Incorporating both late/post-Glacial bottom-water warming and eustatic sealevel rise into a MH (methane hydrate) stability model explains why at least some major submarine landslides (e.g., the Storegga Slide, Norwegian margin) may have been triggered by MH dissociation during the early Holocene (rising sealevels), not during the lowest sealevels of the LGM (Last Glacial Maximum, 18–20 ka). In the case of Storegga, failure may have been initiated either below the upper slope or under the shelf (for fresh pore water, especially with admixed ethane). At water depths below ca. 800m, persistent low bottom water temperatures allowed the sealevel rise to thicken the MH stability zone over time, ruling out Holocene initiation of failure. MH release by Holocene mass wasting cannot have initiated deglaciation.
Carpenter  first noticed a spatial association between submarine slide scars and seismic evidence for subbottom gas hydrate, primarily methane hydrate (MH), offshore the eastern United States. This led him to suggest that MH dissociation could have facilitated mass wasting. Later authors hypothesized that MH dissociation during glacial sealevel lowstands provided the negative feedback to end glaciations: If the methane liberated by MH dissociation found its way into the atmosphere, a strong greenhouse effect would then have initiated ice sheet meltback [e.g., Nisbet, 1990; Paull et al., 1991]. Tests for the above hypotheses include accurate age-dating of slides, to verify a correlation between sediment failure and sealevel lowstands. If the time of failure could be established, the ice core data from Greenland and Antarctica could be searched for correlative increases in atmospheric methane, and perhaps climate warming events. No such correlations have been demonstrated to date, although few slides have so far been accurately dated.
 We focus here on the continental slope and rise west of Norway, and the Barents Sea [Figure 1; Mienert et al., 1998; Vogt et al., 1999]. The largest submarine landslide decorating this margin is the Storegga Slide, a compound (three-event) slide that dislodged ca. 5580 cubic kilometers of sediment. The three separate Storegga slides [Bugge et al., 1988] were first believed to have occurred at different times. Recent work shows the three main events to have occurred essentially at the same time, ca. 8.15 ka [calendar years, Haflidason et al., 2001]. One tsunami caused by the Storegga slide washed the coasts of Scotland and Norway [e.g., Bondevik et al., 1997].
 The early Holocene date of the Storegga slide presents a problem for the hypothesis that MH dissociation, caused by low sealevels of the LGM, was responsible. At the time of this slide (8.15 ka), global sealevels had already risen to within 25 m of modern values [e.g., Fairbanks, 1989]. The Andøya slide off northern Norway (Figure 1) occurred during the mid-Holocene, probably even later than Storegga [Laberg et al., 2000]. The other major slides have not yet been dated accurately, but a “Late Weichselian” age for the Bear Island Fan (BIF) slide [Hald and Aspeli, 1997] makes it possible that this slide also postdates the 18–20 ka LGM.
 The occurrence of major submarine landslides during post-LGM times is inconsistent with theories fingering MH dissociation as the “shut-off” valve for ice ages [e.g., Nisbet, 1990; Paull et al., 1991]. However, we propose that Holocene sliding is NOT inconsistent with MH dissociation as a trigger for mass wasting—but may be a consequence, not cause, of deglaciation. We test the hypothesis that sliding was delayed into the Holocene by modeling the time it took for late glacial or early post-glacial ocean warming to penetrate to the base of the MH stability zone. We include the combined effects of sealevel rise—which thickens the MH stability zone (MHSZ) without delay—and bottom water warming, which takes time to penetrate the sediments. The physical and chemical processes of hydrate formation and dissociation [e.g., Davie and Buffett, 2001; Xu and Ruppel, 1999] are beyond the scope of this paper.
 Our models consider time variations of both pressure (P) and temperature (T) since the 18–20 ka LGM. We assume heat transport was solely by molecular conduction, and apply solutions of Turcotte and Schubert . Due to the low horizontal temperature gradients, we assume heat is transferred only in the vertical direction. MH dissociation is endothermic, while formation is exothermic. However, because the concentration of MH in sediments is highly variable and not well known, we ignore the effects of latent heat. In and below the zones of MH formation or dissociation, temperature changes will be slower than calculated, depending on the amount of MH actually present. However, only a thin layer of MH dissociation would suffice to reduce shear strength, triggering sediment failure.
