1. Motivation
For 30 years, climate models have projected amplified Arctic warming and sea ice loss in response to increased greenhouse gas forcing (Manabe and Stouffer 1980). While the sign of the Arctic response is known, the magnitude has not been constrained by climate models. Mean sea ice thickness, winter cloud increases, and enhanced poleward ocean heat transport have all been identified as factors that explain intermodel spread in the Arctic response to the greenhouse gas forcing (Holland and Bitz 2003). Even more notable than the large intermodel spread, observed rates of Arctic sea ice loss and associated Arctic warming (Serreze et al. 2009) have exceeded most climate model projections (Stroeve et al. 2007). Why are climate models too conservative? Natural variations in atmospheric circulation patterns affect observed Arctic sea ice loss and warming and complicate climate model evaluation, but model biases can also help explain the underprediction of observed Arctic change (e.g., sea ice thickness biases; Bitz 2008).
Though the atmospheric boundary layer response to sea ice loss affects the magnitude and geographic distribution of projected Arctic warming, most existing climate-modeling studies have focused on the large-scale atmospheric circulation response to sea ice loss (e.g., Higgins and Cassano 2009 and references therein). Two recent studies show that boundary layer processes are critical to understanding and modeling Arctic climate change. Deser et al. (2010, hereafter referred to as D10) document the atmospheric response to twenty-first-century sea ice loss in the Community Atmosphere Model, version 3 (CAM3) and the Community Climate System Model, version 3 (CCSM3; Collins et al. 2006). While the maximum sea ice loss was prescribed in summer [June–August (JJA)] and fall, D10 found the largest local net energy budget response in winter. During winter, enhanced turbulent heat fluxes over the ocean led to widespread Arctic warming below 850 mb, an erosion of the low-level temperature inversion, and increased snowfall over Siberia and northern Canada. D10 found the greatest winter warming in areas with strong low-level temperature inversions. Boé et al. (2009, hereafter referred to as B09) show that present-day winter near-surface stability is negatively correlated with the projected mixed layer ocean warming in climate models that contributed to the Coupled Model Intercomparison Project phase 3 (CMIP3). Because present-day near-surface stability in climate models is often stronger than observed, B09 postulate that excessive negative temperature feedbacks cause many climate models to underpredict Arctic warming in response to increased greenhouse gas forcing.
In addition to boundary layer stability, boundary layer clouds have important and complex effects on Arctic climate. Specifically, the magnitude and sign of the Arctic cloud influence on surface radiation budgets is a complicated function of latitude, season, surface albedo, and cloud height and optical depth (e.g., Shupe and Intrieri 2004). CMIP3 models project increases in low-level Arctic clouds under increased greenhouse gas forcing (Vavrus et al. 2008). Projected cloud increases reflect incoming solar radiation, mask underlying sea ice loss, and reduce the strength of positive surface albedo feedbacks, but they also enhance surface greenhouse warming. Several climate models project negative Arctic cloud feedbacks, indicating projected cloud increases can have a stronger influence on shortwave feedbacks than on longwave feedbacks (Soden et al. 2008). Given the intricacies of Arctic cloud processes and the paucity of observations that constrain cloud influence on surface radiation budgets, it is not surprising that climate models struggle to represent Arctic cloud processes (e.g., Gorodetskaya et al. 2008; Gettelman et al. 2010).
The documented importance of the boundary layer to Arctic climate projections, and recent dramatic reductions in observed Arctic sea ice extent, motivate a new assessment of the local response to sea ice loss in climate models. This study evaluates the boundary layer and energy budget response to recent sea ice loss in the Community Atmosphere Model version 4 (CAM4). CAM4 is the atmospheric component of the Community Climate System Model version 4 (CCSM4), which is being used for climate change projections as a part of the fifth Intergovernmental Panel on Climate Change (IPCC) assessment report (AR5). Thus, results discussed here are relevant to future climate-modeling studies that use CCSM4. Because most boundary layer and cloud parameterizations remain unchanged in CAM4, this work also has relevance to all studies that used CAM3–CCSM3 to evaluate Arctic climate change and sea ice loss (e.g., Teng et al. 2006; Vavrus et al. 2008; Holland et al. 2006; D10; Higgins and Cassano 2009). The paper is organized as follows. In section 2, we describe CAM4 and our model experiment design. We evaluate the boundary layer response to prescribed 2007 ocean conditions in 1-day observationally constrained CAM4 forecasts and in 10-yr CAM4 runs with a freely evolving atmosphere (ATM). Section 3 contains the results. We begin by documenting the influence of sea ice loss on CAM4’s boundary layer in July 2007 and September 2007 forecasts. Through comparison of the modeled and observed Arctic cloud response to sea ice loss, we identify a deficiency in CAM4’s stratus cloud parameterization. We describe and implement a physically motivated modification to the stratus cloud parameterization, requiring a well-mixed boundary layer for stratus clouds to be diagnosed. We then contrast the influence of stability-based cloud parameterizations on Arctic clouds in the forecasts and freely evolving CAM4 runs. In section 4, we discuss the relevance of our findings for high-latitude climate feedbacks. We conclude with a summary of our most important findings in section 5.
2. Model description and methods
a. CAM4 description
Convective clouds are of negligible importance in the Arctic. As described in this paper, CAM4’s parameterization of stratus clouds (herein called CLDST) has an important influence on the Arctic cloud response to sea ice loss.
b. Short-term forecasts to evaluate CAM4’s response to sea ice loss
1) CAM DART description
The use of short-term forecasts to assess and improve climate models is an emerging research field (e.g., Hannay et al. 2009; Rodwell and Palmer 2007; Phillips et al. 2004). When data assimilation techniques are used to produce initial conditions, short-term forecasts enable evaluation of climate model performance during periods of known climatic importance or when abundant observations are available. For this study, we used the Data Assimilation Research Testbed (DART) to produce initial conditions for CAM4 forecasts. DART is a state-of-the-art ensemble filter data assimilation software package developed at the National Center for Atmospheric Research (NCAR; Anderson et al. 2009). Because CAM4 was used both to generate initial conditions in DART and to run forecasts, no foreign model error was introduced in our forecasting experiments. In the generation of CAM4 initial conditions, observations that constrain the large-scale atmospheric circulation were assimilated; however, few boundary layer and no surface or cloud observations were assimilated. Thus, comparison of the modeled and observed boundary layer response to sea ice loss provided a fruitful strategy for identifying CAM4 boundary layer parameterization deficiencies.
