Machine learning models are a relatively new system for automation of data monitoring and are now being considered for the typically time intensive task of evaluating seismic data in real time. Seismic networks currently possess automated real-time systems that pick phases and detect events as they occur. Newer machine learning models, however, could potentially improve the systems we have now, by increasing the pick accuracy and number of picks, improving event locations, detecting more low magnitude events, and decreasing the workload of analysts. The machine learning model chosen for this study is PhaseNet, a deep-learning neural-net model that was trained on Northern California data, primarily small magnitude earthquakes (3.0 and below), and is currently used in post-processing within the Southern California Seismic Network. Before PhaseNet can be used in real-time monitoring, either in tandem with other systems or by itself, its performance with moderate-to-large magnitude earthquakes must be evaluated. To do this, I have compared phase picks from 50 Southern California earthquakes with magnitudes of 3 to 7.2 that were made by PhaseNet, the current real-time system (AQMS/Earthworm), and seismic analysts. For this analysis, I mainly focused on arrival time differences and whether PhaseNet was able to make a pick using the analyst picks as a ground-truth. Preliminary results indicate that on average PhaseNet makes fewer picks within a short distance of an earthquake, as was expected based on experience from previous applications. In addition, PhaseNet on average seems to pick events slightly earlier in the waveform than analysts for all analyzed earthquakes. I also compare the performance of the current real time system and PhaseNet, as well as looking for any performance trends relating to magnitude and location. Overall, the analysis suggests that PhaseNet could be useful in real-time seismic monitoring when used in conjunction with other pickers or potentially after being retrained specifically for larger magnitude earthquakes.

We have developed a novel method for retrieving the single-point lower ionospheric density profile, utilizing observations from Loran-C and a ray-theory-based Earth-Ionosphere WaveGuide (RT-EIWG) model. Compared to other signals probing the D layer, such as VLF transmitters which provide path-averaged information on the upper boundary, and lightning sferics, which are less stable and dependent on the unpredictable occurrence of lightning, Loran-C offers significant advantages. The reflections of the Loran-C signal are clearly resolved over distances of 300-600 km. Additionally, the 1-hop reflection does not overlap with subsequent hops, and the signal is consistently emitted along a fixed path (interval ranges from 40 to 100 milliseconds). The RT-EIWG model effectively simulates ionospheric reflections at distances less than 1000 km. By combining the RT-EIWG model with Loran-C observations, we have established a new method for determining the electron density profiles of the D layer. Initial results indicate that the nighttime lower ionosphere varies with a period of a few minutes consistently. The method would be a powerful tool for the research of lower ionosphere interaction in the future.

Volatility is a critical factor in agri-commodity derivative trading, especially as the contract maturity date approaches. This study explores how volatility affects future price variations in the Indian agri-commodity derivatives market. We analyzed daily futures prices, spot prices, trading volumes, and open interest data for Farmer’s Producer Organization FPO-traded agri-commodities from January 2016 to December 2023. Our methodology included the Jonckheere–Terpstra test and Ordinary Least Squares (OLS) regressions with realized volatility to assess the impact of contract maturity on commodity prices. We also applied the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) (1, 1) model to examine how volatility changes with the time to maturity of the derivative contracts. Additionally, we tested the negative covariance hypothesis to explore whether futures volatility in agri-commodities increases as maturity decreases. Our analysis, controlling for seasonality and time to maturity, found limited support for this hypothesis in the Indian futures market, except for barley and turmeric. We noted a positive correlation between close price volatility and trading volume, indicating that the rate of information arrival significantly affects volatility. These findings underscore the impact of volatility, time to maturity, and contact premium on hedging strategies and highlight opportunities to enhance trade policies for effective agri-derivative trading, benefiting rural farmer producers and other market participants.

The Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) user facility deploys a mix of long-term fixed atmospheric observatories and short to mid-term mobile facilities with a complex array of instrumentation from surface meteorology and flux stations to lidars, radars, and more. Over 30 years and 4 petabytes of data are publically available in ARM’s data discovery tool. ARM has long developed code to support quality control, data product development, and various other tools for working with the data but these have mostly been developed internally in silos across the various groups of ARM infrastructure. Twenty-one years into existence, ARM released its first open-source community software and has been engaging with the open-science community more and more as the years pass.

With increasing climate variability, natural disaster frequency, and populations there is an emerging need for digital awareness of the built environment. Knowing who and what is where is critical to better allocating resources in emergencies, planning for the future, and translating the outcomes of different environmental models into tangible impacts that help bridge the policy-science gap. To date, a range of open source projects have sought to represent the built environment, but each has focused on either spatial completeness (e.g. AI methods), accuracy (hand digitization in OpenStreetMap), or attribute richness (OpenAddresses). The reality is that all of these are needed for different uses, and integrating these datasets offer a unique path towards a spatial and attribute rich data product. This work describes a method for developing a best available Built Environment Asset Registry (BEAR) from a suite of open - yet disparate - data sources. To date, BEAR conflates commonly utilized, continental-scale footprint and address databases covering the contiguous US. The resulting product is developed as a cloud-native database that conforms to an extensible schema. Using this core representation, we highlight efficient access patterns for retrieving subsets of BEAR from the Lynker Spatial data portal, and examples of applying BEAR to hazard modeling and social vulnerability assessment. We show how this product can increase the skill and utility of forecasts related to flooding and fires.

A document by David M. Tratt. Click on the document to view its contents.

CO2 injection into Permian Basin reservoirs faces challenges due to low-permeability rocks and the complex interactions between CO2, brine, and rock. This paper examines the impact of relative permeability on carbon storage, offering practical insights into storage capacity, pore structure, distribution, and capillary pressure. It compares reactive transport simulation using the geochemical model, aided by PetraSim/Toughreact software, with laboratory experimental data to validate the geochemical reactions, providing practical solutions for carbon storage in reservoirs. This research presents a comprehensive approach that integrates geochemistry and petrophysical properties during core flooding of Permian core samples.  The methodology encompasses a comprehensive petrophysical characterization and analytical chemistry of the rock fabric, utilizing XRD and SEM. Coreflooding experiments were conducted to observe dispersion and chemical changes, with in-situ density variation monitoring using industrial and micro-CT scanners. Geochemical data for model input were obtained through Inductively Coupled Plasma (ICP) analysis. Thin-section measurements enabled a detailed assessment of mineralogy, pore structures, and changes in geochemistry resulting from CO2 interactions. A 30-day CO2 storage experiment revealed evidence of mineral dissolution, supported by increased concentrations of Ca, Mg, and Mn in post brine samples. Micro-CT imaging and nuclear magnetic resonance (NMR) confirmed changes in pore geometry and validated mineralization, precipitation, and dissolution (MPD) effects. Relative permeability measurements captured changes due to CO2 flow and residual trapping.  This study provides new insights into CO2-brine-rock interactions across various timescales, thereby enhancing predictive capabilities for long-term storage. Reactive transport modeling aligned with laboratory studies, elucidating long-term storage dynamics, including porosity, permeability, mineralization, precipitation and dissolution (MPD), and reaction changes using PetraSim. In this research, the additional use of nuclear magnetic resonance (NMR) validated the pore structure of the core sample, both before and after carbon storage, in terms of MPD and geochemistry. Using geochemical reactive transport simulators such as Toughreact, we created a geochemical model by combining laboratory and simulation studies. Our findings aim to enhance the future monitoring and management of CO2 sequestration systems, thereby contributing to climate change mitigation efforts in the Permian Basin.

