LETTERSPUBLISHED ONLINE: 16 FEBRUARY 2015 | DOI: 10.1038/NGEO2350

Tide-mediated warming of Arctic halocline byAtlantic heat fluxes over rough topographyTom P. Rippeth1*, Ben J. Lincoln1, Yueng-Djern Lenn1, J. A. Mattias Green1, Arild Sundfjord2

and Sheldon Bacon3

The largest oceanic heat input to the Arctic Ocean resultsfrom inflowing Atlantic water. This inflowing water is warmerthan it has been in the past 2,000 years1,2. Yet the fateof this heat remains uncertain3, partly because the water isrelatively saline, and thus dense: it therefore enters the ArcticOcean at intermediate depths and is separated from surfacewaters by stratification. Vertical mixing is generally weakwithin the Arctic Ocean basins, with very modest heat fluxes(0.05–0.3Wm−2) arising largely from double di�usion4–8.However, geographically limited observations have indicatedsubstantially enhanced turbulent mixing rates over roughtopography9–14. Here we present pan-Arctic microstructuremeasurements of turbulent kinetic energy dissipation. Ourmeasurements further demonstrate that the enhanced con-tinental slope dissipation rate, and by implication verticalmixing, varies significantly with both topographic steepnessand longitude. Furthermore, our observations show thatdissipation is insensitive to sea-ice conditions. We identifytides as the main energy source that supports the enhanceddissipation, which generates vertical heat fluxes of morethan 50Wm−2. We suggest that the increased transfer ofmomentum from the atmosphere to the ocean as Arctic sea icedeclines is likely to lead to an expansion of mixing hotspots inthe future Arctic Ocean.

Atlantic water (AW) enters the Arctic Ocean at depths between40 and 200m, and its core is∼4 ◦C warmer than the overlying cold,fresher halocline and surface mixed layer water, as illustrated bythe temperature profiles in Fig. 1. As the boundary currents of theArctic Ocean circulate cyclonically around the basin, the interfacebetween the AW and the overlying waters is progressively erodedby mixing1,15. The density difference across this interface, the AWthermocline, is the major barrier to upward heat flux, such that inthe absence of vertical turbulent mixing, only a small fraction of theAWheat can reach the sea surface or sea ice directly1. This paper willfocus on vertical mixing across the AW thermocline. Measurementsof profiles of the rate of dissipation of turbulent kinetic energy (ε)are presented, fromwhich cross-gradientmixing rates are estimated,for locations covering much of the seasonally ice-free Arctic Ocean:see map in Fig. 1. The measurements include transects taken acrossthe shelf break north of Svalbard through dense ice cover in autumn2008, and spring and early summer of 2010 and 2011, and moreopen-water conditions in the summers of 2012 and 2013 (transect 1in Fig. 1). Further transects were made across the continental slopenorth of Severnaya Zemlya (transect 2), the Laptev Sea (transects3 and 4) and the East Siberian Sea (transect 5) during the ice-freeSeptember 2007, and through 100% first year ice cover in October2008. Cross-slope transects weremade in the largely ice-free Canada

Basin (transects 6 and 7) in the summer of 2012. Themeasurementsspan the upper 500m of the water column and fully resolve theAW thermocline.

The average dissipation rate across the AW thermocline variessignificantly with bathymetry (Fig. 2a). Within the central ArcticOcean (over bathymetry >2,000m) average values of between∼5×10−10 and 2×10−9 Wkg−1 are observed: close to (but above)the instrument noise level, and in agreement with previouslypublished estimates4,6,8,9. It is interesting to note that these verylow levels of ε seem insensitive to sea-ice cover. The observedvalues are substantially lower than background values reportedfor intermediate depths in the central Atlantic and Pacific Oceanbasins (∼10−8 Wkg−1; refs 16,17), and are often not sufficientlyenergetic to drive significant turbulent mixing4,6. A consequence ofthe very low levels of turbulence is the formation of thermohalinestaircases that link the AW to the colder, fresher overlying waters(Fig. 1—grey profile). The staircases arise from the competingimpacts of the temperature and salinity gradients on density, andthe greatly differing molecular diffusion rates for heat and salt.These unique phenomena support vertical exchange of heat throughdouble diffusion but with relatively weak resulting across-gradientheat fluxes (0.05–0.3Wm−2; refs 4–7). The double-diffusive fluxestogether with weak dissipation associated with the internal wavefield8 are insufficient to explain the observed cooling of the AW asit passes through the Arctic Ocean1,6,18.