2) CAM forecast initial conditions and testing
We used DART to produce observationally constrained initial conditions for short-term CAM4 forecasts. We ran 80 CAM4 ensemble members within DART during time periods of interest for forecasting. Wind and temperature observations from radiosonde, aircraft, and satellite observations were assimilated every 6 h to produce an analysis, a blended observation–model state. The assimilated observations are identical to those used to produce the National Centers for Prediction (NCEP)–NCAR 50-Year Reanalysis (Kistler et al. 2001). Ocean boundary conditions were prescribed using observed sea surface temperature (SST) and sea ice distributions (Hurrell et al. 2008). To produce an initial atmospheric condition for CAM4 forecasts, we averaged analyses from the 80 observationally constrained ensemble members. Because of complications arising in averaging the land state from individual ensemble members, a single ensemble member analysis was used to initialize the land component of the CAM4 forecasts.
We used standard metrics to evaluate the CAM4 forecasts for quality and physical reasonableness. Time derivatives of surface pressure and 700-mb air temperature were used to identify the relaxation time scale for perturbations that result from initialization. A significant reduction in the time derivatives occurred through forecast hour 12 (FH12), but we found little change after forecast hour 24 (FH24). We also used forecast verifications, the difference between the forecast and the analysis as a function of time, to assess how quickly the CAM4 forecasts deviated from an observationally constrained model state. We found the forecasts steadily diverged away from the analyses. Forecast errors grew quickly in the Southern Hemisphere, reflecting the lack of observations in this region. Based on these tests, we decided that to minimize model initialization shocks, but also to ensure that the model state is as close to the observed state as possible, the optimal forecast hour for comparison with observations was FH24. Earlier (later) forecast hours would have more (less) impact from initialization shock, but they would be closer to (farther from) an observationally constrained model state.
3) CAM4 forecast modeling strategy
After evaluating several CAM4 forecasts, we designed a forecasting strategy that enabled us to directly compare hundreds of CAM4 forecasts with observations. The goal of our forecasting was to produce monthly-averaged CAM4 values that could be directly compared with monthly-averaged observations or used to evaluate the climate significance of model modifications. Figure 1 shows the forecast cycle and averaging that we implemented. In individual months of interest, we started CAM4 forecasts every 6 h and ran them for 24 h. We averaged FH24 to produce monthly-mean values. No atmospheric data assimilation was done during the forecasts, but observed SST and sea ice distributions (Hurrell et al. 2008) were prescribed.
Our modeling strategy enabled evaluation of the fast atmospheric response to sea ice loss, but it did not permit a complete assessment of coupled feedbacks or the full large-scale atmospheric response. In both the analyses and the forecasts, sea ice with a thickness of 2 m and sea surface temperatures were prescribed and were therefore uncoupled to the overlying atmosphere. Prognostic surface energy budget calculations did allow some surface–atmosphere coupling. Surface temperature, snow depth, and surface albedo coevolved with atmospheric conditions over both land and sea ice.
Our analysis of the cloud and boundary layer response to sea ice loss in CAM4 forecasts focused on July 2007 and September 2007. Record low sea ice extent during these two months provided a large sea ice loss signal. July was selected to evaluate conditions in midsummer when significant incoming solar radiation and surface albedo reductions lead to positive net surface energy budgets over the Arctic Ocean. September was selected to evaluate conditions in early fall when sea ice extent is the lowest of the annual cycle and the Arctic has negative surface energy budgets over the Arctic Ocean.
We used two strategies to document and evaluate the boundary layer response to the 2007 sea ice loss in CAM4 using forecasts (Table 1). The first strategy was to run forecasts in both 2006 and 2007 and examine 2007 minus 2006 differences. The advantage of this strategy is that we could make direct comparisons between observed and modeled 2007 minus 2006 differences. The challenge of this strategy is that we have to discriminate between changes resulting from sea ice loss and the substantial 2007 minus 2006 differences in large-scale atmospheric circulation patterns. The second strategy was to run July and September forecasts with 2007 initial conditions but with climatological SST and sea ice boundary conditions. These climatological forecasts document the transient response to imposing climatological SST–sea ice distributions. The advantage of this strategy is that by comparing forecasts with observed and climatological sea ice conditions, we were able to isolate the CAM4 boundary layer response to sea ice loss. There are two limitations with this strategy. The first is that the forecasted response is applicable only to 2007 atmospheric conditions. Because the atmospheric circulation pattern during the 2007 melt season was anomalous, the forecasted response may not be representative of the response in all years. The second is that we had to consider the degree of equilibration in the transient response to prescribed ice-covered conditions in the climatological forecasts. To evaluate equilibration, we compared temperature and moisture time derivatives from forecast hour 0 (FH0) through FH24 in the climatological forecasts. As expected, near-surface temperature decreased and specific humidity decreased in response to prescribing ice-covered conditions. In the bottom atmospheric layer, the average temperature change over ocean areas from 70° to 90°N was −0.11 K (−0.12 K) at FH12 (FH24) in July and −0.46 K (−0.69 K) at FH12 (FH24) in September. In the bottom atmospheric layer, the average specific humidity change over ocean areas from 70° to 90°N was −0.03 g kg−1 at both FH12 and FH24 in July but 0.09 g kg−1 at FH12 and −0.15 g kg−1 at FH24 in September. Greater absolute change and rates of change suggest longer equilibration time scales in September than in July. In all cases, the change over the first 12 h of the forecast was much greater than the change over the second 12 h, indicating that some degree of equilibration had been achieved by FH24. Even so, the boundary layer was likely more unstable and the relative humidity higher in the climatological forecasts than in forecasts with a fully equilibrated boundary layer, especially in September. Although using FH24 differences will underestimate the atmospheric response to sea ice loss, selection of FH24 was a subjective choice based on many factors, including the desire to have a model state that is close to a 2007 large-scale atmospheric state [see section 2b(2)]. FH24 differences are sufficient to accomplish our goals of isolating model tendencies and response magnitudes.
c. CAM4 runs with a freely evolving atmosphere
We also evaluated CAM4’s boundary layer response to sea ice loss in CAM4 runs with a freely evolving atmosphere and prescribed SST–sea ice distributions (Table 2). Unlike the CAM4 forecasts described in section 2b, the atmospheric state within the freely evolving CAM4 runs was unconstrained by assimilation of 2007 atmospheric observations. We used 10-yr model runs to identify differences resulting from parameterization changes and boundary condition changes. Decadal runs were necessary to reduce the noise resulting from substantial natural variability in high-latitude atmospheric circulation patterns.
3. Results
a. Evaluation of cloud and boundary layer processes in CAM4 forecasts
We begin our assessment of the boundary layer response to sea ice loss in CAM4 by examining maps from the July 2006 (Jul06) and July 2007 (Jul07) forecasts (see Table 1). The Jul06 and Jul07 forecasts had large differences in prescribed sea ice conditions and atmospheric circulation patterns (Figs. 2a and 2b). Record low sea ice extent in July 2007 was evident mainly in the Pacific sector. While July 2006 had relatively weak sea level pressure (SLP) gradients, July 2007 had a strong SLP dipole pattern with a high pressure center north of Canada and a low pressure center over the Laptev Sea.