Heatwave (HW) events are expected to worsen due to climate change, resulting in increased societal impacts such as human illness, crop failures, wildfires, power outages, infrastructure disruption, and damage. The southwest region of Bangladesh, particularly the Khulna division, is anticipated to experience severe consequences of climate change, with increased frequency and intensity of HWs causing numerous deaths due to heatstroke. To understand and mitigate the severity of HWs in Khulna, various aspects were analyzed, such as frequency, duration, and magnitude. This involved calculating the yearly number of HW events (HWN), the yearly sum of HW days (HWF), the length of the longest yearly event (HWD), the average size of all yearly events (HWL), the hottest day of the hottest yearly event (HWA), and the average magnitude of all yearly events (HWM). This analysis used daily maximum temperature records from four observation stations during the summer months (March to May or MAM) from 1990 to 2014, as well as data from thirteen CMIP6 models for three time periods: historical (1990-2014), near future (2026-2050), and far future (2076-2100) for two emission scenarios SSP2-4.5 and SSP3-7.0. Observation data were used to correct the bias of CMIP6 models using empirical quantile mapping (EQM) and to evaluate these models to characterize HW characteristics. HW events were identified using the 90th and 75th percentiles of the distribution of daily maximum temperature values for the reference period. Based on the Pearson Correlation Coefficient (PCC), the values of indices from GCMs are not strongly correlated with the values from the observed dataset. Additionally, indices like HWA and HWM are calculated from GCMs, which are significantly less correlated with the results from the observed dataset, that denotes GCMs cannot project extreme values for particular HW events . The frequency, duration, and magnitude of HWs showed abrupt changes for both emission scenarios compared to the historical period. The historical pattern of HW indices of the Khulna Division has not changed much over 25 years except HWF. HWs are proving to be the deadliest disaster in modern existence. Therefore, a comprehensive understanding of the characteristics of HWs in a particular area is crucial for raising awareness and potentially saving many lives.

Precipitation downscaling is essential for applications such as climate modeling, hydrological simulations, and renewable energy planning. Stochastic downscaling methods, which aim to generate realistic fine-scale precipitation patterns from coarse atmospheric predictors, are particularly important due to the high spatial and temporal variability of rainfall. Despite significant advances in modeling techniques and computational power, precipitation downscaling still remain less accurate and reliable compared to most other meteorological variables, highlighting the need for continued research. In this work, we explore how AI Foundation Models can improve downscaling under the hypothesis that these very large models can better encode physical aspects of weather phenomena. We investigate precipitation downscaling at 0.1° resolution using diffusion models conditioned on ERA5 atmospheric variables and ERA5-Land static fields via embeddings from the Prithvi Weather Foundation Model. The conditioning data include surface and vertical (pressure-level) variables from ERA5 aggregated to 1° daily resolution, combined with ERA5-Land static fields. The target dataset is ERA5-Land daily precipitation at 0.1°. Unlike other approaches, we do not use ERA5 precipitation at coarse resolution; instead, we generate precipitation from noise guided by context variables, so called, channel-synthesis. We explore multiple conditioning strategies, including concatenation, cross-attention with a convolutional encoder, and feature embeddings derived from the Prithvi model to enhance performance. Experimental results show that diffusion models conditioned on Prithvi embeddings improve high-resolution precipitation generation compared to baseline conditioning models.

The detection of biosignatures in exoplanet atmospheres through transmission spectroscopy is essential for assessing planetary habitability, yet traditional atmospheric retrieval techniques are computationally intensive and not well-suited for the large amounts of data expected from upcoming missions. The proposed solution is a machine learning pipeline that leverages Convolutional Neural Networks (CNNs) for more efficient atmospheric composition determination from transmission spectra. A training dataset is generated directly in Python by simulating absorption-like features with added Gaussian noise and additional spectral clutter to mimic realistic artifacts. The spectra are represented as transit depth or intensity ((Rp/Rs)2) plotted with respect to wavelength, then standardized and formatted as inputs for the CNN. Data processing followed the Eureka! pipeline: calibration and reduction steps correct detector/systematic effects, transit light curves are extracted and fit, and the resulting wavelength-dependent transit depths are assembled into transmission spectra suitable for model input. However, because fully processing many targets end-to-end is computationally expensive, the majority of real spectra used in this project were sourced from the NASA Exoplanet Archive, providing a larger set of already-published transmission spectra for training and evaluation. The molecules considered in this work were H2O, CO2, and CH4. The model was trained to predict the presence of these compounds directly from the processed spectra. The trained models were then applied to real transmission spectra sourced primarily from the NASA Exoplanet Archive and their predictions were validated against published atmospheric results to assess model performance. Results show that the CNN can reliably detect H2O, CO2, and CH4 signatures, achieving >90% region-level accuracy across molecules with many predictions made at high confidence (p > 0.65). A one-sample proportion (binomial) test yields p levels p << 0.1 which further indicates performance as significantly better than chance for each molecule-specific model.  