In contrast, the new observations in Fig. 2a show that profile-averaged ε within the AW thermocline is enhanced by up to twoorders of magnitude over the continental slope regions (that is,depths between 200mand 2,000m), when comparedwith the valuesfor the central Arctic Ocean. The largest values are found overthe continental slope to the north of Svalbard with no apparentdependence on sea-ice conditions. Here, the observed dissipationrates of 3–20 × 10−8 Wkg−1 are sufficiently energetic to drivesignificant turbulent mixing and thus prevent the formation ofthermohaline staircases (Fig. 1, red and black profiles). Similarlyenhanced values have been reported for the Yermak Plateau regionfurther to the west9,11, as well as the region north of Svalbard10.Figure 2b shows that the enhancement of the AW thermocline-averaged ε seems to be related to the local topographic slope, withthe largest values of ε observed above the steepest topography.Enhanced ε values (∼10−8 Wkg−1) are also observed further tothe east, over the continental slope poleward of the SevernayaZemlya islands. Here the slope-enhanced values of AW thermoclineε are comparable to those observed at intermediate depths atlower latitudes over rough topography, such as the mid-oceanridge systems16,17 and the continental shelf breaks19. There is alsoa modest enhancement over the Canada Basin continental shelf

1School of Ocean Sciences, Bangor University, Bangor LL59 5AB, UK. 2Norwegian Polar Institute, Fram Centre, 9296 Tromsø, Norway.3National Oceanography Centre, Southampton SO14 3ZH, UK. *e-mail: t.p.rippeth@bangor.ac.uk

NATURE GEOSCIENCE | VOL 8 | MARCH 2015 | www.nature.com/naturegeoscience 191

© 2015 Macmillan Publishers Limited. All rights reserved

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO2350

60o

W

30 oW

1

2

34

5

76

Canada

Basin

150° W

120°

W 120° E

150° E

180°

0°

Dep

th (m

)

90° W

90° E

60° E

30° E

75° N

80° N

85° N

70° N

FramStrait

KaraSea

LaptevSea

EastSiberianSea

Svalbard

SevernayaZemlya

−2 0T (°C)

2

0

50

100

150

200

Figure 1 | Map of the Arctic Ocean showing the bathymetry and the location of the microstructure profiler measurements. Marker shape indicates thesea-ice conditions during those measurements (circle—open water/low ice cover, triangle—significant (>70%) ice cover). The colours refer to thegeographical location of the measurements (consistent with Figs 2 and 3). The numbers refer to transects shown in Fig. 3. Typical temperature profiles areshown for the continental slope region (black in open water and red under sea ice) and for the central Arctic Ocean (grey) to the north of Svalbard.

break (average ε∼3×10−9 Wkg−1 compared with central CanadaBasin average values of <10−9 Wkg−1). Previous observations ofwater-column structure in this region have revealed the absence ofthermohaline staircases over this slope, an indicator of significantturbulent mixing at intermediate depths5. The new observationsreveal that, although the AW thermocline-averaged ε seemsinsensitive to sea-ice cover and bathymetry, there is significantvariation with both local topographic slope and location around theArctic Ocean margins.

Two sources of kinetic energy are usually implicated in drivingturbulent mixing in the ocean: wind and tide20. Recent studieshave suggested increased momentum transfer from the wind to theocean associated with declining seasonal sea-ice cover potentiallyleading to increased mixing21–23. The results presented here provideno evidence that the AW thermocline-averaged ε is sensitive to sea-ice conditions in locations where observations weremade in varyingice conditions, implying that the wind is of lesser importancein supplying energy to mixing at intermediate depths. However,the observed longitudinal variation in transect-mean dissipation(εAW—obtained by averaging the profile-integrated ε, for the AWthermocline, for all profiles taken over the continental slope ineach transect) does correlate with the tidal energy dissipation rate,Dtide (Fig. 3). This is computed as the difference between therate of work by the tide-generating force and the divergence of

