the latent flux and to 20 W m−2 or less for the sensible flux. The density of observations is relatively high over the extratropics of the North Pacific and North Atlantic, with many 5°squares having 100 or more observations per month. However, the density is marginal-to-poor over the tropics and in the Southern Hemisphere, with fewer than 20 observations per month at many 5° squares. In these sparsely sampled regions, random errors in the monthly flux estimates are likely to be substantial. Consequently, in forcing the OGCM in the region between 20°N and 20°S and data-poor regions south of 20°S, we have resorted to a Newtonian damping scheme for the net flux anomalies.

Systematic errors are caused by biases in the bulk formulae and in the observations, and are not automatically reduced by averaging. Time-varying biases in the fundamental observations caused by changes in instrumental practices are probably a significant source of error. These errors are difficult to quantify, but the marine data have "global" (over the available COADS-covered region) trends in w and ΔT, which translate to changes of approximately 0.8 m s−1 and −0.2°C over 1950 to 1986. Because these seem to be at least partially caused by instrument and calibration changes over the period, the fluxes were adjusted to remove the effects of the global average linear trends. This adjustment reduced the linear change in the latent and sensible fluxes over 1950 to 1986 by about 20 W m−2 and 5 W m−2, respectively (DC1 and DC2). This adjustment has only minor influence on the high-frequency variability, but it does significantly affect the multi-year numerical model run, because the effects are cumulative. More discussion of the systematic errors in the marine data is provided below.

Anomalous Variability of the Heat-Flux Components

The dependence of the flux anomalies on the mean conditions and the anomalies of the fundamental surface variables was explored in DC1. For the monthly latent flux anomaly

where an overmark indicates the long-term monthly mean and a prime indicates the monthly anomaly. For monthly means, most of the variation comes from w and Δq, not from the exchange coefficients (DC1). The last term on the right-hand side of (3) does not contribute strongly to latent flux anomalies, because w and Δq monthly anomalies are not locally well correlated over most of the oceans (DC1). A similar relationship involving ΔT instead of Δq describes the monthly sensible flux anomalies.

Equation (3) shows that the joint behavior of mean values and anomalies of wind speed w, the ocean-surface saturation humidity-air humidity difference Δq, and the sea-air temperature difference ΔT must be included to determine the flux anomalies. For atmosphere-ocean models, this interplay between the mean and anomaly components is a challenge. It is not enough to simulate anomalies of w, Δq, and ΔT; their mean fields must also be properly represented.

There are strong seasonal and geographical modulations of the flux anomalies. In the extratropics the variance is greatest in fall and winter when w, Δq, and ΔT have largest mean values and strongest variability. The largest contributions to the extratropical variance of Q1 anomalies involve . Anomalies of Qs become important north of about 35°N, where their variance is dominated by the term involving . In the extratropics, the variance of each of these components is maximum along the western side of the basin, because strong ocean boundary currents and large air-mass differences between the upwind land mass and the sea cause great contrasts. Latent flux anomalies in the tropics often involve greater contributions from than do those in mid-latitudes, although the eastern tropical Pacific has relatively large contributions from .

The Δq and ΔT monthly anomalies are correlated (cooler air is usually drier), especially in the extratropics, so the latent and sensible anomalies usually reinforce each other, yielding relatively large net heating anomalies. In the extratropics, flux anomalies are linked to the local wind direction, but in the tropics they are usually not (DC2). North of about 15°N, the largest positive anomalies are associated with northerly to northwesterly winds; they probably result from meridional advection of atmospheric humidity and temperature, and also from changes in the strength of the westerly winds. In the tropics, there is little relationship between wind direction and the flux anomalies, since horizontal gradients of humidity and temperature are weak and the wind direction is relatively steady.

How does the variability of the anomalous latent and sensible fluxes compare with that of the radiative fluxes? The balance of the heat-flux anomaly terms is distinctly different from that of the mean terms. Combined anomalies of Ql and Qs often exceed 50 W m−2 over regions several hundred kilometers in extent (DC1 , DC2, and DC3). These estimates indicate that monthly anomalies of Ql and Qs are usually larger than those of the radiative components. Comparing the variances of Ql and Qs, with those of bulk-formulae-estimated net solar radiative flux QSW and infrared flux QIR, DC1 found that: (1) Ql and Qs, anomalies dominate the monthly anomalous surface heat budget during winter in the extratropics north of 30°N, (2) Ql anomalies dominate from about 15°N to about 30°N, and (3) Ql and QSW anomalies are about equally large from 15°N to 15°S.

Flux Anomalies and Atmospheric Circulation

Ql and Qs anomalies have regional-to-basin-scale coherence both the North Atlantic and North Pacific. The first four empirical orthogonal functions (EOFs) of the sum of the Ql and Qs anomalies (Q'l+s) account for about half of the total variance of this field in each basin (DC1). That these



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement