Groundwater impacts on surface water quality and nutrient loads in 1 lowland polder catchments : monitoring the greater Amsterdam area 2

The Amsterdam area, a highly manipulated delta area formed by polders and reclaimed lakes, struggles with high 11 nutrient levels in its surface water system. The polders receive spatially and temporally variable amounts of water and 12 nutrients via surface runoff, groundwater seepage, sewer leakage and via water inlet from upstream polders. Diffuse 13 anthropogenic sources, such as fertilizer use and atmospheric deposition, add to the water quality problems in the polders. 14 The major nutrient sources and pathways have not yet been clarified due to the complex hydrological system in such lowland 15 catchments combined with both urban and agriculture areas. In this study, the spatial variability of the groundwater seepage 16 impact was identified by exploiting the dense groundwater and surface water monitoring networks in Amsterdam and its 17 surrounding polders. Twenty-three variables (concentrations of Total-N, Total-P, NH4, NO3, HCO3, SO4, Ca, and Cl in 18 surface water and groundwater, seepage rate, elevation, land-use, and soil type) for 144 polders were analysed statistically 19 and interpreted in relation to sources, transport mechanisms and pathways. The results imply that groundwater is a large 20 source of nutrients in these mixed urban/agricultural catchments, given the higher nutrient levels in groundwater compared 21 with surface water. The groundwater nutrient concentrations exceeded the surface water Environmental Quality Standards 22 (EQSs) in 93 % of the polders for TP and in 91 % for TN. Groundwater outflow into the polders thus adds to nutrient levels 23 in the surface water. High correlations (R 2 up to 0.88) between solutes in groundwater and surface water, together with the 24 close similarities in their spatial patterns, confirmed the large impact of groundwater on surface water chemistry, especially 25 in the polders that have high seepage rates. Our analysis indicates that the elevated nutrient and bicarbonate concentrations in 26 the groundwater seepage originate from the decomposition of organic matter in subsurface sediments coupled to sulfate 27 reduction and possibly methanogenesis. The large loads of nutrient rich groundwater seepage into the deepest polders 28 indirectly affect surface water quality in the surrounding area, because excess water from the deep polders is pumped out and 29 used to supply water to the surrounding infiltrating polders in dry periods. The study shows the importance of the connection 30 between groundwater and surface water nutrient chemistry in the greater Amsterdam area. We expect that taking account of 31 Hydrol. Earth Syst. Sci. Discuss., doi:10.5194/hess-2017-99, 2017 Manuscript under review for journal Hydrol. Earth Syst. Sci. Discussion started: 17 March 2017 c © Author(s) 2017. CC-BY 3.0 License.