the energy flux24 using tidal elevations and velocities from theTPXO8 inverse solution25,26. On the basis of this correlation, theobserved transect-mean εAW accounts for 12% of the total tidalenergy dissipation rate (r=0.67). The conclusion is that the energysupporting much of the enhanced dissipation observed along thecontinental slopes, poleward of the Svalbard and Severnaya Zemlyaarchipelagos, is of tidal origin. The most significant deviationbetween εAW and Dtide is found for transect 6, in the ice-freeCanada Basin. These measurements were made in the immediateaftermath of the unprecedented ‘GreatArcticCyclone’ of 2012wherewind forcing may be an additional contributory factor. Previousgeographically limited studies have suggested enhancedmixing nearrough topography9–14, but these new observations provide the firstcircumpolar evidence for the control of AW mixing rates by theinteraction between the tide and rough topography.

At lower latitudes the cascade of energy from tides to turbulenceis facilitated by the generation of a freely propagating linear internaltide that results from stratified tidal flow over rough topography.However, the newmeasurements were taken poleward of the criticallatitude (74.5◦ N) beyond which the rotation of the Earth prohibitsfreely propagating waves at the dominant semi-diurnal (M2) tidalfrequency. Consequently tidally generated internal waves at theselatitudes are thought to have properties inherent to lee waves27.These waves have short temporal and spatial scales, related to the

192 NATURE GEOSCIENCE | VOL 8 | MARCH 2015 | www.nature.com/naturegeoscience

© 2015 Macmillan Publishers Limited. All rights reserved

NATURE GEOSCIENCE DOI: 10.1038/NGEO2350 LETTERS

Open waterIce-covered

Bathymetric slope10−2 10−1

Water depth (m)0 1,000 2,000 3,000 4, 000

10−9

10−8

10−7

a

b

10−9

10−8

10−7

AW

(W k

g−1)

εA

W (W

kg−1

)ε

Open waterIce-covered

Figure 2 | Variation of profile-average dissipation with bathymetry andtopographic slope. a, Profile-average dissipation rate, ε, in the AWthermocline, plotted against bathymetry. Variability is indicated by the 95%confidence interval. For clarity the confidence interval is not shown when itis smaller than the plotted symbol used. Following Fig. 1 the colour indicatesthe geographical location of the measurements and the shape shows thesea-ice conditions. b, Profile-averaged dissipation in the AW thermoclineplotted against the slope of the topography below.

local topography and stratification, and tend to be dissipated rapidlyleading to local turbulentmixing28. It is likely that large variability inεAW north of Svalbard (as indicated by the 95% confidence intervalin Fig. 3) is a consequence of the temporal and spatial variability ofthe tidal processes generating the turbulence19,29.

The dissipation rates are combined with the observed water-column stratification in the calculation of a diffusion coefficientthat is then used to estimate the magnitude of the heat fluxesacross the AW thermocline resulting from turbulent mixing. Usinginformation from 84 profiles, collected during 5 observationalcampaigns in the region between 2008 and 2013, an average heatflux across the AW thermocline of 22±2Wm−2 is estimated for thecontinental slope poleward of the Svalbard and Severnaya Zemlyaarchipelagos. Heat fluxes of more than 50Wm−2 are calculatedfrom individual profiles. These heat fluxes are over 2 orders ofmagnitude greater than those reported for the central Arctic Ocean,although they operate over amore limited area. The localized natureof the enhanced AW heat fluxes will therefore lead ultimately tospatial inhomogeneity in the Arctic Ocean–sea-ice system responseto climate change with impacts on sea-ice cover and heat transferwith the atmosphere enhanced in the vicinity of rough topography.

The new observations confirm the paradigm of a predominatelydouble-diffusive central Arctic with weak mixing, and contrast it

Longitude

1

2

34

5 67

Dtide (W

m−2)

10−2 10−1

10−2

10−3

10−3

10−4

Open waterIce-covered

0 90° E 180 90° W 0

AW

(W m

−2)

∋

Figure 3 | Transect-mean integrated AW dissipation, εAW, againstlongitude. The colours and the transect numbers refer back to Fig. 1, andthe marker shape indicates the sea-ice conditions (as per Fig. 1). Thevariability is indicated by the 95% confidence interval. The blue line is therate of tidal energy dissipation, Dtide, computed as the di�erence betweenthe work done by the tide-generating force and the divergence of the tidalenergy flux.

