This study has simulated the terrestrial hydrology associated with different climate, landscape, and permafrost regime scenarios for the field case example of the relatively well characterized coastal catchment of Forsmark, Sweden. The regime scenarios were selected from long-term simulation results of climate, topographical, shoreline, and associated Quaternary deposit and vegetation development in this catchment with a time perspective of 100,000 years or more and were used as drivers for hydrological simulations with the three-dimensional model MIKE SHE. The hydrological simulations quantify the responses of different water flow and water storage components of terrestrial hydrology to shifts from the present cool temperate climate landscape regime in Forsmark to a possible future Arctic periglacial landscape regime with or without permafrost. The results show complexity and nonlinearity in the runoff responses to precipitation changes due to parallel changes in evapotranspiration, along with changes in surface and subsurface water storage dynamics and flow pathways through the landscape. The results further illuminate different possible perspectives of what constitutes wetter/drier landscape conditions, in contrast to the clearer concept of what constitutes a warmer/colder climate.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Hence, the inland water integrates and further propagates and distributes the effects of different causes of change. It is then unclear what are the hydrological effects of climate change and those of other, parallel changes. To investigate, distinguish and quantify the drivers of different hydrological changes, and not least those related to permafrost change, the present study utilizes available results from scenario simulations of the possible long-term climate, landscape, and permafrost development in the catchment area of Forsmark, situated on the Baltic Sea coast, approximately 120 km north of Stockholm in central Sweden [Svensk Kärnbränslehantering AB (SKB), 2010a, 2010b]. The Swedish Nuclear Fuel and Waste Management Company (SKB) conducted these scenario simulations as part of their safety assessment of a geological storage concept for high-level nuclear waste, with Forsmark being the proposed final repository site for such waste in Sweden [SKB, 2011]. The present study does not address the specific nuclear waste repository problem but uses the reported simulation results of future climate-landscape-permafrost changes in Forsmark [SKB, 2010a, 2010b] as drivers in detailed hydrological modeling with the general aim to understand, distinguish, and quantify the hydrological responses to the different change drivers.
 The hydrological simulations in this study are carried out with the distributed three-dimensional (3-D) model MIKE SHE [Graham and Butts, 2005; Refsgaard et al., 2010] and with the current conditions and simulated future scenarios of climate-landscape-permafrost states in Forsmark [SKB, 2010a, 2010b] as driving boundary conditions. The driving states differ in surface climate (surface temperature and precipitation), subsurface climate (permafrost existence and thickness), and landscape conditions (topography above sea level and associated shoreline extent, Quaternary deposits, and vegetation). The main objective of the hydrological simulations is to investigate and quantify the effects of these different driving states, and in particular to distinguish the permafrost effects, on the different water flow and water storage components of the terrestrial water cycle in permafrost environments.
2. Materials and Methods
2.1. The Forsmark Catchment and Climate-Landscape-Permafrost States
 The Forsmark catchment has a flat topography, with the present study area being almost entirely below 20 m above the present sea level (m asl). There are no major watercourses, and only small streams connect the lakes, which are generally small and shallow, in this coastal catchment. Fine-grained, low permeable sediments underlie most of the lakes. Approximately 70% of the catchment area is covered by forest, and till is the dominating type of the mostly shallow Quaternary deposits (QD). The site measurements indicate a strong correlation between surface topography and the groundwater table in the QD and an anisotropic hydraulic conductivity in the till, with the vertical hydraulic conductivity being generally smaller than the horizontal conductivity and with a decrease in conductivity with depth. Granitic rock is the dominating bedrock type in the area. The upper ∼200 m of the bedrock is, in some areas within the catchment, highly fractured compared to the deeper bedrock. Horizontal sheet joints are present in the upper bedrock, which are interconnected hydraulically over long distances [Follin, 2008]. The horizontal hydraulic conductivity of the fractures/sheet joints in the upper bedrock is very high, leading to relatively fast, essentially horizontal groundwater flow.
