Search the Community
Showing results for tags 'climate variability; cloud radiative feedbacks; ocean heat transport; sea surface temperature; clouds'.
Found 1 result
Vice-Regent posted a topic in Climate ChangeThe Roles of Cloud Radiative Feedbacks and Ocean Heat Transport in Climate Variability Eleanor Middlemas, University of Miami https://scholarlyrepository.miami.edu/cgi/viewcontent.cgi?article=3225&context=oa_dissertations Abstract Contributions from dynamical ocean heat transport and cloud radiative feedbacks to internal variations in sea surface temperature (SST) are explored using various climate modeling configurations in conjunction with intermodel comparison. The experimental modeling configurations include those with ocean heat transport and/or cloud radiative feedbacks disabled, resulting in a different, freely-evolving climate. By comparing the variability between experiment and control simulations, it is determined that the influence from either ocean dynamics or cloud feedback depends on the region and timescale. To begin, global-scale decadal changes in temperature do not require dynamical ocean heat transport, but instead, low-frequency variations in global mean temperature (GMT) may arise from atmospheric noise interacting with the mixed layer of the ocean. This contradicts numerous studies suggesting that anomalous deep ocean heat uptake is responsible for decadal cooling in GMT that led to the global warming hiatus. On the other hand, it is well- known that El Nino Southern Oscillation (ENSO) is driven by ocean dynamics and may lead to decadal variations in global climate. We find that around half of model disagreement in the magnitude of decadal variability can be explained by model spread in ENSO magnitude, which suggests other processes are contributing to decadal variability in climate models, like cloud radiative feedbacks, for example. Cloud radiative feedbacks as a source of decadal variability is largely overlooked in the literature and could explain some of the low frequency SST variability produced in a climate model configuration without ocean dynamics. To understand cloud radiative feedbacks’ impact on SST variability, we implement cloud-locking in the National Center for Atmospheric Research (NCAR) Community Atmosphere Model, version 5 (CAM5). This method disables cloud radiative feedbacks by prescribing clouds in the radiation module, preventing them from interacting with their environment, but allows the climate to evolve freely and, therefore, enables analysis of the variability without cloud radiative feedbacks. There are many practical uses for cloud- locking, but this is one of the few studies that utilizes the tool for understanding the relationship between cloud radiative feedbacks and internal variability. SST variability changes in the tropics and subtropics due to cloud-locking are the focus of half of this dissertation. Disabling cloud radiative feedbacks produces the largest changes in global SST variability via ENSO. Without cloud radiative feedbacks, the periodicity of ENSO events shifts from 2-7 years to approximately 10 years. Damping by negative shortwave feedback, which aids in the termination of ENSO, is not active in the cloud-locking configuration, so each ENSO event continues to grow unabated until the ocean “discharges”. Furthermore, in the northeastern subtropical ocean basins, a region important for decadal modes of variability, we find that positive cloud radiative feedback enhances SST variability by up to 40%, but that the magnitude depends on the role of the ocean. In the Atlantic Ocean, dynamical ocean heat transport enhances positive SST anomalies, leaving a smaller role for positive cloud radiative feedback. In the Pacific Ocean, ocean heat transport damps positive SST anomalies, so positive cloud radiative feedback competes with heat flux from the ocean. Altogether, the results of this dissertation suggest that more attention should be paid to the role of cloud radiative feedbacks when examining mechanisms of variability. The climate is very sensitive to cloud radiative feedbacks in the model used in this study, and different models and configurations exhibit diverse sensitivities. The spatially varying contribution of ocean dynamics further complicates the importance of cloud feedbacks. This study motivates an intermodel comparison of cloud-locking experiments to better constrain the role of cloud radiative feedbacks on the climate system. Conclusions Intermodel spread in low-frequency SST variability confounds our projections of future climate change and requires deeper understanding of mechanisms of variability. Through the use of different configurations of a state-of-the-art model, this dissertationclarifies the respective roles of two potential contributors to SST variability: dynamical ocean heat transport and cloud radiative feedbacks. The main finding of this dissertation is that cloud radiative feedbacks can significantly enhance or damp SST variability and that their role can be modulated by ocean dynamics. Their corresponding contributions depend on the region. This is one of only a few studies to consider both roles of ocean dynamics and cloud radiative feedback at once. In chapter 2, we pose that dynamical ocean heat transport is not required