone to specific environmental challenges [49]. The Riparian Zones (RZ) is a local CLMS product that supports the objectives of European legal acts and policy initiatives, such as the Biodiversity Strategy to 2020 [27], the Habitats [25] and Birds Directives [26], the Water Framework [36] and the Floods Directives [37].
RZ consists of three products: Delineation of Riparian Zones (DRZ), Land Cover/Land Use (RZ LC/LU) and Green Linear Elements (GLE) [47]. Moreover, the DRZ product consists of three components [48] of which the Delineation of Potential Riparian Zones (DRZP) was used. DRZP is derived from weighing hydrological and geomorphological parameters, among other input data (Table 1), to express the likelihood of an area to host riparian features and hence to provide riparianrelated benefits.
RZ LC/LU and DRZP ( Figure 2) are the main input data used in the approach to identify GI and its potential to provide riparian-related ES respectively, since they provide very detailed information of the riparian environment (LC/LU classes and its characteristics) along large and medium-sized river streams (Table 1).
Corine Land Cover (CLC) consists of an inventory of land covers classified in 44 overall classes. It is a pan-European product initiated in 1985 (reference year 1990) and updated every 6 years. The RZ LC/LU product was performed using CLC 2006/2012, among other inputs (Table 1) Table 2) were used for an internal crosscheck validation approach of the main input data. The goal was updating the LC/LU classes, avoiding false positives in GI identification due to LC/LU changes between 2012 and 2018, before performing the subsequent spatial and temporal analyses on the vegetation condition. Afterwards, the misclassified GI was photo-interpreted using recent Sentinel-2 satellite images. Previous scientific research analysed the difference in precipitation, temperature and daily flux of Po river by comparing forecast data (2021-2050) and recorded data . The comparison showed a significant decrease in annual average water availability and a higher frequency and intensity of extreme events [40,41,72]. This makes the region interesting to study GI as a nature-based solution for mitigating water stress effects [70].
Floods and droughts affect the river basin more intensely in its delta area due to higher pressure on water resources [68,72] (Figure 1). Also, Po delta is the largest wetland in Italy and over a third of its surface is protected under the Birds Directive [26] (Figure 1). Therefore, it was selected as the study area to analyse GI condition using data from the Sentinel-2 satellite platform [46].

Copernicus Land Monitoring Service: Riparian Zones and Corine Land Cover
CLMS local component focuses on different hotspots, i.e., areas that are prone to specific environmental challenges [49]. The Riparian Zones (RZ) is a local CLMS product that supports the objectives of European legal acts and policy initiatives, such as the Biodiversity Strategy to 2020 [27], the Habitats [25] and Birds Directives [26], the Water Framework [36] and the Floods Directives [37].
RZ consists of three products: Delineation of Riparian Zones (DRZ), Land Cover/Land Use (RZ LC/LU) and Green Linear Elements (GLE) [47]. Moreover, the DRZ product consists of three components [48] of which the Delineation of Potential Riparian Zones (DRZP) was used. DRZP is derived from weighing hydrological and geomorphological parameters, among other input data (Table 1), to express the likelihood of an area to host riparian features and hence to provide riparian-related benefits.
RZ LC/LU and DRZP ( Figure 2) are the main input data used in the approach to identify GI and its potential to provide riparian-related ES respectively, since they provide very detailed information of the riparian environment (LC/LU classes and its characteristics) along large and medium-sized river streams (Table 1).     Corine Land Cover (CLC) consists of an inventory of land covers classified in 44 overall classes. It is a pan-European product initiated in 1985 (reference year 1990) and updated every 6 years. The RZ LC/LU product was performed using CLC 2006/2012, among other inputs (Table 1) Table 2) were used for an internal cross-check validation approach of the main input data. The goal was updating the LC/LU classes, avoiding false positives in GI identification due to LC/LU changes between 2012 and 2018, before performing the subsequent spatial and temporal analyses on the vegetation condition. Afterwards, the misclassified GI was photo-interpreted using recent Sentinel-2 satellite images. The DRZP layer is not validated yet (by September 2019) due to lack of reference data and characteristics of riparian zones in sufficient detail [48]. To assure that the layer was suitably characterizing riparian areas in the case study area, it was cross-checked with available local ancillary data, i.e., products offered by the Emilia-Romagna region in an open-source catalogue: the hydro-ecoregions (HERs) [73]  Po HERs have been defined according to the implementation of the Water Framework Directive [36]. Each area is characterized based on: (i) the lithological structure and properties of the rocks (hardness, permeability and influence of water chemistry); (ii) relief (altitude and slope) and (iii) climate, depending on the precipitation and temperature (yearly average and seasonal variation) [73]. On the other hand, the FHRM are defined according to the Floods Directive [37] and delimitate hazard risk areas depending on: (i) scenarios of low, medium or high probability of flood; (ii) return period and (iii) information associated to all the exposed elements [74].
Firstly, these vector and alpha-numeric datasets were interpreted with respect to the aquatic ecosystem functioning and its benefits for the water balance. Subsequently, the outcome was compared with the DRZP buffers and percentage ranges. The correlation is visible in Figures 1 and 2.
The Natura 2000 network, obtained from the same catalogue, defines rich habitats that play a significant role as natural corridors within the wider landscape [75]. It was used to determine its distance to the identified GI. The goal was increasing the conservation priority accordingly [39].

