UC Merced

Leaf-cutter ants are one of the most conspicuous inhabitants of New World forests and plantations. They amaze visitors and worry farmers when thousands of them march in endless parades carrying leaf fragments to their massive underground nests, or when they go back in the opposite direction to collect more. However, rather than eating the hundreds of kilograms of vegetation they harvest each year, they shred them to create a substrate to feed a fungus that has been their fundamental diet for 50 Ma. They are indeed the first farmers on Earth’s natural history and, as any farmer, they have learned to optimize the conditions required by their gardens by engineering their surroundings. Here we present a series of studies designed to shed light on the effects of leaf-cutter ants on soil CO2 dynamics in Neotropical soils, an important part of the rainforest carbon cycle. We studied soil CO2 concentrations at different depths in several nonnest, nest, and abandoned nest soils for three years to understand the seasonal effects of the nest structure in soil CO2. In two selected locations, we monitored soil CO2 concentrations at high frequency (every 30 minutes) along with soil moisture and soil temperature to understand the effect of weather in the short-term, and how the nest presence impacts their dynamics. In addition, we measured soil surface CO2 efflux with closed chambers, and nest vent efflux with our own novel flow-through chambers, which we describe for the first time, that we equipped with thermocouples to monitor temperature gradients. We present statistical and conceptual models to account for differences in soil CO2 and to understand the fluid dynamics of CO2 in nests. Nest soils exhibited lower CO2 accumulation than nonnest soils for the same precipitation amounts. During wet periods, soil CO2 concentrations increased across all depths, but were significantly less in nest than in nonnest soils. Differences were nonsignificant during drier periods. In the short-term, precipitation events impacted soil CO2 concentration more than any other variable, and dramatically increased tortuosity, which leaded to the observed seasonal increases of soil CO2 concentration during wet periods. Surface efflux was equal across nest and nonnest plots (5 μmol CO2 m-2 s-1), suggesting that nest soils do not have enhanced surface emissions. However, vent efflux was substantially (103 to 105 times) greater and followed a diel pattern driven by free convection (warm and moist, less dense air rises out the nest more markedly at night). Episodic wind-forced convection events also provide supplemental ventilation during the day. Nest tunnel CO2 concentrations were less than in soil, suggesting CO2 efflux from the soil matrix into the nest. This is supported by the short-term diel pattern showed in nest soil CO2 concentration that did not occur in nonnest soils, except for a very dry period (El Niño, 2016). Thanks to the nest structure, the nest air is better ventilated than the soil, and CO2 produced in the soil matrix finds a faster way out of the soil through the nest tunnels. The diel pattern in nest vent CO2 efflux seems to regulate the diffusion of CO2 from the soil matrix by affecting the CO2 concentration gradient. These findings indicate that leaf-cutter ant nests provide alternative transport pathways to soil CO2 that increase total emissions and decrease soil CO2 concentrations, and have a lasting impact. We estimate average greenhouse gas emissions of about 78 kg CO2eq nest-1 yr 1. At the ecosystem level, leaf-cutter ant nests can account for 0.2% to 1% of the total forest soil emissions. However, balancing vegetation inputs and emissions, and considering their carbon cycle, these ant nests are a net carbon store in the soil that can persist for a decade or more.

The storage manager, as a key component of the database system, is responsible for organizing, reading, and delivering data to the execution engine for processing. According to the data serving mechanism, existing storage managers are either pull-based, incurring high latency, or push-based, leading to a high number of I/O requests when the CPU is busy. To improve these shortcomings, this thesis proposes a push-based prefetching strategy in a column-wise storage manager. The proposed strategy implements an efficient cache layer to store shared data among queries to reduce the number of I/O requests. The capacity of the cache is maintained by a time access-aware eviction mechanism. Our strategy enables the storage manager to coordinate multiple queries by merging their requests and dynamically generate an optimal read order that maximizes the overall I/O throughput. We evaluated our storage manager both over a disk-based redundant array of independent disks (RAID) and an NVM Express (NVMe) solid-state drive (SSD). With the high read performance of the SSD, we successfully minimized the total read time and number of I/O accesses.