in order to produce decadal-timescale swings in global mean temperature. Numerous studies suggest that the global warming hiatus was caused by enhanced oceanic heat uptake by anomalous ocean heat transport (Balmaseda et al. 2013, Chen and Tung 2014, Meehl et al. 2011, Meehl et al. 2013, Watanabe et al. 2013, Trenberth et al. 2014, Trenberth and Fasullo 2013, Palmer et al. 2011, Lee et al. 2015). We find that an AGCM coupled to a slab ocean model without ocean dynamics can produce changes in GMT of the same magnitude of the hiatus, albeit less frequently than a fully-coupled model. Also, many assume that ENSO drives decadal variability and that intermodel spread in ENSO may explain the disagreement among models in decadal variability. We find that the spread in ENSO variance only explains half of the spread in decadal variance across CMIP5 models. These two findings together motivate further investigation into atmospheric processes that may contribute to decadal variability and the intermodel spread in decadal variability. We hypothesized that cloud radiative feedbacks may contribute to models’ decadal variability based on the fact that cloud radiative feedbacks dominate intermodel spread in climate sensitivity and are tied to low-frequency variability in eastern ocean basins. We isolate their role in SST variability by turning them off with cloud-locking, or prescribing clouds in the radiation module of an AGCM. Chapter 3 describes the implementation of cloud-locking for the first time in NCAR’s AGCM, CAM5. Here, we test the model’s responsiveness to different ways of cloud-locking in the radiation module of CAM5. We find that cloud-locking induces a systematic -2 W/m2at the top of atmosphere (TOA), which leads to climate drift in model configurations with interactive sea ice and SST. Due to the large changes in climatological SST along the sea-ice lines with cloud-locking, we focus on understanding cloud feedbacks’ role in SST variability in the tropics and subtropics where the drift was less than one degree Celsius. We subsequently discovered that, by locking CLDFSNOW in addition to the eight cloud parameters locked for experiments in Chapters 4 and 5, we eliminate the TOA imbalance and, thus, the climate drift in the fully-coupled simulation as well.Chapter 4 shows that cloud-locking in CESM results in an enhanced ENSO on decadal timescales. This is a completely different ENSO response than cloud-locking in a different fully-coupled model, MPI-ESM-LR of the Max Planck Institute. In MPI-ESM-LR, cloud-locking reveals a weaker ENSO due to a lack of positive longwave cloud feedback on circulation. Cloud-locking in CESM, on the other hand, reveals a stronger ENSO because negative shortwave cloud feedback was not able to damp SST variability. Shortwave feedback is the largest source of intermodel spread in ENSO magnitude among atmospheric flux feedbacks on ENSO, and CESM produces one of the most realistic shortwave feedbacksamong climate models. This means that shortwave damping is probably important for damping ENSO in reality, but the role of shortwave feedback in climate models depends on other ENSO feedbacks, like the Bjerknes feedback and convection scheme. The results of this chapter call for further investigation of cloud-circulation feedbacks and their impact on ENSO variability across models. The fact that we obtained a completely different result using a different model highlights the vast differences in ENSO atmospheric feedbacks that exist among models. Accounting for these intermodel differences may lead to improved ENSO predictions.In Chapter 5, we investigate another region where observations show a role for cloud radiative feedback in SST variability: the northeastern subtropical ocean basins. Here, we find that ocean dynamics can alter the role of cloud radiative feedback. Positive cloud radiative feedback from low-lying stratocumulus cloud decks is well-documented in observations and has been identified in climate models, though the extent that SST variability depends on cloud feedback is unknown. We cloud-lock in an AGCM-slab configuration and compare to a cloud-locked fully-coupled simulation to find that the role of cloud feedback on SST is modulated by dynamical ocean heat transport by mixing or advection. The influence from ocean heat transport depends on the region, and the differences result in a slightly different magnitudes of cloud radiative feedback. In the Pacific, the ocean damps SST variability, while in the Atlantic, ocean heat transport enhances SST variability. Therefore, positive cloud radiative feedback is a larger contributor to SST variance in the Pacific than in the Atlantic. The atmospheric feedbacks are analogous, but the role of the ocean differs between basins. Given the importance of the northeastern subtropics in connecting the extratropics to the tropics, the mechanisms presented here may shed light on understanding low frequency modes of variability.With every experiment comes some limitations. In this body of work, we only analyze results from one climate model out of tens of other state-of-the-art CMIP5 models. The role for ocean dynamics and cloud radiative feedback for SST variability that we find using different versions and configurations of the NCAR model may be unique