impact is still reflected by the presence of brackish groundwater in the shallow subsurface (Schot, 1992 b). 23 The construction of the Amsterdam Rhine Canal separated the study area into two parts (Fig. S1): the Central Holland in the 24 west and the Vecht lakes area in the east. In the Central Holland polders, relatively thick peat layers and pyrite rich clays are 25 still present in the shallow subsoil, as described by Van Wallenburg (1975). The Vecht lakes area is characterized by large 26 open water areas and a number of wetland nature reserves. The rest of the Vecht lakes area is mainly grassland used for dairy 27 farming. Soils in this area are generally wet and rich in organic matter and clay (Schot, 1992 b). 28 Mainly during the 20 th century, the urban areas have been growing from the historic city-centres on river and tidal channel 29 levees into the surrounding low-lying polders. To facilitate the construction of buildings, a 1-5 meter-layer of sand was often 30 supplied on top of the original sediments. The thickness of this suppletion sand layer is extremely variable even at a small 31 scale. The sand suppletions are either calcite-poor without shell fragments or calcite rich with shell fragments that indicate 32 water Zuiderzee, which was dammed in the 1930s and transformed into the fresh water Lake IJssel (connected to Lake IJ), 27 which is now the biggest fresh water reservoir of The Netherlands; (2) the deep polders Groot-Mijdrecht and Horstermeer, 28 which are reclaimed lakes with clayey lake sediments at the surface. groundwater quality patterns, the variation of concentrations in and between the 5 regions was visualized using boxplots 1 (Helsel and Hirsch, 2002). 2 Because our dataset contains both fresh and brackish to saline water, we used the mass SO 4 /Cl ratio of the samples as an 3 indicator of sulfate reduction. SO 4 /Cl ratios lower than the sea water ratio of 0.14 (Morris and Riley, 1966) point to the 4 occurrence of sulfate reduction (Appelo and Postma, 2005; Griffioen et al., 2013). Ratios above 0.14 point to the addition of 5 sulfate relative to diluted sea water through processes like pyrite (FeS 2 ) oxidation or through input via atmospheric inputs, 6 fertilizers, manure, or leakage and overflow of sewer systems. 7 Average concentrations in groundwater for each polder were mapped to be compared with average annual surface water 8 concentrations (See section 2.2.2). The potential relationship between the solute concentrations in groundwater (TN, NO 3 , 9 NH 4 , SO 4 , TP, Ca, HCO 3 , and Cl) and the landscape variables (paved area percentage, elevation, seepage rate, surface water 10 area percentage, lutum, humus and calcite percentages of top soil) were explored using the Spearman correlation, which 11 reduces the influence of outliers and yields a robust correlation statistic (Helsel and Hirsch, 2002). 12 To further explore the statistical relations in our data set, scatter plots were made to evaluate HCO 3 , SO 4 , Cl, and nutrient 13 (NO 3 , NH 4 and PO 4 ) concentrations in groundwater. We also explored the links between alkalinity (over 99 % of our 14 groundwater alkalinity was dominated by HCO 3 , (Stuyfzand, 2006)), Cl concentration, SO 4 /Cl ratio and nutrients (TN, NH 4 15 and TP) concentrations. For our interpretation, we also used the calculated amount of consumed or produced SO 4 in mg L -1 16 relative to the SO 4 /Cl ratio of diluted seawater, using Eq. (3): 17 18 In order to understand the impact of cation exchange processes involving Ca and Na exchange during salinization and/or 19 freshening of aquifers (Griffioen, 2004;Stuyfzand, 2006) we defined the amount of exchange Na ex as: 20 21 Where, Na ex is the amount of Na exchange; gw, ground water; seaw, seawater. Na ex > 1 points to freshening, Na ex < -1 to 22 salinizing conditions. 23

Surface water data 24
Loads represent the contribution of polders to surface water quality of the regional water system in weight per time unit. To 25 eliminate the impact of the size of polders, we calculated daily load per area in kg ha -1 d -1 . This was calculated using the 26 daily average loads of each solute divided by the polder areas using Eq. (5): 27 oad per area Where L is daily load kg d -1 , A is polder area (ha), C is daily solute concentration in mg L -1 and Q is daily discharge in m 3 d -28 quality measurements for the period 2006-2013 of 144 polders were extracted from the Waternet database. The 1 measurements were converted to daily time series by stepwise interpolation between the monthly measurements. We 2 assigned a concentration of zero to measurements below the detection limits. Discharge data Q are daily measurements over 3 the same time period. An average over multiple pumps, when present, was taken for each polder. For further details about 4 the data processing we refer to Table S2. 5 The pumping discharge is regulated to respond to water surplus or deficiency conditions in the polder catchments. Using the 6 pumping frequency data, we proved that solute concentrations in pumped water are usually higher at the beginning of each 7 pumping activity (Van der Grift et al., 2016). The pumping rates may also influence water quality in the polder. To eliminate 8 differences caused by pumping rates, we used the normalized concentration calculated using Eq. (6): 9 In this equation, C is the normalized concentration (mg L -1 ), Load per area is from Eq. (1), Q is the pumping discharge (m 3 y -10 1 ), and A is the polder area (m 2 ).The statistical methods that were used for groundwater quality (described in section 2.2.1) 11 were also applied to the surface water normalized concentrations. 12 Based on a national assessment on ecosystem vulnerability, Environmental Quality Standards (EQSs) were set by the Water 13 Boards (Heinis and Evers, 2007). For most ditches and channels in the clay and peat regions, EQSs of TN and TP are 2.4 mg 14 L -1 and 0.15 mg L -1 , respectively (Rozemeijer, 2014). We used these most common EQSs as reference concentration values. 15 For example, the EQSs of TN and TP were used for the legend classifications in our surface water quality maps and were 16 added as reference lines in our concentration boxplots. Percentages of polders exceeded these standards were calculated in 17 this paper. 18

Surface water compared with groundwater solute concentrations 19
We statistically analysed the groundwater and surface water quality data and landscape characteristic variables by (1) 20 calculating the correlation coefficients between averaged groundwater solutes concentrations and normalized concentrations 21 of surface water using the Spearman method, and (2) by selecting variables (based on the correlation matrix above) to be 22 integrated into multiple linear regression models for predicting surface water solute concentrations. Again, the Spearman 23 method was applied and linear regression was based on ranks in order to avoid outliers to determine the outcomes.

