Search for: All records

Remote sensing-based measurements of environmental covariates have proven effective at modeling soil moisture at high spatio-temporal resolutions, but often struggle to produce accurate readings in forested areas, where closed canopies and complex topography can interfere with surface reflectance observations. Spatial autocorrelation is another–often neglected–component of digital soil mapping (DSM), which has the potential to inflate model performance and obscure variable contributions. In this study, we integrated in situ, 10 m resolution measurements of soil moisture from a 1-hectare plot within the Lamont Sanctuary Forest (LSF) experimental site in the northeastern U.S. with remotely-sensed environmental covariates and a digital elevation model. Our goals were to (1) construct accurate machine-learning models of soil moisture; (2) assess and mitigate autocorrelation to improve our models’ generalization ability; (3) understand the relative influence of factors on forest soil moisture. We evaluated random forests (RF), eXtreme Gradient Boosting (XGBoost), and Gaussian Process Regression, using a buffered block cross-validation strategy. The RF models demonstrated the best performance, with an R2 of 0.794 and a root mean squared error (RMSE) of 0.046 m3/m3. Adjusting for spatial autocorrelation improved our model’s ability to handle novel samples by reducing data leakage. Downscaled land surface temperature (LST) was the overall most important predictor, followed by topographical features like slope and biotic indices like the normalized difference moisture index (NDMI). These results highlight key controls on soil moisture that could be used to inform targeted forest management. 

n this study, we investigate the source of an ash layer found in sediment core VM33-116 (Lat: -2.9, Long: 148.583), located in the Bismarck Sea. Originally, this research aimed to understand the impact of volcanic ash on biological productivity and climate. However, our findings suggest that the ash layer, dated to approximately 3000–4000 years ago, is most likely from a submarine volcanic eruption similar to the Hunga Tonga eruption of 2022. The ash layer (39–42 cm) contains an unusual spherule aggregate and black foraminifera. Experimental work has related the formation of spherule aggregates to the intense lightening in the high-altitude volcanic plume (58 km) produced by submarine eruptions like Hunga Tonga.) The black foraminifera are colored by black carbon on the inside of the foraminiferal test- a feature that might imply increase biological productivity. Elemental analysis using ITRAX and XRF revealed changes in Fe/Al and Ba/Al ratios, indicating volcanic ash deposition. However, no significant spikes in productivity-related elements (e.g., P, S, or Si) were observed, suggesting minimal impact on biological productivity. Corrected radiocarbon dating and geochemical analysis points to a submarine rhyolitic volcano located about 90 km from the core site as the likely source. Other nearby volcanoes, such as Tavui and Rabaul, are too distant to account for the ash. Our findings contribute to understanding the source of volcanic ash in the region and suggest that Hunga Tonga-type underwater eruptions can leave lasting geochemical markers, even across centuries. The presence of spherule aggregates and black foraminifera provides valuable insights into the effects of such eruptions on surrounding marine sediments Further research is needed to fully understand the relationship between submarine volcanic activity and its impact on biological productivity. 

The paleogeographic position of Laurentia during the Neoproterozoic to Middle Cambrian remains debated, with some reconstructions suggesting a subpolar setting and others placing it in tropical latitudes. The Cambrian-aged Leithsville Formation, its subdivisions (Califon, Hamburg, and Wallkill members), and the underlying Hardyston Formation, all offer insight into this period of tectonic transition. This study uses detailed stratigraphic, sedimentological, and potential paleomagnetic analysis to evaluate Laurentia’s Early Cambrian latitudinal position and the depositional history of these formations. Fieldwork included creation of a detailed stratigraphic section, documentation of lithologic and structural features, and collection of samples for petrographic and paleomagnetic analysis. Preliminary results from the Leithsville Formation reveal diverse sedimentary features—including oncoids, pisoids, stromatolites, thrombolites, Skolithos, ripple marks, and cut-and-fill channels—alongside diagenetic features such as chert-replacement, dolomitization, and stylolites. The presence of Archaeocyathus in both the Califon and Hamburg members supports a Lower Cambrian age. Environmental interpretation suggests a range of depositional settings: the Califon Member represents a shallow, open carbonate platform with thrombolites and rubble zones; the Hamburg Member reflects episodic transgressions and regressions, with arid sabkha and tidal-flat conditions; and the Wallkill Member records high-energy tidal to peritidal conditions marked by pisoidal shoals and stromatolitic mats. These features collectively suggest Laurentia was located in a low-latitude, arid, tropical to subtropical environment during the Early Cambrian, inconsistent with the 45 to 50 degree temperate zone that current models place this portion of Laurentia. Ongoing analysis of thin sections and paleomagnetic data will help further constrain the continent’s paleolatitude during this critical interval in Earth’s history, allowing for an overall better understanding of Earth’s climate in deep time. 

Ice rises are slow moving areas of grounded ice adjacent to ice shelves, where ice flows outwards from a flow center that is independent of the rest of the ice sheet. Not only do these grounded regions provide buttressing forces that enhance ice-shelf stability, but their relative flow stability and isolation also make them ideal locations for extracting ice-core climate records and for reconstructing regional millenial-scale ice flow changes from internal radiostratigraphy. However, dating ice cores and flow changes recorded in radiostratigraphy can require knowledge of ice-rise flow and rheology, both of which are poorly constrained; models employing a standard rheological parameterization often fail to reproduce observed vertical velocities derived from ice-penetrating radar. These discrepancies highlight the need for improved constraints on the rheological properties of ice within ice rises. We present an inversion technique built around physics-informed neural networks (PINNs) using the DIFFICE.jax package to infer ice viscosity directly using observations from a phase-sensitive ice-penetrating radar observations and the mass and momentum balance equations. First, we use a mass balance equation to obtain velocity and ice density profiles from the radar observations. These velocities and densities are subsequently used with a full-Stokes momentum balance equation to obtain estimates for ice pressure and viscosity. We demonstrate the functionality of our inversion approach using synthetic data, and apply it to ice-penetrating radar data from the Weddell Sea Sector of Antarctica. This work will reveal the spatial variability in ice viscosity within these ice rises, for the first time, improving ice rheology parameterizations and ultimately improving how we reconstruct past climate change with ice cores and glaciological change with ice-rise internal stratigraphy. 

Extraterrestrial oceanic impact craters are key to understanding Earth’s geological and biological evolution, yet they remain under-documented due to low core coverage, tectonic activity, and limited access to deep-sea data. Given that over 70% of the Earth’s surface is covered by seawater, many impacts likely occurred underwater. This study evaluates and applies a range of detection methods, including magnetic susceptibility profiling, X-ray fluorescence (XRF) analysis, and petrographic and electron microscopy, to investigate a potential impact site near the Indian Ocean Ridge Triple Junction. On an analog seismic line, the prospective impact crater appears as a flat, transparent sediment layer with disturbed acoustic basement beneath it. Marine core RC17-95 was selected for analysis based on its proximity to the crater candidate, its high topographic elevation, relatively high magnetic susceptibility, and coarse fraction features suggestive of or promoting deposition by an impact-related tsunami. The core was processed by sieving and grain-size separation, revealing Fe and Al-rich spherules with quench textures, spherule aggregates, flow-textured and micro-vesicular impact glass at multiple depths. At 282 cm depth, both the base of this layer and the core there are cemented angular breccias and meteorite fragment candidates. ITRAX XRF scans confirmed the presence of significant amounts of Palladium, a Platinum Group Element (PGE), from 0-282 cm subbottom depth, further supporting the hypothesis of an extraterrestrial origin. These markers are consistent with impact ejecta and tsunami deposits. By integrating geophysical, geochemical, and mineralogical analyses, this research provides preliminary evidence for an extraterrestrial impact tsunami layer in the Indian Ocean. It contributes to improved methods for identifying oceanic impact craters in the future. 

The presence of meltwater or brine in icy systems may affect the flow of ice. For example, ice flow speed is known to rapidly increase close to and above its melting point, but it is not yet clear whether this is an effect of temperature, the presence of melt alone, or the amount of melt. Ice is also often found in contact with brine, both in warming glacial environments and at the ice-ocean interface in outer solar system moons, where the existence of secondary liquid phases may further modify flow. We are therefore conducting two series of experiments to determine how melt fraction and melt distribution affect the dynamic microstructure of temperate ice. The first experimental series tests grain growth over time, varying both salinity and temperature to show that both the presence and amount of melt control ice grain size. Preliminary results suggest that grain growth depends strongly on the amount of melt, not just the temperature, and that high salinities may impede grain growth above eutectic temperatures. The second series of experiments tests how the deformation of temperate ice may be enhanced by fabric formation in both solid and melted phases, using a custom direct-shear setup in a cryogenic servo-controlled deformation apparatus. Through these tests, we trace the development of both melt preferred orientations (MPO) and crystallographic preferred orientations (CPO), analyzing whether MPO formation should be accounted for in commonly used flow laws for solid temperate ice. 