Sentinel-2 Multispectral Imagery
The presented work used also Sentinel-2 (S2) satellite data because it can monitor large surfaces with high spatial, temporal and radiometric resolutions. This may explain its worldwide use as input data for land cover/land use monitoring and decision-making applications [50,51]. Table 3 shows the main characteristics of the optical sensor on-board S2 (the Multispectral Instrument-MSI) and the band set used.  3 10 days using one satellite, 5 days using two.
Just 4 bands were needed to calculate the biophysical variables applied to assess GI condition in terms of its vegetative health stage and water content [52]. Vegetation indices are calculated using the spectral bands that capture the Red and Near-Infrared (NIR) reflectance, as this part of the electromagnetic spectrum shows a higher sensitiveness to the leaf chlorophyll content [78,79]. On the other hand, leaf water content largely controls the spectral reflectance in the Short-Wave Infrared (SWIR) interval of the electromagnetic spectrum [80].

Methodology
The proposed steps to identify the existing NWRM and estimate its disposition and condition for delivering riparian regulative ES ( Figure 3) are based on the benchmarks developed for mapping and assessing ecosystems and their services [39,[57][58][59][60][61][62][63][64][65]. This is mainly required by the EEA to fulfil the targets of the 2020 Biodiversity Strategy in this regard [27].

Input Data Acquisition
Specifically, the Delimitation Units DU018A and DU005A, that catch the study area, were downloaded from the Delineation of Potential Riparian Zones (DRZP) and the Riparian Zones Land Cover/Land Use (RZ LC/LU) products in the local CLMS. Then, they were clipped using the basin boundaries.
As for the satellite data used, Sentinel-2 Level 1C (S2 L1C) scenes of the tile T32TQQ that covers Po delta were acquired from the French Sentinel collaborative ground segment PEPS-CNES [81], an operating platform mirroring all Sentinel products provided by the European Space Agency (ESA). Just one S2 tile was analysed in order to observe an area in which the climate conditions, phenology types and development trends could be considered almost the same [51]. 36 scenes for the period of 1st January-30th October 2018, with a cloud cover below 50%, were downloaded and pre-processed. This year was selected to perform an intra-annual assessment of the variability in the phenological trend of the selected GI. Also, it was selected since the scenes were less affected by atmospheric effects and cloud cover.

Pre-Processing
To assure that the main input Copernicus products were suitably characterizing the riparian areas in the case study area in the analysed year, they were cross-checked with auxiliary data, i.e., available local ancillary data and the updated version of CLC 2018. This way, the analysis was based on LC/LU datasets and satellite images of the same period.