Solutes redistribution in surface water 1
Loads were used to assess the impact of different polders as sources of solutes for the boezems and the receiving water 2 bodies further downstream. In general, the spatial patterns can be distinguished through maps of the surface water solute 3 loads per area if there are no other influences. However, there are exceptions such as the seepage water which is pumped out 4 of the two upconing polders Groot Mijdrecht and Horstermeer, which is discharged into the Boezem system and used as inlet 5 water for the surrounding polders during summer. To show the impact of this inlet water on the receiving polders' water 6 quality, we analysed the inlet solute loads and the resulting surface water concentrations for polder Botshol. Polder Botshol 7 (part of polder # 104 Noorderpolder of Botshol (zuid and west)) with an area of 1.3 km 2 receives inlet water from the Amstel 8 boezem system that has a significant contribution of seepage water that is pumped out of the polder Groot Mijdrecht. 9 Two models were applied for simple solute concentration calculations based on inlet water quality. Model 1 calculates the 10 accumulation of solutes in the water body, with evaporation as the only output for water (leaving the solutes behind). Model were also for the two deep polders Groot Mijdrecht (# 80) and Horstermeer (Upconing area) with known upconing of salt 1 groundwater. The Central Holland area was dominated by fresh groundwater with low Cl and Ca concentrations, but with 2 considerable amounts of HCO 3 . Polders with relatively high chloride (>1000 mg L -1 ) are distributed along the former 3 Zuiderzee margin, plus the Upconing area which is two deep polders with known upconing of brackish water. Relative to the 4 regions above, the Vecht lakes area and the Ice pushed ridge showed significantly less mineralized waters with lower HCO 3 5 and Cl concentrations. For example, the P75s of Cl in these two regions are below 150 mg L -1 and the P75s of HCO 3 below 6 350 mg L -1 . The groundwater HCO 3 concentrations (Fig. 6) show an east-west increasing trend with highest concentrations 7 in both the fresh and brackish areas west of the Amsterdam Rhine Canal. 8

SO 4 and SO 4 /Cl 9
The Zuiderzee margin and the Upconing area showed large ranges of SO 4 concentrations (P25 and P75: 7-125 mg L -1 and 7-10 250 mg L -1 , respectively) with the SO 4 /Cl mass ratios generally lower than the 0.14 ratio for diluted seawater. The polders in 11 the eastern Zuiderzee margin showed the highest average SO 4 levels ( Fig. 6). The Central Holland area exhibited the lowest 12 SO 4 concentrations with the smallest variability, with SO 4 /Cl P75 typically lower than 0.14. However, some outliers in this 13 region reached quite high sulfate concentration levels (>200 mg L -1 ). The Vecht lakes and the Ice pushed ridge showed 14 intermediate sulfate concentrations and typically have a SO 4 /Cl ratio clearly above 0.14. 15