Consistent with observations (Kay et al. 2008; Kay and Gettelman 2009), we found that the July 2007 large-scale circulation pattern had a significant influence on the CAM4-forecasted 2007 minus 2006 differences in cloud cover (Fig. 2c). For example, CAM4 forecasted fewer clouds in 2007 than in 2006 over the Arctic Ocean north of Canada, a region that was ice covered in both years. Because our main goal was to evaluate the boundary layer response to sea ice loss, we focused on changes over “newly open water,” that is, regions that were ice free in July 2007 but ice covered in July 2006 (Fig. 2a). A key finding was that CAM4 produced unrealistic ubiquitous low cloud increases over newly open water. In the Beaufort, Chukchi, and East Siberian Seas, CAM4-forecasted cloud increases did not agree with satellite cloud observations (Fig. 2d). Over the Laptev Sea, CAM4-forecasted cloud increases did qualitatively agree with satellite observations but the July 2007 increases did not extend toward the North Pole as in the observations.
Circulation-driven July near-surface stability differences were broadly similar in the forecasts and the satellite observations, though the magnitude of the modeled 2007 minus 2006 differences was greater than observed (Figs. 2e and 2f). Like B09 found in winter, we found excessive near-surface stability over the Arctic Ocean in the July CAM4 forecasts as compared to the Atmospheric Infrared Sounder (AIRS) satellite observations. In regions with large near-surface stability (e.g., along a strong SLP gradient in July 2007), CAM4 biases approached 7 K.
We next assessed the boundary layer response to September 2007 sea ice conditions in maps from the September 2006 (Sep06) and September 2007 (Sep07) forecasts (Table 1). Like the July forecasts, the September forecasts also had significant differences in their prescribed sea ice conditions and atmospheric circulation patterns (Figs. 3a and 3b). When compared to September 2006, September 2007 had less sea ice over the central Arctic Ocean. September 2007 minus 2006 circulation difference maps had low pressure centers over the Aleutian low and Icelandic low and high pressure centers over Siberia and Canada.
To assess the influence of September 2007s record low sea ice conditions on the CAM4 forecasts, we compared modeled and observed 2007 minus 2006 difference maps (Figs. 3c–f). Forecasted low cloud increases and near-surface stability decreases over newly open water in September 2007 were qualitatively consistent with physical expectations and recent satellite observations (Kay and Gettelman 2009). Near-surface stability biases were smaller in September forecasts than in the July forecasts. Interestingly, the forecasted near-surface stability was up to 4 K weaker than AIRS observations over newly open water. As a result, 2007 minus 2006 reductions in near-surface stability over the newly ice-free central Arctic Ocean were greater in the forecasts than in the AIRS observations (Figs. 3e and 3f).
To isolate the boundary layer response to the 2007 sea ice loss in CAM4, we next examined maps from the July and September 2007 forecasts with climatological SST–sea ice boundary conditions (Jul07_clim and Sep07_clim, respectively) and compared them to maps from the forecasts with 2007 SST–sea ice boundary conditions (Jul07 and Sep07). As described in section 2b(3), the climatological forecasts document the transient response to 2007 SST–sea ice conditions. Like the Jul07 − Jul06 and Sep07 − Sep06 difference maps (Figs. 2c and 3c), the Jul07 − Jul07_clim and Sep07 − Sep07_clim difference maps (Figs. 4c and 4d) show cloud increases over the newly ice-free Arctic Ocean. The similar cloud response to sea ice loss in Figs. 2 –4 confirms that the CAM4-forecasted cloud increases over newly open water in July and September resulted from prescribed sea ice differences. While July near-surface stability was minimally affected by ocean conditions, September near-surface stability decreased by up to 6 K over areas of newly open water (Figs. 4e and 4f).
To provide a detailed look at the boundary layer in the CAM4 forecasts, we next contrast monthly-mean-forecasted boundary layer properties and radiation budgets over the newly ice-free Arctic Ocean during July 2007 and September 2007 (Table 3). Forecasted values in 2007 were qualitatively consistent with physical expectations and provided a baseline for comparison with the other forecasting experiments (Table 1). When compared to Sep07, Jul07 boundary layer heights were lower and near-surface stability was higher. Over newly open water, Jul07 had net heating at the top of atmosphere (TOA; +53.8 W m−2) and the surface (+148.0 W m−2), while Sep07 had net cooling at the TOA (−156.1 W m−2) and the surface (−26.8 W m−2). As expected, shortwave radiation terms were much larger in Jul07 while surface turbulent fluxes were greater in Sep07. Forecasted low cloud amounts over newly open water were similar in Jul07 (72%) and Sep07 (71%), which is inconsistent with observations that show more low Arctic cloud over open water in early fall than in midsummer (e.g., Kay and Gettelman 2009). Sep07 clouds had more in-cloud water and ice than Jul07 clouds. To assess the impact of clouds on radiative fluxes, we compute total (shortwave + longwave) cloud forcing values by differencing net all-sky and net clear-sky fluxes. Positive total cloud forcing values indicate that clouds warm the surface or TOA. At the TOA, cloud forcing was negative in both Jul07 (−77.1 W m−2) and Sep07 (−11.4 W m−2). At the surface, the net longwave contributions increased total cloud forcing. The resulting surface cloud forcing was still negative in Jul07 (−55.7 W m−2) but positive in Sep07 (28.2 W m−2).
With the 2007 monthly-mean values as a baseline, we next evaluate the boundary layer and energy response to the 2007 sea ice loss in the CAM4 forecasts (Jul07 − Jul07_clim and Sep07 − Sep07_clim; Table 1). Only small changes in SLP or 500-mb geopotential heights (Z) resulted from sea ice loss (not shown). There were, however, substantial boundary layer and radiative flux changes associated with the transition from an ice-free to an ice-covered state. Over newly open water in both July and September, boundary layer heights were higher, the near-surface inversion strength was weaker, and cloud amount and water paths increased (Table 3). Changes in boundary layer structure and surface turbulent fluxes were greater in September than in July, but low cloud fraction and net shortwave flux increases were greater in July than in September. In July, Arctic sea ice loss increased the net TOA (surface) energy budget by +21 W m−2 (+19.4 W m−2). In September, net TOA flux changes were small (+1.4 W m−2), but the enhanced surface heat loss of −17.9 W m−2 led to a net atmospheric heating response of +19.3 W m−2. In both months, surface downwelling shortwave radiation was reduced over newly open water because of cloud increases and reduced multiple scattering between the surface and the clouds. Cloud increases over newly open water also increased surface downwelling longwave radiation in both months.
