¹ Institute of Marine Sciences and Earth Sciences Department, University of California, Santa Cruz, CA
[afisher@earthsci.ucsc.edu, 831-459-4089, 831-459-3074 (fax)]
² Marine Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA
³ Ocean Drilling Program, Texas A & M University, College Station, TX
⁴ Department of Chemistry and Geology, Mankato State University, Mankato, MN
⁵ Pacific Geoscience Center, Geological Survey of Canada, Sidney, BC

NOTE: a shorter version of this paper appeared in Spring 1999 in EOS, along with a sidebar comment from one of the reviewers. This site was constructed for scientists interested in learning about the details of data analysis described in the EOS paper, since space limitations (and limitations in the patience of most EOS readers) prevented a lengthy discussion in print.

If you wish to see plots of individual station records, go to the bottom of this page and follow the links to data from Site 902 and Site 903. If you are not familiar with the sites or with the curious evidence for recent changes in shallow bottom water offshore New Jersey, please read on. Your comments will be sincerely appreciated.

Introduction

Thermal data from boreholes on land have been used to infer recent changes in land-surface temperatures, as these surface temperature changes propagate downward into the Earth and leave a measurable record [e.g., Pollack and Chapman, 1993]. We have used thermal data from the upper 150 m of sediment below the seafloor to estimate the temperature history of bottom water offshore New Jersey, and the results are surprising, with temperature changes on the order of 6-10 °C apparently required during the last 50-150 years within a restricted area. Data were collected during Ocean Drilling Program (ODP) Leg 150 [Fig. 1; Mountain et al., 1994], with the idea that shelf and slope waters might vary in temperature on decadal to centennial time scales, and that such variations should be recorded in subseafloor sediments. The seafloor (like Earth's surface on land) acts as a low-pass filter, with high-frequency temperature variations being filtered out in the shallowest sediments, and longer-period variations penetrating to greater depths. Given typical oceanic sediment properties, seasonal signals penetrate to depths of a few meters below the seafloor (mbsf), while centennial signals may penetrate several tens to over a hundred meters.

Fig. 1. Site map showing locations of ODP Sites 902 and 903 and local bathymetry, with depth contour in meters (modified from Miller et al., 1994; Mountain et al., 1994).

Thermal diffusion reduces the amplitude of these signals as they propagate, so longer-term variations must be relatively larger to be preserved. Given depth limitations and typical sampling intervals of ODP piston coring (100-150 mbsf and 10 m, respectively), as well as uncertainties in in-situ temperature and thermal conductivity determinations, we had anticipated that bottom water temperature variations on the order of 1-4°C within the last several hundred years might be resolvable during this study. The results were quite surprising, and in fact, we are not able to explain our observations with confidence in terms of directly observed oceanographic or geological processes. We have prepared this report to make others aware of our findings, to solicit ideas as to what these observations may indicate, and to suggest that additional investigation of temperature changes in shallow bottom water may be worthwhile. A more detailed discussion of data collection, processing, interpretation, and modeling methods can be found at http://emerald.ucsc.edu/~afisher.

In-situ sediment temperatures were measured at ODP Sites 902 and 903 (Fig. 1), on the slope of the U. S. eastern continental shelf (see Miller et al., [1994], Mountain et al. [1994], and references therein). Site 902 was drilled in 802 m of water, while Site 903 was drilled 4.4 km upslope in 453 m of water. Both sites were located on the edges of submarine canyons (Fig. 1). Surface waters in this region are dominated by shelf, slope, and Gulf Stream sources. The main thermocline is seasonally stable at depths <400 m, and Gulf Stream eddies ("warm core rings") are common. Surface sediments grade from silt to clay (up slope to down slope), with little deposition of river discharge since the early Holocene due to the rise in sea level. Most modern sedimentation is hemipelagic, with some resuspended riverine mud and shelf sediments. Modern sedimentation at these shallow depths seems to be dominated by in-situ deposition rather than by downslope transport. Piston cores, grab samples, and submersible observations indicate that both large- and small-scale transport is limited mainly to blocks found at the foot of the slope (>2000 m water depth).

Experimental Methods

Sediment temperatures were measured during ODP Leg 150 using the Advanced Piston Coring (APC) temperature shoe (APC tool), comprising a power supply, data logger, and temperature sensor fit within a sediment coring shoe. Precruise calibration of these tools provided absolute accuracy of ³0.006 °C over a range of 0 to 30°C. The tool is run as part of regular coring operations, but after the coring assembly is fired into the sediments, it is left in place for 8-10 minutes to allow partial thermal equilibration of the sensor. Temperature data were collected at subseafloor depth spacing of 0.5-9.5 m, while thermal conductivity data were collected every 0.5-1.0 m.

