Thurnherr is an observational physical oceanographer interested in how the ocean works. A functional understanding of the ocean can often best be obtained by combining observational and modeling (both analytical and numerical) approaches; Thurnherr focuses on observations and enjoys collaborating with modelers in particular to help plan and interpret data from observational programs. His most important research interests concern processes acting near topography (including hydrothermal circulation), horizontal and vertical dispersal, as well as the large-scale circulation, with emphasis on the return limb of the overturning circulation (mixing and upwelling). Additionally, he is interested in oceanographic instrumentation and methods, in particular those related to the measurement of currents, internal waves and turbulence with acoustic methods (especially ADCPs). In the following sections, several of Thurnherr's main research interests are discussed in some detail.
Most of Thurnherr's work so far has been carried out in the context of multidisciplinary projects, involving ecologists, geophysicists and chemical oceanographers. He considers multidisciplinary collaborations particularly fruitful because of the broader perspectives they entail. Furthermore, data from other disciplines, such as geochemical gradients near hydrothermal vent fields or biogeographic patterns inferred from genetic studies, often provide useful constraints on the physical oceanography.
The topographically rough flanks of the mid-ocean ridge system cover more than half of the seafloor, at least in the Atlantic and Indian Oceans. Additionally, it has been estimated that there are approximately 100,000 seamounts rising more than 1000m above the surrounding seafloor. Thus, "topographic roughness" is the norm, rather than the exception, and the effects of this topographic roughness on the ocean are varied and complex.
It has been known for many decades that topography affects the regional hydrography. In particular, isopycnal surfaces tend to dome upward over the crests of mid-ocean ridges and isolated seamounts and dip downward into the flanks further down on the topographic structures. The primary mechanisms usually put forward to account for these patterns are bottom-intensified diapycnal mixing over topographic slopes [e.g. Wunsch, DSR 1970; Phillips, DSR 1970; Thompson and Johnson, DSR 1996] and potential-vorticity conservation over topographic highs [Holloway, JFM 1987]. Given that subinertial motions in the ocean are largely geostrophically balanced these hydrographic patterns suggest the existence of horizontal circulations that are anticyclonically sheared over the crests and cyclonically sheared over the flanks. Available observations, however, are only partially consistent with this conceptual picture. In the subtropical South Atlantic, for example, all available zonal sections crossing the Mid-Atlantic Ridge show isopycnals dipping into the ridge flanks but the vertical scale of these "ridge-flank boundary layers" is more than an order of magnitude too large to be consistent with diffusive boundary layers on slopes [Thurnherr and Speer, JPO 2003]. Additionally, float [Hogg and Owens, DSR 1999] and tracer [Ledwell, cruise report 2000] observations from the western ridge flank do not show the expected equatorward drift. It turns out that both these observations are related to the presence of deep and more-or-less regularly spaced fracture-zone canyons that corrugate the flanks of all slow-spreading mid-ocean ridges. Detailed analysis of hydrographic observations reveals that the dipping isopycnals in the Brazil Basin are largely restricted to the canyons where the resulting pressure gradients lead to persistent and strong up-canyon (rather than geostrophically balanced along-ridge) flows [Thurnherr and Speer, JPO 2003; Thurnherr et al., JPO 2005]. In 2009 two additional hydrographic data sets were collected in deep canyons on the global mid-ocean ridge system: one in a ridge-flank canyons on the Southeast Indian Ridge and the other in the rift valley of the Cayman spreading center in the Caribbean. While taken in "arbitrary" valleys, they both show the same hydrographic patterns that are observed in the ridge-flank canyons in the South Atlantic and in the rift valley of the Mid-Atlantic Ridge (MAR), implying that "valley flows" are extremely common.
In contrast to the cyclonically sheared along-flank flows that are apparently much less common in the ocean than previously thought, anticyclonic flow associated with isopycnal doming over seamounts and ridge crests seems to be the norm, rather than the exception [e.g. Thurnherr et al., DSR 2011]. These flows, which arise from topographic rectification, are associated with typical velocities and lateral scales on the order of 10cm/s and 10km, respectively, i.e. too narrow to be sampled during typical WOCE-style hydrographic sections and probably too weak to impact the ocean on a large scale. However, the anticyclonic ridge-crest currents likely play an important role in the dispersal of hydrothermal organisms [McGillicuddy et al., DSR 2010], the larvae of which must settle at hydrothermal sources, which are predominantly found along the crest of the global mid-ocean ridge system.