to that model. We know that CAM produces more realistic cloud radiative feedbacks with respect to ENSO compared to other models (Lloyd et al. 2012, Bellenger et al. 2014, Ferrett et al. 2018), but the accuracy ofNCAR model’s representation of ocean dynamics is unclear. Furthermore, decadal variability is difficult to study due to the limited length and spatial coverage of the observational record. Additional coverage of observations of the deep ocean (below 700 meters) is limited, so decadal changes in the deep ocean are inferred through energy estimates rather than directly measured (Trenberth et al. 2014), leaving the precise role of ocean heat transport an open question. Though, even with more complete ocean observations at shallower depths in the Atlantic, for example, there is still debate about the role of ocean dynamics in the Atlantic Multidecadal Variability (Clement et al. 2015, Zhang et al. 2015). Based on the simulations conducted, ocean dynamics and cloud radiative feedbacks are likely both contributing to variability on interannual-and-longer timescales. On interannual timescales, ENSO is certainly driven by ocean dynamics, but negative shortwave cloud feedback is important for maintaining the oscillation’s magnitude and timescale. The balance between cloud feedback and ocean dynamics on sea surface temperature probably depends on the strength of other feedbacks which vary among models. But the spread in ENSO only explains half of intermodel spread in decadal variability, suggesting that other processes are important for decadal variability in climate models. Positive cloud radiative feedbacks in northeast subtropical ocean basins enhances SST variability by up to 40%, but its precise contribution depends on the region. SST in the Pacific, hosting the Interdecadal Pacific Oscillation, is merely damped by ocean processes, and so positive cloud radiative feedback is very important for low-frequency variations in SST there. In the Atlantic, on the other hand, ocean dynamics drive SST variations, so clouds have less of an impact on SST variability. More work is required to understand respective contributions of ocean and atmosphere to extratropical variability, but this dissertation elucidates atmosphere-ocean processes that can result in internal variations in SST.The work put forth in this dissertation has resulted in an additional useful tool for understanding cloud radiative feedbacks as well as highlighted the importance of simplifying climate models to understand mechanisms. A grand challenge in climate modeling is to understand the relationship between cloud radiative feedbacks and circulation (Bony et al. 2015), and cloud-locking provides an opportunity to pursue this challenge. This new implementation in widely-used CESM can open many doors for research concerning cloud radiative feedbacks and circulation. Furthermore, this dissertation demonstrates the significance in using simpler modeling configurations to understand mechanisms of climate change and variability. By deactivating various climate system components, one can clearly see the function that these components fulfill, which, in some cases, provides a different answer than a multimodel analysis could reveal. ----------------------------------------------------------------------------------------------------------------------- Future Work Upon completion of this work, questions still remain about the role of cloud radiative feedbacks in climate sensitivity and climate variability. This chapter outlines the unanswered questions and proposes a few suggestions that utilize cloud-locking as a tool to tackle these questions. First and foremost, now that we understand that the source of the TOA imbalance induced by cloud-locking described in Chapter 3 is due to not locking CLDFSNOW, our next step is to reproduce the presented experiments with that additional parameter locked. The climate drift in the presented experiments is a limitation for potential users because they cannot separate whether their results are due to cloud-locking or due to climate drift, unless they focus their analysis on regions that have small drift, like the tropics and subtropics (as in Chapters 4 and 5). Furthermore, there are additional cloud-locking experiments to be conducted to understand the sensitivity of CESM’s climate to cloud radiative effects: (1)Prescribing various cloud levels individually to determine if a certain type of cloud leads to drift (2)Prescribing individual cloud fields as an extension of section 3.3.3 to isolate the contribution of processes associated with each cloud field to climate drift(3)Shuffling cloud fields from multiple years in time to account for potential bias introduced by selecting a particular cloud year as well as decorrelating cloud fields in time. For example, maybe the year selected for cloud may be a particularly stormy one. No radiative impacts from storms will be included if cloud fields have no temporal autocorrelation.