 As the proposed final repository site for spent nuclear fuel in Sweden, the hydrological state of the Forsmark catchment is well investigated and characterized under current climate and landscape conditions [Johansson, 2008; Johansson and Öhman, 2008; Jarsjö et al., 2008; Destouni et al., 2008a]. However, the time perspective that must be considered in the safety assessment of such a repository is 100 kyr and more into the future. Over this time the climate, as well as the landscape, and with them also the hydrology and water-ice manifestations in the catchment will change. The landscape, for instance, will change as a net result of both isostatic and eustatic changes [Påsse, 2001; Brydsten and Strömgren, 2010]. To account for this and other change aspects, SKB has conducted long-term scenario simulations of future climate and landscape change in Forsmark; detailed descriptions of these simulations are reported by SKB [2010b]. A basic reference scenario assumes then only natural climate variability and change, with no influence from human activity. This scenario is based on a simulated repetition of reconstructed climate conditions for the last glacial cycle, and yields a first periglacial climate period with permafrost occurring at Forsmark at around 10,000 AD [SKB, 2010b]. An alternative climate scenario considers also effects of anthropogenic global warming and yields the first period of periglacial conditions occurring at around 60,000 AD.
 For the present study, scenario results with systematic differences in their surface climate (temperature and precipitation), landscape and permafrost conditions were selected from SKB's different long-term change simulations for Forsmark [SKB, 2010a, 2010b]. This selection includes a range of different climate-landscape-permafrost states, from the present, cool temperate climate-landscape regime at the Forsmark site, to potential future Arctic periglacial climate-landscape regimes, with or without continuous permafrost of various possible thickness (Table 1). The selection was made with regard to this study's main objective to investigate and quantify the (separate and combined) climate-landscape-permafrost state effects on different water flow and water storage components. The selected climate-landscape-permafrost scenarios were used as boundary and internal permafrost conditions for hydrological simulations with the distributed 3D hydrological model MIKE SHE [Graham and Butts, 2005; Refsgaard et al., 2010], which includes both groundwater and surface water flow, and soil-atmosphere exchange of water (see further model description in the auxiliary material Text S1, section 1).
Table 1. Simulation Cases for Different Surface Climate, Landscape, and Permafrost Regimesa
Condition Differences Between Cases
Surface climate is given in terms of temperature, T, precipitation, P, and associated potential evapotranspiration, PET, conditions at the land surface.
Cool temperate climate with present landscape and shoreline
Between 1a and 1b: Landscape (catchment geometry, QD, land cover, shoreline)
Cool temperate climate with future landscape and shoreline
Between 1b and 2a: Surface climate (T, P, PET)
Arctic periglacial climate with future landscape and shoreline and without permafrost
Between 2a and 2b: Permafrost, existing or nonexisting
Arctic periglacial climate with future landscape-shoreline and with 100 m permafrost
Between 2b and 2c: Permafrost thickness
Arctic periglacial climate with future landscape and shoreline and with 240 m permafrost
2.2. Hydrological Simulations
Figure 1 shows the hydrological model area, which is in total 180 km2, and the landscape conditions associated with the different simulation cases in Table 1. Case 1a regards the present landscape and its shoreline, while all other cases regard the future landscape and its associated shoreline [Påsse, 2001; Brydsten and Strömgren, 2010], developed QD (erosion and sedimentation processes) and land use (e.g., lake terrestrialization) [Brydsten and Strömgren, 2010], and vegetation with related model properties influencing evapotranspiration (ET) [Löfgren, 2010; Bosson et al., 2010].
 When evaluating the results, and in the calculations of water fractionation into different water balance components only the part of the model area that constitutes land (green for the present landscape, green and yellow for the future landscape in Figure 1) is considered in each case. In case 1a only the mainland part (with land area of 34 km2) of the Forsmark catchment is considered, i.e., the island (green color) to the right in Figure 1 is not a part of the result evaluation. Currently, the dominating part of the model area is covered by the Baltic Sea (case 1a, with sea and island area of 146 km2), while in the future landscape cases (cases 1b–2c), large parts of the model area are located above the present sea level (with land area of 178 km2, and sea area of 2 km2). The shoreline, i.e., the part of the model land area that is exposed to the sea, is then 66 km in case 1a (only the mainland considered) whereas it is only 8 km in the other cases.
 To simulate the present hydrology at the site (case 1a), the current cool temperate climate regime (with surface temperature and precipitation as shown by red lines in Figure 2) was used as driver, and all relevant available data from the SKB site investigations were included in the simulation and calibration, and the testing of the calibrated MIKE SHE model. The testing was made against independent data, not used in the calibration process, as described in more detail by Bosson et al. [2008, 2010] (see also detailed description in Text S1, section 2). Generally, simulated groundwater and surface water levels, as well as surface water discharges agreed well with local measurements at the site.