The global overturning circulation re-distributes heat and fresh water around the Earth and, thus, affects climate both regionally and on a global scale. Large buoyancy fluxes from the ocean to the atmosphere take place primarily at high latitudes where they cause densification and sinking of near-surface waters. In order to close the circulation, both diapycnal mixing and upward vertical motion (upwelling) are required. In addition to the two major cells of the global overturning circulation, there are overturning circulations in all closed ocean basins.
The presence of the canyons corrugating the flanks of slow-spreading ridges with their associated along-canyon flows discussed above has important consequences for turbulence and mixing in the abyssal ocean. Microstructure measurements and dye-dispersal observations from the Brazil Basin in the western South Atlantic indicate that much of the mixing there takes place near the topographically rough flank of the MAR. The increased mixing near the ridge has been inferred to be caused primarily by breaking internal waves, forced by the tides [Polzin et al., Science 1997; Ledwell et al., Nature 2000; St. Laurent et al., JPO 2001]. This interpretation forms the basis of at least two mixing parameterizations for circulation and climate models [Jayne and St. Laurent, GRL 2001; Polzin, JPO 2004]. However, this interpretation is apparently inconsistent with the observation that most of the observed turbulence occurs in the deep fracture zone canyons, rather than above the topography, and it does not take the effects of the numerous overflows along the path of the up-valley flows into consideration [Thurnherr et al., JPO 2005]. In Thurnherr's interpretation, the Brazil Basin data imply that it is the mixing within and the flow up these ridge-flank canyons that close the buoyancy and mass budgets of the bottom water in the subtropical South Atlantic. The dynamics in the ridge-flank canyons, where along-valley advection of buoyancy is balanced by diapycnal buoyancy fluxes, is surprisingly similar to the hydrography and dynamics in sloping rift-valley segments of slow-spreading mid-ocean ridges [Saunders and Francis, Progr. Oceanogr. 1985; Thurnherr and Richards, JGR 2001; Thurnherr et al., JPO 2002; Thurnherr, DSR 2006; St. Laurent and Thurnherr, Nature 2007; Thurnherr et al., JMR 2008].
Wind forcing in the Southern Ocean is another process that helps closing the global overturning circulation, in particular that of the upper NADW cell. Thurnherr is involved in the DIMES project that aims to quantify horizontal and vertical mixing in the Southern Ocean, which is required in order to understand the relative contributions of the different processes.
Dispersal in the ocean is important in a variety of different contexts, including dispersion of accidentally (e.g. oil spill) or purposely (e.g. mining operations) released pollutants, gene flow and biogeography [e.g. Mullineaux et al., 2002], the design of marine protected areas [e.g. Vandover et al., 2012], as well as in the interpretation of natural (e.g. terrigenic helium) and anthropogenic (e.g. CFCs, bomb tritium, etc.) tracers as constraints for the ocean circulation and mixing [e.g. Thurnherr et al., DSR 2011]. Dispersal in the ocean is a 3-dimensional process that involves scales ranging from the basin scale to a few millimeters (dissipation scale), which makes it very difficult to study. In particular, dispersal near topography is strongly affected by topographic effects on the currents and mixing [e.g. Speer et al., 2003] such as the ones discussed above. The most direct method for studying dispersal in the ocean consists in releasing a tracer substance and studying the evolution of the resulting cloud over time [e.g. Jackson et al., DSR 2010]. While this yields direct estimates for advection and eddy diffusion in all three dimensions, such experiments are difficult and expensive to carry out and the results can be affected by the often considerable time required for sampling. Additionally, the result from a tracer-release experiment reveals but one of many possible dispersal pathways [e.g. Lavelle et al., JGR 2010] and it is often difficult to draw general inferences, e.g. regarding the processes affecting dispersal, from single tracer releases [e.g. Jackson et al., DSR 2010]. Therefore, dispersion is best studied with a combined methodology involving tracer-release experiments, moored measurements of circulation and hydrography, regional surveys and numerical models, as was done during the LADDER project of which Thurnherr was a co-PI.
Thurnherr's interest in dispersion in the ocean is at least partially motivated by the difficulty in studying this topic. Too often, inferences about dispersal in the ocean are drawn from "progressive vector diagrams" (hodographs) constructed from the records of single point current meters even though the assumptions underlying this method are rarely valid [e.g. Thurnherr, 2004]. Which methods are better suited to investigate dispersal in the ocean, on the other hand, depends sensitively on the regional setting, as well as the time and space scales of interest. Of particular interest to Thurnherr in this context is the question how dispersal affects the biogeography of hydrothermal (and other) organisms [e.g. Mullineaux et al., 2002; McGillicuddy et al., DSR 2010] and the design of marine protected areas [e.g. Vandover et al., 2012].
