Comments on “Examination of the Decadal Tropical Mean ERBS Nonscanner Radiation Data for the Iris Hypothesis”

Lin et al. (2004, hereafter LWWH) examined the Iris hypothesis of Lindzen et al. (2001, hereafter LCH) using the variations of the Earth Radiation Budget Satellite (ERBS) radiation fluxes at the top of the atmosphere (TOA) (Wielicki et al. 2002) and the sea surface temperature (SST) taken from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996) based primarily on satellite retrievals (Reynolds and Smith 1994). They applied the 3.5-box climate model of LCH to evaluate the Iris hypothesis using the tropical radiation parameters of both LCH and Lin et al. (2002). They concluded that there was no evidence for the strong negative feedback of the Iris effect and that the strong variation of the high-level clouds with the SST (taken from LCH’s observations of the Iris effect) was likely a major factor for causing the deviation between the predicted and observed shortwave (SW) and longwave (LW) variations at TOA. We find that the conclusions of LWWH are contradictory to the data given in their tables and figures, which actually suggest a strong negative feedback between TOA radiation and the surface temperature, consistent with the Iris hypothesis.

Two conclusions are given in the abstract of LWWH. One is that “the predicted tropical mean radiative flux anomalies are generally significantly different from those of the ERBS measurements, suggesting that the decadal ERBS nonscanner radiative energy budget measurements do not support the strong negative feedback of the Iris effect.” Throughout the work of LWWH, only anomalous temperature and radiative fluxes, relative to the 1985–89 values, are presented. The feedback between temperature and radiation is never defined. Without defining the feedback and explicitly computing the feedback from the temperature and radiation data, one cannot arbitrarily claim that the ERBS data do not show strong negative feedback of the Iris hypothesis. Actually, the data given in Table 2 of LWWH imply a strong negative feedback for all ERBS, Iris, and National Aeronautics and Space Administration (NASA) Langley Research Center (LaRC) cases.

As shown in Fig. 7 of LCH, the nonfeedback response of temperature ΔT_o to an external forcing ΔQ can be written as

where G_o is a nonfeedback gain. When a feedback process is present, an additional forcing proportional to the response ΔT is produced. This forcing is written as FΔT and is added to the external forcing. The response is now

and

where f = G_o F is the feedback factor. The feedback is positive for 0 < f < 1 and negative for f < 0.

For the Tropics as a whole, the mean outgoing longwave radiation (OLR) taken from the Earth Radiation Budget Experiment (ERBE) (Barkstrom 1984) is ∼255 W m⁻², which is equivalent to an effective emitting temperature of ∼259 K. The gain without water vapor and cloud feedbacks is the inverse of the derivative of the Planck function with respect to the temperature at 259 K,

where B is the Planck flux integrated over the entire IR spectrum. From Table 2 of LWWH, the change of the net radiation at TOA is −0.65 W m⁻² for ERBS. Assuming that this change of the net TOA radiation is because of the change in the SST, we have

With the change of the surface temperature ΔT by 0.144 K from the 1980s to the 1990s, we have F = −4.5 W m⁻² K⁻¹, f = G_o F = −1.1, and

The feedback factor of −1.1 is exactly the same as the case γ = 1 of LCH. For the Iris and LaRC cases shown in Table 2 of LWWH, the factor f ranges from −0.75 to −2.9, and ΔT/ΔT_o ranges from 0.26 to 0.57. Thus, the data shown in LWWH imply a strong negative feedback for all the ERBS, Iris, and LaRC cases. This is contrary to the claim by LWWH that the decadal ERBS data do not support the strong negative feedback of the Iris effect.

In the above analysis using the tropical mean radiation and SST, we assume that the heat transport out of the Tropics does not affect the feedbacks. We do the simple calculations just to demonstrate that the data provided by LWWH cannot be used to claim that there is no evidence of strong negative climate feedback. Indeed, in considering the impact of the tropical Iris effect on global climate, LCH did consider the sharing of heat with the extratropics, and such sharing does reduce the global feedback. However, it is not clear that such sharing must be included when considering observed interdecadal changes where such sharing is automatically present. Even if it were, the overall feedback would remain large. In the above analysis we also do not take into consideration the decadal change of ocean heat storage in evaluating the climate feedback. One must distinguish the difference between the Iris feedback (which occurs on fast time scales and is dependent only on the SST), the resulting climate feedback, and the calculation of equilibrium climates. Only the last depends significantly on the take up of heat by the ocean.