to the more turbulent continental slope regions where the tidesinteracting with steep topography act to control the rate of turbulentmixing. We note that the observed turbulent mixing is a resultof relatively modest tidal currents. Model studies suggest thatretreating seasonal sea-ice coverage will result in an increasedtransfer of momentum from the atmosphere to the ocean22,23 andrecent observations reveal the spin-up of the wind-driven Beaufortgyre circulation21 and enhanced currents due to near-inertialwaves30 in response to declining sea-ice cover. The accelerationof the currents in response to sea-ice retreat has the potential togrow the geographic extent of turbulent mixing to other regions ofrough topography where at present flows are too weak to induce leewave formation. Hence, the coupling of these large and small-scaleprocesses will probably drive the expansion of mixing hotspots, andso increase local AW heat fluxes, which will in turn feedback on thealready declining Arctic sea ice and increase momentum transferbetween the atmosphere and the ocean.

MethodsTurbulent dissipation rate measurements. The profiles of the rate of dissipationof turbulent kinetic energy (ε) are made using a loosely tethered free-fall velocitymicrostructure profiler (Rockland VMP500 model) deployed by the BangorUniversity team and a loosely tethered ISW MSS 90L microstructure profiler bythe Norwegian Polar Institute (NPI; Tromsø) team. The instrument is deployedfrom a ship and takes profiles of velocity microstructure together withtemperature and salinity down to near the sea bed, or up to 500m depth indeeper water. The rate of dissipation of turbulent kinetic energy is then calculatedfor depth bins (1z) of approximately 1m size using assumptions of stationarityand homogeneity29. At each station reported, 1 or 2 profiles were made over aperiod of approximately 1 h with longer time series (up to 6 h) collected by theNPI in the vicinity of the shelf break at 30◦ E.

The profile-averaged ε values shown are the station averages for the AWthermocline. The vertical depth limits for the AW thermocline are defined, foreach cast, as the region from the temperature minimum in the cold overlyingwater down to the AW temperature maximum (for example, see bold section ofthe temperature profile on Fig. 1). The number of profiles used ranged from 1 to12. Full details of the number of profiles collected for each location, during eachvisit, are given in Supplementary Table 1.

The transect-mean AW thermocline dissipation, εAW, shown in Fig. 3 isobtained by averaging the profile-integrated dissipation, for the AW thermocline,for all profiles (n) taken over a continental slope transect. That is,

εAW=1n

n∑1

∑AWt

ε1z

The number of profiles (n) used for each point ranges from 2 to 23, withaverages representing temporal means over several days, for each transect. Ameasure of variability in both station-average ε (Fig. 2) and transect-mean εAW

NATURE GEOSCIENCE | VOL 8 | MARCH 2015 | www.nature.com/naturegeoscience 193

© 2015 Macmillan Publishers Limited. All rights reserved

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO2350

(Fig. 3) is obtained by bootstrapping the individual 1m depth bin values of εacross the AW thermocline, for appropriate stations. The variability is shown inFigs 2 and 3 as the 95% confidence interval.

Topography. Topography used in the comparisons presented in Fig. 2 wasextracted from the GEBCO (General Bathymetric Chart of the Oceans) databasehttp://www.gebco.net.

The GEBCO database incorporates IBCAO (International Bathymetric Chartof the Ocean) data for the Arctic region.

Heat flux calculation method. Eddy diffusivity, Kz, and heat flux estimates werecalculated for the AW thermocline using a layer-averaged method. For eachprofile an average ε is calculated for the layer together with the buoyancyfrequency N , where N 2

=−(g/ρ0)∂ρ/∂z . The layer mean values are thencombined to form a turbulent diffusivity, Kz=0.2ε/N 2. Heat fluxes (Fh) are thenestimated using Kz and the layer mean temperature gradient, 1T/1Z , whereFh=−ρ0cpKz1T/1Z (ref. 29), ρ0 is a reference density, cp is specific heatcapacity and 1Z is layer thickness.

The AW thermocline heat flux for the slope region north of Svalbard andSevernaya Zemlya (longitudinal range 16◦–31◦ E) is an average value made fromestimates for 84 profiles, collected during 5 observational campaigns in theregion between 2008 and 2013. The largest values for individual profilesexceed 50Wm−2.