 The model of the present hydrological state (case 1a) was further used as initial condition for the simulation of the future regime scenarios (cases 1b–2c). In case 1b, the present cool temperate climate was used as driver, while a possible future Arctic periglacial surface climate was used in cases 2a–2c; Figure 2 shows the characteristics of both surface climate regimes. The surface climate data used in the MIKE SHE simulations are time varying T, P and potential evapotranspiration (PET). Actual evapotranspiration, ET, and its different components (snow sublimation, interception, evaporation from soil and ponded water, transpiration) are calculated in the simulations.
 For the hydrological simulations of the cool temperate surface climate we used measured meteorological data from a selected year of the SKB site investigation period, considered to represent the current long-term average situation at the site [Bosson et al., 2010]. The mean annual air temperature (MAAT) for the selected year of the current cool temperate climate regime is +6.4°C and the annual P is 583 mm. The corresponding meteorological time series used to represent the possible future Arctic periglacial surface climate was that simulated for such a climate regime at Forsmark by Kjellström et al. , with resulting MAAT of −7°C and annual P of 412 mm. In all simulation cases, the representative driving climatic year for each case was repeated until a long-term annual hydrological steady state was reached. The simulation results for all cases 1a–2c are transient over the year, and describe the seasonal and smaller temporal scale dynamics throughout the representative resulting hydrological year.
 The offshore sea-covered areas in each simulation case were accounted for by a prescribed time-varying hydraulic head, representative for the locally measured sea level fluctuations of today. The bottom boundary of the model domain, placed at 600 m depth below the present sea level, constituted a no flow boundary. The groundwater divides were assumed to coincide with the surface water divides, thus a no flow boundary was applied along the land parts of the model boundary (catchment water divide), and a time varying head boundary, representing the sea level fluctuation, along the shoreline of each simulation case (Figure 1). In cases 1b–2c, representing a possible future landscape, the shoreline is situated at −31.42 m asl relative to today's shoreline.
2.3. Permafrost Representation in the Hydrological Simulations
 Even though MIKE SHE does not support thermal modeling, simulation scenarios were also set up to investigate the hydrological effects of the possible existence of continuous permafrost with different thickness in cases 2b and 2c. The approach to account for permafrost effects in MIKE SHE simulations has been described in more detail by Bosson et al.  and is shortly summarized here. Several model parameters important for the description of freeze and thaw processes, as well as permafrost conditions, have been identified. Hydraulic properties are then chosen to mimic a low-permeable permafrost layer, with the hydraulic conductivities in both the saturated and unsaturated zone reduced so that water cannot infiltrate or flow through the permafrost. The Manning number at the surface is further changed so that water on the land surface becomes immobile when the air temperature is below 0°C.
 In the present simulations, an active layer of 1 m depth was assumed to overlie the permafrost, with its freezing and thawing depending on the weather conditions during the year. The hydrological year was divided into seven parts, starting with two periods of freezing, when the hydraulic conductivity and other permafrost related parameters were of the stepwise changed toward fully frozen conditions [Bosson et al., 2010]. These parameters values were applied throughout the frozen period, followed by three thawing periods when the hydraulic conductivities were increased in a stepwise way toward representing unfrozen conditions, and finally, an entirely unfrozen period was simulated at the end of the hydrological year.
 In case 2b, the thickness of the permafrost layer was set to 100 m and in case 2c to 240 m. According to Forsmark-specific permafrost simulations for the SKB safety assessment [SKB, 2010b], with the same driving periglacial climate regime as in the present cases 2a–2c [Kjellström et al., 2009], the permafrost thickness for an average ground surface temperature of −4°C would be on the order of 240 m at the Forsmark site, as assumed here in case 2c. The 100 m thick permafrost case 2b is considered as an additional comparative case, in order to investigate the hydrological effects of permafrost thickness.