2. Model Constraints
 We used the eustatic sealevel curve of Fairbanks  to calculate the time-dependent component of overburden pressure. The headwalls of the Storegga and BIF slides are located near the outer edge of grounded LGM ice sheets, which, although grounded on the shelf, were nearly afloat at their distal edges. Thus, post-glacial rebound, if any, was probably small; therefore the sealevel history in the slide areas was probably nearly eustatic. Landward of our profiles (Figure 1) a thicker ice sheet and low basal temperature probably caused an expansion of the MHSZ. However, any MH formed subglacially would have decomposed under the rapid pressure drop as the ice sheets disappeared. The sealevel rise following the LGM then expanded the MHSZ, particularly during the periods 12.5–11.5 and 10–9ka, when global sealevels rose most rapidly [Fairbanks, 1989].
 The precise time variation of bottom water temperature in the slide areas is not well known. Based on the paleoceanographic reconstructions of Koc et al.  and Miller et al. , we place the earliest possible time of emplacement of near-modern water temperatures at 15 ka, and we calculated models (not shown) based on this date. However, we consider the end of the Younger Dryas (11 ka) as the most probable time of significant warming, and use this date here.
 We assume a modern-like (historical) water temperature structure [See Figure 3 of Vogt and Sundvor, 1996] was applied to the seafloor instantaneously at 11 ka. However, the subbottom temperature history several millennia later is not very sensitive to the exact times and rates of water warming. We assumed a constant −1°C for the glacial-age ocean prior to warming.
 The equilibrium function (phase boundary in P,T space) for hydrate depends on porewater composition [Sloan, 1990, 1998; Peltzer and Brewer, 2000]. Admixtures of heavier hydrocarbons tend to increase the range of stability, whereas porewater brackishness has the opposite effect [Dickens and Quinby-Hunt, 1997]. Our models were calculated for fresh water with no higher hydrocarbons, and with 1% and 2% ethane. We also explored the case of porewater with seawater salinity. Posewang and Mienert  found that seawater and 1% ethane predicted a good fit to the lower BSR (Bottom Simulating Reflector) they detected north of the Storegga Slide, in 1000 m water.
 Models for steady-state subbottom temperature e.g., [Ruppel, 2000] require surface heatflow, the variation of conductivity with depth, and subbottom heat sources, the latter neglected here. We assume a steady-state temperature distribution vs depth (the “geotherm”) existed in the slide area until the time of bottom-water warming. Subbottom warming rates depend on the thermal diffusivity.
 Regional heatflow averages ca. 40–60 mW/m2 in both slide areas [Sundvor et al., 2000]. Typical conductivities in the top few meters of sediment are ca. 1.29 W/m-°K for the Storegga area and ca. 1.13 W/ W/m-°K for the BIF area. There are few thermal data from the continental shelves bounding the slide headwalls. Posewang and Mienert  used a gradient of 50° K/km for an 880 m deep site just north of the Storegga slide; they based this value on borehole temperature as well as seismoacoustic data on BSR depth. Although a global compilation of heatflow shows typical shelf values of ca.80–120 mW/m2, very rapid LGM sedimentation (up to 1000m/Ma) along shelf edge depocenters around the Nordic Basin margins would have depressed surface heatflow [Louden and Wright, 1989].
 Given the above data and the various uncertainties, we computed subbottom temperatures for heat flows of 40, 50 and 60 mW/m2, based on a conductivity of 1.1 W/m-°K. The thermal diffusivity (3.697 × 10−7 m2/sec) was calculated from this conductivity by the relation of Villinger and Davis . Model results (Figures 1 and 2) are shown for the most probable heat flow (50 mW/m2). Given the uncertainties in parameters, Posewang and Minert's (1999) BSR at 1000 m water depth (285 m subbottom) is not greatly less than our model prediction at this depth (370–410 m). In addition, the temperature at 285 m may be colder than predicted from stability curves [Ruppel, 2000; p. 36].
