Climate change and a growing global population pose ongoing threats to critical resources. As resources required by the agriculture sector continue to diminish, it is critical to leverage the emerging technologies and new solutions within the sector. New cultivation practices have emerged over the years, allowing food to be grown within urban areas. Greenhouses are versatile in the resources needed for their operation, as well as the foods that can be grown. While greenhouses provide a potential for a more constant food supply, there is a lack of optimization between the components. There are benefits to having modular components of a greenhouse, allowing for adjustments or repairs to singular pieces. However, there is inefficiency in the entire system, since each component functions without considering the others. To improve greenhouse efficiency, a closed-loop system can be introduced. A greenhouse is a closed system, and by repurposing, reusing, and recirculating resources, a greenhouse can evolve to have a closed-loop system. This enables the components of a system to share resources more effectively, communicate any systems changes that are required, and minimize waste outputs.This research explores the current technology in the space of agriculture and computer science to create a fully closed-loop system. The most noticeable system components are food waste, nutrient systems, water systems, growing media, and heating and energy. Not all components within a greenhouse can leverage the same artificial intelligence methods and techniques based on existing findings. There are methods in place that allow the components to interpret data gathered from the greenhouse and alter its operational patterns. There remains a lack in communicating this information to other aspects of the system to have it make informed data-driven decisions as well. One can optimize singular components thereby reducing resource reliance, to a certain threshold until it impacts the plant’s development and yield. When all the systems components’ resource needs and outputs converge the functionality of the system can be optimized to utilize resources at a higher efficiency. Results are indicative of very siloed and isolated research, exploring closed-loop systems within greenhouses, but not leveraging its full capabilities.

A document by Antony Joh Moothedan. Click on the document to view its contents.

The Hunga Tonga Volcano eruption launched a myriad of atmospheric waves that have been observed to travel around the world several times. These waves generated Traveling Ionospheric Disturbances (TIDs) in the ionosphere, which are known to adversely impact radio applications such as Global Navigation Satellite Systems (GNSS). One such GNSS application is Precise Point Positioning (PPP), which can achieve cm-level accuracy using a single receiver, following a typical convergence time of 30 mins to 1 hour. A network of ionosondes located throughout the Australian region were used in combination with GNSS receivers to explore the impacts of the Hunga-Tonga Volcano eruption on the ionosphere and what subsequent impacts they had on PPP. It is shown that PPP accuracy was not significantly impacted by the arrival of the TIDs and Spread-F, provided that PPP convergence had already been achieved. However, when the PPP algorithm was initiated from a cold start either shortly before or after the TID arrivals, the convergence times were significantly longer. GNSS stations in northeastern Australia experienced increases in convergence time of more than 5 hours. Further analysis reveals increased convergence times to be caused by a super equatorial plasma bubble (EPB), the largest observed over Australia to date. The EPB structure was found to be ~42 TECU deep and ~300 km across, traveling eastwards at 30 m/s. The Hunga Tonga Volcano eruption serves as an excellent example of how ionospheric variability can impact real-world applications and the challenges associated with modeling the ionosphere to support GNSS.

A document by Chang Liao. Click on the document to view its contents.

Thawing of permanently frozen ground (permafrost) has increased in recent decades with negative implications for human and non-human adaptation to climate change. Impacts include reduced ground stability, increased transportation risk, and changes in water availability. Direct measurements of permafrost active layer thickness (the depth of thawed ground overlying permafrost) are sparse. Measurements currently exist for a few hundred sites located primarily in the Northern Hemisphere supported by the Circumpolar Active Layer Monitoring (CALM) Program. The sparsity of direct active layer thickness measurements limits broad-scale understanding of changes in permafrost thaw and confidence in future projections. To address the sparsity of direct active layer thickness measurements, we developed a method to estimate active layer thickness change from streamflow measurements, which integrate processes over broad spatial areas and are more common than point-scale active layer thickness measurements. The method uses classical principles of hydraulic groundwater theory and nonlinear baseflow recession analysis, which sets it apart from prior methods based on linear recession analysis. The method is applied to catchments in the continuous and discontinuous permafrost zone of the North American Arctic containing co-located streamflow and CALM active layer thickness measurements. We find good agreement in the magnitude and direction of measured and predicted active layer thickness trends. This suggests that regional-scale estimates of active layer thickness change can be obtained from streamflow measurements, which may open the door to retrospective estimation of active layer thickness change in data sparse Arctic regions with short, sporadic, or even nonexistent ground-based active layer measurements.