For decades scholars have investigated the role of amicus curiae briefs in Supreme Court decision-making. Existing work on the influence of these briefs on opinion content focuses exclusively on the use of “borrowed language” where the justices take language directly from the briefs and incorporate it into their majority opinions. Most of the time justices borrow language without attribution. However, much less often, they decide to formally cite the amici. This presents an interesting puzzle; why do justices sometimes borrow language without attribution while at other times they explicitly cite amici while using little of their language?

In this dissertation I argue that borrowing language from an amicus brief and citing it are two distinct uses, done for different reasons, with different implications. Borrowing language is unique in that it is discreet in nature and is unlikely to be revealed to the reader. Therefore, the justices have leeway when it comes to borrowing language as there should be limited influence on perceptions of the Court and its decisions (i.e. the Court’s legitimacy). Citing amicus curiae briefs, however, is much different in that it is clearly revealed to the reader. As such, there can be implications for the Court’s legitimacy depending on what types of interests the justices cite.

I test the implications of this theory using data on over 1,600 cases where amicus briefs were filed in the 1988-2008 terms. I find that the justices borrow language when they need information, and they borrow from ideologically congruent actors. I also find that the evidence on whether they deliberately avoid citing ideologically extreme interests is mixed. On the one hand, they cite less frequently and are less likely to cite in salient cases, but they do still cite ideologically overt interests. Finally, I implement a survey experiment using a high quality, census balanced sample of 3,000 respondents to test whether citations can influence acceptance of Supreme Court decisions. I find that the public is less accepting of citations to ideologically extreme interests and that they are less accepting of decisions that cite interests that are ideologically incompatible with their own preferences.

The studies within this dissertation use a suite of long-term flux-tower, remotely sensed, and spatially distributed data to more accurately assess the withdrawal of subsurface plant-accessible water storage during multi-year dry periods, more accurately represent measurements of evapotranspiration across the landscape, and examine how vegetation use of plant-accessible water storage varies along latitudinal and elevation gradients, and with time. First, a suite of flux towers from across the arid and semi-arid western United States were used to assess the response of evapotranspiration under varying climates and vegetation types to drought. Here we found that regions experiencing a Mediterranean climate are substantially more dependent on subsurface storage than those receiving a summer monsoon, but available plant-accessible subsurface water storage in the Mediterranean climates can support evapotranspiration for the entirety of a multi-year dry period at some locations. It was also discovered that a transition from snow to rain could increase dependency vegetation on plant-accessible subsurface water storage by as much as 20\% at energy-limited, snow-dominated sites. Next, measurements of evapotranspiration were distributed across the 14 river basins draining into California$'$s Central Valley. This was performed by expanding on current remotely sensed-based methods to include climatic data and consider vegetation type. This novel approach decreased the root-mean-square error by 31-50\% when compared to methods only using NDVI and was insensitive to the spatial resolution of data used. This product showed that evapotranspiration was greatest in the northern basins, peaking at lower elevations, and decreased in magnitude while peaking at higher elevations as latitude decreased. It was also revealed that runoff was derived in primarily one of two ways in this region, the rain-dominated north where annual rainfall grossly exceeds annual evapotranspiration; and the snowmelt-driven south where most precipitation contributes to high-elevation snowpack in energy-limited areas. Finally, the 14 basins draining into California$'$s Central Valley could be binned into four groups based upon what water-balance components and climatic variables were most highly correlated with changes in subsurface water storage, the northernmost, northern, mid-range and southern basins. The results showed that the southern basins may have already reached a critical threshold in storage drawdown, explaining why tree mortality is so widespread in the region, and that the northern and northernmost basins will likely follow a similar path if measures are not taken to reduce evapotranspiration. The studies in this dissertation provided comprehensive analyses of how evapotranspiration spatially varies and how its response to climate extremes alters the hydrologic cycle. Spatial products are in high demand for water resources and forest management applications, and although quantifying uncertainties remain a challenge, these products provide substantial value to improving our understanding of the water cycle.

