Date of Award

January 2018

Document Type


Degree Name

Doctor of Philosophy (PhD)


Atmospheric Sciences

First Advisor

Aaron D. Kennedy

Second Advisor

Xiquan Dong


The cloud and radiative properties simulated in an assortment of global climate models (GCMs) and reanalyses are examined to identify and assess systematic biases based upon comparisons with multiple satellites observations and retrievals. The global mean total column cloud fraction (CF) simulated by the 33-member multimodel mean is 7% and 17% lower than the results from passive (Moderate Resolution Infrared Spectroradiometer, MODIS) and active (CloudSat and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation, CALIPSO) satellite remote sensing platforms. The simulated cloud water path (CWP), which is used as a proxy for optical depth, on global average, has a negative bias of ~17 g m–2. Despite these errors in simulated cloud properties, the simulated top-of-atmosphere (TOA) radiation budgets match relatively well with Clouds and the Earth Radiant Energy System (CERES) measurements. The biases in multimodel mean global TOA reflected shortwave (SW) and outgoing longwave (LW) fluxes and cloud radiative effects (CREs) are less than 2.5 W m–2. Nevertheless, when assessing models individually, some physically inconsistent results are evident. For example, in the ACCESS1.0 model, the simulated TOA SW and LW fluxes are within 2 W m–2 of the observed global means, however, the global mean CF and CWP are underpredicted by ~10% and ~25 g m–2, respectively. These unphysical model biases suggest tuning of the modeled radiation budgets.

Two dynamically-driven regimes, based on the atmospheric vertical motion at 500 hPa (ω500), are identified to provide a more quantitative measure of error in the radiation fields determined separately by biases in CF and CWP. These error types include the regime-averaged biases, biases in the sensitivity of TOA CREs to CF/CWP, and their co-variations. Overall, the biases in simulated CF and CWP are larger in the descent regime (ω500 > 25 hPa day–1) than in the ascent regime (ω500 < –25 hPa day–1), but are better correlated with observations. According to CERES observations, the sensitivity of LW CRE to CF is stronger in the ascent regime than in the decent regime (0.82 vs. 0.23 W m–2 %–1) and the multimodel mean overestimates this value by ~40%. The difference in sensitivity of SW CRE to CF between the two regimes is less drastic (–1.34 vs. –1.12 W m–2 %–1). TOA CREs rely independently on CWP in regions of large scale ascent and decent, as their sensitivities are similar between these two regimes (e.g., SW CRE/CWP = –0.28 W g–1 for both regimes). In general, the total TOA CRE errors are heavily weighted by their biases in simulated sensitivity and biases in the simulated CF.

A new observationally-constrained, data product is generated that can be used as a process-oriented diagnostic tool to further identify errors in simulated cloud and radiation fields. Based on the CloudSat and CALIPSO Ice Cloud Property Product (2C-ICE), and through one-dimensional radiative transfer modeling, a global database of radiative heating rate profiles is produced for non-precipitating single-layered ice clouds. Non-precipitating single-layered ice clouds have a global occurrence frequency of ~18% with considerable frequency in the tropical upper troposphere (13–16 km). A variety of ice cloud types exist in the sample of single-layered ice clouds developed here, which is determined by the distribution on cloud-top temperatures (CTT). For example, a peak in the distribution near 190 K (260 K) suggests the existence of cirrus (glaciated ice) clouds. The ice cloud microphysical properties responsible for having the largest impact on radiation (e.g., ice water content [IWC] and effective radius [Re]) are largest in the tropics and mid-latitudes according to 2C-ICE. Accordingly, this is where the strongest TOA SW absorption, and subsequently, the strongest upper tropospheric net radiative heating (> 1.5 K day–1) occurs. This newly generated product will provide the data for which new ice cloud parameterizations can be developed in global models.