 Permafrost thickness influences in particular the potential for through taliks, unfrozen areas in the permafrost [French, 2007], to form in the landscape. The thicker the permafrost the fewer through taliks can be maintained, as reported for instance from Forsmark-specific, detailed permafrost simulations [SKB, 2006; Hartikainen et al., 2010]. Site-specific simulations performed in particular to study the relation between lake geometry, permafrost thickness and talik formation [SKB, 2006] considered two types of circular lakes with flat bathymetry as representative for (1) shallow lakes with bottom temperature above 0°C and (2) deep lakes with bottom temperature above 4°C. The depth of the simulated representative shallow lake was then set to 2 m and the depth of the representative deep lake was set to 8 m. In the present study, the Forsmark-specific permafrost simulation results from SKB  were used so that lakes with mean depth of 0.5–4 m were considered as shallow (bottom temperature above 0°C for the Arctic periglacial climate with MAAT of −7°C), and lakes with mean depth >4 m were considered as deep (bottom temperature above 4°C under the periglacial climate). SKB  showed that a through talik can develop beneath the lakes if (1) the radius of a shallow lake (with its surface area interpreted as a circle) exceeded the thickness of the surrounding permafrost and (2) the radius of a deep lake (interpreted as for a shallow lake) was ≥0.6 times the thickness of the surrounding permafrost.
 If a lake in the case 1b landscape fulfilled condition 1 or 2 relative to the different permafrost thickness in cases 2b and 2c, an unfrozen column was simulated in the QD and bedrock under the lake to represent a talik. These talik formation conditions in the MIKE SHE modeling of Forsmark have also been used and described previously by Bosson et al. [2010, Appendix 1]. The area under the sea bay was assumed unfrozen; that is, the hydraulic properties for sea bottom sediments and underlying bedrock were kept the same in cases 2b and 2c as in case 2a.
Table 2 summarizes some mean annual hydrological results for all simulation cases (as outlined in Table 1), while Figure 1 shows the resulting number of through taliks in the permafrost cases 2b and 2c. In the following, we discuss the results further in relation to different water storage and water flow components, and their dynamics over an average year at the different climate, landscape and permafrost states represented by the different simulation cases.
Table 2. Driving Climate and Resulting Hydrological Variables for the Simulation Cases in Table 1a
The listed variables are temperature, T; potential evapotranspiration, PET; precipitation, P; evapotranspiration, ET; runoff, R; and groundwater recharge Rgw. The percent values for wetlands and lakes refer to their surface area in relation to the total land area.
Water in storage is at the end of each simulation, when the storage change is zero. Wetlands are areas with surface water depth (D) 0.01 m ≤ D < 0.3 m, and lakes are areas with surface water depth (D) D ≥ 0.3 m.
3.1. Water in the Landscape
Table 2 shows that the landscape shift from case 1a to case 1b decreases the relative lake and wetland area in the landscape from 15% to 13.1%. The added shifts in climate and permafrost, from case 1b to cases 2a, 2b, and 2c, then increase the surface water area up to a range of 16–18.5%, with the higher values applying to the permafrost cases 2b and 2c (Table 2).
 Regarding the climate shift from case 1b to case 2a, the mean surface temperature (T) clearly determines case 2a as representing a colder climate than case 1b. However, with regard to water, case 2a could be viewed as both wetter and drier than case 1b. There are multiple hydrological parameters that determine different wetness and dryness aspects, with complex relations between them. Case 2a has greater surface water storage (in lakes and wetlands), as well as greater mean annual runoff (R) than case 1b, mainly due to the smaller mean annual evapotranspiration (ET), both in absolute terms and in relation to P, in case 2a than in case 1b (Table 2). The greater P and ET fluxes, which link the landscape and atmospheric water, may be a basis for considering case 1b as wetter than case 2a from an atmospheric perspective. The greater water storage (lakes and wetlands) and R in 2a, however, mean that this case can be viewed as wetter than 1b from a landscape perspective.
 The surface water storage in lakes and wetlands, and R are even greater in the permafrost cases 2b and 2c than in the nonpermafrost case 2a. This is both because ET is smaller and because less water can infiltrate due to the permafrost, with more water then remaining at the ground surface, in cases 2b and 2c than in case 2a. The two permafrost cases 2b and 2c, with essentially the same P, ET, R and mean annual groundwater recharge (Rgw), but different permafrost thickness, have similar surface water storage in lakes and wetlands (17.9–18.5%). The main permafrost thickness effect is to regulate the number of through taliks, and through them the amount and direction of water exchange between the surface and the deeper groundwater under the permafrost. The total net flow contribution through the inland through taliks is then a relatively small (about 1 mm in 2b and 0.1 mm in 2c) net recharge flux into the deep groundwater below the permafrost. This contribution is greater in 2b because it has more inland through taliks than 2c (Figure 1). The inland surface water bodies that form through taliks in case 2b are also present at the land surface in case 2c, but they do not in 2c, with its thicker permafrost, fulfill the conditions for through talik formation. Also the sea bay constitutes a through talik in the permafrost simulation cases 2b and 2c, yielding a total net discharge flux contribution (of about 1 mm in 2b and 0.1 mm in 2c) from the deep groundwater to the sea.