Tidal dissipation calculation. The tidal energy dissipation rate, D, can beexpressed as a local balance between the work rate by the tide-generating forces(W ) and the tidal energy flux (P ; ref. 24):

Dtide=W−∇ ·P

where W and P are defined as

W= gρ〈U ·∇(ηeq+ηSAL)〉

P=−gρ〈Uη〉

Here, 〈 〉 denote time averages, U is the tidal transport vector, η is the tidalelevation, ηeq is the equilibrium tidal elevation, ηSAL is the self-attraction andloading elevation, g is gravity and ρ is a reference density. The tidal amplitudesand currents from the TPXO8 database are combined with the astronomicalforcing to calculate the tidal energy flux and work rate from which the tidalenergy dissipation rate for the principle tidal constituents in the region, thesemi-diurnal M2 and S2 and the diurnal K1 and O1 constituents, is calculated.

Microstructure data. Access to the microstructure data used in the paper maybe requested. The Bangor VMP data presented in this paper may be requestedfrom the British Oceanographic Data Centre, National Oceanography Centre,Liverpool, UK (http://www.bodc.ac.uk). Request ‘ASBO and TEA-COSImicrostructure data’. The MSS data used may be requested from the NorwegianPolar Institute.

Received 4 August 2014; accepted 18 December 2014;published online 16 February 2015

References1. Polyakov, I. V. et al. Arctic Ocean warming contributes to reduced polar ice cap.

J. Phys. Oceanogr. 40, 2743–2756 (2012).2. Spielhage, R. et al. Enhanced modern heat transfer to the Arctic by warm

Atlantic water. Science 331, 450–453 (2011).3. Rudels, B. et al. Circulation and transformation of Atlantic water in the

Eurasian Basin and the contribution of the Fram Strait inflow branch to theArctic Ocean heat budget. Prog. Oceanogr.http://dx.doi.org/10.1016/j.pocean.2014.04.003 (2014).

4. Fer, I. Weak vertical diffusion allows maintenance of cold halocline in thecentral Arctic. Atmos. Oceanic Sci. Lett. 2, 148–152 (2009).

5. Timmermans, M-L., Toole, J., Krishfield, R. & Winsor, P. Ice-Tethered Profilerobservations of double diffusive staircases in the Canada Basin thermocline.J. Geophys. Res. 113, C00A02 (2008).

6. Lenn, Y. D. et al. Vertical mixing at intermediate depths in the Arctic boundarycurrent. Geophys. Res. Lett. 36, L05601 (2009).

7. Padman, L. & Dillon, T. M. Vertical heat fluxes through the Beaufort Seathermohaline staircase. J. Geophys. Res. 92, 10799–10806 (1987).

8. Guthrie, J. D., Morison, J. H. & Fer, I. Revisiting internal waves and mixing inthe Arctic Ocean. J. Geophys. Res. 118, 3966–3977 (2013).

9. Padman, L. & Dillion, T. Turbulent mixing near the Yamak Plateau during theCoordinated Eastern Arctic Experiment. J. Geophys. Res. 96, 4769–4782 (1991).

10. Steele, M. & Morison, J. D. Hydrography and vertical fluxes of heat and saltnortheast of Svalbard in autumn. J. Geophys. Res. 98, 10013–10024 (1993).

11. D’Asaro, E. A. & Morison, J. D. Internal waves and mixing in the Arctic Ocean.Deep-Sea Res. 39, S459–S484 (1992).

12. Fer, I., Skogseth, R. & Geyer, F. Internal waves and mixing in the marginal icezone near the Yermack Plateau. J. Phys. Oceanogr. 40, 1613–1630 (2010).

13. Shaw, W. & Stanton, T. Vertical diffusivity of the Western Arctic Oceanhalocline. J. Geophys. Res. 119, 5017–5038 (2014).

14. Rainville, L. & Winsor, P. Mixing across the Arctic Ocean: Microstructureobservations during the Beringa 2005 Expedition. Geophys. Res. Lett.35, L08606 (2008).

15. Carmack, E. &Melling, H. Warmth from the deep.Nature Geosci. 4, 7–8 (2011).16. Ledwell, J. R. et al. Evidence for enhanced mixing over rough topography in the

abyssal ocean. Nature 403, 179–403 (2000).17. Rudnick, D. L. et al. From tides to mixing along the Hawaiian Ridge. Science

301, 355–357 (2003).18. Holloway, G. & Proshutinsky, A. Role of tides in Arctic Ocean/ice climate.