The other conclusion, as stated in the abstract of LWWH, is that “Poor agreements between the satellite and the model predictions even when the tropical radiative properties from CERES observations (LaRC parameters) are used imply that besides the Iris-modeled tropical radiative properties, the unrealistic variations of tropical high cloud generated from the detrainment of deep convection with SST assumed by the Iris hypothesis are likely to be another major factor for causing the deviation between the predictions and observation.” To argue the fact that neither LCH nor LaRC matches the data means the assumed relation between cloud cover and SST is incorrect is totally illogical, given the plethora of alternative hypothesis. In point of fact, the data given in Table 2 of LWWH suggest that the negative relation between cloud cover and SST of the Iris hypothesis is correct, and that the magnitude of the negative slope (−0.22 K⁻¹) is, in fact, not strong enough if the ERBS and the SST data presented by LWWH are correct. Justifications of these arguments are given below.

The radiative fluxes given in Fig. 2 of LWWH are the basis for their conclusion that the ERBS data do not support the strong negative feedback of the Iris hypothesis. The table shows the anomalous tropical mean radiative flux during the 1994–97 period, relative to the mean values of the period 1985–89, for ERBS, Iris, and LaRC. With an anomalous SST of +0.144 K, the positive values of anomalous LW and SW, and the negative values of the anomalous net radiation, are consistent with the Iris hypothesis. For the Iris hypothesis, the precipitation efficiency increases as SST increases. The enhanced precipitation reduces the amount of cloud condensate reaching the upper troposphere, resulting in reduced detrainment, and, hence, the reduced production of cirrus clouds and a drier upper troposphere. The reduced high-level clouds and upper-tropospheric water vapor cause both the outgoing LW radiation and the absorption of SW radiation to increase. Corresponding to an increase of 0.144 K in SST, the ERBS data show that the Tropics emit extra LW radiation of 3.05 W m⁻² and absorb extra SW radiation of 2.40 W m⁻². The net is a loss of radiative energy by 0.65 W m⁻². The signs of these numbers are all consistent with the Iris hypothesis, that is, a negative climate feedback.

Compared to the anomalous LW of the ERBS (3.05 W m⁻²), the predicted anomalous LW of Iris and LaRC (0.9–2.1 W m⁻²) is much smaller by a factor of 1.5–3. The discrepancy is even larger for the anomalous SW; the predicted anomalous SW is smaller than the ERBS by a factor of 3–8. To be consistent with ERBS, the Iris sensitivity of cloud cover to SST of −0.22 K⁻¹ needs to actually be greatly enhanced. Because there are no data to support this, it is reasonable to question the suitability of the ERBS and SST data used by LWWH.

The ERBS SW and LW radiative fluxes are derived from nonscanner wide field-of-view measurements (Wielicki et al. 2002). The spatial resolution of the ERBS data archive is 10° latitude × 10° longitude, and the temporal resolution is 1 month. Because of the low spatial resolution, mean SW and LW fluxes cannot be computed separately over land and ocean. On the other hand, satellite retrievals of surface temperature are only available for the oceanic region. Thus, it is not feasible to relate the variation of the tropical averaged ERBS radiative fluxes to that of the surface temperature in a meaningful manner. Regardless of this limitation, LWWH used the ERBS radiative fluxes and the SST averaged over the Tropics to investigate the Iris hypothesis. The effect of land temperature is totally ignored. The LW, SW, and net radiation anomalies of Iris and LaRC, as shown in Figs. 2–4 and Table 2 of LWWH, are computed using only the mean tropical SST. The land temperature is not included. On the other hand, the anomalous ERBS radiation is the mean value averaged over both oceanic and land regions. Therefore, the comparisons of the ERBS anomaly with the Iris- and LaRC-predicted anomalies are inappropriate. To address this problem, LWWH stated that the tropical mean ERBS anomalies over oceans are very similar to those calculated over all scene types. This statement is irrelevant to the question of how the results might be altered if both land and ocean are included in computing the tropical mean surface temperature. It might well be the case that the surface temperature variation shown in Fig. 1 of LWWH would change significantly if the land temperature were included.

