We propose a hydrologic-specific Explainable AI (XAI) framework to extract nonlinear and nonstationary Impulse Response Functions (IRFs) used by Long Short-Term Memory (LSTM) models to simulate streamflow. IRFs reveal how LSTMs emulate streamflow generation and celerity propagation processes in response to impulses of precipitation (P), temperature (T), and potential-evapotranspiration (PET), enabling the evaluation of LSTMs' functional realism under short-term and long-term varying climate/weather conditions. Applying this framework to 672 catchments in USA and Canada, we found that while LSTMs achieve exceptionally high predictive accuracy, their extracted functionality often contradicts established hydrologic principles. Unexpectedly, the isolated effects of T or PET short-term variations on LSTM's simulated streamflow and celerity rate are often positive in direction. For example, in over 70% of raindominated catchments, particularly along the Pacific Coast, increased T within 1-14 days prior to streamflow events is associated with higher streamflow and celerity rates. Similarly, in the southeastern USA and California, LSTMs predict increased streamflow solely in response to PET rises. In snow-dominated catchments, particularly in the Rockies, LSTMs misinterpret PET (representing evaporation function) as a proxy for temperature, the primary snowmelt's driver. These behaviors, likely influenced by seasonality in dynamic inputs, data non-homogeneity, simplicity bias, and missing causal factors, undermine LSTMs' scientific reliability for streamflow forecasting and climate impact assessments. Our XAI framework integrates deep learning with hydrologic context through Differentiable modeling, providing a screening tool for the evaluation of hydrologic models' functional realism in short-and long-term forecasting, climate change studies, and generalization to ungauged basins.

Effective assessment of potential climate impacts requires the ability to rapidly predict the time-varying response of climate variables. This prediction must be able to consider different combinations of forcing agents at high resolution. Full-scale ESMs are too computationally intensive to run large scenario ensembles due to their long lead times and high costs. Faster approaches such as intermediate complexity modeling and pattern scaling are limited by low resolution and invariant response patterns, respectively. We propose a generalizable framework for emulating climate variables to overcome these issues, representing the climate system through spatially resolved impulse response functions. We derive impulse response functions by directly deconvolving effective radiative forcing (ERF) and near-surface air temperature time series. This enables rapid emulation of new scenarios through convolution and derivation of other impulse response functions from any forcing to its response. We present results from an application to near-surface air temperature based on CMIP6 data. We evaluate emulator performance across 5 CMIP6 experiments including the SSPs, demonstrating accurate emulation of global mean and spatially resolved temperature change with respect to CMIP6 ensemble outputs. Global mean relative error in emulated temperature averages 1.49% in mid-century and 1.25% by end-of-century. These errors are likely driven by state-dependent climate feedbacks, such as the non-linear effects of Arctic sea ice melt. We additionally show an illustrative example of our emulator for policy evaluation and impact analysis, emulating spatially resolved temperature change for a 1000 member scenario ensemble in less than a second.

The Arctic tundra is a significant source of CH4 emissions that are generated through anaerobic methanogenesis, and vegetation has explained much of the heterogeneity in CH4 fluxes across the Arctic landscape. Wet sedge-dominated tundra exhibits particularly high rates of CH4 flux due to the presence of aerenchyma and therefore, plant-mediated transport of CH4. Yet the response of CH4 to future CO2 concentrations is still largely uncertain. Model predictions assumed that elevated CO2 (eCO2) will increase primary production and, as a result, CH4 emissions, but this has been poorly tested given the challenges of conducting eCO2 experiments in the Arctic. To address this knowledge gap, we tested the response of CH4 fluxes in soil and vegetation cores from a wet-sedge tundra ecosystem in Alaska to eCO2. The eCO2 stimulated plant productivity, but did not increase CH4 emissions. We think that carbon substrate availability is not limiting CH4 emissions from tundra soils because there is already substantial soil carbon present. Therefore, even an increase in plant productivity will not necessarily result in increased tundra CH4 emissions. Model predictions should consider the soil carbon storage in different ecosystems when predicting the increase in CH4 loss with elevated CO2.

A document by Meinrat Andreae. Click on the document to view its contents.

Magnetic flux ropes are an important area of research in space physics. Among the various magnetic field enhancement events in the solar wind, some remain puzzling due to their multiple possible origins, with interlinked flux ropes being a key candidate. In this study, we analyzed two distinct events to develop a classification framework. Using high-resolution Hall MHD models of interlinked flux ropes, we reproduced the observed magnetic field and velocity components. Our findings show that one event involved flux ropes with a chirality combination that favors magnetic reconnection, while the other had a configuration that suppressed reconnection, consistent with observations. By applying standard data analysis methods, we identified chirality combinations consistent with our models. This study provides direct evidence of 3D flux rope interactions and offers data products to facilitate future event classification.