At the surface (TOA), forecasted July and September total cloud forcing decreased over newly open water by 35.9 W m−2 (43.3 W m−2) in July and 2.3 W m−2 (8.2 W m−2) in September. These decreases in total cloud forcing were influenced both by surface albedo decreases and low cloud increases. Recall that even with no change in clouds, the cooling effect of clouds as measured by the shortwave cloud forcing is enhanced when surface albedo decreases. Therefore, low cloud increases over newly open water in the July and September CAM4 forecasts led to an even greater reduction in cloud forcing. Interestingly, polar maps show that the sign of the September surface total cloud forcing response was a function of latitude (Fig. 5). September surface cloud forcing was positive (warming) over the entire Arctic Ocean; however, cloud increases equatorward of 82°N reduced cloud warming, and cloud increases poleward of 82°N enhanced cloud warming.
Finally, we document the vertical structure of the forecasted boundary layer response to sea ice loss by comparing atmospheric profiles over observed 2007 ice-free conditions and climatological sea ice cover (Fig. 6). The profiles provided useful clues to the reasons behind the ubiquitous low cloud increases over newly open water in the Jul07 and Sep07 forecast experiments.
In July 2007 forecasts, temperature and relative humidity profiles were largely unaffected by underlying surface ocean conditions. Nevertheless, stratus cloud fraction and in-cloud liquid water content increased over newly open water. Forecasted cloud fraction and water content increases associated with the transition from an ice-free to an ice-covered ocean occurred primarily in the bottom two atmospheric levels. Because changes in humidity and atmospheric stability were small, changes in the atmospheric state cannot explain the modeled July cloud increases. Instead, we found that July cloud increases over open water occurred because CAM4’s stratus cloud parameterization CLDST only diagnoses stratus clouds over open water, not over sea ice. As discussed above, CAM4’s July ubiquitous cloud increases over newly open water are inconsistent with observations.
Sea ice loss had a greater influence on the temperature and humidity profiles in the September forecasts than in the July forecasts. September air temperature increases associated with the transition from an ice-covered ocean to an open ocean were especially evident near the surface. Over newly open water, relative humidity was lower in the bottom atmospheric level, but it was higher in the next three overlying atmospheric levels. September relative humidity profile changes produced changes in the cloud fraction profile. Cloud fraction decreased in the bottom atmospheric level because warming outpaced moistening but increased aloft because moistening outpaced warming. Because of the relatively low atmospheric stability in the September forecasts, CLDST had a negligible influence on Arctic cloud amounts. September in-cloud liquid water content decreased in the two model levels close to the surface but increased aloft.
b. Improving the CAM4 forecasts
1) Description of CAM4’s CLDST
Using CAM4 forecasts, we identified unrealistic Arctic cloud increases over newly open water resulting from CAM4’s stratus cloud parameterization CLDST. This discovery led us to evaluate if CLDST was consistent with physical expectations for cloud formation over the ice-free Arctic Ocean. We examined the conceptual model underlying CLDST and found an assumption that is inconsistent with cloud formation in stable boundary layer regimes such as those that exist in the Arctic. Here, we motivate and describe a physically motivated change to CLDST to remove this assumption. We begin by discussing the rationale for including CLDST in CAM4.
Wood and Bretherton (2006) review the physical setting for stratus cloud formation atmospheric regimes with a well-mixed boundary layer capped by a strong temperature and humidity inversion. As expected, the strong inversion prevents penetration of dry free tropospheric air into the boundary layer. The inversion also caps moisture mixed up from the surface ocean and allows stratus clouds to form. Once a cloud forms, cloud-top radiative cooling destabilizes the boundary layer, enhances the mixing, and helps maintain the cloud. Like many global atmospheric models, CAM4 has difficulty reproducing this physical setting (Hannay et al. 2009). Poor vertical resolution and a turbulence closure that is based on dry static energy are two candidates for explaining CAM4’s inability to reproduce observed cloud-topped turbulent boundary layers.
Because CAM4 cannot accurately predict stratus clouds using model physics, CLDST was implemented to empirically diagnose stratus clouds. Based on the cloud-topped turbulent boundary layer physical setting, three criterion were included for CLDST to diagnose stratus: 1) a stable atmospheric regime (θ700 mb − θsurface > 10 K), 2) a moisture source (open water fraction > 0.01), and 3) a capping inversion to trap the moisture that produces the stratus cloud (Γcap < −0.125 K mb−1 between 700 mb and the surface). When these three conditions are met, CLDST uses observed correlations between low cloud fraction and lower-tropospheric stability (Klein and Hartmann 1993) to diagnose low clouds over open water. CLDST has an important influence on subtropical cloud feedbacks. For example, stability-based cloud formulations are responsible for producing negative cloud climate feedbacks in CCSM3 (Stephens 2005) and in box models of the tropics (Miller 1997). Although CLDST was implemented to ameliorate a subtropical cloud deficit, it affects low cloud formation over the open ocean in stable atmospheric regimes at all latitudes.
We evaluated the global influence of CLDST in CAM4 by comparing runs with and without CLDST (CAM4_2007repeat and CAM4_nocldst_2007repeat; see Table 2). Without CLDST, CAM4 underestimated subtropical low cloud amount, cloud liquid water path, and shortwave cloud radiative forcing off western continental boundaries. Including CLDST reduced TOA net shortwave radiation and resulted in better agreement between the modeled and the Clouds and the Earth’s Radiant Energy System (CERES; Loeb et al. 2009) TOA net shortwave fluxes in subtropical stratus regimes. For example, including CLDST reduced July shortwave cloud forcing by −53 W m−2 in the California stratus region, resulting in a value (−94 W m−2) that was much closer to the CERES observations (−100 W m−2).
In the Arctic (70°–90°N), CLDST produced the greatest low cloud increases and net TOA net shortwave decreases (−10 W m−2) in July because of the coincident occurrence of large lower-tropospheric stability, open water, and appreciable incoming solar radiation. Although CLDST had a smaller influence in the Arctic than it did in the subtropics, CLDST enhanced a well-known Arctic shortwave radiation deficit bias in the physics used in CAM4 (e.g., Collins et al. 2006; Gorodetskaya et al. 2008). When compared to CERES observations, CAM4 had excessive summer shortwave cloud cooling over the Arctic. For example, CERES reported a July shortwave cloud forcing of −67 W m−2, while CAM4_2007repeat had a July shortwave cloud forcing of −80 W m−2.
Revisiting the assumptions underlying CLDST revealed a shortcoming for its application in the Arctic. Unlike subtropical boundary layers, observed Arctic boundary layers often have surface and near-surface inversions. As a result, the well-mixed boundary layer assumption embedded in CLDST is violated in the Arctic. To assess how often this well-mixed boundary layer assumption was violated, we examined histograms of the maximum stability underlying diagnosed stratus clouds (Γmax) (Fig. 7). While the California stratus region had relatively weak stability underlying diagnosed stratus clouds, the Arctic Ocean frequently had very stable layers (Γmax < −0.2 K mb−1) underlying stratus clouds diagnosed by CLDST.