3. Storegga and Bear Island Fan Slides: Model Results
 A select subset of calculated MH stability profiles (Figure 1) predicts the changes in thermal and MH stability fields for both the Storegga and BIF slides between 11 ka (the time of bottom water warming) and 8.15 ka, immediately prior to failure. (The Bear Island Fan slide probably occurred earlier). Any MH present in the orange or red subbottom fields (from a few tens to ca. 100 m thick) must have dissociated between 11 ka and 8.15 ka.
 The intersection of the failure surfaces (B) with these dissociation belts is the most likely site of initial failure, i.e. under the shelves. However, for the case of salty porewater, the Storegga MHSZ did not extend under the shelf and does not intersect the failure surface. (This is not the case for the BIF slide, due to lower temperatures and greater shelf depths). Under the salty porewater scenario, Storegga failure could have been initiated in water depths between the shelf edge and ca. 600–800m water depths, and then propagated backwards into the shelf. At greater depths, both failure surfaces again intersect the base of the MHSZ (Figure 1), but the perpetually cold bottom water allowed sealevel rise to thicken the MHSZ by ca. 5–10 m over the same time. Therefore, if MH dissociation helped trigger these two slides, post-Glacial failure (demonstrated for Storegga) originated on the upper slope or under the shelf. (By contrast, any LGM slides could have originated at any water depth where MH existed at the time, because low pressure was the only factor).
 We also calculated the trajectories with increasing time, of particular parcels of sediment in (P,T) space, relative to MH stability curves (Figure 2). All points share an “upward” displacement over time, reflecting the pressure increase of the rising sealevel (Figure 3, inset). The trajectory bend is the effect of bottom water warming at 11 ka, and is most dramatic and nearly instantaneous at shallow depths (e.g., points 1 and 2). Depending on porewater salinity and amount of admixed ethane, a number of “tracks” cross into the MHSZ during the period 18–11 ka, and then once more exit out of the zone, showing that MH could form in the late Weichselian, only to dissociate after bottom water warming. Points 1–8 show the (P,T) conditions that existed at the failure surfaces at 8.15 ka. Four situations existed along this surface; I: #1 enters and exits the MHSZ, but lies within the upper sediments which, due to oxidation of methane by sulfate, lacks MH. We made this zone 115 m thick [Posewang and Mienert, 1999]. II: Storegga point 2 and BIF points 2–4 enter and exit the MHSZ, and are prime sites of failure. III: Storegga points 3–5 and adjacent BIF points remain out of the MHSZ. IV: Points 6–8 remain inside the MHSZ or move into it by pressure increase.
 Calculation of the post-slide (P,T) history of the subbottom in the slide scars is the subject of a later paper, so the final point (0 ka) on all trajectories represents modern conditions under the margins adjacent to the slides. Subbottom points near Storegga #9 and 10, and BIF # 3, 4 and 9 have moved significantly farther out of the MHSZ (depending on salinity and ethane content) from 8.15 ka to the present, suggesting an increased modern vulnerability to failure.
 Using published hydrate stability criteria (fresh and saline porewater, and admixed ethane (0, 1 and 2%)), we explain how major submarine landslides along the Norwegian margin (Storegga, Andϕya, and perhaps Bear Island Fan) could have been triggered by MH dissociation subsequent to the low sealevels of the LGM. MH dissociation was “delayed” because bottom water warming occurred after sealevels had begun to rise, and by the time it took for this warming to “diffuse” (conduct) down to the base of the MHSZ. Holocene sliding may be more widespread than supposed, and, if so, cannot have been the “shutoff valve” for glacial ages. Along the Norwegian margin the MHSZ models suggest that MH may have formed under the shelves after deglaciation (due to pressure rise), only to dissociate later due to bottom water warming. Slides triggered at water depths greater than ca. 600–800m could only have been triggered by LGM dissociation, due to the permanently cold water at such depths. If MH dissociation can promote sliding, sediments along upper continental slopes may still be vulnerable to failure by MH dissociation today.
 We thank W. Dillon, J. Gettrust, M.D. Max, H. Haflidason, and W. Wood for discussions and the Office of Naval Research for support. Two anonymous reviews were valuable. Irene Jewett helped prepare illustrations.