Methodology
The proposed steps to identify the existing NWRM and estimate its disposition and condition for delivering riparian regulative ES ( Figure 3) are based on the benchmarks developed for mapping and assessing ecosystems and their services [39,[57][58][59][60][61][62][63][64][65]. This is mainly required by the EEA to fulfil the targets of the 2020 Biodiversity Strategy in this regard [27]. Figure 3. Flowchart of the proposed methodology to identify Natural Water Retention Measures (NWRM) in riparian areas of a river network, its disposition to provide regulative ES and its condition (input data explained in Section 2.2). 1 Comparison between the land cover/land use classes assigned to the identified GI due to the launch of an updated version of Corine Land Cover in 2018. 2 Due to lack of reference data on riparian characteristics in a sufficient level of detail, the authors checked the correlation between the DRZP layer's modelled area in the case study and regional products. Figure 3. Flowchart of the proposed methodology to identify Natural Water Retention Measures (NWRM) in riparian areas of a river network, its disposition to provide regulative ES and its condition (input data explained in Section 2.2). 1 Comparison between the land cover/land use classes assigned to the identified GI due to the launch of an updated version of Corine Land Cover in 2018. 2 Due to lack of reference data on riparian characteristics in a sufficient level of detail, the authors checked the correlation between the DRZP layer's modelled area in the case study and regional products.
The Maccs-Atcor Joint Algorithm (MAJA) [82] was used right after acquiring the S2 L1C scenes from the PEPS-CNES segment [81]. The selection of this atmospheric correction algorithm was based on its unique method for detecting clouds and shadows using multi-temporal series of data input instead of a single image [82]. This improves the correction of atmospheric, shadows and even slope effects in comparison with SNAP or Sen2Cor-derived S2 L2A [83], which could affect the vegetation and water indices that are calculated afterwards. Thus, time series of the T32TQQ tile were processed together since it did not represent a massive quantity of data. Afterwards, 7 scenes were selected, one per month from March to September 2018 (Table 4). This selection was made due to their lower cloud cover and hence fewer missing values. Also, since this period of the year catches the most prominent stage of the vegetation development cycle (according to the growing stages specified by the Food and Agriculture Organization of the United Nations, FAO [50,51]). Thus, this period catches the highest values of the analysed bio-geophysical indices if the detected GI is being adequately maintained [78][79][80]. Table 4. Scenes used of the T32TQQ tile, caught by the S2 MSI sensor, to test the approach; downloaded and pre-processed on the PEPS-CNES collaborative ground segment [81]. Lastly, band 11 (SWIR 1 ) was resampled from 20 to 10 m using the Semi-Automatic Classification Plugin 6.2.5 [84] in QGIS 3.2.1 in order to calculate the indices with the finest spatial resolution.

Identification of Agriculture and Forest NWRM in Riparian Areas
The proposed approach focused first on detecting vegetation and forestry GI sites in the riparian areas of Po river basin. More specifically, nature-based measures for water retention (NWRM). The RZ LC/LU product is the main input used [47]. As it follows the MAES nomenclature (levels 1 to 4) for defining the LC/LU classes, level 4 was consulted due to its higher level of detail.
These LC/LU classes were linked to the type of agriculture and forest NWRM using the catalogues developed by the European Commission Directorate-General Department for Environment Policies (DG-ENV) [42] (Table 6). An area of 4040 km 2 of agriculture and forest NWRM was detected in the riparian areas of Po river basin, finding a significant appearance of forest riparian buffers (66%), followed by meadows and pastures (20%). Table 6. Types of agriculture and forest GI for natural water retention (NWRM) identified in the riparian areas of Po river basin and corresponding area extent.

Pastures
Managed grasslands without trees and scrubs with a Tree Cover Density (TCD) of less than 30% and over or equal 30% Dry, mesic and alpine and subalpine grasslands without trees with a TCD of less than 30% After that, spatial operations were carried out in a model using ArcGIS 10.2 raster calculator algorithms ( Figure 3).

Spatial Model of GI Disposition to Deliver Regulative ES
Clipping the identified NWRM (1st output) by the DRZP model [47] allows to detect the disposition of each site to deliver the associated regulative ES in the riparian system [42]. This disposition can be expressed by weighing different hydrological and geomorphological parameters that affect the appropriate functioning of riparian ecosystems, especially during extreme events, such as floods or droughts [28,30,42,[72][73][74], and that the DRZP product takes into account (Table 1): (i) distance to water bodies, (ii) slope, (iii) flood hazard risk areas and their return period and (iv) soil type (i.e., erosion and permeability features) [47,48].
As a result, a spatial model of the identified agriculture and forest GI shows its disposition to deliver NWRM-related regulative ES (2nd output), measured from 0% to 100% and with a spatial resolution (SR) of 100 m due to the SR of the Copernicus product [47].
Pixel-based Assessment of GI Condition

•
Buffering of the Natura 2000 network The Natura 2000 areas of the river basin [75] were buffered every 10 m until 100 m [39,85]. Then, the identified GI was classified according to 10 distance ranges. This parameter was used to prioritise adequate management and conservation of those GI sites that belong or are close to Natura 2000 areas due to their contribution to the ecosystem's appropriate functioning, being hence significantly vulnerable items to changing climate consequences [39,[57][58][59][60][61][62][63][64][65].