In order to contribute to the deliberation of medium- to long-term local climate policies towards Carbon Neutral by 2050, which is a statutory goal in Japan, the CO2 emissions by prefecture before fiscal year 1990 for each sector such as manufacturing, residential households/businesses, and transportation (passenger/freight) were estimated.

Particulate nitrate (PN) is a critical component of fine particulate matter (PM2.5), especially during wintertime over agricultural regions. In recent years, the contribution of wintertime PN to PM2.5 mass over the Midwestern United States (MWUS), an agriculturally intensive region, has stayed relatively high over time (up to 40%). PN formation is controlled by nitrogen oxides (NOX = NO + NO2), ammonia (NH3), and volatile organic compounds (VOCs). To best control wintertime PM2.5 burden, it is critical to determine PN sensitivity to precursor gases. Despite increases in NH3 concentrations and the flattening of previously decreasing PN trends since 2013 over the MWUS, the factors controlling local PN formation are not well constrained. Satellite tropospheric column NH3/NO2 ratios have been shown to be effective in diagnosing PN sensitivity over East Asia, Europe, and the Eastern United States. Here, we expand this approach to quantify spatially and temporally resolved multidecadal wintertime PN formation sensitivity in the MWUS. Using GEOS-Chem sensitivity simulations and satellite observations from the Ozone Monitoring Instrument (OMI) and the Infrared Atmospheric Sounding Interferometer (IASI), we diagnose wintertime PN sensitivity to NH3, NOx, and VOCs from 2007 to 2023. The NOx-sensitive regime dominates over rural areas, and the NH3-sensitive regime dominates over major cities. VOCs do not control PN formation. Recently, the NOx-sensitive regime is becoming more prominent over the MWUS. NO2 satellite column densities increase from 2007 to 2017 by 0.7 ± 1.1% yr-1 and accelerate from 2018 to 2023 (7.0 ± 1.5% yr-1). NH3 satellite column densities also slowly increase by 0.8 ± 0.6% yr-1 from 2007 to 2017 and accelerate to 2.1 ± 1.2% yr-1 from 2018 to 2023. Despite the recent increasing trends for both precursor gases, PN is still highly NOx-sensitive over the MWUS because of the high abundance of NH3. Interestingly, trends in surface NO2 concentrations oppose satellite observations, decreasing by -1.8 ± 2.0% yr-1 from 2018 to 2023, suggesting the influence of aloft NOX emissions in satellite retrievals. Our work indicates that targeting NOX emissions over rural areas can reduce wintertime PN and PM2.5, while controlling NH3 is more effective over cities.

Saturn’s moon Titan presents as one of the most compelling planetary bodies for exploration due to its dense and replenishing atmosphere, stable surface liquids, and potential for complex organic chemistry. Building upon data from the Cassini-Huygens mission and in anticipation of the Dragonfly mission, our research explores a prototype of an Earth-based high-altitude balloon airship to simulate a future Titan mission. The airship design includes a custom-3D modeled gondola incorporating temperature, pressure, gas composition, and spectral sensors. Similarly, the design prioritized maintaining stable flight and passive buoyancy to maximize energy efficiency and atmospheric data collection. Computational Fluid Dynamics (CFD) simulations were used to evaluate performance in both Earth and Titan environments. CFD models revealed stronger turbulence and flow separation in Titan’s atmosphere, informing a future proposal for control surfaces to manage unstable airflow. We propose two instrumentation methods: Spectrometer-Camera Atmospheric Notation (SCAN) and Gathering Atmospheric Sensing Observations (GAS-O). These scientific payloads investigate surface reflectivity and atmospheric variability. Our mission profile involves a 10-month aerial path across Titan’s troposphere and stratosphere, collecting temporal and spatial data to characterize the hydrocarbon cycle and surface features. This project serves as a proof-of-concept for long-duration airship missions on Titan and highlights future directions, including autonomous control, physics-based modeling, and atmospheric profiling for optimal airship travel.

Coastal communities situated at the intersection of rivers and coastal plains depend on levees, dykes, and barrages for crucial flood protection. However, despite the local flood risk reduction benefits, these structures can exacerbate flooding and associated damages in other parts of the city. The magnitude and spatial distribution of the economic impact of these dynamics are seldom explored. Here, we use a combination of hydrodynamic and economic models to assess the shared risk of adaptation across the entire city of Surat, India. We simulate the 1D-2D hydrodynamic model to generate inundation maps of individual and compound flood inundation scenarios while considering protection structures. Subsequently, we utilize the available flood depth damage function to calculate expected flood damages. Our assessment of shoreline and riverine bank protection scenarios reveals that while these measures protect areas in the proximity of adaptation, they can exacerbate flooding in far inland regions of the city, resulting in greater flooding than experienced prior to adaptation. We also demonstrate that strategic flooding of certain regions, particularly in areas with space for water storage helps to reduce the flooding and damage in the city. The initial assessment highlights that considering shared risk in an economic evaluation of adaptation can uncover previously unaccounted costs associated with uncoordinated adaptation actions. Our findings highlight the necessity of integrating shared risk into urban floodproofing strategies, revealing hidden costs and enabling more effective economic evaluations.

The large-scale issues of mining and deforestation in the Brazilian Amazon Rainforest have emerged as critical global concerns, particularly due to their impacts on weather patterns and surface temperatures. This study aims to investigate the correlations among mining activities, deforestation area, local and regional surface temperatures and precipitation patterns. We employed the Earthrise Enterprise mining detector model, a lightweight convolutional neural network (CNN), which was trained using hand-labeled data from Sentinel-2 satellite imagery spanning the years 2017 to 2023. This trained model effectively differentiates mining areas from other topographical features with an accuracy of 99.11% at a threshold value of 0.9. Other numerical climate data for the Brazilian Amazonas region from 2017 to 2023 were extracted, including: 1) surface temperature, 2) deforestation area, and 3) precipitation. These data were sourced from Hansan Global Land Discovery and Analysis and ECMWF Reanalysis v5 (ERA5) imagery, utilizing Google Earth Engine for processing. We employed the Pearson correlation analysis to examine the relationships among these variables. Our findings indicate a strong positive correlation between mining and deforestation areas (r = 0.81), a weak correlation between deforestation areas and Earth’s surface temperature (r = 0.27), and a negative correlation between surface temperature and precipitation amount (r = -0.52). This study demonstrates that CNNs can be effective tools for analyzing satellite images to explore mining activities. The correlation between mining and deforestation illustrates the role of mining plays in the deforestation of the Amazon Rainforest. The weak correlation between deforestation and temperature indicates that surface temperature changes may be dynamic and multifactorial. Finally, the negative correlation between surface temperature and precipitation suggests that surface temperature increase can be an important drought-predicting factor. We will also discuss the limitations of this study due to the limited number of data points used for the correlation study, which impacts the strength and accuracy of correlations.