• Synthetic, holistic review of tracer Green's function techniques in ocean circulation models • Emphasize common underlying theory and ability to customize tracers to address diverse questions • Illustrate power of ocean tracer Green's functions for diagnostics of transport timescales and pathways

A document by Hesam Saeidi. Click on the document to view its contents.

Enhanced Weathering (EW) of silicate rocks is viewed an essential methodology for scaling carbon dioxide removal (CDR) to reach goals of negative global CO2 emissions. EW has an equally important potential for improving agricultural soil, increasing crop yields, and reducing traditional fertilizer use and associated waste. Of all potential silicate raw materials for EW basaltic (mafic) rocks have perhaps the greatest potential, even though the total amount of CDR per unit is less than pure minerals like olivine. Basaltic rocks are widespread compared to other lithologies, are proven to benefit degraded agricultural soils, have significant CDR potential, and minimal negative environmental impacts including heavy metals. Systematic approaches to select basaltic raw material for agricultural and CDR benefits include petrographic and XRD analyses of minerals, XRF and ICP-MS analyses of major and trace elements and grainsize analysis of primary milled material or byproduct material from other aggregate uses. Analysis of commercially available “basalt” shows a wide range in relative quality in terms of primary mineralogy, chemical composition, degree of alteration and metamorphism, achievable grain size, proximity to low carbon transportation and application sites, climate, soil type, and other economic factors. Some products advertised as basalt are dark colored felsic rocks (andesite, dacite) or metamorphosed mafic rocks all of which have lower potential for CDR. These factors limit potential sources of basalt to a small percentage of quarry locations. Mafic, basaltic rocks have SiO2 contents between 45 and 52 percent and vary widely in percentages of mineral phenocrysts– e.g. olivine may range from 0 to >5%, which are also factors in selection of raw material. From an agricultural perspective, analyses of water extractable and Mehlich-3 nutrients reveal values less than the totals of primary unweathered basaltic material; these values increase with degree of EW. Widespread adoption of EW methods for improving soil health will require local and regional crop specific trials and realistic guidelines for application rates and grain sizes that are practical and economic for producers. Characterizing raw material is paramount to instilling confidence in CDR potential and benefits for crops and soil health.

A document by Md Shadman Sakib. Click on the document to view its contents.

A document by Chandra P Dubey. Click on the document to view its contents.

Applying a nested watershed framework to concentration-discharge (cQ) relationships offers valuable insights into the spatial and temporal variability of solute transport. We examined nitrate-N cQ relationships at two nested spatial scales in a small agricultural watershed using three years of 30-minute data, analyzing both composite records and individual storm events. Metrics including the flushing index (FI), hysteresis index (HI), cQ slope (b) and the ratio of coefficients of variation for concentration and discharge (CVC/CVQ) were used to compare event-scale and composite cQ patterns, evaluate the influence of metric choice, assess spatial consistency during 97 concurrent storm events, and identify controls on inter-event variability. Composite-scale analyses showed nitrate-N enrichment, while event-scale patterns were dominated by dilution or chemostasis, depending on the metric used. Across scales, nitrate-N enrichment on the falling limb, or post-storm enrichment, emphasized the need for an additional metric to characterize falling limb dynamics. Spatial consistency in cQ behavior varied by metric: CVC/CVQ and FI showed higher consistency during concurrent storms, while cQ slope and HI often revealed contrasting patterns. Controls on event-scale cQ patterns varied across spatial scales. Indicators of seasonality, antecedent moisture, and storm event magnitude influenced the FI and cQ slope only at the subcatchment, while the HI was primarily driven by storm magnitude across scales. These findings highlight the importance of scale, both temporal and spatial, and metric choice in interpreting nitrate-N export dynamics.

Despite the complexity of the physical mechanisms behind precipitation, many studies linking precipitation to weather types or atmospheric circulation focus on a single variable, oversimplifying the interactions involved. An alternative, multivariate approach to examining various aspects of precipitation generation is through the use of air masses (AMs). The gridded weather typing classification (GWTC-2) is a synoptic weather typing/air mass classification scheme that captures the holistic, multivariate nature of weather at any given time and location, utilizing multiple atmospheric variables. This study examined the association between precipitation and GWTC-2 air masses (AMs) on regional to global scales-the first use of the GWTC-2 AMs to study precipitation. Although regional differences exist, AMs modulate precipitation over land more than in oceanic areas. There is also a noticeably weaker association between AMs and tropical precipitation compared to extratropical precipitation, where up to 60% of summer precipitation variability is explained. Also, land precipitation is better modulated by AMs than oceanic precipitation. Regarding individual air mass influences, Humid Warm AMs surprisingly poorly correlate with land precipitation, likely because water vapor does not increase linearly with higher temperatures over land. Also, compared to the other two humid AMs, the Humid Cool AM contributes 5x more precipitation in deserts. More interestingly, the frontal AMs, defined to capture migrating extratropical cyclones, strongly contribute to tropical precipitation and capture the West African Monsoon remarkably well. The association between GWTC-2 AMs and precipitation has potential applicability in ensemble precipitation forecasting and downscaling GCM model simulations of daily precipitation.

