Lazaros Oreopoulos (NASA-GSFC) What new have we learned about cloud radiative effect and feedback in the last decade? Lazaros Oreopoulos (NASA-GSFC)

Effect vs feedback Cloud Radiative Effect (CRE) can be observed (TOA) by satellites CRE = F – Fclr To obtain CRE for SFC and ATM some modeling is often necessary Cloud Radiative Feedback is associated with ΔCRE ΔCRE = ΔF – ΔFclr and is much harder to observe and relate to ΔTs. We mostly use CGCMs to quantify it F is the net flux = down – up.

Cloud Radiative Effect (CRE)

Global CRE from various sources (no uncertainty bars) SW LW net TOA SFC Sources: ERBE: Harrison et al. (1990) ISCCP FD: Zhang et al. (2004) CERES EBAF: http://ceres.larc.nasa.gov/cmip5_data.php CloudSat/CALIPSO: Henderson et al. (2013). Scales have been expanded to highlight the differences. Note that according to CloudSat/CALIPSO clouds significantly warm the atmosphere! ATM

Vertical variability of CRE (expressed as cloud effect on HRs) DJF JJA LW Radiative heating rates were calculated from the R04 2B-FLXHR-LIDAR product [Henderson et al ., 2012; L’Ecuyer et al ., 2008], which uses the CloudSat microphysical retrievals and combined CloudSat/CALIPSO cloud mask as inputs to a broadband, two-stream, plane-parallel, adding and doubling radiative transfer model. The model produces upward and downward fluxes at 240m vertical increments. There is a crude correction to convert instanenous SW fluxes to diurnal means. The SW CRH is, in the annual mean, 2.1 times smaller than the LW CRH (note the differing scales). High clouds tend to reduce SW absorption by the underlying atmosphere, while low clouds tend to enhance it by increasing the geometric path length of photons, thus increasing the probability of absorption. The Cloudsat/CALIPSO-derived SWCRH is consistent with both these ideas: it is maximum in the tropical upper troposphere, where cirrus are ubiquitous, and also large in the subtropics and midlatitudes of the summer hemisphere, where low cloud occurs in the presence of a low summertime solar zenith angle. This zone of stronger SW CRH also extends into the lower troposphere, especially in the higher latitudes where low clouds are particularly prevalent. A maximum in LW CRH can be seen in the tropical lower and middle troposphere. As demonstrated in L08, this is a result of enhanced infrared cloud radiative warming by high-topped clouds, particularly cirrus that prevent LW radiation from escaping to space, while also radiating toward the surface. The resulting atmospheric heating is maximized in the lower and upper troposphere, but the tropical region near 550 hPa actually experiences cloud-related cooling. The more convectively active portions of the tropics, such as the West Pacific and Indian Ocean basins, have been observed to contain a peak in midlevel clouds, probably associated with a climatological mean stable layer near the height of the freezing level. The area of strong infrared cooling (and negative LW CRH, i.e., cloud radiative cooling) near 775 hPa is dominated by emission from cloud tops, especially over the SHEM where the maximum mean JJA cooling is 3.4K/d . Haynes et al . [2011] reported a 3 year, area-averaged cloud fraction from Cloudsat/CALIPSO of 0.81 for the region between 30 S and 65 S and showed that clouds with average tops near 2.5 km (or approximately 750 to 800 hPa) dominate this region. Low clouds are also common in the middle to high latitudes of the NHEM [e.g., Mace , 2010], but they cover a smaller fraction of the total area and therefore have a reduced radiative impact. It is not surprising, therefore, that there is a hemispheric asymmetry in radiative cooling at 775 hPa, with the SHEM midlatitudes to high latitudes undergoing radiative cooling during the entire year, while the corresponding latitudes in the NHEM cool during the boreal winter and warm during the boreal summer. Another feature is an area of reduced cooling in the low levels (enhanced LW CRH) apparent in both hemispheres in the vicinity of 925 hPa. This is a result of infrared warming from cloud bases and is largest when a cold boundary layer overlies a relatively warm surface. During the NHEM summer, the region of enhanced low-level LW CRH persists even where the overlying cloud radiative cooling is largely absent. This occurs because the clear-sky component of the cooling is also large, which we speculate is due to relatively high water vapor concentration in the upper-boundary layer during the NHEM summer. SW From Haynes et al. (2013)

Atmospheric LW CRE from CERES (Wm-2) (Courtesy of Norman Loeb)

