the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A Novel Model Hierarchy Isolates the Effect of Temperature-dependent Cloud Optics on Infrared Radiation
Abstract. Clouds exert strong influences on surface energy budgets and climate projections. Yet, cloud physics is complex and often incompletely represented in models. For example, temperature-dependent cloud optics parameterizations are rarely incorporated into the radiative transfer models used for future climate projections. Prior work has shown that incorporating these optics in downwelling longwave radiation calculations results in increases of as much as 1.7 W m−2 for Arctic atmospheres. Here we examine whether implementing these optics in climate models leads to significant climate impacts. We use a novel methodology based on a hierarchy of models. In two-stream radiation and single-column models, incorporating temperature-dependent optical properties had a small impact (< 1 W m−2). Similarly, impacts were statistically insignificant on infrared radiation within freely evolving atmospheric model simulations. In contrast, there was a much larger effect (1–3 W m−2) from optics changes when the winds within our atmospheric model experiments were nudged towards reanalysis winds. This new application of wind-nudging experiments helped to isolate the effect from temperature-dependent cloud optics changes by reducing the internally generated atmospheric variability. In summary, we found a signal from temperature-dependent optics, but this effect is small compared to climate variability and didn't impact long term Arctic temperature trends. More broadly, this work demonstrates a new framework for assessing the climate importance of a physics change.
- Preprint
(12550 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 12 Jan 2025)
-
RC1: 'Comment on egusphere-2024-2043', Anonymous Referee #1, 04 Oct 2024
reply
This manuscript seeks to understand how a change in the specification of the optical properties of liquid clouds might affect the simulation of Arctic climate. The change is motivated on physical grounds - the index of refraction of water is temperature dependent but this dependence is usually ignored when mapping cloud physical to cloud optical properties in broadband codes. Mie calculations are performed for drop size distributions consistent with the CESM2 climate model using indexes of refraction valid at 273, 263, and 240K; these are used in an off-line radiative transfer model across limited spectral ranges in the infrared, as well as in single-column simulations, uncoupled and coupled freely-running ensembles, and in an ensemble in which winds are nudged towards reanalysis at high latitudes. Small differences in spectrally-resolved fluxes in the offline simulations; differences in integrated fluxes are lost in the variability in single-column and free-running model simulations, becoming large enough to be distinguished from noise, if still small, in the nudged simulations.
The manuscript has two goals: 1) to assess the possible impact of an elaboration of cloud optics on simulations in the Arctic, and 2) to develop methods for such an assessment. Both goals might be reached more effectively by revisiting the experimental design to more clearly delineate the perturbation that might be expected from such a change from any subsequent impact on simulations.
The motivation for the study is well-grounded: the index of refraction of liquid water does indeed depend on temperature, so the degree to which fluxes might be systematically biased in some circumstances is not know a priori. What can be anticipated, however, is that the impact on fluxes will be restricted to thin clouds, since (band-wise) fluxes will only change when (band-wise) optical thickness is in the range of roughly 0.5 - 3 i.e. where the clouds are neither optically thin or thick. (That this is not illustrated using the two-stream model is a missed opportunity.) The impact of changes in the index of refraction thus depends on the population of clouds and atmospheric states. The magnitude of this change for a given population of clouds, such as those produced by a particular climate model, could be evaluated with off-line broadband radiation calculations.
The title and framing of the manuscript is misleading: tests in the dynamical models do not use temperature-dependent cloud optics; rather they replace cloud optics computed with the index of refraction used at a single temperature with optics computed at a different temperature. Whether accounting for the temperature dependence of cloud optical properties would impact fluxes and/or other simulation characteristics can not be assessed with the current information. The authors assert that the computational cost of implementing cloud optics that depend on temperature would be “immense” (lines 282-287) but this is unlikely to be true: cloud optics are usually a tiny portion of the time spent in radiation calculations.
Interpretation
The motivation for the “model hierarchy” is not made clear. The answers to the questions on lines 74-83 are tautological, i.e. the single column model is motivated by asking what the impact is at a given location during a finite time frame. Linking each set of simulations to a testable hypothesis will help readers make sense of results.
Parametric sensitivity studies, as in Rowe et al. 2013 and as might be done with the spectrally-resolved model are useful in motivating the work. A missing step is broadband calculations analogous to those used in the global model simulations - say, offline calculations with RRTMG over the distribution of Arctic clouds produced by CESM - to understand how those parametric sensitivities convolve with the population of Arctic clouds in the model to be examined and whether one might expect systemic differences in interactive simulations. It is unclear what is gained from the wind-nudged simulations, which are motivated by trying to constrain internal variability, that wouldn’t emerge more clearly from calculations applying changed optics to the clouds produced by the unperturbed model.
Single-column model simulations can be expected to diverge somewhat in response to even tiny changes, making the interpretation of changes on particular days in a long simulation ambiguous. (Did the authors consider doing ensembles of single-column model simulations to see if the cloud optics change can be teased out?).
What is the motivation for simulations with global models? Such simulations are useful when scales interact - here, if the change in cloud optics might be expected to systematically impact interactions between the Arctic and rest of the world. Is that expected? If not why wouldn’t assessment with regional model be more informative?
The claim (repeated seven times) that the approach represents a “novel model hierarchy” is not well-founded. “Model hierarchy” refers to sets of equations representing the same underlying system with different levels of complexity. It’s a stretch to call a configuration in which winds are relaxed to time-varying empirical values a separate element and the idealized radiative transfer calculations are clearly a different beast. As the authors note the use of wind nudging is not novel. The work can stand on its own without claims to greater generality than are supported.
Some more minor comments
Most figures could be more carefully crafted to emphasize the narrative points being made. Some figures (2, 12) illustrate concepts that emerge clearly from the text. Others contain information that’s visually hard to parse. Figure 3, for examples, requires readers to mentally subtract lines from two different panels, in addition to showing variations with respect to two related by hard-to-interpret quantities, while the information density of figure 4 is low. The authors might fruitfully review each figure and refine those that do not advance the story being told.
The captions of Figure 7 and later note that statistical significance is assessed “following Wilks (2016)” but the text provides no elaboration. Is significance computed accounting for false discovery rate? If so this should be noted more clearly in the main text.
The authors are quite free with advice to others (e.g. line 238, line 249, line 279). This may be worth revisiting given the nuanced results obtained.
The formulation of the offline radiative transfer model in appendix A is confusing. The offline model is used to compute longwave fluxes. Liquid clouds do not scatter longwave radiation so it’s not clear why one would use two-stream equations representing multiple scattering (A7-A11). It would be far simpler to use Schwartzchild’s equation, potentially accounting for intra-layer temperature gradients as in section 2.1 of Clough et al. 1992 (doi:10.1029/92JD01419). Indeed that’s what models like RRTMG do.
The equations in both appendicies are well-known and the tables are available in the original literature. Since neither sheds light on the problem at hand they can be safely omitted.
Citation: https://doi.org/10.5194/egusphere-2024-2043-RC1
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
242 | 63 | 69 | 374 | 11 | 9 |
- HTML: 242
- PDF: 63
- XML: 69
- Total: 374
- BibTeX: 11
- EndNote: 9
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1