A document by Robert L. McGraw. Click on the document to view its contents.

A document by Zahra Zali. Click on the document to view its contents.

The Laccadive-Chagos Ridge (LCR) is a prominent aseismic ridge in the Indian Ocean. The origin and nature of the crust beneath the LCR have been debated. Based on Ar-Ar geochronology of the volcanic basement rocks from ODP wells 713 and 715 from Chagos and Maldives ridges, a hotspot trail model was proposed for the genesis of the LCR. On the other hand, based on geophysical studies, the LCR has been inferred as a continental sliver/hyperextended continental crust which has undergone heavy underplating and Reunion hotspot volcanism. Even though these two different hypotheses differ on the genesis and nature of the LCR, both agree that the ridge experienced extensive Reunion hotspot trail volcanism, making the crust's nature challenging to decipher. We report Ar-Ar ages of two rhyolite rock samples from a well drilled on the Padua bank located on the northern extent of the Laccadive ridge. These acidic rocks are at a depth of ~1700m and ~2300m and underlie the Tertiary sediments forming the basement in this well. The previously determined K-Ar geochronology ages are 60.2 Ma and 102 Ma. The precise Ar-Ar dates of these rocks determined in the present study are 76.2 ± 0.4 Ma and 77.5 ± 0.5 Ma (2σ) respectively. These acidic rocks (78-76 Ma) coupled with onshore acidic rocks of the St-Mary's island (87-84 Ma) and Ezhimala (95-93 Ma) suggest continuous extension between the India and Laccadive ridge after the Indo-Madagascar continental breakup (88 ma). Contrary to the earlier hypothesis, the acidic rocks of Laccadive Ridge are older than the Reunion Plume trail volcanism (62-60 Ma). The geophysical and geochronological evidence suggests that the Laccadive ridge is continental in nature, and the volcanism predates the Reunion plume volcanism. These evidences suggest that the Laccadive Ridge is a continental thinned crust that has undergone rift-related volcanism rather than a hotspot trail.

Numerous deterministic and probabilistic techniques have been utilized extensively to examine the velocity distributions for open channel flows under the various flow and channel characteristics. It is generally known that traditional approaches such as logarithmic velocity laws and power laws furnish vertical velocity profiles for the wide channels (width-to-depth ratio > 5) only. Also, the accuracy of the traditional methods is highly parameter dependent which, in turn, is evaluated using empirical or semi-empirical relationships only. Owing to the limitations of the conventional approaches, the entropy concept is formulated probabilistically and applies to all aspect ratios (width-to-depth ratio). The prominent entropy types explored to date are Shannon entropy, Tsallis entropy, and Renyi entropy. Recently, Fractional entropy has been demonstrated to study the velocity distributions for the open channel flows. Laboratory measurements have been used to demonstrate the applicability and correctness of all the techniques in adverse channel bed slope conditions (Singh & Khosa, 2022). The measurements comprise the velocity data at the different adverse slope values. Finally, a rigorous comparative analysis for the estimated velocity profiles using the four generally used velocity laws and four entropy-based velocity distributions have been conducted, which showed that the information entropy-based methods are far better than the traditional techniques.

In Iran, both local and regional scale land-surface deformation has resulted from the decline in groundwater levels \cite{Motagh_2008}. Moreover, the gap between groundwater extraction and renewal is so large that the resulting short-term impacts are likely to be irreversible \cite{Olen_2021}.Here we use Earth Observation (EO) data to calculate vertical subsidence rates due to groundwater extraction in Tehran, Iran’s capital city. This data includes Sentinel-1 Interferometric Synthetic Aperture Radar (InSAR); very high-resolution (VHR) Pléiades optical stereo imagery; and ICESAT-2 laser altimetry data.The Centre for Observing and Monitoring Earthquakes and Tectonics (COMET) Looking into Continents from Space (LiCSAR) automated processing system is used to process six years (2015-2021) of Sentinel-1 SAR acquisitions \cite{Lazeck__2020} for interferometric (InSAR) analysis. The system generates short baseline networks of interferograms. We also correct for atmospheric noise using the GACOS system \cite{Yu_2018a,Yu_2018} and perform time-series analysis using open-source LiCSBAS software \cite{Morishita_2020}. Vertical and horizontal (east-west) velocities are calculated and mapped to produce nationwide and regional velocity fields \cite{Watson_2022}. Preliminary results indicate maximum subsidence rates in Tehran exceed 100 mm/year.VHR Digital Elevation Models (DEMs) and laser altimetry ground returns are also used to calculate land subsidence rates in Tehran. By comparing rates calculated using all three EO techniques we aim to validate InSAR velocities whilst investigating and constraining the benefits, drawbacks, and biases associated with each technique. Phase bias \cite{De_Zan_2015}, for example, may be introduced to calculated InSAR velocities when using the LiCSBAS short baseline network strategy as subsiding regions in Iran are often vegetated cropland.Publishing the open-source COMET-LiCS Subsidence Portal was the focus of previous work (https://comet-subsidencedb.org/). The portal presents automatically processed LiCSAR Sentinel-1 interferograms and LiCSBAS velocity timeseries for 99 subsiding regions across Iran. Interactive tools allow stakeholders to make quick, critical assessments related to extents and rates of subsidence. Validating portal data using DEMs and laser altimetry is essential before expanding the portal to have a global focus.

A document by Fanny Lehmann. Click on the document to view its contents.

Large earthquakes, especially those occurring in a city or population centers, create devastation and havoc, and often times kindle several deaths and injuries, and significant infrastructure damage that lead to several billions of dollars in losses. Marine earthquakes are the leading cause of large tsunamis which cause deaths, destruction, displacement of population, and a possible nuclear meltdown. Thus, prediction of earthquake or its aftershocks or earthquake early warning system has a great potential to mitigate the loss of life as well as different kinds of damage. Earthquake prediction would mean forecasting the occurrence of an earthquake by providing both its magnitude estimate and accurate location. Earthquake prediction has been an important area of seismology research for quite a while, and it looks like it will continue to be an important area of research. Recently, with the implementation of deep learning in seismology, scientists have been able to detect, predict, and model seismic waves and earthquake aftershocks. Earthquake aftershocks are generally triggered by changes in stress formed by large earthquakes that happen within, or surrounding a given fault network system. The main goal of this study is to investigate the improvement of aftershock pattern predictions with the implementation of tuning and optimizing of deep learning parameters. To achieve these goals, we have developed an algorithm that can help first gather mainshock-aftershock sequence data. Some of the criteria used in identifying earthquakes that initiate an aftershock is to look at earthquakes that happen within a certain radius, the values we attempted are within about 0.5 degrees range, and within a certain period, from few seconds to several weeks of the occurrence of the main shock. For the sequence identification, we have been using seismic data from the United States Geological Survey (USGS)-National Earthquake Information Center (NEIC). We are also looking at different open-source data gathered by researchers for a similar study. The deep neural networks we are implementing make use of Keras python Toolkit, and Theano and Tensorflow libraries, with a plan to use PyTorch python library instead of Theano library in the future because of some maintenance issues. To this point our attempts have shown a good progress.

