Reply to comment by X. Zhang on “Measurements of air-sea gas exchange at high wind speeds in the Southern Ocean: Implications for global parameterizations”

Authors


1. Introduction

[1] We thank Zhang [2007] for recognizing the significance and importance of our work. In his comment, he questioned whether we have correctly assessed the wind speed/gas transfer velocity relationship derived from the NZ SOLAS Air-Sea Gas Exchange (SAGE) experiment [Ho et al., 2006]. Specifically, he assumes that we only analyzed the gas transfer velocity data from SAGE using a quadratic wind speed enhancement and not a cubic wind speed enhancement. In fact, we concluded that the quadratic fits of Wanninkhof [1992] and Nightingale et al. [2000] were more consistent with the SAGE data than the cubic relationship of Wanninkhof and McGillis [1999], on the basis of statistical comparisons using corresponding quadratic and cubic wind speed enhancements. Also, he has misinterpreted a sentence in the abstract of Ho et al. [2006]: “The results clearly reveal a quadratic relationship between wind speed and gas transfer velocity, rather than a recently proposed cubic relationship” to mean that the data derived from the SAGE Experiment excludes all cubic relationships, when in fact, we were merely assessing existing wind speed/gas exchange relationships. In the following, we respond to the comment by Zhang [2007] and address both of his misunderstandings.

2. Wind Speed Enhancement

[2] The assumption by Zhang [2007] that we used a quadratic wind speed enhancement when we conclude that the quadratic fits of Wanninkhof [1992] and Nightingale et al. [2000] were better than the cubic relationship of Wanninkhof and McGillis [1999] is incorrect. While Ho et al. [2006] did not present cubic enhancement–corrected gas transfer velocity data because of space constraints, a cubic enhancement correction was made when comparing the SAGE data to the cubic relationship of Wanninkhof and McGillis [1999]. The enhancement corrections were made following Wanninkhof et al. [2004, equation (3)], which was referenced. We compared the various published wind speed/gas exchange relationships to the SAGE data based on the χ2 values of the respective fits, and the χ2 values for the quadratic fits of Wanninkhof [1992] (χ2 = 1354) and Nightingale et al. [2000] (χ2 = 666) were a factor of 5 and 10 smaller, respectively, than that for the Wanninkhof and McGillis [1999] (χ2 = 7059) cubic relationship. While space constraints did not permit us to show all the details, a more comprehensive discussion is in preparation for publication elsewhere as stated by Ho et al. [2006].

[3] Furthermore, a significant portion of Zhang [2007] deals with deriving an approximation for the cubic wind speed enhancement factor. This is an interesting exercise that would be of benefit where wind speed data are not available. However, it is not necessary for SAGE and other gas exchange experiments where direct time series measurements of wind speed are obtained.

3. Cubic Relationships

[4] Zhang [2007] also misinterprets a sentence in the abstract by Ho et al. [2006]: “The results clearly reveal a quadratic relationship between wind speed and gas transfer velocity, rather than a recently proposed cubic relationship” to mean that the data derived from SAGE excludes all cubic relationships. Ho et al. [2006] compared the SAGE data to existing relationships between wind speed and air-sea gas exchange and stated that the SAGE data are inconsistent with the recently proposed cubic relationship of Wanninkhof and McGillis [1999]. This is clearly apparent here [Ho et al., 2006, p. 3]:

“A comparison of the SAGE data with existing parameterizations between wind speed and gas exchange indicate that gas transfer velocities at high wind speeds are significantly lower than predicted by the cubic parameterization of Wanninkhof and McGillis [1999] but higher than the piecewise linear one of Liss and Merlivat [1986]. The SAGE 3He/SF6 data are best described by the Nightingale et al. [2000b, hereinafter referred to as N2000] and Wanninkhof [1992, hereinafter referred to as W92] relationships (Figure 2).”

[5] Nightingale et al. [2000] and Wanninkhof [1992] are both quadratic relationships.

4. Quadratic and Cubic Fits to the SAGE Data

[6] Although Ho et al. [2006] was less concerned with whether a quadratic or cubic relationship best fit the SAGE data, this is a central point in the comment by Zhang [2007]. The gas transfer velocities k(600) of Ho et al. [2006, Table 1] are slightly different than those of Ho et al. [2006, Figure 2] because Ho et al. [2006, Table 1] was not updated during the manuscript revision. The correct data are shown here in Table 1. The analysis by Ho et al. [2006] was made using the correct k(600) data shown in Ho et al. [2006, Figure 2], and the discrepancy in k(600) between Table 1 and Figure 2 of Ho et al. [2006] was not the source of the misinterpretation by Zhang [2007].