The APC tools were carefully tested prior to ODP Leg 150 at the Woods Hole Oceanographic Institution (WHOI) CTD facility, and the resulting calibration coefficients were used to process all data. Calibration provided absolute accuracy greater than the digital resolution of the tools (±0.006 °C), with the tools held submerged in a thermally-stable (±0.001 °C), large volume circulating bath for 20-30 minutes at individual temperatures over a range of 0 to 30°C.

Extrapolation of temperature records from the time that the tool was in the sediment during each deployment was based on a comparison between observations and an idealized, one dimensional, radial heat conduction model that simulates the thermal decay of the coring shoe following frictional heating from tool insertion [Horai, 1985; Horai and Von Herzen, 1985]. As is standard practice, a segment of temperature versus time data from each deployment was compared to a theoretical decay function, and the penetration time of the tool was shifted relative to the model in order to get the best statistical (least-squares) fit. The trend of the observations was extrapolated to infinite time, when the decay function equals zero, to estimate the equilibrium in-situ temperature.

Thermal conductivity measurements were made on recovered cores using the needle-probe method [Von Herzen and Maxwell, 1959] approximately once per meter. Performance of the thermal conductivity instrument was checked using reference materials that were rotated between needles during the measurements. Uncertainties in point thermal conductivity values are estimated to be 5-10%.

Initial interpretations of all APC tool measurements were presented in Mountain et al. [1994], but we have reprocessed all downhole data, significantly revising some equilibrium sediment temperatures, and rejecting others completely because they were unreliable due to motion of the coring shoe in the sediment. In an effort to process the data consistently, the first 150 s of data following tool penetration were generally neglected in fitting to the idealized theoretical model, and a range of reasonable thermal conductivity values was used for each deployment. The resulting range in equilibrium temperatures should be considered to be minimal uncertainty, but systematic errors and deviations from the idealized model used to extrapolate observed temperatures may have caused additional bias. Corer depth errors result from uncertainty in the depth of the drill string (±1 m) and incomplete corer stroke and recovery.

Temperature Profiles and Heat Flow

Figure 2 shows estimated in-situ temperatures from the most reliable deployments plotted versus cumulative thermal resistance. The thermal resistance can be understood as subseafloor depth corrected for differences in thermal conductivity [e.g., Davis, 1988]. If the thermal regime is at steady state and is purely conductive, this plot should yield a straight line. This is the case for Site 902 (Fig.2A).

Fig. 2. In-situ sediment temperatures versus cumulative thermal resistance (depth corrected for differences in thermal conductivity) determined from data collected during ODP Leg 150. Horizontal bars indicate ranges in temperatures estimated with a range of reasonable sediment thermal conductivities. Vertical bars indicate depth uncertainties. A. Site 902. Filled symbols are from data collected with APC Tool #12, open symbols are from data collected with Tool #18. Diamonds are from Hole 902A, squares are from Hole 902C, circles are from Hole 902D. Solid line is a least-squares best fit to all data. B. Site 903. All data are from Hole 903A, diamonds are from APC Tool #12, filled squares are from Tool #17, open squares are from Tool #18. Dashed line is projection of deep gradient to the seafloor.

The seven deepest measurements at nearby Site 903 indicate similar steady-state, conductive conditions (Fig. 2B), but the shallowest six measurements document a negative thermal gradient. The magnitude and consistency of this departure from steady-state, conductive conditions is striking. Extrapolation of the deep thermal gradients from Sites 902 and 903 to the seafloor suggests regional bottom water temperatures around 4.5±0.2 °C, somewhat cooler than the lowest value measured during ODP Leg 150. The shallowest measurement at Site 903 provides only an upper bound on in-situ temperature because of probe motion during deployment (Fig. 2B).

While the Site 902 thermal data indicate a steady-state, conductive thermal gradient, the data from Site 903 indicate nonsteady-state and/or nonconductive conditions. Obvious possible causes of the apparent thermal structure at Site 903 include recent mass movement or fluid flow. To explain the Site 903 thermal profile by sediment slumping, the upper 80 mbsf would need to have moved from higher on the slope quite recently (ca. 100 yrs, as dicussed below), becoming thermally homogenized while preserving complete internal sedimentary and geochemical structure. Instead, the upper 120 mbsf of the Site 903 sediment section is essentially in tact as deposited, with only minor apparent slumping very close to the seafloor [Shipboard Scientific Party, 1994b; Christensen et al., 1996; McHugh et al., 1996].