3.2. Water Storage Dynamics
 Regarding the temporal pattern of both surface and subsurface water storage dynamics, Figure 3 (top) shows that it is essentially the same for the different landscapes of the two warmer climate cases, 1a and 1b. Groundwater storage increases during the autumn rains, with increasing groundwater table and saturated zone (SZ) extent, and correspondingly decreasing unsaturated zone (UZ) extent, as consequences. Also the surface water storage increases somewhat during this period. In spring, after the snowmelt, the groundwater storage and level decrease again and the UZ (SZ) extent increases (decreases), primarily due to increasing ET as the vegetation grows.
 The main water storage effect of the landscape shift from case 1a to case 1b is that the autumn increase in groundwater (SZ) storage is smaller in 1b, and as a consequence the groundwater table is lower during the whole period October–June in case 1b than in case 1a. The change in groundwater storage over the year is then also smaller in case 1b, even though Rgw is somewhat greater in this case than in case 1a (Table 2).
 Comparison between Figure 3 (top) and Figure 3 (bottom) shows that the shift in surface climate from case 1b to case 2a fundamentally changes the seasonal pattern of water storage dynamics. The low temperatures during autumn and winter in case 2a lead to continuous snow accumulation from October to the end of April. The groundwater level and storage, and the associated SZ and UZ extents are almost constant over this period, and exhibit only relatively small changes in response to the snowmelt in spring. The surface water storage exhibits greater increase in response to the snowmelt.
 The presence of permafrost in cases 2b and 2c has some, relatively small influence on the water storage dynamics, regarding details in the UZ and SZ relation after snowmelt compared to the corresponding nonpermafrost case 2a. Since the active layer is only 1 m thick in the present permafrost case simulations, and no water can infiltrate below this level, the available depth extent for UZ and corresponding SZ/groundwater storage changes are quite limited in cases 2b and 2c. The active layer becomes fully saturated much faster, almost directly after the snowmelt starts, and the surface water storage exhibits a larger snowmelt response in cases 2b and 2c than in case 2a. Less water can infiltrate and more water is therefore present at the ground surface, which contributes to the greater lake and wetland area in cases 2b and 2c than in case 2a (Table 2). However, the permafrost thickness (case 2b or 2c) has no water storage influence.
3.3. Water Flux Partitioning at the Land Surface
Figure 4 shows that the transience of the main water fluxes ET, R and Rgw, and their partitioning at the land surface over the year are strongly influenced by the climate shift from cases 1a and 1b to cases 2a–2c. The main difference between the cases is due to the larger snow accumulation over a much longer period of time in the periglacial climate of cases 2a–2c than in the cool temperate climate of cases 1a and 1b. In cases 2a–2c, it is mainly only after the snow has melted in May that water becomes available for flow by ET, R and Rgw. The rate of snowmelt exceeds then largely the soil infiltration capacity and water flows mostly out from the catchment in more or less equal amounts by ET and R in the colder climate cases 2a–2c, whereas the outflow by ET is about twice that by R in the warmer climate cases 1a and 1b.
 Also the mean (and total cumulative) annual ET – both in absolute terms and relative to P –shifts mainly due to the shift in climate from case 1b to case 2a (Table 2 and Figure 4). In contrast, the absolute mean (and total cumulative) R shifts more or less similarly in all case shifts, but the relative runoff coefficient R/P shifts mostly, as does ET/P, in the climate shift from case 1b to case 2a. The mean (and cumulative) annual Rgw, however, shifts most considerably (decreasing) in the permafrost shift from case 2a to cases 2b and 2c, even though it is to somewhat smaller degree also affected by the climate shift from 1b to 2a (decreasing Rgw), and the landscape shift from 1a to 1b case (increasing Rgw). However, the shift in permafrost thickness between 2b and 2c does not affect Rgw much.
