A study of the ocean's water masses using data and models

Abstract

Water masses are collections of water parcels that share similar histories and fate. The validity and value of the concept of a water mass depends on scale, spatial and temporal, as well as purpose. In spite of pragmatic and conceptual limitations, this dissertation shows that the treatment of the ocean as composed of a finite set of water masses can provide a practical analytical framework in which to describe aspects of the ocean's physical and biogeochemical structure. In particular, the case is made that, given the current biases of ocean models, model diagnostics developed in a water mass reference frame are useful tools to identify, understand and attribute causes for the specific biases. Water masses also provide an intuitive basis for analyzing the response of ocean models to many types of perturbations, be they numerical experiments or analogues of real climatic phenomena.

This dissertation is divided in three parts. The first part provides a detailed account of the ocean's water mass system. The second part utilizes the information gathered in part I to develop an analytical framework that is used to evaluate global coupled model simulations. Finally, part III investigates a methodology designed to account for the confounding role of natural variability for the purpose of measuring the change in the oceanic anthropogenic carbon distribution from repeat hydrography sampling programs.

<bold>Part I</bold> A synthesis of the ocean's major water masses is presented that compares and contrasts the water masses of the Northern and Southern polar, subpolar and subtropical regions. The world's most prominent bottom waters, deep waters, intermediate waters and central waters are reviewed sequentially, including a discussion of formation mechanisms, thermohaline and biogeochemical properties, variability, formation rates and their role in ocean circulation or climate.

Water masses originating at high polar latitudes are discussed first, with a comparative description of the freshwater and sea ice conditions in the Arctic and Antarctic regions. Low salinity intermediate waters from subpolar regions are described next, followed by high salinity intermediate waters that originate from evaporation basins. The central waters that ventilate the thermocline are discussed last, including a presentation of the five types of mode waters.

One recurrent theme of the water mass synthesis presented here is that overflows, entrainment at the overflows, water mass formation and export is often impacted critically by small bathymetric features, such as the Maud Rise in the Weddell Sea, the Orphan Knoll in the Labrador Sea, the depth and width of continental shelves or the presence of canyons cutting through the shelf break. Mode and intermediate water formation are critically dependent on the winds and on the cycling of the mixed layer. Some water masses, such as Eastern Subtropical Mode Waters in the North Pacific, can also depend critically on peculiarities of the seasonal cycle of the buoyancy flux, making these waters an interesting potential early warning diagnostic for climate change. As water mass formation often requires cold winter air bursts, small zonal or meridional shifts in the winds can have a dramatic impact on the formation characteristics of water masses.

<bold>Part II</bold> While global coupled climate models do, for the most part, resolve the main structure of the ocean, that is, models possess bottom waters, deep waters, intermediate waters and central waters, global climate models vary greatly in their ability to resolve the water masses regionally. The temperature and salinity distribution simulated by models also show significant biases. Water masses, although difficult to define quantitatively, can provide a useful framework to evaluate and discuss model simulations and biases.

This portion of the dissertation is composed of two chapters. Chapter 3 describes the structure of the overturning streamfunction in a set of global coupled climate models (Climate Model Inter-comparison Project 3, CMIP3). The main result of this chapter indicates that the different strengths of the meridional overturning circulation of the North Atlantic cell in the CMIP3 model set are mostly due to model parameterization issues associated with effective diapycnal mixing or viscosity. In the Southern Ocean, winds take on additional importance. A tight relationship is in fact found between the strength of the abyssal cell and wind intensity in the Southern Ocean across the CMIP3 models.

Chapter 4 addresses the effects of stratification. This chapter looks for relationships between specific hydrographic measures of stratification and aspects of the circulation. This study relies on the information presented in part I to analyze the water mass system simulated by global coupled circulation models. Using classical (volumetric T/S analysis, core layers) water mass analysis techniques, the model experiments are compared to the data and to one-another, revealing systematic relationships between model biases found among the CMIP3 models and these models' circulations. It is found that temperature and salinity biases result from differences in the model's circulations more than the differences in the models' circulation depend on the biases of the thermohaline properties.

<bold>Part III</bold> The third section of the dissertation utilizes the concept of water mass more indirectly: as a means to evaluate and select linear regression models. This chapter focuses on the use of statistical regression methods to isolate the anthropogenic carbon decadal change signal from hydrographic data. A synthetic data set derived from ocean biogeochemistry and circulation model simulations, in which the true signals are known, is used as a simulator to optimize algorithms and evaluate the accuracy of the particular methods in their ability to recover the true estimate of the change in the interior anthropogenic carbon concentration. It is shown that regression analysis on repeat hydrographic data sets such as WOCE and CLIVAR can account for water mass variance and short-term variability, and estimate the true 10-year change in anthropogenic carbon inventory with a better than 10% accuracy. More importantly, this chapter outlines a strategy that can be used to select appropriate regression models in the context of non-ideal repeat hydrographic samples.