Physical factors influencing the distribution of a top predator in a subtropical oligotrophic estuary

We used longline fishing to determine the effects of distance from the ocean, season, and short-term variation in abiotic conditions on the abundance of juvenile bull sharks (Carcharhinus leucas) in an estuary of the Florida Everglades, U.S.A. Logistic regression revealed that young-of-the-year sharks were concentrated at a protected site 20 km upstream and were present in greater abundance when dissolved oxygen (DO) levels were high. For older juvenile sharks (age 1+), DO levels had the greatest influence on catch probabilities followed by distance from the ocean; they were most likely to be caught at sites with .3.5 mg L21 DO and on the main branch of the river 20 km upstream. Salinity had a relatively small effect on catch rates and there were no seasonal shifts in shark distribution. Our results highlight the importance of considering DO as a possible driver of top predator distributions in estuaries, even in the absence of hypoxia. In Everglades estuaries hydrological drivers that affect DO levels (e.g., groundwater discharge, modification of primary productivity through nutrient fluxes) will be important in determining shark distributions, and the effects of planned ecosystem restoration efforts on bull sharks will not simply be mediated by changing salinity regimes and the location of the oligohaline zone. More generally, variation in DO levels could structure the nature and spatiotemporal pattern of top predator effects in the coastal Everglades, and other tropical and subtropical estuaries, because of interspecific variation in reliance on DO within the top predator guild.

Text analytics of social media: Sentiment analysis, event detection and summarization

In the last decade, large numbers of social media services have emerged and been widely used in people's daily life as important information sharing and acquisition tools. With a substantial amount of user-contributed text data on social media, it becomes a necessity to develop methods and tools for text analysis for this emerging data, in order to better utilize it to deliver meaningful information to users. ^ Previous work on text analytics in last several decades is mainly focused on traditional types of text like emails, news and academic literatures, and several critical issues to text data on social media have not been well explored: 1) how to detect sentiment from text on social media; 2) how to make use of social media's real-time nature; 3) how to address information overload for flexible information needs. ^ In this dissertation, we focus on these three problems. First, to detect sentiment of text on social media, we propose a non-negative matrix tri-factorization (tri-NMF) based dual active supervision method to minimize human labeling efforts for the new type of data. Second, to make use of social media's real-time nature, we propose approaches to detect events from text streams on social media. Third, to address information overload for flexible information needs, we propose two summarization framework, dominating set based summarization framework and learning-to-rank based summarization framework. The dominating set based summarization framework can be applied for different types of summarization problems, while the learning-to-rank based summarization framework helps utilize the existing training data to guild the new summarization tasks. In addition, we integrate these techneques in an application study of event summarization for sports games as an example of how to better utilize social media data. ^

Text Analytics of Social Media: Sentiment Analysis, Event Detection and Summarization

In the last decade, large numbers of social media services have emerged and been widely used in people's daily life as important information sharing and acquisition tools. With a substantial amount of user-contributed text data on social media, it becomes a necessity to develop methods and tools for text analysis for this emerging data, in order to better utilize it to deliver meaningful information to users. Previous work on text analytics in last several decades is mainly focused on traditional types of text like emails, news and academic literatures, and several critical issues to text data on social media have not been well explored: 1) how to detect sentiment from text on social media; 2) how to make use of social media's real-time nature; 3) how to address information overload for flexible information needs. In this dissertation, we focus on these three problems. First, to detect sentiment of text on social media, we propose a non-negative matrix tri-factorization (tri-NMF) based dual active supervision method to minimize human labeling efforts for the new type of data. Second, to make use of social media's real-time nature, we propose approaches to detect events from text streams on social media. Third, to address information overload for flexible information needs, we propose two summarization framework, dominating set based summarization framework and learning-to-rank based summarization framework. The dominating set based summarization framework can be applied for different types of summarization problems, while the learning-to-rank based summarization framework helps utilize the existing training data to guild the new summarization tasks. In addition, we integrate these techneques in an application study of event summarization for sports games as an example of how to better utilize social media data.

The Role of Teleost Grazers in a Relatively Pristine Seagrass Ecosystem

Trophic downgrading of ecosystems necessitates a functional understanding of trophic cascades. Identifying the presence of cascades, and the mechanisms through which they occur, is particularly important for seagrass meadows, which are among the most threatened ecosystems on Earth. Shark Bay, Western Australia provides a model system to investigate the potential importance of top-down effects in a relatively pristine seagrass ecosystem. The role of megagrazers in the Shark Bay system has been previously investigated, but the role of macrograzers (i.e., teleosts), and their importance relative to megagrazers, remains unknown. The objective of my dissertation was to elucidate the importance of teleost macrograzers in transmitting top-down effects in seagrass ecosystems. Seagrasses and macroalgae were the main food of the abundant teleost Pelates octolineatus, but stable isotopic values suggested that algae may contribute a larger portion of assimilated food than suggested by gut contents. Pelates octolineatus is at risk from numerous predators, with pied cormorants (Phalacrocorax varius) taking the majority of tethered P. octolineatus. Using a combination of fish trapping and unbaited underwater video surveillance, I found that the relative abundance of P. octolineatus was greater in interior areas of seagrass banks during the cold season, and that the mean length of P. octolineatus was greater in these areas compared to along edges of banks. Finally, I used seagrass transplants and exclosure experiments to determine the relative effect of megagrazers and macrograzers on the establishment and persistence of three species of seagrasses in interior microhabitats. Teleost grazing had the largest impact on seagrass species with the highest nutrient content, and these impacts were primarily observed during the warm season. My findings are consistent with predictions of a behaviorally-mediated trophic cascade initiated by tiger sharks (Galeocerdo cuvier) and transmitted through herbivorous fishes and their predators.