NH 4 , TN, NO 3 , and TP 16
The higher groundwater NH 4 and TP concentrations generally locate in the western part of the study area (Zuiderzee margin, 17 Holland (10.6 mg L -1 ) were far higher than in the Vecht lakes (2.1 mg L -1 ) and Ice pushed ridge regions (0.07 mg L -1 ). The 19 same was observed for TP (0.7, 1.6, 0.2 and 0.06 mg P L -1 , respectively). Nutrient concentrations in the Upconing area 20 (medians 5.7 mg NH 4 L -1 and 0.14 mg P L -1 ) were relatively low compared with the groundwater in the Zuiderzee margin 21 and Central Holland areas, although we consider the NH 4 concentration levels to be substantial given the surface water EQS 22 of 2.4 mg N L -1 . TN showed the highest median concentration levels in the Zuiderzee margin and Central Holland regions, as 23 well as in the Ice pushed ridge (7.3 mg N L -1 ). The Ice pushed ridge region also showed the highest level of NO 3. In the latter 24 region, nitrate is the main component of TN, while NH 4 is the main component in the other regions. 25 Groundwater quality varied from fresh, low mineralized in the eastern parts (Vecht lakes and Ice pushed ridge, Figure 4) 26 towards brackish, highly mineralized and nutrient rich groundwater in the northwest (Zuiderzee margin and Central Holland, 27 In the more mineralized groundwater systems, sulfate reduction is a potential cause of the significant relationship between 1 HCO 3 , TP, and NH 4 . From using the SO 4 /Cl ratio of the samples and comparing them with the SO 4 /Cl ratio in seawater (Eq. 2 3), it appears that most of the brackish groundwater showed signs of sulfate reduction. Figure 7 shows that the amount of 3 SO 4 consumed in the sulfate reduction process increased with the chloride concentration of the groundwater, and that sulfate 4 reduction was complete only in part of the groundwaters. Note that groundwater below polders with excess SO 4 are all in 5 water with Cl<1000 mg L -1 . It follows from Figure 8 that high HCO 3 , TP, and NH 4 concentrations mostly occurred in 6 groundwater with a SO 4 /Cl ratio lower than 0.14, indicating sulfate reduction which induces the release of N and P from the 7 mineralized organic matter in the subsurface and the production of alkalinity during that process. Therefore, these waters and Vecht lakes. Due to insufficient surface water quality data, no results are shown for several polders in the Amsterdam 15 city area (see Fig. 4) and the Ice pushed ridge region. The first is related to the monitoring priorities of the Waternet water 16 board, the latter is related to the almost absence of surface water in this region. polders in the Zuiderzee margin and Central Holland regions ( Fig. 9 and 10). The high Ca and HCO 3 concentrations in these 20 polders are also related to the occurrence of brackish water. However, most of the surface water in the Zuiderzee margin and 21 the Central Holland area is fresh with relatively low Cl concentrations (Fig. 10). The Vecht lakes area exhibits the most fresh 22 and least mineralized surface water. 23

SO 4 and SO 4 /Cl 24
The highest SO 4 concentration levels and SO 4 /Cl mass ratios mostly occurred in the Central Holland area, especially the 25 western part. The elevated SO 4 and SO 4 /Cl ratios indicate the presence of sulfate sources other than (relict) seawater in this 26 area, probably atmospheric deposition, agriculture and/or oxidation of pyrite exposed in the upper soils which developed in 27 marine clay deposits and are denoted as "cat clays" (Wallenberg, 1975). In the Zuiderzee margin and the two upconing 28 polders, the median SO 4 levels are 64 and 62 mg L -1 , respectively, and SO 4 /Cl mass ratios of the two upconing polders are 29 below 0.14. A generally lower SO 4 with SO 4 /Cl ratios far exceeding the 0.14 were found in the Vecht lakes region. Similar to the results of groundwater, higher nutrient levels also exist in higher mineralized surface waters, which is also 10 indicated by the correlation results (Table 1,  This indicates that groundwater is the probable source of the water and nutrients in the surface water of the polders. This 13 groundwater impact was further supported by the correlations between the following pairs of solutes in surface water: Cl 14 with Ca (R 2 0.55), HCO 3 (R 2 0.52), SO 4 (R 2 0.49) and NH 4 (R 2 0.51), as well as SO 4 with TN (R 2 0.57) and NO 3 (R 2 0.50). A 15 more direct indication for the groundwater impact is that NH 4  For the soil variables (lutum, humus and calcite), only humus showed correlations with TN, NH 4 , Ca, and Cl in surface water 20 (Table 1). Paved area percentage, surface water area percentage, calcite and clay percentages did not show correlation 21 coefficients above 0.4 with surface water quality. Surface water TN correlated more closely to NH 4 (0.77) than to NO 3 22 (0.57), which reflects that NH 4 is generally the main form of TN in the study area. 23