Referring to the numbers of Table 2, LWWH stated that “Because of the exaggerated and dominant LW effect in the Iris hypothesis, the predicted net incoming anomaly, that is, SW anomaly − LW anomaly, from the hypothesis (Fig. 3b) has absolute values that are quantitatively 60% ∼ 160% larger than the observed ERBS measurements during the 1990s (Table 2).” This statement is contradictory to what is shown in the table. The Iris LW effect of 1.434–2.066 W m⁻² is significantly smaller than the ERBS effect of 3.05 W m⁻². Thus, the Iris LW effect is underestimated instead of exaggerated. By comparison, the LaRC LW effect of 0.887–1.424 W m⁻² is an even greater underestimate. LWWH further stated that “During the 1990s, the averaged net flux anomalies from the LaRC parameters are quantitatively consistent with ERBS measurements with differences about 0.1 ∼ 0.2 W m⁻² (Table 2) and have a much better agreement with observations than those from the Iris hypothesis.” This statement is not only misleading but is also incorrect. The small difference in the net effect between ERBS and LaRC is a result of the cancellation of a large negative LW discrepancy and a large positive SW discrepancy. Judging from the discrepancies in the LW and SW anomalies, it is obvious that the Iris parameters, when compared with the LaRC parameters, have a much better agreement with those of ERBS.

Finally, LWWH claimed that “the results represent the near-balance of the radiative energy over the Tropics during the decades observed by ERBS and the weak feedback physics of the relatively small net radiation differences among cloudy moist (high clouds), clear moist, and dry regions as observed by CERES.” This statement is hard to comprehend. Why is the radiative energy over the Tropics in near balance just because the difference in the net radiation between the two decades (1980s and 1990s) is small (0.65 W m⁻²)? What is the meaning of the near-balance of the radiative energy? LWWH failed to explain why the small difference in the net radiation implied weak feedback physics. On the contrary, the data presented in LWWH imply a strong negative feedback as we have already shown.

LWWH cited Hartmann and Michelsen (2002) and claimed that the strong negative relationship between the high cloud amount and SST obtained by LCH was because of the use of a faulty data analysis. They also cited Fu et al. (2002) and Lin et al. (2002) and claimed that the strong negative feedback of the Iris hypothesis is a result of the incorrect specification of the cloud and radiative parameters in the Tropics. These criticisms are, we feel, incorrect, as we showed in Lindzen et al. (2002) and Chou et al. (2002a, b). The failure of LWWH to even cite our responses is unacceptable as a matter of normal scientific practice. Even if they hold the criticisms to be correct and the responses to be wrong, the readers should be supplied with the information to decide for themselves.

In the LCH 3.5-box model for studying the Iris effect, the Tropics are divided into dry and moist regions, each covering half of the Tropics. The moist region is further separated into a region covered with high-level clouds (cloudy moist) and a region without cirrus clouds (clear moist). The former is assumed to cover 22% of the Tropics and the latter covers 28%. The low-level boundary layer clouds are assumed to cover 50% of the Tropics. In specifying the areal coverage, the SW albedo, and the LW emission of these tropical regions, it is required that the model radiation budgets at the top of the atmosphere are consistent with the overall ERBE radiation budgets.

To examine the validity of the albedo and the LW emission (or effective emission temperature) specified by LCH, Lin et al. (2002) derived the albedo and the LW emission of the dry region by averaging the OLR and albedo of 50% of the total Tropical Rainfall Measuring Mission (TRMM) Cloud and the Earth’s Radiant Energy System (CERES) pixels in the Tropics that have the largest OLR. Following LCH, they used a threshold temperature T₁₁ of 260 K, measured in the TRMM Visible and Infrared Scanner (VIRS) 11-μm channel, to infer high-level clouds. The mean albedo and OLR of those high-cloud pixels were then taken as the albedo and the effective emission of the cloudy moist region. The mean albedo and LW emission of the rest of the pixels are then assumed to be the albedo and LW emission of the clear moist region. The approach of Lin et al. (2002) to specifying the albedo and LW emission of the cloudy moist region was disputed by Chou et al. (2002a). LCH used the threshold temperature of 260 K to merely serve as a surrogate for the extent of the detrainment of cumulus anvil clouds. It is not meant to be the total coverage of high-level clouds, which include thin cirrus clouds. With the cloud identified with T₁₁ < 260 K, it can be expected that these clouds are mostly high and thick. The albedo of 0.51 and the cloud cover of 0.1 of the cloudy moist region as shown in Table 2 of LWWH (LaRC parameters) are representative only for high and thick clouds. As stated in Chou et al. (2002a), the mean albedo of the high-level clouds should be smaller than 0.5 and the areal coverage should be greater than 0.1 if the thin cirrus clouds are taken into consideration.