Motivated by the existence of stable layers underlying diagnosed Arctic stratus clouds, we modified CLDST to require a well-mixed boundary layer. We refer to our modified parameterization as CLDST_MIXBL. CLDST_MIXBL retains all three of the existing criterion for stratus cloud diagnosis from CLDST. To enforce a well-mixed boundary layer, we added a fourth criterion to CLDST_MIXBL: no layer underlying the capping inversion can be more stable than the underlying stability threshold [Γund (k mb−1)]. We appended “cldst_mixbl” to model run names to indicate when CLDST_MIXBL parameterization was used (see Tables 1 and 2).
The physical rationale for adding a well-mixed boundary layer criterion to CLDST is clear: stratus clouds should not be diagnosed if their surface moisture supply is cut off by strong atmospheric stability. Adding a well-mixed boundary layer criterion removes a regime-specific assumption embedded in CLDST and makes CLDST more appropriate globally. Based on our examination of differences between the Arctic and subtropical stability and cloud profiles, and Γmax histograms (Fig. 7), it seemed possible to impose a well-mixed boundary layer criterion that would reduce the unrealistic Arctic stratus cloud response to sea ice loss.
Adding a well-mixed boundary layer criterion to CLDST is physically motivated, but its implementation requires specification of a single value for Γund. Thresholds are a difficult part of any parameterization development effort, especially when strong observational constraints are not available. Our goal was to select a Γund value that enforced the desirable well-mixed boundary layer criterion in the Arctic, without degrading CLDST_MIXBL performance in the subtropics. Based on histograms of the maximum stability underlying diagnosed stratus clouds (Fig. 7), we tried a range of Γund values. Our preferred value of Γund was −0.05 K mb−1 because with this threshold CLDST_MIXBL produced the desired effects in the Arctic, but it did not degrade the stability-based cloud diagnosis in the subtropics.
c. Impact of CLDST_MIXBL on CAM4
We evaluated the impact of CLDST_MIXBL in July CAM forecasts and in CAM4 runs with a freely evolving atmosphere (Tables 1 and 2). We did not evaluate the impact of CLDST_MIXBL on September CAM forecasts because stability-based cloud parameterizations had a negligible impact on the diagnosed cloud fraction in that month (Fig. 6). As expected, the main impact of implementing CLDST_MIXBL was a reduction in low clouds over open water in the regions with stable layers near the surface.
We first describe the impact of CLDST_MIXBL in the July 2007 forecasts (Jul07, Jul06, and Jul07_clim; Table 1). We were very encouraged that CLDST_MIXBL improved the spatial distribution of CAM4-forecasted 2007 minus 2006 July cloud differences as compared to the Moderate Resolution Imaging Spectroradiometer (MODIS) observations (Fig. 8). With CLDST_MIXBL, the unrealistic cloud increases over newly open water in the Beaufort Sea, Chukchi Sea, and the eastern portion of the East Siberian Sea in the July 2007 forecasts were removed. CLDST_MIXBL did not improve forecasted cloud cover in the Laptev Sea and western portion of the East Siberian Sea. Remaining differences between CAM4 and MODIS clouds could be explained by unexplored deficiencies in CAM4’s humidity-based cloud parameterization or by regime-dependent deficiencies in the MODIS cloud retrievals (e.g., overestimation of cloud amount over sea ice; Liu et al. 2010).
Table 4 shows the changes in monthly values that resulted from implementing CLDST_MIXBL in the July 2007 forecasts. Over newly open water, July 2007 low cloud amounts decreased by 31%. As a result, the net energy at the surface (TOA) increased by 11.0 W m−2 (14.9 W m−2) to 159 W m−2 (68.7 W m−2). CLDST_MIXBL also reduced the unrealistic cloud response to sea ice loss in the July 2007 forecasts. Low cloud increases in response to sea ice loss were reduced by 19%, which increased net energy at the surface (TOA) by 8.7 W m−2 (11.6 W m−2) to 28.1 W m−2 (32.6 W m−2). Thus, CLDST_MIXBL reduced, but did not remove, the automatic stability-based cloud response to newly open water in the Arctic.
Encouraged by improved July forecasts, we next evaluated the global influence of CLDST_MIXBL in 10-yr CAM4 runs with a freely evolving atmosphere (Table 2). CLDST_MIXBL did not degrade CAM4’s representation of subtropical cloud amounts or cloud forcing. Monthly-averaged changes in net radiative fluxes over key subtropical cloud regions were small when compared to their absolute magnitude (<10 W m−2 on a 200–300 W m−2 signal; not shown). Unlike CAM4 runs with no CLDST, global changes in low cloud amount and total cloud forcing produced by implementing CLDST_MIXBL were locally small (Fig. 9). Decreases in the subtropical low clouds very close to the western continental coasts improved CAM4 cloud amounts as compared to observations (C. Hannay 2010, personal communication); however, the subtropical cloud decreased but did not produce large changes in radiative fluxes because these clouds had very low liquid water paths. While local changes produced by CLDST_MIXBL were small, the globally averaged annual shortwave cloud forcing increased by +2 W m−2 from −52.7 to −50.6 W m−2. With CLDST_MIXBL, the globally averaged annual CAM4 shortwave cloud forcing value remained within observational uncertainty, improving comparison with CERES data (−47.0 W m−2) but not with the Earth Radiation Budget Experiment (ERBE) data (−54.2 W m−2).
In the Arctic, CLDST_MIXBL had small but desirable effects on clouds and radiation budgets in the freely evolving CAM4 runs. Over the Arctic (70°–90°N), low cloud reductions resulting from implementing CLDST_MIXBL peaked in May (−6%), while enhancements in net shortwave fluxes were greatest in July (+5 W m−2). Over the newly ice-free Arctic Ocean in 2007, low cloud and surface energy budget changes produced by removing CLDST and implementing CLDST_MIXBL were small but similar (not shown).
With the same prescribed SST–sea ice boundary conditions, CLDST_MIXBL had a much greater effect on Arctic clouds in the 2007 forecasts than in the freely evolving runs. For example, implementing CLDST_MIXBL reduced July low cloud amounts over the ice-free Arctic Ocean by a mere 7% in freely evolving runs versus 25% in the 2007 forecasts. Comparing the cloud response to the 2007 sea ice loss in the July 2007 forecasts and the freely evolving run confirmed that the stability-based cloud formulations did not have a consistent influence on Arctic cloud amounts (Fig. 10). In contrast to the July 2007 forecasts, July cloud changes with a freely evolving run were not spatially coherent with July 2007 sea ice loss. Thus, we presume low cloud changes in the freely evolving runs were unrelated to sea ice conditions in July. The net energy response to prescribed July 2007 sea ice loss was much greater in the 2007 forecasts than in the freely evolving runs. For example, with no CAM4 code modifications, the surface (TOA) energy response was +19.4 W m−2 (+21.0 W m−2) for the July 2007 forecasts but only +0.02 W m−2 (+8.2 W m−2) in the freely evolving runs. With CLDST_MIXBL, the surface (TOA) energy response was +28.1 W m−2 (+32.6 W m−2) for the July 2007 forecasts but −6.2 W m−2 (+6.6 W m−2) in the freely evolving runs.