J. Geophys. Res. 112, C04S06 (2007).19. Mead Silvester, J., Lenn, Y. D., Polton, J., Rippeth, T. P. & Morales Maqueda, M.

Observations of a diapycnal shortcut to adiabatic upwelling of AntarcticCircumpolar Deep Water. Geophys. Res. Lett. 41, 7950–7956 (2014).

20. Munk, W. &Wunsch, C. Abyssal recipes II: Energetics of tidal and windmixing. Deep-Sea Res. I 45, 1977–2010 (1998).

21. Giles, K., Laxon, S., Ridout, A., Wingham, D. & Bacon, S. Western ArcticOcean freshwater storage increased by wind driven spin-up of the BeaufortGyre. Nature Geosci. 5, 194–197 (2012).

22. Martin, T., Steele, M. & Zhang, J. Seasonality and long-term trend of ArcticOcean surface stress in a model. J. Geophys. Res. 119, 1723–1738 (2014).

23. Tsamados, M. et al. Impact of variable atmospheric and oceanic form drag onsimulations of Arctic sea ice. J. Phys. Oceanogr. 44, 1329–1353 (2014).

24. Egbert, G. D. & Ray, R. D. Estimates of M2 tidal energy dissipation from TopexPoseidon altimeter data. J. Geophys. Res. 106, 22475–22502 (2001).

25. Egbert, G. D. & Erofeeva, S. Efficient inverse modeling of barotropic oceantides. J. Atmos. Oceanic Technol. 19, 183–204 (2002).

26. Padman, L. & Erofeeva, S. A barotropic inverse tidal model for the ArcticOcean. Geophys. Res. Lett. 31, L02303 (2005).

27. Vlansenko, V., Stashcuk, N., Hutter, K. & Sabinin, K. Nonlinear internal wavesforced by tides near the critical latitude. Deep-Sea Res. 50, 317–338 (2003).

28. Watson, A. J. et al. Rapid cross-density ocean mixing at mid-depths in theDrake Passage measured by tracer release. Nature 501, 408–411 (2013).

29. Lenn, Y. D. et al. Intermittent intense turbulent mixing under ice in the LaptevSea Continental Shelf. J. Phys. Oceanogr. 41, 531–547 (2011).

30. Rainville, L. &Woodgate, R. A. Observations of internal wave generation in theseasonally ice-free Arctic. Geophys. Res. Lett. 36, L23604 (2009).

AcknowledgementsData collection and analysis by the Bangor team was funded through the UK NaturalEnvironmental Research Council ASBO and TEA-COSI Consortia (PI SB). TheNorwegian Polar Institute data collection was funded by the NPI Centre for Ice, Climateand Ecosystems (ICE). Y-D.L. was financially supported by a NERC Fellowship. VMPtechnical support was ably provided by B. Powell. The authors wish to thank the captain,crew, principal scientists and participating scientists of the Viktor Buynitsky, I/B KapitanDranitsyn, CCGS Louis S St-Laurent , RRS James Clark Ross and the RV Lance for theircooperation in the data collection. The authors gratefully acknowledge cruise fundingfrom NERC, the US NABOS programme (http://nabos.iarc.uaf.edu), and the jointUS-Canadian Beaufort Gyre Exploration Project (http://www.whoi.edu/beaufortgyre).

Author contributionsS.B., T.P.R., Y-D.L., B.J.L. and A.S. planned and directed the measurements presented inthis paper. The microstructure data collection and analysis was undertaken by B.J.L.,Y-D.L., A.S. and T.P.R. The inverse tidal modelling was undertaken by J.A.M.G. T.P.R.wrote the first draft of the paper with all the authors contributing to its revision.

Additional informationSupplementary information is available in the online version of the paper. Reprints andpermissions information is available online at www.nature.com/reprints.Correspondence and requests for materials should be addressed to T.P.R.

Competing financial interestsThe authors declare no competing financial interests.

194 NATURE GEOSCIENCE | VOL 8 | MARCH 2015 | www.nature.com/naturegeoscience

© 2015 Macmillan Publishers Limited. All rights reserved