Circadian clocks are an integral part of cellular and molecular regulation. The protein KaiC is the central player in the well-characterized circadian clock system of cyanobacteria. The simplicity of the cyanobacterial cellular structures makes this protein system highly favorable for the study of the circadian rhythm. KaiC forms an overall structure that looks like two stacked homohexamer rings. The two hexameric domains are called the C1 and C2 ring. The clock’s 24-hour period is driven by sequential phosphorylation of residues in the C2 ring and the choreographed binding of KaiA and KaiB to the C2 and C1 rings, respectively. Although the specific steps have been mapped, the structural changes causing, and caused by, the phosphorylation and protein binding are not fully understood. This project will aim to

perform 500ns time-scale all-atom and coarse-grained MD of multiple experimental structures of the KaiC.

In this dissertation, I present my work towards exploring temporal information for better video understanding. Specifically, I have worked on two problems: action recognition and semantic segmentation. For action recognition, I have proposed a framework, termed hidden two-stream networks, to learn an optimal motion representation that does not require the computation of optical flow. My framework alleviates several challenges faced in video classification, such as learning motion representations, real-time inference, multi-framerate handling, generalizability to unseen actions, etc. For semantic segmentation, I have introduced a general framework that uses video prediction models to synthesize new training samples. By scaling up the training dataset, my trained models are more accurate and robust than previous models even without modifications to the network architectures or objective functions.

Along these lines of research, I have worked on several related problems. I performed the first investigation into depth for large-scale video action recognition where the depth cues are estimated from the videos themselves. I further improved my hidden two-stream networks for action recognition through several strategies, including a novel random temporal skipping data sampling method, an occlusion-aware motion estimation network and a global segment framework. For zero-shot action recognition, I proposed a pipeline using a large-scale training source to achieve a universal representation that can generalize to more realistic cross-dataset unseen action recognition scenarios. To learn better motion information in a video, I introduced several techniques to improve optical flow estimation, including guided learning, DenseNet upsampling and occlusion-aware estimation.

I believe videos have much more potential to be mined, and temporal information is one of the most important cues for machines to perceive the visual world better.

Intracellular transport in eukaryotic cells is a process in which cargo, carrying various materials and attached to molecular motors, moves around the cell. The cargos' transport consists of phases of passive, diffusion-based transport in the bulk cytoplasm and active, motor-driven transport along filaments that make up the cell's cytoskeleton. Because of it's role in the active phase of transport, the cytoskeletal geometry is an important factor. In this dissertation, we consider network parameters such as filament length, number, polarization direction, and location and examine their effect on the transport process. This can be achieved by computationally determining cargo transport through simulation and numerical analysis techniques.

We present this research by first demonstrating an approach that evolves a distribution of cargos in time using numerical integration. To do this, we use two coupled differential equations that enforce the distribution movement on and off filaments. An interesting finding here is that the distribution can become ``trapped" at what we consider intermediate filament lengths.

Although we mostly use a simplified model where normal diffusion governs the passive phase of transport, we also consider the effects of incorporating anomalous subdiffusion in the bulk. This means that the entire transport process can be described as anomalously diffusive, with the active transport phase being superdiffusive and the passive transport phase being subdiffusive. One thing we found by taking this approach is that filament length, rather than filament number, has a greater influence on the domination of overall superdiffusive transport at relatively early times compared to the domination of subdiffusive transport at later times. We were able to extend this observation to model the biphasic release of insulin out of cells in which there is a large spike in insulin release, followed by a slower, more sustained release.

In the final chapter, we consider the possibility of cargos capable of switching to different filaments. If multiple motors are attached to a cargo, it can switch from one filament to another, provided one is nearby. In this phase of our research, we took real images of networks of microtubule bundles and extracted network parameters from them in order to run our simulations. We compared our simulation data, where cargos had different switching probabilities, with experimental data, where cargos had different numbers of motors, and were able to draw a correlation between cargo switching probability and motor number. The network images and the experimental data were provided by our collaborator, Professor Jennifer Ross at UMass, Amherst.