What explains the differing cloud and net energy response to July sea ice loss in the freely evolving runs and the 2007 forecasts? The impact of stability-based cloud parameterizations on Arctic clouds and radiation budgets depends on the atmospheric state. In the July 2007 forecasts, the atmospheric state was largely controlled by initial conditions produced by assimilating 2007 atmospheric observations. In contrast, the freely evolving runs were unconstrained by observations and therefore were more representative of the average July atmospheric state in CAM4. As a result of the differing atmospheric circulation patterns (Figs. 11a and 11b), July near-surface conditions in the 2007 forecasts were more stable and drier than in the freely evolving runs (Figs. 11c and 11d). As a result, the freely evolving runs had more humidity-based low clouds but fewer stability-based clouds when compared to the forecasts. Because the cloud amount is taken as the maximum of the clouds diagnosed based on humidity and stability [Eq. (1)], dry and stable atmospheric conditions made the forecasted July 2007 cloud response especially susceptible to unrealistic stratus cloud increases over newly open water.
4. Discussion
This study documents and evaluates the atmospheric boundary layer response to 2007 sea ice conditions in a state-of-the-art climate model. Many qualitative consistencies between the forecasted response, physical expectations, and available observations were found. The most striking discrepancy found is that because of a regime-specific assumption, CAM4’s stability-based cloud parameterization CLDST diagnoses excessive low clouds over the ice-free Arctic Ocean. Because the excessive cloud cover occurs in midsummer when shortwave terms are large in Arctic radiation budgets, it leads to an underprediction of the net energy absorbed at the surface and TOA. In July 2007 CAM4 forecasts, the impact of the unrealistic clouds diagnosed by CLDST was 11 W m−2 (14.9 W m−2) at the surface (TOA). Although a coupled model is required to assess the impact of unrealistic cloud compensation on shortwave feedbacks, these results show that models using CLDST contain an unrealistic negative cloud feedback. Namely, when sea ice extent decreases, stratus cloud amounts automatically increase; and as a result, net energy into the system decreases and sea ice melting and surface warming are reduced from what they would otherwise be.
Of importance to climate projections, the influence of CLDST on Arctic clouds and cloud compensation for sea ice loss depends on the atmospheric state. Unrealistic cloud increases resulting from CLDST dampen sea ice loss most when dry and stable atmospheric conditions exist. That the unrealistic feedback produced by CLDST only manifests itself in specific conditions is encouraging for studies that have used CCSM3 and will use CCSM4 to project Arctic climate change and sea ice loss. Nevertheless, it is disconcerting that the 2007 atmospheric conditions, the conditions that led to the lowest sea ice extent on record, are the conditions in which we expect CLDST to lead to an underestimation of the shortwave feedbacks that enhance sea ice loss. To the extent that 2007-like extreme events determine the trajectory of sea ice loss, unrealistic cloud compensation produced by CLDST will lead to an underprediction of Arctic sea ice loss and warming. Satellite observations since 1979 show that September sea ice extent has little year-to-year memory and recent CCSM experiments show little long-term predictability after extreme events (Holland et al. 2011), so it is possible that the unrealistic negative feedback produced by CLDST will only be important during melt seasons with dry and stable atmospheric conditions. Future work is required to assess the influence of CLDST in a coupled modeling context, but this work is beyond the scope of this paper.
This study illustrates the classic problem of developing and testing a parameterization for a specific climate regime and then applying it outside of the conditions in which it was designed to operate. Luckily, we found it relatively straightforward to remove the regime-specific assumption in CLDST by requiring a well-mixed boundary layer to diagnose stratus clouds. This simple modification improved the modeled year-to-year differences in Arctic cloud amount and resulted in a model with a cloud response to sea ice loss that is more consistent with physical expectations. We do not expect that many future parameterization improvements for high latitudes will be this straightforward and easy to implement. Large seasonal changes and significant variability distinguish the high-latitude climate from lower latitudes, making it an interesting environment to study but a challenging environment to model.
5. Summary
This study used observationally constrained forecasts and runs with a freely evolving atmosphere to document, evaluate, and improve CAM4’s boundary layer and radiation response to record low sea ice extent in 2007. While many aspects of CAM4’s forecasted response to sea ice loss were consistent with physical expectations and available observations, we did find biases in CAM4 near-surface stability and low cloud amounts over the ice-free Arctic Ocean. In most regions, CAM4 had excessive near-surface stability: however, CAM4’s near-surface stability was too weak over the newly ice-free Arctic Ocean in September 2007. Of importance to CAM4’s shortwave climate feedbacks, we found unrealistic cloud increases over newly open water in July 2007 forecasts. Requiring a well-mixed boundary layer in the stability-based cloud parameterization CLDST improved the forecasted cloud response but had a much smaller influence on runs with a freely evolving atmosphere. Interestingly, the influence of stability-based cloud parameterizations on Arctic cloud amounts was a strong function of the mean atmospheric state. Because CAM4’s unrealistic Arctic stratus are most prominent in dry and stable atmospheric conditions, we expect that the influence of unrealistic cloud compensation for sea ice loss will be expressed only when specific atmospheric circulation regimes occur.
Our results highlight the immense utility of observations in a rapidly changing Arctic environment for evaluating and improving climate models. Comparison of observationally constrained forecasts with satellite data during a record low sea ice year was the key to unearthing the cloud parameterization flaw addressed in this study. When model evaluation efforts leverage observations and process understanding, they are in the best position to identify model biases and address obvious errors in climate model parameterizations. This work also shows the importance of evaluating climate models during extreme events and as a function of atmospheric state. While discovering an unrealistic Arctic cloud response produced by a subtropical cloud parameterization was alarming, the influence of the error on projected sea ice loss in coupled models using CLDST is likely confined to specific melt years with dry and stable atmospheric conditions.
While this study successfully identified, fixed, and evaluated the climate significance of a flawed cloud parameterization, many unsolved and unknown parameterization problems remain. Modeling, data assimilation, and physical intuition will provide important guidance for Arctic boundary layer parameterization development and evaluation; however, sustained high-quality boundary layer and surface radiation observations, especially over the newly ice-free Arctic Ocean, are essential for future climate model evaluation and improvement efforts.
Acknowledgments
JEK thanks Peter Caldwell, Gunilla Svensson, and Clara Deser for fruitful scientific discussions, Cecile Hannay and Dave Bailey for help in running and understanding model output, and Chris O’Dell for processing the MODIS data. This work was funded by the U.S. National Science Foundation through NCAR and by NASA Grant 08-MAP-117.