Large-scale ice processes depend heavily on microstructural characteristics, especially the size and growth of individual ice grains. Ice in polar Earth environments and the interiors of icy moons is often in contact with saltwater, but little data exists on how saltwater may affect grain growth at warm temperatures. In order to study these effects, we created pure ice samples that we flooded with saltwater at two different salinities, then allowed grains to grow in temperatures of -5 Celsius and -13 Celsius, both above the H2O-NaCl eutectic point, for periods of 3, 7, 31, 100, 310 hours. We then analyzed the microstructure through visual imaging and calculated effective mean grain size as a function of time, which we then compared to the growth of pure ice over the same time periods. We found that both salinity and temperature affect grain growth parameters in temperate, salty glacial environments. This suggests that any grain-size-dependent flow of glaciers and convection of icy moon crusts may be controlled by local temperature and salinity. 

he upper ocean plays a critical role in the climate system as the interface for the exchange of heat, momentum, and carbon between the atmosphere and the deep ocean. This exchange is mediated through a relatively well-mixed region, which acts as a staging area before properties are moved to the interior or returned to the atmosphere. This well-mixed region has traditionally been quantified in two ways: the mixed layer (a region with uniform density) or mixing layer (a region with high turbulence levels). We propose a new approach for quantifying the properties of the surface ocean. We introduce a new concept of “mixable layers,” a region of the upper ocean that can be mixed easily relative to the water below. A mixable layer is estimated using an approach we call the “energy knee method.” For a given profile, we compute the depth to which the density could be homogenized for a given amount of energy. Repeating this calculation across a range of energy values yields a curve of energy versus mixable depth, and the knee point—the location of sharpest bend—is identified as the mixable layer. This represents the depth where there is a sudden transition to stronger density stratification, where mixing deeper requires much more energy. Our method also introduces a secondary metric for characterizing the upper ocean, called the mixed layer energy (MLE), which is the energy required to completely homogenize a region of the upper ocean. The MLE may also be used with traditional mixed layer or mixing layer depths estimates. Using temperature and salinity profiles from the Argo dataset, we find that our method yields mixable layer depths that closely align with depths of well-mixed water, providing an alternative, potentially more robust approach to estimating the mixed layer depth. In regions with weak pycnoclines, the mixing layer approach generally outperforms the threshold method, conventionally used to estimate mixed layer depths, in identifying the region of homogenous density near the surface. We present global statistics of mixing layer depths and energies, including their regional and temporal mean and variability. Our energy knee method for calculating MLD and MLE provides a more robust and interpretable framework for studying the structure of the upper ocean. 

We sought to develop a longer and more robust history of pollution in the Hudson River by studying LWB1-8, a high sedimentation rate core (~1 cm/yr) retrieved near Yonkers, NY. The sediment has been affected both by industrial pollution and natural disasters such as distant volcanic eruptions. In order to study this core, we analyzed its elemental composition in a variety of ways. Previous data from ITRAX scanning of the core years ago was lined up with new elemental analyses done with an XRF machine in order to pick which layers may be the most likely to contain volcanic ash. If ash particles were deemed likely in these layers, samples were run through a Franz, or magnetic materials were separated out using a Nb magnet. Then, particles of potential ash were picked out by hand. These ash candidates were then run through an SEM machine to provide a more in-depth elemental analysis of the particles as well as obtain high-resolution photos of them. Peaks in uncalibrated Ni, Ti, and SI (peaks in counts) from the ITRAK can be used to locate the depths of prospective volcanic ash layers. Ni peaks were especially good at identifying which layers may contain volcanic ash. .We found at least four layers containing volcanic ash , but there is still uncertainty about their source volcanoes. Many of the volcanic ash particles have very high Fe and very low K contents. These likely come from explosive Icelandic eruptions like those of Hekla. Other ashes have very low Fe, higher K and higher Si. These ashes likely come from volcanic arcs located at high latitudes, such as the Cascade and Aleutian arcs. This experiment has shown that it is possible to find volcanic ash in Hudson River cores. However, the number of ash particles we have retrieved so far is very small, from one to nine per age horizon. We do best at finding ash below 100 cm, where there is little industrial pollution. In future, we need to refine our methods of segregating ash from industrial debris. We must also analyze our ash particles on a microprobe and an ICPMS to determine their source volcanoes. Only then can we convert our measurements of metals versus depth into a pollution history. 

The Long Island Sound (LIS), located between Connecticut and Long Island, New York, is a crucial estuarine ecosystem that is affected by human activities in the surrounding areas. This study uses sediment grab samples from central Long Island Sound that have been collected as part of a larger habitat mapping effort in collaboration with several research groups including the Lamont-Doherty Earth Observatory of Columbia University. We determine element, and especially metal content of the sediment grab samples using x-ray fluorescence (XRF), and map their distribution in the study area, which is essential for characterizing the LIS sediment environment. By comparing the current metal contamination data with historical records, and grain size information we seek to uncover environmental changes over time. The analysis reveals an increase in metal concentrations from east to west, particularly in finegrained sediments. Historical industrial activities significantly contribute to elevated metal levels, with higher concentrations of copper and zinc observed due to restricted water circulation and industrial runoff. Other factors such as weather, human activity and ocean currents probably also influenced the detailed metal distribution.