Photochemistry and transport of tropospheric ozone and its precursors in urban and remote environments

Tropospheric ozone (O3) adversely affects human health, reduces crop yields, and contributes to climate forcing. To limit these effects, the processes controlling O3 abundance as well as that of its precursor molecules must be fully characterized. Here, I examine three facets of O3 production, both in heavily polluted and remote environments. First, using in situ observations from the DISCOVER-AQ field campaign in the Baltimore/Washington region, I evaluate the emissions of the O3 precursors CO and NOx (NOx = NO + NO2) in the National Emissions Inventory (NEI). I find that CO/NOx emissions ratios derived from observations are 21% higher than those predicted by the NEI. Comparisons to output from the CMAQ model suggest that CO in the NEI is accurate within 15 ± 11%, while NOx emissions are overestimated by 51-70%, likely due to errors in mobile sources. These results imply that ambient ozone concentrations will respond more efficiently to NOx controls than current models suggest. I then investigate the source of high O3 and low H2O structures in the Tropical Western Pacific (TWP). A combination of in situ observations, satellite data, and models show that the high O3 results from photochemical production in biomass burning plumes from fires in tropical Southeast Asia and Central Africa; the low relative humidity results from large-scale descent in the tropics. Because these structures have frequently been attributed to mid-latitude pollution, biomass burning in the tropics likely contributes more to the radiative forcing of climate than previously believed. Finally, I evaluate the processes controlling formaldehyde (HCHO) in the TWP. Convective transport of near surface HCHO leads to a 33% increase in upper tropospheric HCHO mixing ratios; convection also likely increases upper tropospheric CH3OOH to ~230 pptv, enough to maintain background HCHO at ~75 pptv. The long-range transport of polluted air, with NO four times the convectively controlled background, intensifies the conversion of HO2 to OH, increasing OH by a factor of 1.4. Comparisons between the global chemistry model CAM-Chem and observations show that consistent underestimates of HCHO by CAM-Chem throughout the troposphere result from underestimates in both NO and acetaldehyde.

Learning Binary Code Representations for Effective and Efficient Image Retrieval

The size of online image datasets is constantly increasing. Considering an image dataset with millions of images, image retrieval becomes a seemingly intractable problem for exhaustive similarity search algorithms. Hashing methods, which encodes high-dimensional descriptors into compact binary strings, have become very popular because of their high efficiency in search and storage capacity. In the first part, we propose a multimodal retrieval method based on latent feature models. The procedure consists of a nonparametric Bayesian framework for learning underlying semantically meaningful abstract features in a multimodal dataset, a probabilistic retrieval model that allows cross-modal queries and an extension model for relevance feedback. In the second part, we focus on supervised hashing with kernels. We describe a flexible hashing procedure that treats binary codes and pairwise semantic similarity as latent and observed variables, respectively, in a probabilistic model based on Gaussian processes for binary classification. We present a scalable inference algorithm with the sparse pseudo-input Gaussian process (SPGP) model and distributed computing. In the last part, we define an incremental hashing strategy for dynamic databases where new images are added to the databases frequently. The method is based on a two-stage classification framework using binary and multi-class SVMs. The proposed method also enforces balance in binary codes by an imbalance penalty to obtain higher quality binary codes. We learn hash functions by an efficient algorithm where the NP-hard problem of finding optimal binary codes is solved via cyclic coordinate descent and SVMs are trained in a parallelized incremental manner. For modifications like adding images from an unseen class, we propose an incremental procedure for effective and efficient updates to the previous hash functions. Experiments on three large-scale image datasets demonstrate that the incremental strategy is capable of efficiently updating hash functions to the same retrieval performance as hashing from scratch.

STRUCTURE AND FUNCTION OF THE GROUP III CHAPERONINS, A UNIQUE CLADE OF PROTEIN FOLDING NANOMACHINES

The survival and descent of cells is universally dependent on maintaining their proteins in a properly folded condition. It is widely accepted that the information for the folding of the nascent polypeptide chain into a native protein is encrypted in the amino acid sequence, and the Nobel Laureate Christian Anfinsen was the first to demonstrate that a protein could spontaneously refold after complete unfolding. However, it became clear that the observed folding rates for many proteins were much slower than rates estimated in vivo. This led to the recognition of required protein-protein interactions that promote proper folding. A unique group of proteins, the molecular chaperones, are responsible for maintaining protein homeostasis during normal growth as well as stress conditions. Chaperonins (CPNs) are ubiquitous and essential chaperones. They form ATP-dependent, hollow complexes that encapsulate polypeptides in two back-to-back stacked multisubunit rings, facilitating protein folding through highly cooperative allosteric articulation. CPNs are usually classified into Group I and Group II. Here, I report the characterization of a novel CPN belonging to a third Group, recently discovered in bacteria. Group III CPNs have close phylogenetic association to the Group II CPNs found in Archaea and Eukarya, and may be a relic of the Last Common Ancestor of the CPN family. The gene encoding the Group III CPN from Carboxydothermus hydrogenoformans and Candidatus Desulforudis audaxviator was cloned in E. coli and overexpressed in order to both characterize the protein and to demonstrate its ability to function as an ATPase chaperone. The opening and closing cycle of the Chy chaperonin was examined via site-directed mutations affecting the ATP binding site at R155. To relate the mutational analysis to the structure of the CPN, the crystal structure of both the AMP-PNP (an ATP analogue) and ADP bound forms were obtained in collaboration with Sun-Shin Cha in Seoul, South Korea. The ADP and ATP binding site substitutions resulted in frozen forms of the structures in open and closed conformations. From this, mutants were designed to validate hypotheses regarding key ATP interacting sites as well as important stabilizing interactions, and to observe the physical properties of the resulting complexes by calorimetry.