Groundwater and surface water quality comparison 24
A common spatial pattern in surface and groundwater chemistry is that polders in the Zuiderzee margin area, the two 25 upconing polders, and the Central Holland area suffer from a worse water quality situation than the polders in the Vecht 26 lakes and Ice pushed ridge areas. However, compared with the underlying groundwater quality, surface water in the whole 27 area has much lower chloride, bicarbonate, and nutrient levels, but higher SO 4 concentrations (Fig. 5 and Fig. 9). The polders 28  Table 1 shows that TP, NH 4 , HCO 3 , and Cl concentrations in groundwater correlate with the same components in surface 3 water (R 2 0.53, 0.43, 0.66, and 0.72). In addition, HCO 3 in groundwater showed moderate correlations with nutrient 4 concentrations in surface water (TP (R 2 0.64), TN (R 2 0.50), and NH 4 (R 2 0.46)). HCO 3 concentrations in surface water also 5 correlated with nutrient concentrations in surface water (TP (R 2 0.60), TN (R 2 0.49), and NH 4 (R 2 0.59)). Surface water SO 4 6 weakly correlated to groundwater Cl (R 2 0.47). 7 Based on these correlations, we selected groundwater parameters and landscape characteristics to be integrated in multiple 8 linear regression models to predict concentrations of surface water components. For most solutes (TP, NH 4 , TN, HCO 3 , and 9 Cl, the R 2 of the regression models is around 0.5, which indicates that around 40 ~ 50% of the spatial variance in surface 10 water can be explained by specific groundwater chemistry parameters, seepage, and elevation. For NO 3 and SO 4 , the R 2 of 11 the regression models (inverse with Elevation) are very low, 0.18 and 0.21, respectively. For all other parameters, the 12 groundwater HCO 3 concentration was the best explaining variable for the surface water concentrations. The spatial variation 13 in HCO 3 SW and Ca SW were relatively well explained by only HCO 3 GW combined with Seepage, respectively (Eq. 13 and Eq. 14 15). 15 The regression models were significantly improved by including groundwater concentrations of TP, NH 4 , and Cl (Eq. 9, 11 16 and 16). In regression models Eq. 9, 10, 11, 12, 14, and 15, the elevation of the polders also explained part of the spatial 17 variation in surface water concentrations. When only including polders with net groundwater seepage, the R 2 improved 18 significantly for TP, NH 4 and HCO 3 . 19 The results above strongly suggest that the groundwater composition puts limitations to the compliance of the receiving 20 surface water towards the EQS defined for N and P. 21 The influence of the redistribution of the large water volumes and loads from deep polders was also observed in Fig. 3 and 1 Fig. 11. Polders that receive inlet water from Groot Mijdrecht and Horstermeer (see section 2. 1.1, Fig. 3)  The impact of this redistributed water on polder water chemistry is demonstrated by a simple water and solute mass balance 8 calculation for the receiving polder Botshol (see paragraph 2.2.4). Fig. 12 gives the chloride concentration results of both the 9 'evaporation' and the 'infiltration/outlet' models. Figure 12 shows that a very simple model can easily explain the peak Cl This study aimed at identifying the impact of groundwater on surface water quality in the polder catchments of the greater 17

Surface water solute redistribution
Amsterdam city area. According to the statistical analysis of data over five regions in the study area, a clear influence was 18 identified. Solute concentrations in groundwater and surface water correlated well, although groundwater solute 19 concentrations were generally much higher than normalized concentrations in surface water. The latter seems logical given 20 the dilution of surface water by the precipitation surplus on an annual basis, with the annually discharged surface water being 21 a mixture of seeping groundwater and precipitation. Moreover, similar spatial patterns in solute concentrations were found in 22 groundwater and surface water. Polders that are influenced by groundwater seepage or by redistributed seepage water from 23 nearby deep polders are at risk of non-compliance, as groundwater concentrations exceeded the TN and TP EQSs for surface 24 water in more than 90% of the polders. Consequently, the groundwater nutrients input hinders achieving water quality 25 targets in the surface water in those lowland landscapes. 26