 Regarding the 192 mm smaller ET output flux in case 2a than in case 1b, most of the difference is due to the 171 mm smaller P input flux in 2a than in 1b, while the remaining 21 mm difference is due to the much longer period with snow, which allows less time for relatively high ET to occur in 2a than in 1b. The 21 mm smaller net output flux, ET-P, implies that so much more water is then available for R, which is also about that much greater (18 mm) in case 2a than in case 1b.
3.4. Pathways of Water Flow Through the Landscape to the Sea
Figure 5 shows that the change in landscape from case 1a to case 1b notably affects the distribution of pathways followed by different parts of the total runoff, R, to the sea. In case 1a, with its long shoreline and associated large exposure to the sea, about 30 mm of water (17% of total R) leaves the model area and flows into the sea as submarine groundwater discharge, SGD, occurring mostly in the uppermost QD layers. This quantification regards total SGD, including both its main, fresh groundwater component and a small component of recirculated seawater, which flows mostly into the coastal aquifer when the inland groundwater table is low (and the associated fresh groundwater pressure and flow are small) and flows mostly out to the sea again when the groundwater table is relatively high (and associated fresh groundwater pressure and flow are relatively large). The present calculation of total SGD being 17% of the total mean annual runoff, R, in the long shoreline case 1a is consistent with a previous, independent estimate by Jarsjö et al.  of the diffuse freshwater flow (i.e., mostly groundwater, rather than focused stream flow) being at most 20% of the total freshwater flow from the present-day Forsmark catchment to the Baltic Sea. Furthermore, it is also consistent with an entirely independent estimate by Destouni et al. [2008b] of total SGD being around 15% (with freshwater and seawater components of 13.5% and 1.5%, respectively) of the total mean annual R from another type of coastal catchment, including larger river discharges in addition to SGD and the smaller types of streams that exist in the present-day Forsmark catchment.
 In case 1b, with its much shorter shoreline, and thereby more channelized flow through fewer but larger stream outlets to the sea than in case 1a (see also landscape Figure 1), SGD is only 4 mm, corresponding to 2% of the total R. This situation is consistent with the total distribution of freshwater discharges from Sweden to the Baltic Sea. These include about 80% flow to the sea through the monitored outlets of relatively large rivers and streams, and 20% unmonitored discharges [Hannerz and Destouni, 2006]. According to simulation results of Jarsjö et al.  the latter are then further distributed so that at least 16% (80% of total 20% unmonitored flow) flows to the sea as surface water through relatively small unmonitored streams and at most 4% (20% of total 20% unmonitored flow) as diffuse unmonitored SGD. According to the present simulation results SGD is then essentially unaffected by the climate shift to case 2a, and is even smaller, nearly zero, in the permafrost cases 2b and 2c.
 In case 1b, the smaller SGD goes together with a correspondingly greater groundwater contribution to the streamflow of 131 mm (70% of the total stream flow, which is nearly the total R to the sea) compared to 96 mm (66% of the total streamflow (R-SGD) and 55% of the total R to the sea) in case 1a. The groundwater contribution to streams is even lower in the periglacial climate of cases 2a–2c, and the occurrence of permafrost in cases 2b and 2c further decreases this contribution. The groundwater contribution to streamflow is then 109 mm (53% of total streamflow and R) in case 2a and 57 mm (26% of total streamflow and R) in the permafrost cases 2b and 2c. The difference between the groundwater contribution and the total streamflow is in all cases due to overland flow directly to the streams.
 The pattern of temporal variation over the year is similar for the total streamflow and the groundwater contribution to it (Figure 5). Furthermore, it is consistent with the temporal variation in total R and similarly dependent on the shift in climate from cases 1a and 1b to cases 2a–2c. Between the colder climate cases, there is a delay in the groundwater contribution to the streams in the permafrost cases 2b and 2c compared to the nonpermafrost case 2a. In case 2a with no permafrost present, the groundwater contribution to streams starts as soon as the snow starts to melt and the groundwater zone starts to be recharged. The delay in the permafrost cases 2b and 2c is because the active layer has to thaw, which starts along with the snowmelt in April with thawing being reached first in the beginning of June, in order for groundwater to be recharged and flow into the streams. The thickness of the permafrost does not affect the timing or the total groundwater contribution to stream discharge in the present simulations.