When Acoustic Doppler Current Profilers (ADCPs) were invented they revolutionized the measurement of velocities in the ocean primarily because they collect velocity profiles, rather than the point measurements made by single-point current meters. Thurnherr has several specific research interests related to ADCP measurements (see also Thurnherr's LADCP page):
Full-Depth Horizontal Velocity Profiles. Most regions of the ocean are too deep for shipboard ADCP systems to sample the full depth of the water column. Therefore, it has become routine to lower ADCPs on CTD rosettes for obtaining full-depth profiles of horizontal velocity, which are useful, for example, for process studies and regional circulation surveys [e.g. Thurnherr and Richards, JGR 2001; Thurnherr et al., JMR 2008; Thurnherr et al., DSR 2011]. However, elaborate data processing is required in order to remove the effects of instrument motion from the velocity measurements. The two fundamentally different methods that are used for this purpose [Fischer and Visbeck, J. Tech. 1993; Visbeck, J. Tech. 2002] sometimes yield markedly different results and the uncertainties of the resulting velocities are not well known. In order to address these issues, Thurnherr carried out an error analysis of deep-ocean LADCP velocities based on comparisons with independent simultaneous velocity measurements from moored instruments [Thurnherr, J. Tech. 2010]. Results indicate that rms velocity discrepancies below 3cm/s are achievable with standard LADCP hardware, but only when multiple velocity-referencing constraints are used. The two different processing methods perform approximately equally well, although one of them had to be extended significantly in order to allow the simultaneous application of multiple velocity-referencing constraints.
Internal waves, turbulence and mixing from ADCP Data. Away from boundaries, turbulence and mixing in the ocean are closely related to the internal wave field. This has led to the development of so-called "finescale parameterization methods" that use shear (and often also CTD-derived strain) observations to determine the energy in the internal wave field from which turbulence levels can be estimated [Gregg, JGR 1989; Polzin et al., JPO 1995; Gregg et al., Nature 2003.] (Strong turbulence can also affect the backscatter environment in the ocean, allowing the observation of turbulence from ADCP echo-return data [Visbeck and Thurnherr, DSR 2009].) Given the very large data base of available profiles it seems only natural to apply this method to CTD/LADCP data to enhance the very small data base of more direct turbulence measurements obtained with microstructure instruments. In order to apply finestructure parameterization methods, the LADCP-derived shear spectra must be corrected for attenuation related to details of ADCP sampling and LADCP processing [Polzin et al., J. Tech. 2002] and this method has become very popular recently [e.g. Naveira-Garabato et al., Science 2004; Kunze et al., JPO 2006; MacKinnon et al., Nature 2008; Tian et al., JPO 2009]. However, a comparison between two published estimates of the diffusivities along the WOCE I02 line across the Indian Ocean at 10S based on the finestructure parameterization method shows section-averaged turbulence and mixing levels differing by up to two orders of magnitude, as is apparent when comparing Fig.~6 of Kunze et al. [JPO 2006] to Fig.~2 of Palmer et al. [JPO 2007]. Since uncertainties of this magnitude are not satisfactory I have begun investigating the problem. Results obtained so far indicate that in many of the studies that use the LADCP finescale parameterization method spectral corrections that are inconsistent with the underlying processing are applied, although it does not appear that the erroneous corrections can account for the differences between the two published I02 analyses [Thurnherr, J. Tech. 2011].
Vertical Velocity. In contrast to traditional single-point mechanical current meters, ADCPs measure all three components of the flow field, i.e. including vertical velocities. In spite of the large vertical velocities and accelerations of CTD packages, Thurnherr has been able to develop a method to obtain full-depth vertical-velocity profiles from CTD/LADCP data to an accuracy of better than 1cm/s, which is sufficient for the study of internal waves, convection, gravity currents and other energetic processes [Thurnherr, IEEE 2011]. Preliminary observations suggest that finestructure spectral levels of vertical velocity are correlated with microstructure-derived turbulence measurements. This suggests that it may be possible to derive global and regional climatologies of turbulence and mixing from CTD/LADCP data with an alternative finestructure method.
| © 2012 A.M. Thurnherr (e-mail) | created: Tue Sep 11 09:37:04 2012 | modified: Tue Sep 11 12:15:47 2012 |