•
Calculation and multitemporal analysis of biophysical variables Vegetation and water indices were calculated for the period of March-September 2018 using the spectral bands from the 7 selected S2 images, once corrected from the atmospheric effect and resampled. Also, a filter of GI with a surface of less than 0.1 ha was applied due to S2 spatial resolution.
Several indices were analysed complementarily to obtain a more accurate and reliable characterization of the environment [52]. These spectral indices (Table 5) were selected for being the most significantly used in vegetation and forestry studies, achieving representative and accurate results in previous experiences [52,86,87].
The Normalized Difference Vegetation Index (NDVI), as well as the Enhanced Vegetation Index (EVI), have been the most successful in studying the development stage, healthiness and vigorousness of vegetation [50,51,86]. On the other hand, the Normalized Difference Water Index (NDWI), also called Normalized Difference Moisture Index (NDMI) in some studies, has been used to detect wetness and water content in vegetation [87].
NDVI is calculated using the reflectance from the Red channel (R) and the Near-Infrared (NIR) (Equation (1)) [78]. EVI was selected since it complements the information derived from NDVI, being more sensitive to differences in heavily vegetated areas and less affected by atmospheric noise [79,86]. It is calculated similarly to NDVI, but also considering the reflectance in the Blue channel (B) (Equation (2)) [79].
However, vegetation indices have a limited capability for retrieving vegetation water content due to uniquely providing information on vegetation greenness (chlorophyll), which is not directly nor uniformly related to the quantity of water in vegetation [80,87]. Thus, NDWI was also calculated.
This index is defined using NIR and SWIR reflectance (Equation (3)) and, as NDVI and EVI, shows values in the range of −1 to +1, with higher values corresponding to higher leaf water content and vegetation cover [80]. The main reason for choosing this index was the easier observable monitoring of vegetative and forestry stages when observing their reflectance in the SWIR bands, as well as to identify water stress [52,87]. Finally, a statistical analysis was performed to evaluate the yearly maximum value (and hence the healthiest and most vigorous vegetation stage) per index, pixel and GI type. This is since each vegetation surface displays its specific multitemporal variation of biophysical characteristics. Thus, each surface is defined according to a specific variation pattern during its annual development cycle [50,51]. •

Rating of the ES condition indicators
The value of the indicators selected to assess GI condition (3rd output) was weighed using the following expressions as assumptions based on the existing theoretical approaches (Table 5) Greenness response and water stress Representing i, the data extracted per pixel (10 × 10 m); D, the disposition to deliver the associated regulative ES, resampled to 10 m SR and expressed in percentages; N, the distance range from a GI to the Natura 2000 network; NDVI(BOA) maxi , EVI(BOA) maxi and NDWI(BOA) maxi , the maximum value per pixel for the analysed period (March-September 2018) of the bio-geophysical indices. All the indicators were analysed as integers and expressed in values from 0 to 10. Moreover, if in a pixel no data existed for a parameter, a null value was assigned to that indicator.
The conservation condition, C, (Equation (9)) was obtained per-pixel and per type of NWRM (∀ j). Then, it was dissolved to obtain one single modelled area indicating GI condition in the case study. The condition index was obtained summing the selected indicators, equally scaled, to allow the easy integration of new condition indicators in future assessments. All the evaluated parameters were considered equally significant since the consulted literature did not mention any distinction of priorities in that regard.

Conservation condition, C
Finally, the integer values obtained with Equation (9), from 0 to 50, with higher values representing a better condition, were translated into a ramp colour legend. This eases the interpretation of the spatial model, quickly locating GI playing a major role in the delivery of ES but that, given their compromised condition, would require management interventions.
The developed model assesses GI actual status based on the maximum value of the vegetation and water indices achieved per type of GI in its intra-annual development trend, among other condition indicators. Neither intra-seasonal nor inter-annual changes are assessed.

Results
Appropriately assessing ecosystems' condition must concern individual ecosystems, but also their territorial context [12,27,39]. Thus, the assumptions made and indicators used in the approach, based on the frameworks developed to map and assess ecosystems' conditions and their services (Table 5), include: (i) disposition of the identified NWRM to provide regulative ES in the riparian system, (ii) proximity to protected areas included in Natura 2000 and (iii) remotely-sensed greenness and water stress response as means of the ecosystem's functional attributes. Figure 4 shows the inputs used (1st and 2nd columns) and outputs delivered (3rd column) for the same area extent. The area is part of the Po delta. It mainly consists of lines of trees and scrub, natural and semi-natural broadleaved forests and alpine and sub-alpine natural grassland (Figure 4a). According to the NWRM catalogue [42], these LC/LU classes were translated into "green cover", "forest riparian buffers" and "buffer strips and hedges" ( Table 6) (Figure 4g). Those NWRM located closer to the river streams present a higher disposition to deliver regulative ES (Figure 4h), based on the DRZP dataset (Figure 4b), and a higher proximity to protected areas in Natura 2000 (Figure 4c). These two condition indicators and the analysis of biophysical variables (2nd column of Figure 4) allowed to detect GI in moderate condition (Figure 4i).