3.5. Summary of Main Hydrological Influences
Figure 6 summarizes the main influences of the investigated shift parameters: surface climate, landscape, and permafrost existence and thickness, on different hydrological components. The latter are specified as: external flows (Figure 6, right) into and out from the catchment system (represented by the solid line box); and catchment-internal components of groundwater recharge (left in the catchment box) and different forms of water storage (right in the catchment box). The red arrows in Figure 6 illustrate the main directions of the different external flows, the internal groundwater recharge flow, and the other internal flows between different storage components. The shift parameters written under each hydrological component heading are those shown by the simulations to considerably influence that component, with the shift parameter written at the top/bottom being the most/least influential.
 Surface climate is clearly an important shift parameter, affecting most but not all of the investigated hydrological components. Surface climate alone affects the snow storage dynamics, in addition to the mean annual ET and P; the relatively small evaporation from snow is illustrated with a dashed red arrow in Figure 6. Landscape alone, and in particular its shoreline length, affects mostly the SGD pathway, and through that the general distribution of freshwater pathways to the sea.
 The SGD, which can be relatively large in catchments with long shorelines, bypasses entirely the surface water pathway to the sea and SGD quantification is difficult [Prieto and Destouni, 2011]. The present results clarify the water balance links between the different flow pathways to the sea, which can be used for bounding SGD quantifications. In general, consideration of water balances on problem-relevant catchment scales can bound and provide important reality checks for both historic reconstructions and future projections of hydrological flows and their changes in a changing climate [Shibuo et al., 2007; Destouni et al., 2010; Asokan et al., 2010; Jarsjö et al., 2011]. The present results emphasize that such checks must also account for the subsurface components and surface-subsurface links of terrestrial water change.
 Surface (lake-wetland) and subsurface (UZ-SZ) water storage, groundwater recharge (Rgw), and total runoff (R) with its different pathway components all collect, integrate, and through their changes propagate further through the landscape the effects of different shift parameters. Among these, not only the surface but also the subsurface climate, expressed here in terms of permafrost existence (as permafrost thickness was found to have only small effect on the investigated hydrological components), is an essential control parameter for terrestrial hydrology.
 This study has distinguished and quantified separate and combined influences of surface climate, landscape and permafrost conditions on different water flow and water storage components, through a scenario simulation and analysis approach applied to the Swedish Forsmark catchment area. The latter represents a relatively well characterized field case example. The results show complexity in the connections between different hydrological flow and storage components, and as a consequence in the responses of these components to shifts in climate, landscape and permafrost regimes. For instance, both R and water storage may increase in a catchment where P decreases if ET decreases even more than P.
 Nonintuitive R responses to P change have been found in different studies of hydroclimatic change, for catchments of different scales and in different parts of the world. Reported earlier findings include R change in the opposite direction than P due to concurrent natural or anthropogenic ET changes [Shibuo et al., 2007], in the same direction but considerably more than P due to concurrent water storage changes, e.g., in the terrestrial cryosphere [Bring and Destouni, 2011], or entirely unrelated to P change due to various parallel land and water use changes on different unresolved local regional scales [Koutsouris et al., 2010]. Hence, linear assumptions of a projected P increase (decrease) leading directly to correspondingly wetter (drier) landscape conditions may in many cases be too simplistic.
 Furthermore, the present results show that the definition of what actually constitutes a wetter or drier climate in a landscape may be problematic. Specifically, for smaller P and ET, which may be viewed as drier conditions from an atmospheric perspective, the present simulations yielded greater R and surface water extent, which may be viewed as wetter conditions from a landscape perspective. Projections and assessments of climate-driven water changes must consider such different perspectives and their implications for what water changes society should expect and try to mitigate or adapt to.
 Landscape changes can further affect the shoreline-dependent SGD, which links to another water perspective problem: highly different SGD quantifications are reported from inland-based (hydrological) and marine-based SGD estimation methodologies [Prieto and Destouni, 2011]. The present results are consistent with other hydrological SGD quantifications, and further emphasize the need to seriously consider the SGD quantification gaps and allocate relevant research efforts toward bridging them. In general, the results of this study have illuminated the need to account for and link subsurface hydrology, and not least its permafrost component in cold regions, to changes at the surface in order to accurately understand and assess hydrological change propagation through the whole inland water system.
 G.D. acknowledges financial support from the Swedish Research Council (VR; project 311-2007-8393, contract 70839301).