Outline: High Latitude Surface Fluxes: Requirements and Challenges for Climate Research

Outline: High Latitude Surface Fluxes: Requirements and Challenges for Climate Research

authorship: U.S. CLIVAR High Latitude Flux Working Group plus other contributors

[Some suggested changes by Ross; comments questions in brackets like this.]

1. Introduction

High latitude regions have been marked by rapid climate change in recent years. Perennial sea ice in the Arctic has decreased by at least 20% since the mid-1970s (reference?), the Southern Ocean has warmed (e.g. Gille, 2002, 2008; Boning, 2008), and grounded ice has melted and broken away from the Antarctic continent (e.g. Rignot and Jacobs, 2002; Shepherd et al, 2004; Thoma et al., 2008.) Most Intergovernmental Panel on Climate Change (IPCC) climate models forecast that these high latitude warming patterns are likely to persist at least through the coming two centuries (??, ??). However, the harsh environment of high latitude regions makes in situ monitoring of these changes challenging. This is particularly true of surface fluxes, which are crucial, because they determine how heat, momentum, fresh water, and gases such as CO2 are exchanged between the atmosphere, ocean, and ice.

[We are not explicit about the motivation for this article in the introduction. Fluxes are critical for accurate climate projections; trustworthy projections are required to target adaptation strategies; etc. ]

The observational challenges in measuring fluxes are myriad. High latitude regions are remote, far from ports or major airports, so field programs require daunting logistics, and autonomous instruments cannot easily be serviced. Moreover, winds are among the strongest in the world (e.g. ??; Renfrew, 2008), so oceanographic instruments must be able to withstand high winds and rough seas, as well as cold temperatures and icing conditions. Flux observations were successfully collected from the Surface Heat Budget of the Arctic Ocean (SHEBA) ice camp, but SHEBA was most successful at characterizing fluxes over year-round sea ice without leads (reference ??), and these are the conditions that appear to be disappearing most rapidly in the Arctic.

We expect that fluxes through an ice-free Arctic Ocean should be distinctly different from fluxes through a high-albedo, ice-covered Arctic Ocean. Although new measurement technologies for the ice-covered ocean have evolved, in part stimulated by the International Polar Year, most of those sensors either stay fixed in the ice and thus fail when the ice melts or operate under the ice with instruction to stay well below the surface when they detect ice. Because of their constantly changing ice conditions, marginal sea ice zones that contain ice/water mixes are among the most difficult regions in the world to instrument for year-round flux observations, and fluxes through these regions have proved difficult to characterize (reference ??). Moreover, the rapid warming in high latitude regions is amplified by feedbacks associated with (1) the high albedo of polar snow and ice (xxref), and (2) feedbacks between snow melt, temperature, and longwave emission (xx ref). If ice is lost, extra heat can be stored in these regions and remain through winter and reduce ice thickness the following spring, further accelerating the loss of ice.

Even over land there is no real high latitude flux observing system. In the Arctic, few flux-quality surface instrumentation sites have been established (confirm, reference ??) In the Antarctic, surface meteorological data are used primarily for aviation, and data relevant for assessing fluxes or climate variability are not routinely collected.

In some parts of the world, satellite data and numerical weather prediction (NWP) can provide reasonable estimates of surface fluxes even in the absence of in situ observations, but these products are less successful in high latitude regions because the overwhelming lack of in situ observations means that the satellites are not well calibrated, particularly at high wind speeds. Moreover, few in situ data are available for assimilation into NWP fields. Additionally, parameterizations used in NWP models are rarely validated for polar conditions. As a result, flux products can differ substantially at high latitudes, even in their climatological average, as illustrated in Figure 1.

While a number of new high latitude programs were initiated as part of the 2007-09 International Polar Year, these programs for the most part did not focus on surface fluxes (Southern Ocean GasEx was one notable exception). The objective of this report is two-fold. We describe the current accuracy of flux estimates for momentum, energy, freshwater, and gas fluxes for the space and time scales dictated by these applications [what applications]. Then we evaluate how these current accuracies compare with the requirements for high latitude fluxes for a range of applications. [Why don’t requirements come first?] In this paper Section 2 summarizes methods used to determine fluxes. Section 3 reviews the methods used to measure and parameterize fluxes using in situ data. Section 4 discusses gridded fields from satellite and numerical weather prediction products. Section 5 considers applications requirements, and section 6 summarizes the results. [The following sentence could go earlier in the introduction as part of the motivation.] This report was coordinated by the US CLIVAR Working Group on High Latitude Fluxes, and it is intended to starting point for community discussion focused on how best to improve surface fluxes at high latitudes.