Table 1. Summary of Wind Speeds and k(600) Derived From 3He/SF6 During SAGE
Wind Speed (u10), m s−1equation image/equation imageGas Transfer Velocities k(600), cm h−1QuikSCAT Enhancement-Corrected k(600), cm h−1Error k(600), cm h−1HoursaMeasurementsb
Ship-BasedQuikSCATShip-BasedQuikSCATShip-BasedQuikSCAT
  • a

    Number of hours covered by the 3 stations used to calculate k(600).

  • b

    Number of measurements used to calculate the mean wind speed in each interval.

7.57.41.131.1125.623.10.52327684
9.88.91.211.1415.313.410.124232114
10.59.01.091.1535.931.29.024195114
10.77.71.221.2419.115.410.424248105
11.011.31.161.1033.330.317.52224886
11.411.51.041.0329.528.615.724232106
11.510.41.021.0530.629.24.72423492
12.48.71.081.2531.925.55.734358152
12.910.81.021.0438.537.06.625234102
15.315.51.051.0352.450.93.450458206
16.116.01.031.0380.878.44.359574246

[7] Zhang [2007] argues that if nonzero intercepts were used, then there is no statistical difference between the quality of a quadratic or a cubic fit to the SAGE data. We can confirm this, as we find a2 = 0.242 (χ2 = 587, r2 = 0.822) and a3 = 0.013 (χ2 = 562, r2 = 0.832), with b0 = 3.8 and 9.7 for the quadratic and cubic fits, respectively (Figure 1).

Figure 1.

Quadratic and cubic fits to gas transfer velocity data from SAGE. The data have been corrected for quadratic and cubic wind speed enhancement. The solid lines are fits with a zero intercept, and the dotted lines are fits with a nonzero intercept. The dashed line is the Wanninkhof and McGillis [1999] relationship.

[8] However, the analysis of Zhang [2007] was based solely on statistical methods to determine the quality of fit to the SAGE data, without considering real world applications and constraints. Although it has similar fitting statistics to an unconstrained quadratic fit, an unconstrained cubic fit to the SAGE data might not represent the relationship between wind speed and gas exchange in the real world. While the relationship that Zhang [2007] is advocating shows that k(600) is ca. 10 cm h−1 at u10 = 0 [see Zhang, 2007, Figure 1], existing data suggest that a gas transfer velocity this high is not reached until u10 > 5 m s−1. A zero intercept has been used for all previous wind speed/gas exchange parameterizations because for global CO2 update calculations, it makes very little difference whether the intercept at u10 = 0 was zero or slightly above zero (although it does create significant regional differences). If zero intercepts were used for the fits to the SAGE data, then a quadratic fit to the quadratic wind speed enhancement corrected data (a2 = 0.266, χ2 = 622, r2 = 0.811) is a slightly better fit than a cubic fit to the cubic wind speed enhancement corrected data (a3 = 0.017, χ2 = 979, r2 = 0.707).

[9] It is important to note that the cubic fits described above are substantially different from that proposed by Wanninkhof and McGillis [1999] (Figure 1). The analysis that we have included here, and the analysis by Zhang [2007], both show that the SAGE data are inconsistent with the cubic relationship proposed by Wanninkhof and McGillis [1999]. This is important because the Wanninkhof and McGillis [1999] parameterization has been used to generate alternate budgets of the global ocean carbon uptake, which yields a significant and important difference from other frequently used parameterizations and has significant implications for prediction of future climate change. Using the delta pCO2 disequilibrium method to calculate global mean ocean CO2, the unconstrained cubic fit described above would yield an uptake for 1995 of 1.0 pgC y−1, while Wanninkhof and McGillis [1999] would yield an uptake of 1.9 pgC y−1.

5. Future

[10] These frequently used wind speed/gas exchange parameterizations [e.g., Wanninkhof, 1992; Nightingale et al., 2000] are semiempirical because the function forms are determined a priori. The main contribution of Ho et al. [2006] is that by obtaining measurements of gas transfer velocities from a range of wind speeds from the open ocean, and especially the first measurements at high wind speeds (>15 m s−1), these parameterizations can now be evaluated. Zhang [2007] has misinterpreted aspects of Ho et al. [2006], but we are in agreement that the field of air-sea gas exchange study needs more measurements of k(600) from various parts of the open ocean so that we can further evaluate these existing wind speed/gas exchange parameterizations and so that we can determine whether a universal relationship exists between wind speed and air-sea gas exchange.

Acknowledgments

[11] This is LDEO contribution 6921a.

Ancillary