Key hydro chemical processes 27
In general, the groundwater chemistry corresponds with the geological history of the study area. In the peat land polder depletes the infiltrating groundwater from oxygen and nitrate, leading to an overall low redox potential in groundwater, 1 which enables the further decomposition of organic matter downstream. 2 Our data strongly suggests that sulfate reduction, sometimes in combination with methanogenesis, is the main process 3 releasing nutrients (N, P) and HCO 3 from the organic rich subsurface in the study area, especially in both the fresh and 4 brackish groundwater of the Zuiderzee margin, the Upconing polders, and Central Holland that are charactreized by low 5 SO 4 /Cl ratios (Table 1, Fig. 8). The Holocene marine transgression undoubtedly influenced the chemistry of groundwater by 6 salinizing processes that also increased sulfate availability derived from diluted sea water. Refreshing of the aquifers by 7 infiltration of fresh water from rivers and rain in more elevated polders and lakes further influenced part of the groundwater. 8 We examined the amount of freshening and salinization using the exchange Na (Na ex ) and investigated how this process may 9 have influenced the release of P as was suggested by Griffioen et al. (2004). Figure S2 shows that high P (and HCO 3 , not 10 shown) does occur in both refreshing water (Na ex > 1) and in salinizing water (Na ex <-1), but mainly when the SO 4 /Cl ratio is 11 below 0.14. Therefore, we infer that sulfate reduction/organic matter decomposition is the prime process in releasing P, and 12 is more discriminating high P than cation exchange processes. There is a high probability for sulfate reduction dominated 13 polder catchments to have very high HCO 3 concentration in groundwater according to Eq.  Table 3. 19 The seepage of the alkalized groundwater increases alkalinity of the surface water, which is indicated by the high correlation 20 between groundwater and surface water HCO 3 , and with Ca in surface water (Table 1) (Table 3  22 [2], [6]), is a probable major reason for enhanced surface water HCO 3 in polders with brackish groundwater, like the 23 polders in the Zuiderzee margin and the Upconing polders (Fig. 8). 24 In the urban area of Amsterdam sand suppletion, which varies greatly in thickness and chemical composition, is another 25 source of alkalinity. Some of the sands contain shell fragments because of their marine origin. However, little is known 26 about the distribution of these calcite-rich sands. The poorly registered spatial distribution and sources of the supplied 27 calcite-rich sands might complicate the assessment of their impact on urban polder water quality. 28 Sulfate concentrations are higher in the receiving surface water than in the groundwater. We ascribe the sulfate surpluses 29 (Fig. 7) to additional sources affecting the surface water, including atmospheric deposition, agricultural inputs, sewer 30 leakage (Ellis, et al., 2005), storm runoff, and/or the oxidation of pyrite (FeS 2 ). Pyrite is ubiquitously present in this area 31 (Griffioen et al., 2013) and oxidizes in the topsoil, where either O 2 or NO 3 can act as electron acceptor (Wallenburg, 1975). 32 We suggest that sulfate concentrations are especially high in polders where shallow groundwater flow is enhanced by the 33 presence of tile drains in clay rich polders that needed this drainage system to prevent water tables rising into the root zone in 34

Groundwater contribution to surface water composition 4
The groundwater in the upper 50 m of the subsurface of the study area is an important source of nutrients in the study area's 5 surface waters (Delsman, 2015). Brackish groundwater especially seeps up into the polders of the Zuiderzee margin region 6 and into the Upconing area. The seepage of paleo-marine, brackish groundwater is driven by the low surface water levels 7 after the lake reclamation and the drainage via pumping stations. De Louw et al. (2010) reported that this groundwater 8 seepage predominantly takes place via concentrated boils through the clay and peat cover layer. 9 The excess water in the Upconing area is re-used as inlet water for several downstream polder catchments, which extends the 10 impact of the brackish, alkaline, and nutrient rich groundwater to a larger scale. The water redistribution disturbs the 11 'natural' surface water quality patterns and local groundwater impact in the receiving polders, such as polder Botshol. The 12 redistributed water largely infiltrates and returns with variable travel times via the groundwater system back towards the 13 deep upconing polders. 14

Other sources of nutrients 15
Besides the contribution from nutrient rich groundwater seepage, this study indicated that there are other possible sources of 16 nutrients in the study area. In agricultural lands, a part of the applied nutrients is typically lost towards the surface water via 17 drainage and runoff. The high groundwater NO 3 concentrations in the Ice pushed ridge are caused by the infiltration of 18 agricultural water (Schot et al., 1992b). The high nitrate loads and concentrations in surface water and groundwater of the 19 polders in the southeast (e.g. # 122 (Muyeveld), # 140 ('t Gooi)) originate from agricultural activities in surrounding polders 20 In the urban polders within the Amsterdam city that have no significant seepage (average seepage 0), TP and TN EQSs are 21 frequently exceeded because of intensive human activities such as application of fertilizer, feeding ducks and fish, and point 22 emissions like sewer overflow leakage from the sewer system (pers. Comm. Waternet). 23 In the study area, the most intensively urbanized polders are mainly infiltrating and are more affected by inlet water 24 containing high Cl and HCO 3 concentrations than by groundwater. For deep urban polders, the situation is different. In these 25 polders, the influences of typical urbanization related water quality issues are masked by the large impact of brackish, 26 nutrient rich groundwater exfiltration. Although the paved area percentage in this paper was used as a variable representing 27 urban land cover influences, it seems not be the dominant landscape characteristic that governs the spatial patterns in polder 28 surface water quality. Urban water quality is determined by multiple factors, as was also concluded by several other studies 29 The Vecht lakes polders with high surface water area percentages, representing lakes that are mainly used for recreation 1 purposes, showed relatively low solute concentrations and loads in surface water ( Fig. 10 and Fig. 11). In our study area, 2 many lakes and polders with large surface water areas show large infiltration rates due to their elevation relative to other 3 polders (Vermaat et al., 2010). Moreover, some of these lakes are replenished by inlet water that has passed a phosphate 4 purification unit. In addition, the large open water area retains nutrient transport due to long residence times and ample 5 opportunities for chemical and biological transformation processes like denitrification, adsorption, and plant uptake. 6

Uncertainties and perspectives 7
Due to the disturbance of urban constructions, combined with redistribution of water through artificial drainage corridors, 8 water flow in lowland urban areas is more complex than in rural or non-low-lying and freely draining catchments. Natural 9 patterns of water chemistry might be significantly disturbed and hydrochemical processes are masked. The understanding of 10 urban water quality patterns might improve if the monitoring program would be extended with tracers that are typical for 11 specific sources, such as sewage leakage or urban runoff. Most solutes that are currently measured can originate from 12 various anthropogenic and natural sources. 13 In the statistical analysis, for each pair of variables, only polders with complete data were taken into account, which could 14 result in a loss of information. Seepage data was simulated by a group of models of which the results may deviate from the 15 hard to measure actual seepage. We used averages of groundwater concentrations and soil properties, which caused a loss of 16 information on the spatial variation within the polders. The interpolation of groundwater quality also added uncertainty, for 17 example hidden correlations for groundwater parameters. In addition, differences in sampling methods and analytical 18 procedures between groundwater and surface water quality monitoring programs may add uncertainties. These uncertainties 19 may all have influenced the data characteristics apart from the uncertainties in the concentration measurements caused by the 20 sampling, transport, and analytical procedures. 21 In future studies, urban lowland catchments with and without seepage could be studied separately and more detailed land use 22 or paved area categories could be included. The drainage and/or leakage from sewage systems and the drainage via tube 23 drains should be taken into consideration. Drainage systems can provide a short-cut for solute transport towards surface 24 water (Rozemeijer and Broers, 2007), leading to higher solute concentrations in surface water. High groundwater levels may 25 induce groundwater discharge via the sewage or drainage systems (Ellis, 2005). In addition, studying the temporal variation 26 of surface water quality will give more insights into how the groundwater impact on surface water quality functions, as well 27 as on solutes transport and pathways in urban hydrological systems. A detailed monitoring network in several urban polder 28 catchments, which is anticipated as further work, could yield a more complete insight into water and contaminant flow routes 29 and their effects on surface water solute concentrations and loads.

Conclusion 1
In this paper, a clear groundwater impact on surface water quality was identified for the greater Amsterdam area. It was 2 concluded that this groundwater seepage significantly impacts surface water quality in the polder catchments by introducing 3 brackish, alkaline, and nutrient rich water. In general, nutrient concentrations in groundwater were much higher than in 4 surface water and often exceeded surface water Environmental Quality Standards (EQSs) (in 93 % of the polders with 5 available data for TP and in 91 % for TN) which indicates that groundwater is a large potential source of nutrients in surface 6 water. Our results strongly suggest that organic matter mineralization is a major source of nutrients in the subsurface of 7 coastal peat land areas. High correlations (R 2 up to 0.88) between solutes in groundwater and surface water confirmed the 8 effects of surface water-groundwater interaction on surface water quality. Especially in seepage polders, groundwater is a 9 major source of Cl, HCO 3 , Ca and the nutrients N and P, leading to general exceedances of EQS's for N and P in surface 10 waters. Redistribution of these high nutrient seepage waters in dry periods seems to lead to EQS exceedances in adjacent 11 boezem systems and in the receiving polders. Surface water quality in the Amsterdam urban area is also influenced by 12 groundwater seepage, but other anthropogenic sources, such as leaking and overflowing sewers might amplify the 13 eutrophication problems.