REFERENCES
Anderson, J., T. Hoar, K. Raeder, H. Liu, N. Collins, R. Torn, and A. Avellano, 2009: The data assimilation research testbed: A community facility. Bull. Amer. Meteor. Soc., 90 , 1283–1296.
Bitz, C. M., 2008: Some aspects of uncertainty in predicting sea ice retreat. Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications, Geophys. Monogr., Vol. 180, Amer. Geophys. Union, 63–76.
Boé, J., A. Hall, and X. Qu, 2009: Current GCMs’ unrealistic negative feedback in the arctic. J. Climate, 22 , 4682–4695.
Boville, B. A., P. J. Rasch, J. J. Hack, and J. R. McCaa, 2006: Representation of clouds and precipitation processes in the Community Atmosphere Model Version 3 (CAM3). J. Climate, 19 , 2184–2198.
Collins, W. D., and Coauthors, 2006: The Community Climate System Model version 3 (CCSM3). J. Climate, 19 , 2122–2143.
Deser, C., R. Tomas, M. Alexander, and D. Lawrence, 2010: The seasonal atmospheric response to projected Arctic sea ice loss in the late twenty-first century. J. Climate, 23 , 333–351.
Gettelman, A., V. P. Walden, L. M. Miloshevich, W. L. Roth, and B. Halter, 2006: Relative humidity over Antarctica from radiosondes, satellites, and a general circulation model. J. Geophys. Res., 111 , D09S13. doi:10.1029/2005JD006636.
Gettelman, A., and Coauthors, 2010: Global simulations of ice nucleation and ice supersaturation with an improved cloud scheme in the Community Atmosphere Model. J. Geophys. Res., 115 , D18216. doi:10.1029/2009JD013797.
Gorodetskaya, I. V., L. B. Tremblay, B. Liepert, M. A. Cane, and R. I. Cullather, 2008: The influence of cloud and surface properties on the Arctic Ocean shortwave radiation budget in coupled models. J. Climate, 21 , 866–882.
Hannay, C., and Coauthors, 2009: Evaluation of forecasted southeast pacific stratocumulus in the NCAR, GFDL, and ECMWF models. J. Climate, 22 , 2871–2889.
Higgins, M. E., and J. J. Cassano, 2009: Impacts of reduced sea ice on winter Arctic atmospheric circulation, precipitation, and temperature. J. Geophys. Res., 114 , D16107. doi:10.1029/2009JD011884.
Holland, M. M., and C. M. Bitz, 2003: Polar amplification of climate change in coupled models. Climate Dyn., 21 , 221–232.
Holland, M. M., C. M. Bitz, and B. Tremblay, 2006: Future abrupt reductions in the summer Arctic sea ice. Geophys. Res. Lett., 33 , L23503. doi:10.1029/2006GL028024.
Holland, M. M., D. B. Bailey, and S. V. Vavrus, 2011: Inherent sea ice predictability in the rapidly changing Arctic environment of the Community Climate System Model, version 3. Climate Dyn., in press, doi:10.1007/s00382-010-0792-4.
Holtslag, A., and B. Boville, 1993: Local versus nonlocal boundary layer diffusion in a global climate model. J. Climate, 6 , 1825–1842.
Hurrell, J., and Coauthors, 2008: A new sea surface temperature and sea ice boundary data set for the Community Atmosphere Model. J. Climate, 21 , 5145–5153.
Kay, J. E., and A. Gettelman, 2009: Cloud influence on and response to seasonal Arctic sea ice loss. J. Geophys. Res., 114 , D18204. doi:10.1029/2009JD011773.
Kay, J. E., T. L’Ecuyer, A. Gettelman, G. Stephens, and C. O’Dell, 2008: The contribution of cloud and radiation anomalies to the 2007 Arctic sea ice extent minimum. Geophys. Res. Lett., 35 , L08503. doi:10.1029/2008GL033451.
Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82 , 247–268.
Klein, S. A., and D. L. Hartmann, 1993: The seasonal cycle of low stratiform clouds. J. Climate, 6 , 1587–1606.
Liu, Y., S. Ackerman, B. Maddux, J. Key, and R. Frey, 2010: Errors in cloud detection over the Arctic using a satellite imager and implications for observing feedback mechanisms. J. Climate, 23 , 1894–1907.
Loeb, N. G., B. A. Wielicki, D. R. Doelling, G. L. Smith, D. F. Keyes, S. Kato, N. Manalo-Smith, and T. Wong, 2009: Toward optimal closure of the earth’s top-of-atmosphere radiation budget. J. Climate, 22 , 748–766.
Manabe, S., and R. J. Stouffer, 1980: Sensitivity of a global climate model to an increase of CO2 concentration in the atmosphere. J. Geophys. Res., 85 , 5529–5554.
Miller, R. L., 1997: Tropical thermostats and low cloud cover. J. Climate, 10 , 409–440.
Neale, R. B., M. Jochum, and J. H. Richter, 2008: The impact of convection on ENSO: From a delayed oscillator to a series of events. J. Climate, 21 , 5904–5924.
Phillips, T. J., and Coauthors, 2004: Evaluating parameterizations in general circulation models: Climate simulation meets weather prediction. Bull. Amer. Meteor. Soc., 85 , 1903–1915.
Platnick, S., M. D. King, S. A. Ackerman, W. P. Menzel, B. A. Baum, J. C. Riedi, and R. A. Frey, 2003: The MODIS cloud products: Algorithms and examples from Terra. IEEE Trans. Geosci. Remote Sens., 41 , 459–473.
Rasch, P. J., and J. E. Kristjánsson, 1998: A comparison of the CCM3 model climate using diagnosed and predicted condensate parameterizations. J. Climate, 11 , 1587–1614.
Rodwell, M. J., and T. N. Palmer, 2007: Using numerical weather prediction to assess climate models. Quart. J. Roy. Meteor. Soc., 133 , 129–146.
Serreze, M. C., A. P. Barrett, J. C. Stroeve, D. N. Kindig, and M. M. Holland, 2009: The emergence of surface-based Arctic amplification. Cryosphere Disuss., 3 , 11–19.
Shupe, M. D., and J. M. Intrieri, 2004: Cloud radiative forcing of the arctic surface: The influence of cloud properties, surface albedo, and solar zenith angle. J. Climate, 17 , 616–628.
Soden, B. J., I. M. Held, R. Colman, K. M. Shell, J. T. Kiehl, and C. A. Shields, 2008: Quantifying climate feedbacks using radiative kernels. J. Climate, 21 , 3504–3520.
Stephens, G. L., 2005: Cloud feedbacks in the climate system: A critical review. J. Climate, 18 , 237–273.