2. Approaches to Determining Fields of Fluxes

Surface fluxes fall into three general categories. Radiative fluxes measure include the shortwave electromagnetic radiation from the sun impinging on the ocean (or ice) surface and the longwave electromagnetic radiation emitted from the surface and from within the atmosphere. Freshwater fluxes measure precipitation and evaporation (i.e., latent heat). And turbulent fluxes measure just about everything else, including momentum, sensible and latent heat, and gas exchange.

The surface energy budget includes the following components (King and Connolley 1997):

where, and are the downward and upward long wave fluxes, is the downward shortwave flux, is the surface albedo, is the sensible heat flux, is the latent heat flux, and G is the conductive flux through the snow/ice pack (Pavolonis, et al., 2003).

1.  Basics of Fluxes

A.  Basic definitions of fluxes for the purpose of this paper Bourassa, Pinker

radiative versus turbulent fluxes.

B.  Differences among easily available products Bourassa, Speer (sea ice zone)

types of fluxes, basic heat balance, comparison of different product, definition of accuracy and uncertainty

3. In situ methods and their parameterization of surface fluxes Fairall

[Way too many equations that I can not even see; but no matter we probably shouldn’t have any equations in this paper.]

Turbulent fluxes characterize a major part of ocean-atmosphere exchange. While they may be measured directly with appropriate sensors placed on a suitable platform over the ocean, most applications require estimates distributed over space and time. The principal use of direct in situ flux observations is to advance calibrate and validate indirect methods so that fields of fluxes can be determined from variables, such as wind speed and sea-surface temperature, that are available on the required space/time scales. The indirect methods are known as bulk flux algorithms. Bulk flux algorithms have been in use for nearly a century. They now form the basis for the ocean-surface boundary condition in virtually all climate and NWP models, retrieval of turbulent fluxes from satellite observations, and have been used extensively to estimate the heat balance of the oceans from historic weather observations from volunteer observing ships (WCRP 2000). Advances in understanding of the physical processes involved in air-sea exchange and in observing technologies has promoted steady improvements in the sophistication and accuracy of these algorithms.

In bulk algorithms the turbulent fluxes are represented in terms of the bulk meteorological variables of mean wind speed, air and sea surface temperature, and air humidity:

(1)

where x can be u, v wind components, the potential temperature, q, the water vapor specific humidity, q, or some atmospheric trace species mixing ratio. Here cx is the bulk transfer coefficient for the variable x (d being used for wind speed) and Cx is the total transfer coefficient; ΔX is the sea-air difference in the mean value of x, and S is the mean wind speed (relative to the ocean surface), which is composed of a magnitude of the mean wind vector part U and a gustiness part Ug:

. (2)

In [2] z is the height of measurements of the mean quantity X(z) above the sea surface (usually 10 m) and is the gustiness factor. The gustiness term in [2] represents the near-surface wind speed induced by the BL-scale; it prevents the transfer coefficients from becoming singular at low wind speeds.

Properly scaled dimensionless characteristics of the turbulence at reference height z are universal functions of a stability parameter,, defined as the ratio of the reference height z and the Obukhov length scale, L. Thus, the transfer coefficients in [1] have a dependence on surface stability prescribed by Monin-Obukhov similarity theory:

, (3)

where the subscript n refers to neutral (z = 0) stability, is an empirical function describing the stability dependence of the mean profile, and zox is a parameter called the roughness length that characterizes the neutral transfer properties of the surface for the quantity, x (see also Fairall et al. [2003] for details).