Horizontal variability of (SW) CRE (bias when ignoring tau and re horizontal variations at 1 deg ) Biases for liquid are larger at midlats during their respective summers because of increased illumination. Ice cloud bias peaks re in the tropics and reflect the movement of the ITCZ. These are diurnal averages obtained by using a few assumptions. Horizontal variations are known from the joint histograms of tau and re. From Oreopoulos et al. (2009)

Overlap effects on CRE (from a GCM) SW LW Cloud are assumed inhomogeneous (beta distribution in each layer) in all experiments. Generalized overlap with fixed decorrelation lengths produces the largest cloud fraction (in general) and therefore the largest CREs. Barker has also studied CRE bias using CloudSat observations. top row: max-ran – generalized (fixed parameters) ~4 Wm-2 bottom row: generalized (CloudSat) – generalized (fixed parameters) ~1 Wm-2 Oreopoulos et al. (2012)

Decomposing CRE

Net TOA CRE from CERES Wm-2 (Courtesy of Norman Loeb)

Breakdown of CRE by cloud type Many ways to define cloud type Chen et al. 2000 Hartmann et al. 1992 Chen et al. (2000) used four days only (one for each season) to represent the entire year. Based on RT calculations. Hartmann et al. (1992) uses multi-variable regressions between ERBE CRE and ISCCP cloud fractions for each cloud type.

Spectral/cloud type breakdown of (LW) CRE (tropical oceans) H2O, 0-560 & > 1400 cm-1 CO2, 560-800 cm-1 O3, 800-980 cm-1 H2O continuum, 980-1100 cm-1 Cloud type (high, low middle) presence and impact on CRE can be inferred by normalized band CRE (fCRE). In this ratio he common factor of cloud fraction cancels and fCRE is only sensitive to cloud-top temperature, making it a useful quantity in diagnosing and evaluating modeled CRE. The largest values of fCRE [;(0.25–0.35)] for Band 1 (H2O) are found over the regions with frequent occurrence of high clouds, such as ITCZ, SPCZ, and Indian monsoon regions. The smallest values occur in regions frequently covered by marine stratus (i.e., low cloud), such as the Pacific coast off South America, the Namibia coast, and the ocean region west of Australia. For band 2 (CO2 band;), the contrast in fCRE between high-cloud and low-cloud regions (0.16 vs 0.11) is much smaller than that of band 1 (0.3 vs 0.05). The observed band 3 (H2O continuum) fCRE peaks over the low-cloud regions and drops over the high-cloud regions. The observed spatial map of band 4 fCRE (ozone band) is similar to that of band 3, owing to the fact that the ozone band is sensitive to the cloud and surface thermal contrast in a similar way as for neighboring bands. These maps have been compared with their counterparts from 3 GCMs. Huang et al. (2013)

Breakdown of CRE by“circulation regime” (Tropics, ERBE + ERA-40) LW and SW (x -1) CRE from ERBE (1985-89) composited by ERA-40 vertical velocity at 500 mb. Vertical bars indicate the seasonal deviation within each regime. This is for the tropics only (30 S to 30 N) A more continuous idealization of the tropical circulation was proposed by Bony et al. (2004). This uses the 500-hPa large-scale vertical velocity as a proxy for large-scale vertical motions of the atmosphere and decomposes the Hadley–Walker circulation as a series of dynamical regimes defined using . In the Tropics, nearly all of the upward motion associated with ensemble- average ascent occurs within cumulus clouds, and gentle subsidence occurs in between clouds. Since the rate of subsidence in between clouds is strongly constrained by the clear-sky radiative cooling and thus nearly invariant, an increase of the large-scale mean ascent corresponds, to first order, to an increase of the mass flux in cumulus clouds (Emanuel et al. 1994). Therefore, considering dynamical regimes defined from allows us to classify the tropical regions according to their convective activity, and to segregate in particular regimes of deep convection from regimes of shallow convection. The statistical weight of the different circulation regimes (Fig. 4) emphasizes the large portion of the Tropics associated with moderate sinking motions in the midtroposphere (such as found over the trade wind regions), and the comparatively smaller weight of extreme circulation regimes associated with the warm pool or with the regions of strongest sinking motion and static stability such as found at the eastern side of the ocean basins. These extreme regimes correspond to the tails of the probability distribution function. The atmospheric vertical structure (observed or modeled) can then be composited within each dynamical regime. Illustrations of the dependence of cloud radiative properties and of precipitation on the large-scale circulation are displayed in Figs. 4b,c, showing the satellite-derived precipitation and cloud radiative forcing (CRF) as a function of omega (omega being derived from meteorological reanalyses). These increase as the vigor of the convective mass flux increases. Bony et al. (2004, 2006)