(4)Directly prescribing atmospheric shortwave and longwave cloud radiative heating rates instead of cloud fields. This circumvents any potential perturbations that may be introduced by prescribing individual fields in the radiation code, since cloud radiative heating rates is the output from the radiation module that the rest of the climate model utilizes for other calculations.Considering the fact that cloud-locking in different ways and in different climate models results in varying magnitudes of drift (both positive and negative) suggests that intermodel comparison of cloud-locking may be useful in determining the contribution of cloud feedback on circulation to climate sensitivity. Previous cloud-locking experiments in other climate models exhibit similar climate sensitivities and drift (Zhang et al. 2010, Mauritsen et al. 2013, Radel et al. 2016), but each model resulted in different climate drift. Cloud-locking in the GFDL model, version 2.1, results in a warming in the poles actually warm by around +1-2ºC, while the rest of the planet cools by about 0.5ºC. There is no report of the global mean climate drift. Cloud-locking in the MPI-ESM-LR model also results in a warming of around +1ºC over the course of 250 years. Cloud-locking in CESM1.2, the model used in this dissertation, results in a global cooling of around 1ºC over the course of 400 years. Furthermore, an intermodel comparison of cloud-locking simulations would behugely beneficial in isolating other climate model biases. Preliminary results in CESM2 show that cloud-locking actually alleviates biases in precipitation, wind, and temperature over the western warm pool (personal communication with Jim Benedict). Furthermore, the large signal in the Southern Ocean leads to shifts in the ITCZ, suggesting that cloud-circulation feedbacks may impact models’ ITCZ bias. Both of these examples further motivate the use of cloud-locking as a tool for understanding the role of cloud radiative feedbacks in climate model biases and intermodel spread in climate sensitivity.There is also large climate sensitivity to cloud-locking in the extratropics and polar regions. AGCM-slab models produce significant decadal variability in extratropical and polar regions (fig. 2.3) mainly over land in the northern hemisphere and over ocean in the southern hemisphere, and these regions are large contributors to global-mean decadal variations (Brown et al. 2015). Internal atmospheric variability significantly contributes to decadal variability in these regions (Brown et al. 2015). Due to the increase in sea ice concentration in response to cloud-locking in the AGCM-slab configuration, a sea-ice-cloud-albedo feedback may be impacting the climate sensitivity. Also, the Southern Ocean is one of the locations where negative TOA imbalance induced by cloud-locking is largest in the configuration with prescribed SSTs (fig. 3.11), suggesting that some cloud feedback interact may interact with sea ice to enhance the global cooling in AGCM-slab and fully-coupled simulations. The other regions of largest TOA imbalance include the land in tropical regions. This raises questions about the importance of cloud-circulation feedbacks for land surface temperature, and given the prominence of convection in those regions, cloud-circulation feedbacks are probably important for precipitation as well. An approach to investigate the importance of cloud-circulation feedbacks for the climate in these regions is to limit cloud-locking to certain regions. Mechanisms of internal variability and the sign and magnitude of cloud radiative feedbacks depend on the region. Already, one study has cloud-locked according to latitude in various idealized models (aquaplanet configurations) to understand cloud radiative feedbacks’ impact on atmospheric circulation and the resulting impact on the global warming response (Voigt and Shaw 2016). These authors found that the magnitude of the atmospheric circulation and mean climate response dependson the model. One could apply cloud-locking in various regions to identify which regions lead to largest climate sensitivity, and then motivate future investigations into cloud feedbacks’ impact on climate sensitivity. Ultimately, the goal of understanding mechanisms of internal variability is to better constrain our limits of climate predictability and improve predictions. Some uncertainty in predictions due to internal variability may be irreducible due to the chaotic nature of the climate system, but given intermodel spread in internal variability, we may reduce some of this uncertainty. Applying cloud-locking in different models, for example, may aid in locating climate model biases regarding cloud-circulation feedbacks. Also, I did not have the chanceto pursue questions regarding cloud radiative feedbacks’ direct influence on climate predictability. How does the models’ representation of cloud feedback impact climate predictability? Does the intermodel spread in cloud radiative feedback impact projections of climate change? For example, a positive cloud radiative feedback with increased greenhouse gases would suggest that global mean temperature would warm faster with warming due to cloud radiative feedback. But if positive cloud radiative feedback also means greater SST persistence, and thus, greater SST variability, does this translate to larger ensemble spread in single-model projections of climate? How would this change the signal-to-noise ratio? We have just scratched the surface of the breadth ofknowledge regarding cloud feedbacks and their role in climate projections and related uncertainty.