Stroeve, J., M. M. Holland, W. Meier, T. Scambos, and M. Serreze, 2007: Arctic sea ice decline: Faster than forecast. Geophys. Res. Lett., 34 , L09501. doi:10.1029/2007GL029703.
Teng, H., W. M. Washington, G. A. Meehl, L. A. Buja, and G. W. Strand, 2006: 21st century arctic climate change in the CCSM3 IPCC scenario simulations. Climate Dyn., 26 , 601–616.
Vavrus, S., and D. Waliser, 2008: An improved parameterization for simulating Arctic cloud amount in the CCSM3 climate model. J. Climate, 21 , 5673–5687.
Vavrus, S., D. Waliser, A. Schweiger, and J. Francis, 2008: Simulations of 20th and 21st century Arctic clouds in the global climate models assessed in the IPCC AR4. Climate Dyn., 33 , 1099–1115.
Wood, R., and C. S. Bretherton, 2006: On the relationship between stratiform low cloud cover and lower-tropospheric stability. J. Climate, 19 , 6425–6432.
CAM4 forecasting strategy. Throughout the entire month of interest, CAM4 forecasts were started every 6 h (at the stars) and run for 1 day (ending at the gray line). The 1-day forecasts were averaged to produce monthly-mean values for comparison with observations and sensitivity tests.
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
Observed and CAM4-forecasted (Jul07 − Jul06; see Table 1) July 2007 − July 2006 difference maps: (a) observed sea ice fraction difference (Hurrell et al. 2008); (b) sea level pressure difference from CAM4 forecasts; (c) total cloud difference from CAM4 forecasts; (d) total cloud difference from MODIS satellite data (Cloud_Fraction_Combined Level-3, collection 5 product; see Platnick et al. 2003); (e) near-surface stability difference from CAM4 forecasts; and (f) near-surface stability difference from V4 AIRS satellite data (Gettelman et al. 2006). Near-surface stability is defined as the potential temperature difference between 925 and 1000 mb.
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
As in Fig. 2, but for the observed and CAM4-forecasted (Sep07 − Sep06, see Table 1) September 2007 − September 2006 difference maps.
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
CAM4-forecasted response to prescribed July 2007 and September 2007 sea ice anomalies: (a) July 2007 sea ice fraction anomaly from Hurrell et al. (2008); (b) September 2007 sea ice fraction anomaly from Hurrell et al. (2008); (c) July 2007 low cloud response (Jul07 − Jul07_clim; Table 1); (d) September 2007 low cloud response (Sep07 − Sep07_clim; Table 1); (e) July 2007 near-surface stability response (Jul07 − Jul07_clim); and (f) September 2007 near-surface stability response (Sep07 − Sep07_clim). Near-surface stability is defined as the potential temperature difference between 925 and 1000 mb.
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
CAM4-forecasted cloud forcing response to prescribed July 2007 and September 2007 sea ice anomalies (Figs. 4a and 4b): (a) July 2007 top of atmosphere cloud forcing response (Jul07 − Jul07_clim; Table 1); (b) September 2007 top of atmosphere cloud forcing response (Sep07 − Sep07_clim; Table 1); (c) July 2007 surface cloud forcing response (Jul07 − Jul07_clim); and (d) September 2007 surface cloud forcing response (Sep07 − Sep07_clim). Averages are reported over areas that experienced at least 0.5 reductions in sea ice fraction (see Table 2 for more values). Cloud forcing is defined as the difference between the net all-sky flux and the net clear-sky flux.
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
Profiles of (left to right) temperature, humidity, total cloud fraction, and in-cloud water in CAM4 forecasts (Table 1): (a) July 2007 forecasts and (b) September 2007 forecasts. Profiles over open water are from forecasts with observed sea ice and ocean conditions (Jul07, Sep07), while profiles over sea ice are from forecasts with climatological sea ice and ocean conditions (Jul07_clim, Sep07_clim). Profiles are average values over grid cells that experienced at least a 0.5 sea ice fraction decrease relative to the 1982–2001 climatology.
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
Summer (JJA) histograms of maximum layer stability underlying stratus clouds (Γmax): (a) Arctic Ocean (70°–90°N) and (b) the California stratus region (20°–30°N, 120°–130°W). Values are from a CAM4 run with a freely evolving atmosphere and prescribed 2007 ocean conditions (CAM4_2007repeat; Table 4). Histograms of the maximum stability underlying the strongest inversion derived from diagnostic runs with no CLDST were qualitatively similar (not shown).
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
The July 2007 − July 2006 total cloud differences: (a) CAM4 forecasts (Jul07 − Jul06; see Table 1), (b) CAM4 forecasts with CLDST_MIXBL (Jul07_cldst_mixbl − Jul06_cldst_mixbl), and (c) MODIS observations (Platnick et al. 2003).
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
Influence of stability-based cloud parameterizations on global annual-mean values in freely evolving CAM4 runs (Table 4): (a) CLDST on low cloud fraction (CAM4_ 2007repeat − CAM4_nocldst_2007repeat), (b) CLDST_MIXBL on low cloud fraction (CAM4_ 2007repeat − CAM4_cldst_mixbl_2007repeat), (c) CLDST on TOA cloud forcing, and (d) CLDST_MIXBL on TOA cloud forcing.
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
July low cloud fraction response to prescribed July 2007 sea ice loss (Fig. 4a): (a) CAM4 forecasts (Jul07 − Jul07_clim; Table 1) and (b) CAM4 runs with a freely evolving atmosphere (CAM4_2007repeat − CAM4_clim; Table 4).
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
(left) July atmospheric conditions in 2007 CAM4 forecasts (Jul07; Table 1) and (right) in CAM4 runs with a freely evolving atmosphere (CAM4_2007repeat; Table 4): (a) Jul07 sea level pressure, (b) CAM4_2007repeat July sea level pressure, (c) Jul07 near-surface stability, (d) CAM4_2007repeat near-surface stability, (e) Jul07 near-surface relative humidity, and (f) CAM4_2007repeat near-surface relative humidity. Near-surface stability is defined as the potential temperature difference between 925 and 1000 mb. Near-surface humidity is defined as the average relative humidity in the bottom three CAM4 levels (surface to ∼900 mb at sea level).
Citation: Journal of Climate 24, 2; 10.1175/2010JCLI3651.1
CAM4 forecast experiments. All forecasts were started from observationally constrained initial conditions produced by DART using CAM4 with no code modifications. Climatological sea surface temperature and sea ice extent boundary conditions are from 1982–2001 averages.
CAM4 runs with a freely evolving atmosphere. Climatological sea surface temperature and sea ice extent boundary conditions are from 1982–2001 averages.
Boundary layer properties and radiative fluxes in 2007 CAM4 forecasts.a
Influence of CLDST_MIXBL on boundary layer properties and radiative fluxes in 2007 CAM4 forecasts.a