Spatial Model of GI Disposition to Deliver Regulative ES
A spatial model weighing the capacity of the identified GI to deliver NWRM-related ES, such as protection against flood events, was obtained for the entire river basin. This disposition is measured in percentages from 0% to 100% and translated into a colour ramp from light to dark blue, following the labels of very low, low, medium, high and very high capacity to ease its interpretation ( Figure 5).
The highest capacity to deliver the related regulative ES was found in NWRM located closer to the river flows (dark blue). Moreover, the riparian buffers located closer to the river courses were wider in the central part of the basin than in the delta area ( Figure 5).

Spatial Model of GI Disposition to Deliver Regulative ES
A spatial model weighing the capacity of the identified GI to deliver NWRM-related ES, such as protection against flood events, was obtained for the entire river basin. This disposition is measured in percentages from 0% to 100% and translated into a colour ramp from light to dark blue, following the labels of very low, low, medium, high and very high capacity to ease its interpretation ( Figure 5).

Pixel-Based Assessment of GI Condition
A model representing riparian GI condition in 2018 was obtained in Po delta ( Figure 6). GI condition is illustrated according to a colour ramp from green to red, following the labels of good, acceptable, moderate, severe or critical, depending on the indicators' values (Table 5), assessed through Equations (4)- (9).
The assessment model delivered might serve as an early warning tool for that GI holding a major ecological role (highly weighed capacity to deliver regulative ES and proximity to Natura 2000 sites), but not suitably conserved (showing low vegetation healthiness and exposure to water pressures according to the bio-geophysical indices' values). Figure 6 shows the identified riparian GI condition in two different areas of Po delta: (i) GI buffers closer to the river flow (left), mainly identified as "forest riparian buffers" and "green cover" (Table 6), and (ii) fields located between river flows (right), identified as "forest riparian buffers" and "crop rotation" (Table 6), whose disposition to deliver regulative ES is shown in Figure 5 (right).
The developed model allows to find GI in severe and critical conditions (orange to red colours) in both locations. In the first case, it corresponds to areas that suffer from water stress events (floods and droughts) due to influence of the river flow and, in the second, to stressed vegetation. These results are estimated from very low values of the bio-geophysical indices (which characterise bare soil and water masses) combined with a high capacity of the NWRM to deliver regulative ES and the proximity to, or inclusion in, protected habitats. The highest capacity to deliver the related regulative ES was found in NWRM located closer to the river flows (dark blue). Moreover, the riparian buffers located closer to the river courses were wider in the central part of the basin than in the delta area ( Figure 5).

Pixel-Based Assessment of GI Condition
A model representing riparian GI condition in 2018 was obtained in Po delta ( Figure 6). GI condition is illustrated according to a colour ramp from green to red, following the labels of good, acceptable, moderate, severe or critical, depending on the indicators' values (Table 5), assessed through Equations (4)- (9).
The assessment model delivered might serve as an early warning tool for that GI holding a major ecological role (highly weighed capacity to deliver regulative ES and proximity to Natura 2000 sites), but not suitably conserved (showing low vegetation healthiness and exposure to water pressures according to the bio-geophysical indices' values). Figure 6 shows the identified riparian GI condition in two different areas of Po delta: (i) GI buffers closer to the river flow (left), mainly identified as "forest riparian buffers" and "green cover" (Table 6), and (ii) fields located between river flows (right), identified as "forest riparian buffers" and "crop rotation" (Table 6), whose disposition to deliver regulative ES is shown in Figure 5 (right).

Disposition to Deliver Regulative ES
The capacity to provide regulative ES ( Figure 5) was evaluated in an area of 3077 km 2 of the detected NWRM in Po basin, of which over 40% presented high or very high ranges (Figure 7).  The developed model allows to find GI in severe and critical conditions (orange to red colours) in both locations. In the first case, it corresponds to areas that suffer from water stress events (floods and droughts) due to influence of the river flow and, in the second, to stressed vegetation. These results are estimated from very low values of the bio-geophysical indices (which characterise bare soil and water masses) combined with a high capacity of the NWRM to deliver regulative ES and the proximity to, or inclusion in, protected habitats.

Disposition to Deliver Regulative ES
The capacity to provide regulative ES ( Figure 5) was evaluated in an area of 3077 km 2 of the detected NWRM in Po basin, of which over 40% presented high or very high ranges (Figure 7). "Forest riparian buffers" represented the greater GI area showing a very high disposition to deliver regulative ES, followed by "meadows and pastures" (Figure 8). This outcome could be expected since these classes represent the major riparian GI surface in the basin. However, it seems remarkable that nearly all the surface occupied by natural "riverbanks" presented a very high disposition to deliver regulative ES. This fact might be related to their proximity to the river streams and hence their important role in flood scenarios.