Finally, LWWH attribute the observed changes in radiation to the intensification of the tropical circulation, essentially following Chen et al. (2002). LWWH further quote the work of Del Genio and Kovari (2002) that claimed inconsistency between the TRMM data and the Iris hypothesis. Both the attribution and the claim of Del Genio and Kovari (2002) are inappropriate to the evaluation of the Iris effect. Both Chen et al. (2002) and Del Genio and Kovari (2002) fail to distinguish changes in cloud cover because of increased or decreased cumulus convection from changes that are a result of increased or decreased detrainment from cumulus towers. As emphasized by LCH, only the latter is relevant to the Iris effect. The failure to distinguish these changes can even result in getting the opposite (and totally fallacious) sign of the Iris effect. The point is that circulation acts primarily to organize convection that is fundamentally a result of evaporation. Thus, changes in circulation may act to concentrate convection in one region, while suppressing it in another. However, integrated over the Tropics, the effect of circulation drops out. The most appropriate measure of convection is the cumulus mass flux M_c. Averaged over the Tropics, M_c = E/q, where E the evaporation is generally taken to be proportional to q_s − q, where q_s is the saturation humidity, while q = rh × q_s is the specific humidity near the surface, and rh is the relative humidity. We, thus, see that the average M_c = (1 − rh)/rh; that is, it depends only on rh. As long as rh does not change with changes in mean temperature, then M_c remains unchanged as well. However, if rh increases, M_c should actually decrease. In LCH, the Iris effect referred to the change of detrainment per unit M_c. For simplicity, LCH took M_c to be proportional to the area for which T₁₁ was less than 220 K. Del Genio and Kovari (2002) used TRMM and NCEP data for a period of only 5 days (1–5 February 1998) to study the relation between clouds, precipitation, and SST. They found that in the equatorial band from 15°S to 15°N, both cloud cover and anvil cloud thickness increase with SST. They then concluded that the results are inconsistent with the adaptive Iris hypothesis and suggested a positive feedback between cloud amount (in both horizontal and vertical extent) and SST. In analyzing the TRMM and NCEP data, they grouped clouds, precipitation, and SST data in the entire equatorial band into discrete “bins” and found that both cloud cover and thickness increase with increasing SST (their Fig. 14). The range of the tropical SST is 26°–30°C. The cold oceanic regions are representative of the tropical eastern Pacific and Atlantic Oceans and the western Indian Ocean, whereas the warm oceanic regions are typical of the tropical western Pacific and Atlantic Oceans and the tropical eastern Indian Ocean. Figure 14 of Del Genio and Kovari (2002) merely shows that cumulus convection is concentrated in the warm regions and suppressed in the cold regions. Thus, it is not surprising that cloud cover and cloud thickness are greater, for example, in the Pacific warm pool, where SST is in the range of 28°–30°C, than that in the tropical eastern Pacific, where SST is in the range of 26°–27°C. This is analogous to what LCH show in their Fig. 6a where they relate cloud cover to SST for the region of 0°–15°S. There is increasing cloud cover with increasing SST because the higher SSTs correspond to the Southern Hemispheric (SH) summer when convection is concentrated in this region, while lower SSTs correspond to SH winter when convection is reduced. However, when these results are normalized by a measure of cumulus activity, one sees in LCH’s Fig. 6c that the cloud cover per unit cumulus decreases with increasing SST, as one expects from the Iris effect. Thus, the unnormalized results shown in Fig. 14 of Del Genio and Kovari (2002) are irrelevant to the feedback between cloud and SST, and, hence, the Iris hypothesis. Their conclusion that there is no observational evidence supporting the strong negative climate feedback is unfounded.

Given that dynamics cannot, on average, significantly change the amount of cumulonimbus activity, papers such as LWWH, Chen et al. (2002), and Wielicki et al. (2002), in fact, provide confirmation for the Iris hypothesis. For those concerned with global warming, this should be welcome news.