Determining statistical significance of changes in soil organic carbon (SOC) stocks at specified probabilities of exceedance is challenging when using landscape-scale datasets due to high natural variability, even over short distances. Using 100 resampled sites from the Australian National Soil Carbon Monitoring Network, we assessed SOC stock changes via design-based statistical inference (probabilistic) methods. Each site represents a 25 x 25 m area within a paddock (with the same soil type and land management), where 10 soil cores were randomly selected using a 5 x 5 m sub-grid. Samples were collected at 0–10, 10–20, and 20–30 cm depths and summed to estimate SOC stocks for 0–30 cm. Maintaining protocol and laboratory consistency with baseline data from ~10 years ago allowed for robust comparisons. Soil organic carbon stocks were adjusted to equivalent soil mass before analysis. The mean change in SOC between surveys was calculated, along with the sampling variance within each 25 x 25 m support area. The variance of SOC change, Var(ΔSOC), is given by Var₁ + Var₂ – 2Cov₁₂, where Var₁ and Var₂ represent the sampling variances of SOC stock for the first and second sampling times, respectively, and Cov₁₂ is the covariance term. As sampling events were independent, Cov₁₂ was assumed to be zero. To ensure conservative stock estimation, SOC stock changes were evaluated at exceedance probabilities of 50%, 60%, 70%, 80%, and 90%, corresponding to quantiles of the student’s t-distribution (α = 0.50, 0.40, 0.30, 0.20, and 0.10). The quantile function (inverse cumulative distribution) was used to scale the mean change, incorporating uncertainty into estimates. We present initial results and key lessons from applying this method across diverse Australian agricultural landscapes.

The Atmospheric Radiation Measurement (ARM) Data Center (ADC), at Oak Ridge National Laboratory (ORNL), is a central hub of the U.S. Department of Energy’s ARM user facility and supports global climate and atmospheric science by managing one of the most comprehensive and openly accessible archives of atmospheric observations in the world. With over seven petabytes of curated data spanning more than 30 years, the ADC processes and distributes over 50 terabytes of new data each month from fixed, mobile, and aerial observatories operating in climate-critical regions worldwide.As data grows in volume, complexity, and value, the ADC is advancing new AI- and ML-enabled capabilities to transform how users discover, access, and interact with ARM data. These efforts align with broader DOE and ORNL initiatives in building scalable, FAIR-aligned (Findable, Accessible, Interoperable, and Reusable) data ecosystems supporting the next generation of open science.Central to this transformation are AI-powered, user-centric tools that assist researchers through natural language interfaces, simplifying the process of identifying relevant data products across ARM’s complex metadata hierarchies. These tools reduce the entry barrier for new users while enabling personalized discovery pathways for more experienced users. By bridging and translating across diverse metadata structures, these tools enhance data usability and streamline cross-domain discovery. These tools also leverage a scalable vector-based infrastructure that enables contextual understanding, user-aware recommendations, and integration into future AI-ready APIs.The ADC’s broader portfolio includes AI-powered metadata and abstract generation, anomaly and epoch detection models, a Primary Measurement Type classifier, and a data recommendation engine – each advancing automation, knowledge extraction, and scalable ML pipeline readiness. Strategic collaborations with the USGS, the Earth Science Information Partners (ESIP), and ORNL’s AI Program further strengthen these efforts to accelerate AI-driven workflows. Together, these efforts position the ADC at the forefront of intelligent Earth science data platforms and open science innovation, enabling intuitive, efficient and more impactful scientific discovery.

Land subsidence has been occurring in the Western U.S. since the early 1900s. In southwestern Utah, evidence for land subsidence and earth fissures related to excessive groundwater withdrawal dates back to the early 2000s. However, there has yet to be a survey to identify or monitor land subsidence throughout the state of Utah. In this study, we process InSAR time series across the state, using freely available datasets from ALOS-1 (2006-2011), Sentinel-1 (2015-2025), and ALOS-2 ScanSAR (2015-2025). We identify land subsidence with rates ranging from 0.5 cm/year (Snake Valley) to 4 cm/year (Escalante Desert) over the past 20 years, as well as highlight seasonal signals with amplitudes of up to 2 cm/year (Salt Lake Valley). We demonstrate the feasibility of combining these results with groundwater level trends to estimate aquifer properties and storage change on a regional scale. Finally, we discuss how novel regional datasets, such as those produced by the OPERA North America DISP Product, may be able to inform future groundwater-flow models.