Disposition to Deliver Regulative ES
The capacity to provide regulative ES ( Figure 5) was evaluated in an area of 3077 km 2 of the detected NWRM in Po basin, of which over 40% presented high or very high ranges (Figure 7).  "Forest riparian buffers" represented the greater GI area showing a very high disposition to deliver regulative ES, followed by "meadows and pastures" (Figure 8). This outcome could be expected since these classes represent the major riparian GI surface in the basin. However, it seems remarkable that nearly all the surface occupied by natural "riverbanks" presented a very high disposition to deliver regulative ES. This fact might be related to their proximity to the river streams and hence their important role in flood scenarios.

Condition Assessment in 2018 in Po Delta
The current condition was assessed ( Figure 6) in an area of 102.5 km 2 , which represents the agriculture and forest riparian NWRM in the Po delta. Over 80% was evaluated as presenting either an acceptable or moderate condition in 2018 following the presented approach, 5% showed a good condition and about 12% presented a severe or critical status (Figure 9).

Condition Assessment in 2018 in Po Delta
The current condition was assessed ( Figure 6) in an area of 102.5 km 2 , which represents the agriculture and forest riparian NWRM in the Po delta. Over 80% was evaluated as presenting either an acceptable or moderate condition in 2018 following the presented approach, 5% showed a good condition and about 12% presented a severe or critical status (Figure 9). "Crop rotation" stood out as the riparian NWRM most affected by a critical condition, followed by "forest riparian buffers" and natural "riverbanks" (Figure 10). This might be due to high exposure to vegetation stress and water pressures, but it must be considered that natural "riverbanks" also represented a very high capacity to deliver regulative ES (Figure 8). Therefore, this GI should be managed accordingly. Also, "green cover" was the main GI facing a severe condition (Figure 10).

Condition Assessment in 2018 in Po Delta
The current condition was assessed ( Figure 6) in an area of 102.5 km 2 , which represents the agriculture and forest riparian NWRM in the Po delta. Over 80% was evaluated as presenting either an acceptable or moderate condition in 2018 following the presented approach, 5% showed a good condition and about 12% presented a severe or critical status (Figure 9). "Crop rotation" stood out as the riparian NWRM most affected by a critical condition, followed by "forest riparian buffers" and natural "riverbanks" (Figure 10). This might be due to high exposure to vegetation stress and water pressures, but it must be considered that natural "riverbanks" also represented a very high capacity to deliver regulative ES (Figure 8). Therefore, this GI should be managed accordingly. Also, "green cover" was the main GI facing a severe condition ( Figure 10). On the other hand, a high percentage of agricultural fields holding crop rotation, intercropping or other complex crop patterns (gathered as a sole GI class named "crop rotation") ( Table 6) showed a good or acceptable condition ( Figure 10). As for the GI class "continuous cover forestry", it was the only one not found in Po delta, which could be assumed since it did not represent a significant area in the basin.

Discussion
Decision-makers have mentioned the need for a practical tool that shows values or thresholds characterizing ES supply and condition, facilitating the accomplishment of policy objectives [88,89]. The presented approach can serve as a baseline, helping in understanding the indicator frameworks [55]. It specifically focuses on easing the tasks of Member States (MS) on the Mapping and Assessment of Ecosystems and their Services (MAES) (Action 5 of the 2020 Biodiversity Strategy [27]).
The approach focuses on riparian zones since the potential to positively contribute to the socioeconomic and environmental resilience of their area of influence has been proven [19,20,28,30,42] based on the following effects: improved water quality, positive trend of new natural riparian functionalities, enhanced environmental and morphological quality and increased awareness of stakeholders and citizens.
Specifically, the Riparian Zones dataset was used as the main input of the approach. Very few scientific experiences exist that refer to it [48,91]. In fact, the main product used, DRZP, is not validated by September 2019 due to lack of appropriate validation data comparable to it [48]. Thus, the developed approach contributes to increase the applicability of this dataset, not generally to riparian areas, but specifically to map and assess NWRM, which are important elements for the water sector, especially for regulating extreme events, such as floods or droughts. The method identifies NWRM in riparian regions, evaluates its capacity to deliver regulative ES and assesses its current preservation condition. The approach considered innovative elements to assess GI and allowed to find weaknesses that should be solved in existing datasets.
GI condition was assessed according to the following criteria (selected from the aforementioned indicator frameworks): (i) capacity to provide ES, (ii) membership or proximity to the Natura 2000 On the other hand, a high percentage of agricultural fields holding crop rotation, intercropping or other complex crop patterns (gathered as a sole GI class named "crop rotation") ( Table 6) showed a good or acceptable condition ( Figure 10). As for the GI class "continuous cover forestry", it was the only one not found in Po delta, which could be assumed since it did not represent a significant area in the basin.