Microorganisms are pivotal in mangrove ecosystems, serving as recyclers and transformers of nutrients in these typically nutrient-limited areas. Auckland’s Waitemata Harbor houses several mangrove stands that are polluted with a slew of hydrocarbons and heavy metals. The primary objective of this study is to identify the level of PAH/heavy metals from Waitemata Harbor mangrove stands that may impact fungal growth. Secondarily, this experiment sought to characterize the microorganisms naturally present within the sediments. Two sites were selected from around the Harbor: Hobson Bay (highly polluted) and Herald Island (less polluted), the sediments from which were either autoclaved or non-autoclaved, inoculated with a growth medium (MEA) along a dilution gradient, inoculated with three species of fungi, and incubated on petri dishes for approximately 20 days. Plates with autoclaved sediment were used to determine the sediment to agar ratio that would allow the most successful fungal growth over time. Plates with non-autoclaved sediment were used to culture the microorganisms naturally present in the sediments. Results indicated that Hobson Bay sediments in higher concentrations are detrimental to Ileodictyon cibarium growth. Trichoderma, Bacillus, Talaromyces and Penicillium species grew on plates with non-autoclaved Hobson Bay sediment, while only Talaromyces species grew on Herald Island sediments. This experiment provides insight into the fungal community composition of the two sites, and may serve as an introductory examination of the bioremediation potential of fungi within the Harbor ecosystem.

The term internet of things has evolved by many folds due to the convergence of multiple technologies like machine learning, embedded systems, wireless-sensors, real-time analytics, etc. A growing number of IoT devices are being developed for consumer use, including connected vehicles, home automation, wearable technology, connected health, and remote monitoring devices. We want to apply some of these technologies and techniques to make commercial spaces smarter.

Buildings are responsible for a significant portion of energy consumption in the US, accounting for more than 40% of US primary energy consumption. Heating, ventilation, and air-conditioning (HVAC) accounts for nearly 50% of that use. Conditioning buildings is important since people spend 87% of their time in the place they live (residential) and the place they work (commercial). Despite this massive expense, many users are dissatisfied with the thermal conditions in buildings.

In this thesis, we explore the tradeoff between commercial building HVAC energy consumption and the quality of thermal conditioning provided to users. The framework has several components that help to address the current HVAC control systems shortcomings, including (a) occupancy sensing in real-time (b) occupancy prediction models based on historical occupancy data (c) human-in-the-loop comfort feedback (d) data-driven thermodynamic building models, and (e) weather forecasting data

All these components provide the necessary input to our model predictive control optimization framework that minimizes monetary costs in energy use while maintaining quality comfort bounds for the building's users based on real-time user's feedback. We tested our framework OFFICE in a real LEED Gold certified university building with over 20 workers performing their daily tasks for 4 weeks, and we showed that we could obtain monetary costs savings of more than 10% while at the same time reducing the users' dissatisfaction levels with thermal comfort from 25% to 0% dissatisfaction, significantly improving the quality of thermal service provided to the building's users.

Lawns, also known as turf, cover an estimated 128,000 square kilometers in North America, consuming an estimated 7 billion gallons of freshwater each day. Despite recent developments in irrigation control and sprinkler technology, state-of-the-art irrigation systems are unable to consider localized water requirements across the irrigation system and deliver localized control, preventing efficient irrigation. Inspired by preliminary results in simulation, we introduce a distributed irrigation controller, allowing us to sense moisture data across the space, actuate each sprinkler independently, and perform computation in a distributed way. To efficiently schedule irrigation for these distributed devices, we introduce modeling techniques allowing us to predict future water movement through the space caused by runoff, leaching, and weather effects that will affect the moisture in the system. These models are then used as constraints in optimization to choose schedules for the distributed valves that minimize system water consumption while maintaining optimal plant health. Finally, we show through extensive deployment side by side with state-of-the-art control strategies that our proposed systems are capable of providing significant water savings while simultaneously providing a higher quality of service to the turf compared to the baselines. Furthermore, we find that through clever system design we can achieve a perpetual system lifetime and virtually eliminate manual system configuration requirements, allowing us to bridge the technology gap to the end user to vastly improve system adoptability. In this way, we demonstrate the feasibility of wireless sensor use in turf irrigation systems.