Breakdown of (daytime) CRE by Weather State ISCCP extended topics weather states

TOA CRE, extended tropics (8 WS), Oreopoulos and Rossow (2011) highest RFO lowest RFO The left figure shows that annually averaged TOA LW CREs below 20 Wm−2 characterize WS5, WS6, WS7, WS8 (the least convectively active states), but the SW CREs assume a wide range (as do the net CREs) of values between −35 Wm−2 and −195 Wm−2 that correlate well with the dominant taus WS3 and WS4 are almost indistinguishable in terms of their average LW CRE, but separate very clearly in terms of SW (and net) CRE, both being larger for WS3, the more convectively active state of the two. WS1 and WS2 appear quite apart in LW‐SW CRE space not just from the other weather states, but also from each other (their net CRE differs by ∼150 Wm−2). In addition to having the strongest LW CRE, WS1 also has the strongest SW CRE and net CRE. In contrast, WS8 has the weakest SW, LW and net CREs. WS4 and WS7 representing completely different cloud mixtures have almost indistinguishable SW CREs, but differ in net CRE because of their LW CRE differences. The greatest similarity in net CRE is between WS4 and WS8, even though they correspond to entirely different cloud mixtures with distinct LW and SW CRE components. Seasonal variations of LW CRE are very weak for all weather states (the maximum sdev of ∼9% of the annual mean occurs for WS6). For the tropics, even the SW CRE cycle is rather weak (∼8%, also for WS6). A notable feature of Figure 4 (top) is that only two states approach a nearly zero daytime TOA net CRE, one being the boundary layer cumulus‐dominated WS8 and the other being the cirrus‐dominated WS4, contradicting some claims that tropical deep convection produces this condition. [18] In terms of contribution to the tropical TOA CRE (Figure 4, bottom) we observe the following: The states with the greatest mean LW CRE when present, WS1 and WS2, are two of the strongest contributors to the total LW CRE, but they are surpassed in contribution by WS3 which has a far larger RFO (see Figure 3). The top SW CRE contributor is again WS3, followed by WS1. WS3 achieves this because of its large RFO, while WS1 because of its large CF (Figure 3) and larger CRE when present (due to optically thicker clouds; see Figure 1). Despite the attention given to marine stratus as causing differences of climate model CRE, the weather states where they are prevalent are not the largest contributors to tropical and subtropical CRE; as Figure 4 shows, it is the convectively active states that contribute most of the SW CRE. WS1 and WS3 also stand out in terms of net daytime TOA CRE contributions. A second group of states (WS2, WS5 and WS8) has significant net CRE contributions as well, ranging from ∼10 to ∼14%. WS6 is one of the weakest contributors in all components of CRE, probably because of its low RFO, but WS4 (with many thin cirrus) is the weakest contributor to the overall net TOA CRE. The most seasonally varying contribution to CRE comes from WS5 which exhibits a sdev close to 20% of the mean for all three CRE components. We have previously seen that this weather state also stood out for its seasonal variability of CF contribution because of significant RFO seasonal variations. WS1 CF-1st, RFO-7th WS3 CF-4th, RFO-2nd WS4 CF-6th, RFO-4th WS8 CF-8th, RFO-1st

CERES CRE vs. MODIS cloud regimes ISCCP MODIS-CERES MODIS CRE is diurnal, not daytime

SFC vs TOA LW CRE, extended tropics ATM cool ATM heat This figure compares the LW CRE at TOA and SFC for the tropical region, the sum of which is indicative of the radiative heating of the atmosphere by clouds: weather states falling above the 1‐to‐1 line indicate cloud regimes cooling the atmosphere whereas states below the line indicate regimes warming the atmosphere. Although WS3 plays a major role in the TOA CRE, its net effect on LW heating of the atmosphere is nearly zero as is the contribution from the cloud regime with lots of scattered cumulus (WS8). The two stronger convective states, WS1 and WS2, produce a large heating of the tropical atmosphere, whereas all the boundary layer weather states (except WS8) produce a weaker cooling. These results reflect the general situation in low latitudes where the fair weather atmosphere is cooled by radiation and the surface is heated, even with some clouds present, but the CRE reinforces storm system latent heating of the atmosphere while cooling the surface. WS2 WS5