Seismology is a data-driven science with a huge amount of data gathered for over a century. Though seismic data recording started in 1900, the growth of seismic data has obviously been exponentially in the last three decades. This data growth can be easily noticed if one takes a close look at just one of the largest seismological data centers in the US, the Data Management Center (DMC) of the Integrated Research Institutions in Seismology (IRIS). Data at the DMC grew from less than 10 Tebibytes in 1992 to about 800 Tebibytes in 2022. With the availability of such a large amount of seismic data, it is paramount to develop new seismic data processing and management tools to help analyze and find new and better seismic models. Developing new big seismic data processing and management tools will be helpful to make the best use of such growing big seismological data sets. The main goal of this investigation is the development of efficient data manipulation and processing tools for retrieval, processing, merging, aggregation, and management of big seismic data from disparate data sources. In this study, such big seismic data processing tools are being developed using python programming language and open source python libraries, and the tools we are developing will be helpful to extract, split, and convert, merge and process big seismic data. In addition, python is very suitable for data science and has powerful libraries to process and manage data and applications. Significant contributions have been made in recent python based libraries for seismic data processing, though there are still some rooms for improvement when it comes to seismic applications to merge, convert, manage  and process big seismic data from disparate data sources and converting different file formats. Seismic data from different networks surrounding the Rio Grande Valley have been collected from different data sources. Our attempt is to test the developed tools and evaluate their performance. This study has made important progress in this regard and the results are promising. 

Limited research has assessed the spatio-temporal distribution and chronic health effects of NO2 exposure, especially in developing countries, due to the lack of historical NO2 data. A gap-filling model was first adopted to impute the missing NO2 column densities from satellite, then an ensemble machine learning model incorporating three base learners was developed to estimate the spatiotemporal pattern of monthly mean NO2 concentrations at 0.05° spatial resolution from 2005 to 2020 in China. Further, we applied the exposure dataset with epidemiologically derived exposure response relations to estimate the annual NO2 associated mortality burdens in China. The coverage of satellite NO2 column densities increased from 46.9% to 100% after gap-filling. The ensemble model predictions had good agreement with observations, and the overall, temporal and spatial cross-validation (CV) R2 were 0.88, 0.82 and 0.73, respectively. In addition, our model can provide accurate historical NO2 concentrations, with both by-year CV R2 and external separate year validation R2 achieving 0.80. The estimated national NO2 levels showed a increasing trend during 2005-2011, then decreased gradually until 2020, especially in 2012-2015. The estimated annual mortality burden attributable to long-term NO2 exposure ranged from 305 thousand to 416 thousand, and varied considerably across provinces in China. This satellite-based ensemble model could provide reliable long-term NO2 predictions at a high spatial resolution with complete coverage for environmental and epidemiological studies in China. Our results also highlighted the heavy disease burden by NO2 and call for more targeted policies to reduce the emission of nitrogen oxides in China.

The true bottleneck of artificial intelligence (AI) is not access to the data, but rather labeling this data. We have tons of raw agriculture image data coming from various sources and manual labelling remains to be a crucial step to keep the data well organized which requires considerable amount of time, money, and labor. This process can be made more efficient if we can automatically label the raw data. We propose contrastive learning representations for agriculture images (AgCLR) model that uses self-supervised representation learning approach on unlabeled real-world agriculture field data, to learn the useful image feature representations from the images. Contrastive learning is a self-supervised approach that enables model to learn attributes by contrasting samples against each other without the use of labels. AgCLR leverages the state-of-the-art SimCLRv2 framework to learn representations by maximizing the agreement between differently augmented views of same sample. We have incorporated critical enablers like mixed precision, multi-GPU distributed parallel computing, and use of Google Cloud's Tensor Processing Units (TPU) for optimizing the training process. We achieved 80.2% accuracy while classifying the test data. We further applied AgCLR to unrelated task to determine the alleys and rows in corn field videos for corn phenotyping and we observed two cluster formations for alleys and rows when plotted embeddings in a 3-dimensional space. We also developed a content-based image retrieval tool (pixel affinity) to identify similar images in our database and results were visually very promising.