Geochemical profiles are highly sensitive to fluid flow (more so than thermal profiles), but pore fluids squeezed from the upper 160 mbsf of sediments at Site 903 indicate dominantly diffusive and reactive conditions [Mountain et al., 1994; Hicks et al., 1996]. There is no geochemical evidence for transient lateral fluid flow (which under extreme conditions may cause a negative geothermal gradient) or for pervasive vertical fluid flow. The latter process would tend to reduce the thermal gradient but not make it negative. While we can not completely eliminate the possibility that some combination of these mechanisms is responsible for the unusual thermal structure at Site 903, the observational evidence is inconsistent with these explanations.

If the observed thermal structure at Site 903 resulted from changes in bottom water temperatures, the general scenario is as follows: steady-state heat flow was initially 45 mW/m2 with a bottom water temperature near 4.5 °C. Bottom water temperature increased rapidly and generated the negative thermal gradient above 80 mbsf. The bottom water temperature subsequently decreased back towards the present value.

Inverse Modeling of Bottom Water Temperatures

To estimate the time-temperature history of bottom water at Site 903, we analyzed the sediment temperature and thermal conductivity data using an inverse model [Wang, 1992]. The model estimates the temperature versus time history at an upper boundary of a conductive system. The time-series is constrained to be bounded and smooth, and the model incorporates a multi-layer thermal conductivity structure as well as uncertainties in both thermal conductivities and temperatures [Wang, 1992].

The model was used to estimate the last 2000 years of seafloor temperatures at Site 903. The constraints on the model were twelve subseafloor temperatures (Fig. 2B), and a twelve-layer thermal conductivity structure based on available measurements [Mountain et al., 1994]. We tested a range of numerical parameters with several dozen simulations, and found that large variations in seafloor temperatures were required to match the observed sediment temperatures. Two example results are compared to the estimated sediment temperatures in Figure 3.

In Case A, the mean bottom water temperature during the time of the simulation was assumed to be 4.0 with a standard deviation of 2.0 °C. The temporal variation in bottom water temperatures was set to 8.0 °C, allowing large changes during the simulation, and all sediment temperatures were assigned uncertainties of 0.5 °C. The resulting inverse solution includes a bottom water temperature of about 5.5 °C until 150 years before present (ybp), followed by a rapid warming to 11.3 °C (peaking around 25 ybp), and a subsequent cooling (Fig. 3B). The match to subseafloor sediment temperatures is only fair, however, with a poor match at the two inflection points (above and below the negative thermal gradient; Fig. 3A).

In Case B, we applied lower uncertainties (0.1 °C) to the inflection points and the deepest three measurements. The model results provide a better fit to the observations, but require even larger variations in bottom water temperature, from 4 to 15 °C over the last 80 ybp.

Fig. 3. Selected results of inverse models [Wang, 1992] using data constraints from Site 903 to estimate the recent history of bottom water temperatures. A. Comparison of model results with subseafloor observations. Symbols as in Figure 1. Horizontal error bars around model results indicate the standard deviations of estimated in-situ temperatures. Uncertainties in measured temperatures and measurement depths are not shown. Cases are described in the text. B. Apparent bottom water temperature histories for the same two cases shown in Fig. 3A.

In general, the best fit to the data was achieved when small uncertainties were placed on the estimated temperatures and when the bottom water temperature was allowed to vary over a wide range. When uncertainties were increased and/or the variations in bottom water temperatures were reduced, a less variable temperature-time record was generated, but the fit to the subseafloor data was poorer.

If the shallowest data point at Site 903 is omitted, a reasonably good fit to the remaining data also can be obtained with a model comprising an initial bottom-water temperature of 4.5 °C and a sudden increase of 8-10°C about 50-100 ybp. A subsequent decrease in bottom-water temperature is then required to be consistent with present-day bottom water temperatures.

Discussion

An examination of NOAA hydrographic data in the World Oceanographic Atlas (National Oceanographic Data Center, GOODbase search engine) revealed that here are no direct observations of water column temperatures below 400 m water depth from within 10 km of Site 903 (38.938°N, 72.817°W) prior to 1970, and no data deeper than 200 m prior to 1959. The available hydrographic data from within 10 km of Site 903 also show no consistent trends during this time.

Data from 400-500 m water depth between 30°N and 45°N show an enormous temperature range over the last 85 years. Observations from 400 m made during 1912-1940 within several degrees of Site 903 include many values above 10°C. Temperatures above 15°C at 400 m water depth are found at latitudes >37°N and are common at latitudes <35°N, but temperatures at 400 m depth at 38-40°N are generally 5-8 °C. The historical data are inconsistent, however, and observations of high temperatures in one year are commonly followed by much lower temperatures several years later. In addition, measurements made early in the century are of varying (and often unknown) quality, and there are few repeat measurements at single locations.