Breakdown of CRE by cloud regime in GCMs Note that MODIS and ERBE are NOT combined. MODIS is treated as another model for top panel and ERBE also as another model for bottom three panels. Clusters are not defined from p-tau histograms directly, but using a more simplified method (tot_CF, mean CTP and albedo of original regimes) Williams and Webb (2009)

Cloud Radiative Feedback From leaked IPCC AR5 WGI Second Order Draft: “…likely positive, with a likely range of −0.2 to 1.4 Wm-2K-1.” “The cloud feedback remains the most uncertain radiative feedback in climate models.” Some consensus among CGCMs: High clouds rise, high and middle cloud amounts decrease, storm tracks shift poleward, low cloud amount decreases (especially in subtropics) Based on the above synthesis of cloud behaviour, the net radiative feedback due to all cloud types is judged likely (>66% chance) to be positive. This is reasoned probabilistically as follows. First, since evidence from observations and process models is mixed as to whether GCM cloud feedback is too strong or too weak overall, and since the positive feedback found in GCMs comes mostly from mechanisms now supported by other lines of evidence, the central (most likely) estimate of the total cloud feedback is taken as the mean from GCMs (+0.8 W m–2 K–1). Second, since there is no accepted basis to discredit individual GCMs, the probability distribution of the true feedback must be at least as broad as the distribution of GCM results. Third, since feedback mechanisms are probably missing from GCMs (particularly involving 1 thin high clouds or low clouds) and some CRMs suggest feedbacks outside the range in GCMs, the probable range of the feedback must be broader than its spread in GCMs. We estimate the likely range of this feedback by doubling the spread (quadrupling the variance) about the mean of the data in Figure 7.9, that is assuming an uncertainty 170% as large as that encapsulated in the GCM range added to it in quadrature, and assuming Gaussian errors. This yields a 90% (very likely) range of −0.2 to 1.4 W m–2 K–1, with a 16% probability of a negative feedback.

Why so uncertain? Difficult to constrain from observations: time series too short; signal is weak (many cancellations, SW/LW), observations are imperfect, contamination Clouds imperfectly represented in GCMs Multiple ways used to extract even from GCMs “surface temperature-mediated” change of CRE in response to forcing (e.g. 2xCO2) According to Zelinka et al. (2013) some locations the cloud adjustments act in opposition to and in other locations act in the same direction as the cloud feedbacks, according to CGCMs. CRE also changes in response to rapid changes in land Ts, troposphere (WV, T), and As (“rapid adjustments”); circulation; these CRE changes should be removed changes in clear sky fluxes (due to changed CO2, WV, T, Ts, As), in both terms, also alter CRE; corrections should be applied

Correcting ΔCRE using “Soden” radiative kernels SW LW Corrections to ΔCRE Shell et al. (2008) Corrected ΔCRE vs. uncorrected ΔCRE Corrections to SWCRE seem to come mainly from sfc albedo masking of DCRE (or cloud masking of sfc albedo changes!) and occur of course at high latitudes. Corrections to LWCRE are more substantial and mainly reflect temperature and humidity effects on the clear sky flux components folded into DCRE. The noncloud variables also make CRFLW more negative, except at high latitudes (Fig. 7b). The effect is largest in the tropics. The global-average CRFLW from CAM is quite close to zero (0.06 W m2); CRFkLW 1.2 W m2. The atmospheric temperature contribution is 0.9 W m2; in a warmer climate, clouds are warmer and thus more effective at radiating to space, a cooling effect. The water vapor contribution is 1.3 W m2, and CO2 contributes 0.7 W m2. When no clouds are present, water vapor and CO2 absorb more longwave radiation in the doubled-CO2 case than in the present-day simulation. Thus, the difference between all- and clear-sky fluxes (CRF) is reduced in the doubled-CO2 case, resulting in a negative contribution to CRFLW. The only positive contribution is from the surface temperature (1.7Wm2). By trapping the upwelling longwave radiation from the surface, clouds decrease the outgoing longwave radiation. In a warmer climate, more longwave radiation is emitted by the surface, so clouds trap more radiation, leading to a larger positive CRF in the doubled-CO2 climate.