Discussion
Decision-makers have mentioned the need for a practical tool that shows values or thresholds characterizing ES supply and condition, facilitating the accomplishment of policy objectives [88,89]. The presented approach can serve as a baseline, helping in understanding the indicator frameworks [55]. It specifically focuses on easing the tasks of Member States (MS) on the Mapping and Assessment of Ecosystems and their Services (MAES) (Action 5 of the 2020 Biodiversity Strategy [27]).
The approach focuses on riparian zones since the potential to positively contribute to the socio-economic and environmental resilience of their area of influence has been proven [19,20,28,30,42] based on the following effects: improved water quality, positive trend of new natural riparian functionalities, enhanced environmental and morphological quality and increased awareness of stakeholders and citizens.
Specifically, the Riparian Zones dataset was used as the main input of the approach. Very few scientific experiences exist that refer to it [48,91]. In fact, the main product used, DRZP, is not validated by September 2019 due to lack of appropriate validation data comparable to it [48]. Thus, the developed approach contributes to increase the applicability of this dataset, not generally to riparian areas, but specifically to map and assess NWRM, which are important elements for the water sector, especially for regulating extreme events, such as floods or droughts. The method identifies NWRM in riparian regions, evaluates its capacity to deliver regulative ES and assesses its current preservation condition. The approach considered innovative elements to assess GI and allowed to find weaknesses that should be solved in existing datasets.
GI condition was assessed according to the following criteria (selected from the aforementioned indicator frameworks): (i) capacity to provide ES, (ii) membership or proximity to the Natura 2000 network and (iii) indicators of the ecosystem's functional attributes: greening response and water stress ( Table 5). Obtaining the condition index by summing the indicators' values, equally scaled, allows its scalability in terms of integrating indicators from future frameworks. Considering the distance of GI to protected areas included in Natura 2000 is a significant and novel element of the approach. These areas play a distinct role in the natural ecosystem's functioning [27]. Therefore, the appropriate conservation of GI connected to these areas must be prioritised accordingly [39].
Firstly, the NWRM catalogue developed by DG-ENV [42] was used to identify GI as to ensure consistency with already existing and validated definitions. However, it shows weaknesses in the NWRM class definition (e.g., not including wetland riparian vegetation in the hydro-morphological sector). Also, it is challenging to translate some classes into specific LC/LU (e.g., green cover). Therefore, the process for identifying GI should remain more autonomous to increase the usability of the approach.
The analysis focused on vegetation riparian GI. Hydro-morphological types of GI (e.g., wetlands, saltmarshes or reeds) are also important for flood and drought regulation [42], but the interpretation of biophysical variables differ for "green" and "blue" GI, i.e., the reflectance values, annual developing trends and hence the meaning of the indices [50,51,[78][79][80]. Therefore, appropriate indices to monitor this GI must be thoroughly selected and understood before this GI's assessment could be integrated in the approach. Annual trends also vary between different vegetation GI. Thus, the indices' values were rated per type of GI.
The approach used S2 data to calculate the biophysical variables with a high spatial resolution (10 m) [46] and thereby overcome the weakness of the CLMS [49]. Just Copernicus' global component provides bio-geophysical indices products, by September 2019. These products present 300 m/1 km spatial resolutions, not fulfilling the needs at local scales [52,54].
However, clouds and shadows remain a major inconvenience when processing and interpreting bio-geophysical indices, even after correcting the images from the atmospheric effect [51,92]. Therefore, the indices can sometimes show low or even negative values that do not correspond to the vegetation features.
This was solved by analysing the maximum values of the indices in the period of March-September 2018. Thus, the analysis considers the most vigorous, healthiest and highest leaf water content response of each GI development cycle in the analysed year [50,51]. However, the possibility of having evaluated values that represent a cloud or shadow instead of the natural surface condition must be considered.
Analysing the outcomes, the representativeness of each NWRM, as well as the capacity to provide regulative ES, can be quickly interpreted in the modelled area. In Po river basin, "forest riparian buffers", followed by "meadows and pastures", popped up as the greater GI areas (Table 6). Moreover, most of this surface presented a high or very high capacity to deliver regulative ES (Figure 8). Instead, "continuous cover forestry" was the less presented GI (Table 6).
Also, the delivered model shows GI condition. 12% of the riparian GI area in Po delta presented a high conservation priority (severe or critical condition) in the analysed year ( Figure 9). Most of this GI was also closer to river streams ( Figure 6). Thus, the condition index may result due to stressed or saturated vegetation conditions (i.e., very low values of the bio-geophysical indices) merged with a high capacity to supply regulative ES ( Figure 5) and proximity to protected areas.
"Riverbanks" represent a frequently used nature-based measure for flood protection. This NWRM face severe and critical conditions in Po delta (Figure 10), while having a very high capacity for delivering regulative ES (Figure 8). However, it must be considered that riverbanks sometimes represent fully functional sparsely vegetated areas [42].
Alongside the regulation of water stress events, such as floods or droughts, riparian areas can provide other significant benefits to the environment and society [19,20,28,42]. Therefore, having open access to a tool that quickly highlights GI facing severe or critical conditions, hence not appropriately conserved, should raise awareness, help and motivate decision-makers to take action, supporting restoration and management strategies that improve the environmental quality of that area (e.g., revegetation measures) [23,31,43].
To this end, the interpretation of the model shall be supported by degradation assessments and analyses of the impact of GI presence or absence [93]. Also, inter-annual analyses would allow to interpret the GI condition index depending on previous trends (i.e., considering each GI likelihood to achieve different indices' maximum values). However, the required datasets do not perfectly fit inter-annual analyses, with the consequent impact on the results of using LC/LU datasets from other periods. In this regard, the upcoming Copernicus product CLC+ will ease a more analytic mapping of the Earth's surface and a more flexible combination with other datasets [94].
Finally, the decision of motivating either conservation initiatives on GI in a good condition or recovering strategies on damaged GI will depend on regional policy objectives and cost-effectiveness of the measures [93]. This decision could be further substantiated by applying an ES valuation exercise, supplementing the information derived from the condition indicators. However, this approach remains challenging due to the complexity of standardising ES values, which are strongly dependent on context specific circumstances that would require detailed local datasets [95].