1. Instrumentation

The neutral transfer coefficients (or, equivalently, the roughness lengths) are determined by direct observations of the fluxes and associated mean bulk variables required in [1]. Mean bulk variables may be measured with relatively slow response instruments optimized for accurate means. Turbulence variables require a flat frequency response to fluctuations out to about 10 Hz (somewhat dependent on the height of the sensor and the conditions). A variety of techniques have been used to estimate the fluxes (Smith et al. 1996) but the eddy-correlation method is considered the standard. In the eddy-correlation method u’ and x’ are measured and the flux is estimated as a simple time average of their products (i.e., ). The inertial-dissipation method (IDM) has also seen broad application. IDM is based on the high-frequency part of the variance spectrum. Its principal advantage is insensitivity to motion (hence, no motion corrections). However, an empirical stability function is required to obtain the flux so it is not considered an unbiased standard. Velocity turbulence is typically measured with a sonic anemometer or a multiport pressure system; humidity turbulence is usually measured with a fast infrared absorption hygrometer; temperature turbulence is usually obtained from sonic anemometer speed-of-sound or from micro-thermal wires. Aircraft and ship platforms dominate the transfer coefficient database, and these require motion corrections to obtain the true air velocity (Edson et al., 1998). [Could we show examples of actual data used to make these measurements---a correlation plot and a spectrum showing whatever fitting is needed to estimate the flux.] An example of a cluster of sensors for a ship-based flux observing field program is shown in Fig. 1.

1. Computation of Transfer Coefficients

The reduction of an ensemble of observations of turbulent fluxes and near-surface bulk meteorological variables to estimates of the mean 10-m neutral transfer coefficient is a problem of some subtlety. The straightforward approach is to convert each observation to Cx10n

(4)

then average to obtain

(5)

The 10-m neutral values of the mean profile are computed as

(6a)

, (6b)

where can be or . Note, the sign difference between [6a] and [6b] follows from being defined as in Equation [1].

Because artificial correlation may confuse attempts to compute the mean transfer coefficients via [5], Fairall et al. [2003] computed estimates of mean transfer coefficients as a function of wind speed. Here the fluxes are averaged in wind speed bins and the mean transfer coefficient is the one that correctly returns the mean or median flux

(7)

where the subscript b refers to values computed with the bulk algorithm.

The accuracy of a set of transfer coefficients from a particular field program is extremely difficult to assess. It is clear that uncertainty in the coefficient is the combined result of the flux and bulk mean variable inaccuracies. Sampling uncertainty, sensor bias, frequency attenuations, platform flow distortion, and inadequate motion corrections all degrade the results (Fairall et al. 1996; 2000; McGillis et al. 2001). The actual computation of the transfer coefficient also relies on the application of various 2nd-order physical corrections (Fairall et al. 2003): choice of Ψ functions, a cool-skin correction SST, reduction in the water vapor pressure of seawater by salinity, use of a gustiness parameter, the dilution effect (Webb et al. 1980), or flow distortion corrections for mean wind speed (Yelland et al. 1998). High-quality estimates of the scalar transfer coefficients require conditions where is large. This both improves the signal-to-noise ratio of the flux observation and reduces the measurement fractional error in .

1. Turbulent Flux Parameterizations

Turbulent fluxes may be estimated from specifications (observations) of the basic mean bulk variables in [1] by specifying the height and stability-dependent transfer coefficients, Cx, or the roughness lengths, zox. The momentum (drag) coefficient is known to vary significantly with mean wind speed while the scalar coefficients have weak wind speed dependence. The computation of fluxes should also account for the 2nd-order effects mentioned above, but these are often ignored or assumed to be imbedded in the transfer coefficient. The observing technologies have advance sufficiently in recent years that ignoring the 2nd-order effects makes a noticeable difference. The drag coefficient is often represented as a simple wind speed dependent formula,

(8)

or the velocity roughness length is represented as a Charnock plus smooth flow form (Smith 1988)

(9)

Where u* is the friction velocity, ν the kinematic viscosity of air, α the Charnock parameter, and β the smooth flow parameter. The sensible heat and latent heat transfer coefficients may be given as a constant or the roughness length parameterized by the velocity roughness length Reynolds number,

(10)

A bewildering variety of algorithms and approaches are available (see Brunke et al., 2003 for twelve examples). Brunke et al. (2003) compared the algorithms to a ship-based data set with about 7200 hours of observations and found mean bias magnitudes of 1 to 10 E-3 out of 65 E-3 Ntm-2 for stress and 0.5 to 20 out of 102 Wm-2 for the sum of latent and sensible heat flux. In Fig. 2 we show the wind speed dependence of momentum and moisture transfer coefficients for five sample algorithms; mean estimates from the database used by Brunke et al. (2003) are included. This same basic approach, with a few modifications to account for the frozen surface, can be applied to turbulent flux parameterization over ice (see Brunke et al. 2008 for examples from four climate and weather models).