(Application of ISCCP simulator in CGCMs instrumental) Breakdown of cloud feedback parameter (CMIP5) (Application of ISCCP simulator in CGCMs instrumental) LW, SW, NET Red is LW, black is net, blue is SW. Zelinka kernels give feedback without corrections necessary (as in the DCRE method), but rapid adjustments should be accounted for. Our ability to understand where cloud feedbacks are coming from and how the models differ is enhanced by the breakdowns by cloud type or type of cloud change. Zelinka et al. 2013; Please refer to Zelinka talk on Thursday

Breakdown of cloud feedback parameter by circulation regime Vial et al. (2013) CMIP5 CGCMs Tropics only Tropically-averaged cloud feedback parameter (estimated using the NCAR kernels) plotted as a function of the change in cloud radiative effect (i.e., including cloud adjustments, and without correction of the cloud-masking effect) normalized by the global mean surface temperature change over the tropics. Models that predicts a greater tropically-averaged NET cloud sensitivity (i.e., cloud feedback or change in CRE) than the tropically-averaged multimodel mean NET cloud sensitivity are shown in red (5 models), and those predicting a lower cloud sensitivity than the multi-model mean are in black (6 models) Note that the proper cloud feedback in LW and NET is larger than the uncorrected (for cloud masking effects) feedback (although the latter includes rapid adjustments correction). SW (top), LW (middle) and NET (bottom) cloud feedback composited in each dynamical regime. Results are presented for two groups of models: models that predicts a greater tropically-averaged NET cloud feedback than the tropically-averaged multi-model mean NET cloud feedback (in red, 5 models), and those with a lower cloud sensitivity than the multi-model mean (in blue, 6 models). Vertical bars show the inter-model standard deviation in each group. Cloud feedbacks are estimated using the NCAR model’s radiative kernels

Breakdown of ΔCRE by cloud regime in GCMs Williams and Webb (2009) Change in the regime mean RFO, SWCRE, LWCRE and NCRE in response to doubling CO2 in CFMIP models. As in the Zelinka example, we can trace back where DCRE (not cloud radiative feedback!) is coming from.

“Observed” cloud radiative feedback SW Zhou et et al. (2013) “That this work is an analysis of the cloud response to short-term climate variations is an important caveat. Previous work has shown little correlation between the cloud feedback in response to these short-term (mainly ENSO) climate variations and the response to long-term global warming”. LW Note that the deviations between the two methods are more pronounces in the LW and result in net feedback deviations mainly in the NH. net MODIS uses “Zelinka kernels” CERES uses “Soden kernels”

Short- vs. long-term cloud radiative feedbacks Can Dessler use short term variations for cloud feedbacks? The CGCMs give different results for short vs. long-term feedbacks although there aresome similarities in the patterns. This is a multi-model ensemble of 13 CMIP3 models. Soden kernels have been used to correct DCRE. The response is computed for each model as follows: The long-term climate change response is determined by the difference in temperature, water vapour, surface albedo and global mean surface temperature between global model simulations of future-climate (average over the period 21012130 from the A1B experiments) versus present-day conditions (average over 19702000 from the 20C3M experiments). The response due to short-term interannual climate variations is computed from the slope of the linear least squares fit between the monthly anomalies in temperature, water vapour and surface albedo and the monthly anomalies in global mean surface temperature. The model responses are computed from 30-yr period (1970-2000) anomalies of the coupled model 20C3 experiments. The climate feedbacks for each variable derive from the product of the above climate responses with the three-dimensional kernels from the GFDL model. long-term short-term CMIP3 models, Koumoutsaris (2013)

So, where do we stand? We have expanded our observational sources of CRE, and have moved beyond TOA CRE (incl. better attribution) But climate change-relevant ΔCRE from obs remains elusive (time series too short) When we turn to CGCMs, ΔCRE is not, strictly speaking, enough to derive cloud radiative feedback But most importantly, we don’t know how realistic model ΔCRE is (e.g., CMIP5 better than CMIP3?) Still, we have made progress in tracing model ΔCRE to the nature of cloud changes and in understanding better what contributes to differences among CGCMs

Additional slides

Zhou et et al. (2013) Average cloud fraction in each Ptop-tau bin (%). (b) Slope of the regression of cloud-fraction anomaly in each bin vs. ΔTs (%/K). (c-e) The contribution to the net cloud feedback, SW cloud feedback, and LW cloud feedback, respectively, in W/m2 /K. Note that the multiplication of cloud radiative kernels with cloud fraction anomalies occurs at every location and is then spatially averaged for display in this figure. Bins where the regression slope is statistically significant (>95%) are marked with black crosses.