Conclusions
There is a growing demand for GI and ES assessments due to their significant role in natural hazards mitigation and climate change adaptation. However, information on the condition of, and changes in, Europe's ecosystems dominated by vegetation is still limited. The presented approach represents a new method to overcome the current lack of data on riparian characteristics. The integration of highly detailed products from the CLMS (Riparian Zones) with other datasets (protected areas in Natura 2000) and high-resolution satellite data (S2) demonstrates, through the followed approach, their potential to: (i) identify GI in the riparian areas of a river network, specifically agriculture and forestry measures that serve for natural water retention; (ii) assess its disposition to deliver the related regulative ES; (iii) analyse its condition according to the existing indicator frameworks, such as the 2018 MAES report and (iv) rank its conservation priority. Policy-related factors, such as the latest initiatives of each region to either conserve GI in good condition or recover degraded GI, should be taken into account in the latter case.
GI sites are currently subjected to many pressures caused by both natural and anthropogenic actions, which decrease its capacity for delivering ES. Thus, Copernicus evolution depends on meeting the needs for ecosystems' monitoring and management coming from environmental and socio-economic policies and strategies at global, European and local scales. In this regard, the presented research highlights once more [52] the products' spatial resolution as a significant handicap, by September 2019. On one hand, having full access to bio-geophysical indices based on already processed data from S2 would ease the tasks of downloading and processing this data, as well as dealing with missing values. On the other hand, it would highly increase the spatial resolution of the products to 10 m, fulfilling users' needs for studies at local scales [54]. Then, these indices could be used to develop new products or improve the existing ones in the frame of mapping and assessing green areas. The upcoming CLMS High Resolution Vegetation Phenology and Productivity will help in this sense [96].
In conclusion, the approach allows to: • Provide a scalable tool based on open-source data, mainly from the CLMS, to support environmental and sustainability policies and strategies in the field of mapping GI and monitoring its condition and pressures in riparian areas.

•
Provide a design to account for the constitutive elements of nature-based solutions, such as GI, including its multifunctionality and a simultaneous delivery of environmental and social benefits, based on a multi-stakeholder engagement.
Finally, it can be concluded that it is possible to conjugate environmental protection and territorial development though the coordination of monitoring activities. Prioritising GI's need of restoration depending on its role in the river system, proximity to protected areas and current condition can help raising awareness and implementing actual needs in regional coordination actions. This determines the need of communication with the public and decision-makers to highlight the potential of Copernicus as an upstream service to collect, share, organize and elaborate data on natural resources management, both in real time and historical studies. In the future, a closer collaboration with policy and decision-makers could help in Copernicus uptake at local scale, creating services that suit their needs and requirements and making it a more real-world tool that facilitates the use of remotely-sensed information in policy.