The seafloor topography around Site 903 is irregular (Fig. 1), and if there was variability in bottom water temperatures in this area over the last 50-150 years, it could reflect local processes. Perhaps there was a shift in the thermocline, with sharp vertical and lateral gradients in water temperatures at the depth of the seafloor crossing Site 903. These gradients would need to be quite sharp, as the lateral offset between Sites 903 and Site 902 is only about 4 km, and there is no apparent change in bottom water temperature at Site 902. Another possibility is that variations in bottom water temperatures reflect migration of a channel for cross-shelf transport, whereby warm, sediment-laden bottom water flowed for some years across Site 903, and then this channel was abandoned. At present we have no evidence that any of these processes occurred at Site 903, but the lack of a suitable alternative explanation leaves recent changes in bottom water temperatures as one option.

We are hopeful that other scientists working in this area may have data that will support or refute the hypothesis that bottom water temperatures at Site 903 have changed by perhaps 6-10 °C over the last 50-150 years. If other data from near Site 903 are not available, perhaps there are reliable records from similar settings along the continental shelf and upper slope where variations in bottom water temperatures have been recorded on decadal time scales. It may be worthwhile to collect additional sediment temperatures within the upper 100 mbsf in these settings, to test the idea that subseafloor temperatures along continental margins and in marginal seas can be used to estimate changes during the recent past. If significant changes in bottom water temperatures are found to occur over spatially-restricted areas, this could have implications for estimates of short term climate and/or ocean current variability.

Acknowledgements

This study of sediment temperatures and changes in bottom water temperatures along the New Jersey margin would not have been possible without the support of the co-chief scientists (Greg Mountain and Ken Miller) and the ODP Leg 150 scientific party. John Compton provided an early assessment of pore water profiles at Sites 902 and 903. This work was supported by NSF grants OCE-9416735 [later transferred to UCSC as OCE-9596183] (AF) and OCE-9415301 (RVH). Marlene Noble and Ken Drinkwater provided helpful reviews.

References

Christensen, B., B. Hoppie, R. Thunell, K. Miller, and L. Burckle, Pleistocene age models, Leg 150, in Proc. ODP, Sci. Res., vol. 150, edited by G. Mountain, K. Miller, P. Blum and D. Twitchell, pp. 115-127, Ocean Drilling Program, College Station, TX, 1996.

Davis, E. E., Oceanic heat-flow density, in Handbook of terrestrial heat-flow density determination, edited by R. Haenel, L. Rybach and L. Stegena, pp. 223-260, Kluwer, Amsterdam, 1988.

Hicks, K., J. Compton, S. McCracken, and A. Vecsei, Origin of diagenetic carbonate minerals recovered from the New Jersey continental slope, in Proc. ODP, Sci. Res., vol. 150, edited by G. Mountain, K. Miller, P. Blum and D. Twitchell, pp. 311-323, Ocean Drilling Program, College Station, TX, 1996.

Horai, K., and R. P. Von Herzen, Measurement of heat flow on Leg 86 of the Deep Sea Drilling Project, in Init. Repts., DSDP, vol. 86, edited by G. R. Heath and L. H. Burckle, pp. 759-777, U. S. Govt. Printing Office, Washington, D. C., 1985.

McHugh, C. M. G., J. E. Damuth, S. Gartner, M. Katz, and G. Mountain, Oligocene to Holocene mass-transport deposits of the New Jersey margin and their correlation to sequence boundaries, in Proc. ODP, Sci. Res., vol. 150, edited by G. Mountain, K. Miller, P. Blum and D. Twitchell, pp. 189-228, Ocean Drilling Program, College Station, TX, 1996.

Miller, K., and G. S. Mountain, Introduction, in Proc. ODP, Sci. Res., vol. 150, edited by G. S. Mountain, K. Miller and P. Blum, pp. 5-9, Ocean Drilling Program, College Station, TX, 1994.

Mountain, G. S., K. Miller, and P. Blum, Proceedings of the Ocean Drilling Program, Initial Reports, vol. 150, pp. 883, Ocean Drilling Program, College Station, TX, 1994.

Pollack, H.N., and D.S. Chapman, Underground records of changing climate, Sci. Am., 268, 44-50, 1993.

Von Herzen, R. P., and A. E. Maxwell, The measurement of thermal conductivity of deep-sea sediments by a needle probe method, J. Geophys. Res., 64, 1557-1563, 1959.

Wang, K., Estimation of ground surface temeratures from borehole temperature data, J. Geophys. Res., 97, 2095-2106, 1992.

Go to: Site 902 Data Summary (includes table of interpreted in-situ temperatures and plots from all deployments)

Go to: Site 903 Data Summary (includes table of interpreted in-situ temperatures and plots from all deployments)

Return to Top of Page