Although adequately detailed kerosene chemical-combustion Arrhenius reaction-rate suites were not readily available for combustion modeling until ca. the 1990’s (e.g., Marinov [1998]), it was already known from mass-spectrometer measurements during the early Apollo era that fuel-rich liquid oxygen + kerosene (RP-1) gas generators yield large quantities (e.g., several percent of total fuel flows) of complex hydrocarbons such as benzene, butadiene, toluene, anthracene, fluoranthene, etc. (Thompson [1966]), which are formed concomitantly with soot (Pugmire [2001]). By the 1960’s, virtually every fuel-oxidizer combination for liquid-fueled rocket engines had been tested, and the impact of gas phase combustion-efficiency governing the rocket-nozzle efficiency factor had been empirically well-determined (Clark [1972]). Up until relatively recently, spacelaunch and orbital-transfer engines were increasingly designed for high efficiency, to maximize orbital parameters while minimizing fuels and structural masses: Preburners and high-energy atomization have been used to pre-gasify fuels to increase (gas-phase) combustion efficiency, decreasing the yield of complex/aromatic hydrocarbons (which limit rocket-nozzle efficiency and overall engine efficiency) in hydrocarbon-fueled engine exhausts, thereby maximizing system launch and orbital-maneuver capability (Clark; Sutton; Sutton/Yang). The rocket combustion community has been aware that the choice of Arrhenius reaction-rate suite is critical to computer engine-model outputs. Specific combustion suites are required to estimate the yield of high-molecular-weight/reactive/toxic hydrocarbons in the rocket engine combustion chamber, nonetheless such GIGO errors can be seen in recent documents. Low-efficiency launch vehicles (SpaceX, Hanwha) therefore also need larger fuels loads to achieve the same launched/transferred mass, further increasing the yield of complex hydrocarbons and radicals deposited by low-efficiency rocket engines along launch trajectories and into the stratospheric ozone layer, the mesosphere, and above. With increasing launch rates from low-efficiency systems, these persistent (Ross/Sheaffer [2014]; Sheaffer [2016]), reactive chemical species must have a growing impact on critical, poorly-understood upper-atmosphere chemistry systems.

In light of climate change, wildfires are concurrently becoming more frequent and devastating worldwide. Though we have a good understanding of fire frequency changes in the past using charcoal analysis, records of the characteristics of these fires, such as fire severity, are lacking. This limits our ability to model how fire severity responds to climate change. Boron isotopes in clay minerals show promise as a novel fire severity proxy, where increased 𝛿11B is correlated with higher fire severity. Through reacting boron leached from experimentally combusted plants with clays, we determine that the observed correlation with fire severity is likely caused by input of isotopically heavier boron from combusted leaves. In contrast, combusted barks lower the 𝛿11B of clays upon reaction. Despite the different results, in both experiments with barks and leaves, similar boron isotope fractionation is observed during boron adsorption onto clays, where the lighter 10B is preferentially adsorbed. Therefore, the different results are likely caused by the different boron isotope composition of leaves and barks, where leaves have a much higher 𝛿11B (~30 ‰) than bark (~9 ‰). Combustion temperature can also affect the 𝛿11B of clays: changes to the 𝛿11B of clays were observed only when reacting with bark combusted at >300 °C, or with leaves combusted at >550 °C. This could be because more boron can be leached into solution from materials combusted at higher temperatures, which in turn results in greater adsorption onto clays during reactions. Clays have higher 𝛿11B in soils affected by high severity fires that consume tree crowns, because these fires combust more leaves that then deposit their isotopically heavier boron content into the soil. This relationship could help complete our fire record and improve our ability to predict future fire characteristics.

A document by Annarita Marrano. Click on the document to view its contents.