A document by Saman Aryana. Click on the document to view its contents.

This study presents and quantitatively analyzes a new dataset detailing areas affected by ballistic projectiles generated by major explosions and paroxysms at Stromboli. The dataset includes a total of 67 events over ≈150 years, based on an extensive review of historical, observational, and monitoring data. By quantifying distances, directions, and areas affected by ballistic fallout, we provide the necessary data, together with their uncertainty, to produce probabilistic maps of ballistic hazard as presented in a companion study. Our main findings include: (1) 23%-37% of major explosions do not exceed 500 m in their maximum ballistic distance, while 17%-42% of paroxysms remain under 1,500 m; however, 12%-14% of major explosions can exceed 1,000 m, and 29% of paroxysms extend over 2,000 m; (2) sector-based analysis of ballistic distance distribution produces probabilities up to three times lower compared to those based on the maximum distances; (3) directional analysis of ballistic dispersal shows a predominant direction towards the East half-plane (87%) for major explosions and towards the North half-plane (64%) for paroxysms; (4) the average affected area was 6.9e+04 m² for major explosions and 3.6e+05 m² for paroxysms with a mean amplitude of ≈90 degrees for both categories. Notably, major explosions and paroxysms show a continuous distribution of maximum ballistic distance and area affected, suggesting the absence of a net separation between these two categories with respect to this phenomenon. Results highlight the limited influence of uncertainty in reconstructing the dispersal areas and stress the importance of volcanological monitoring of Stromboli’s explosive phenomena

The Longitudinal Valley Fault (LVF) and the Central Range Fault (CRF) in eastern Taiwan consist of a head-to-head double-vergence structure hosting disastrous earthquakes. It was previously proposed that the fault slip on one of these faults suppresses the earthquake generation on the other. Nonetheless, the 2022 Chihshang earthquake (Mw 7.0) on the CRF occurred soon after the 2022 Yuli earthquake (Mw 6.7) on the LVF. Here, we provide a comprehensive framework of the fault interaction consistently explaining these contradictory findings. First, we estimated the coseismic slip distribution of the Yuli earthquake from Global Navigation Satellite System (GNSS) and L-band satellite interferograms. The results indicate almost pure reverse faulting. Second, with the estimated slip distribution, we calculated changes in Coulomb failure function (ΔCFF) on the CRF due to the Yuli earthquake on the LVF. The ΔCFF reaches +0.25 MPa around the main rupture area of the Chihshang earthquake, which is equivalent to the clock advance of ~50 years, suggesting a large impact on the earthquake generation cycles. Finally, we found that a rake angle of fault slip has a significant effect on the ΔCFF on the other fault in the double-vergence structure: it takes large positive values when 90° like the Yuli earthquake, but almost negative when 45° or less, which comprehensively explains the seismic quiescence previously reported and the positive ΔCFF on the CRF caused by the Yuli earthquake. The strong impact of the rake angle is also supported by the temporal distribution of historical earthquakes in eastern Taiwan.

A document by Abdelhamid Ads. Click on the document to view its contents.

A document by Suma Bhanu Battula. Click on the document to view its contents.

Framed by Connecticut’s rocky shore and New York’s outstretched Long Island, the 180 km tidal estuary known as the Long Island Sound makes up one of many major estuary systems across North America’s eastern coastline. Since the Last Glacial Maximum roughly 21,000 years ago, the formation of the Long Island Sound tells of an intricate geologic history—of a thawing Laurentide Ice Sheet, a draining glacial lake, river-sculpted lakebeds, and the inpour of a rising ocean. The rise of the Atlantic has turned what was once the exposed lakebed of Glacial Lake Connecticut into a flooded estuary, burying away the Sound’s chronologic record held in submerged paleo-features and remnant glacial lake deposits. In an effort to elucidate the glacial to post-glacial history of the Sound, we use high-resolution 2-D CHIRP sonar data to image sub-bottom strata in the Sound’s northeastern region between Westbrook, CT, and Fisher’s Island, where Glacial Lake Connecticut once drained toward a farther Atlantic. Subsurface facies interpretations of digitized profiles were created using Kingdom Suite (Seismic Micro-Technology, Inc.) and then interpolated using ArcGIS Pro, reconstructing a detailed mapping of the region’s sub-bottom. Subsurface facies interpretations were created using Procreate. Substantial erosion from ongoing marine action has exposed considerable portions of lakebed deposits at the sound floor alongside outcropping bedrock, buried promontories, and paleo-shore remnants. Our study revealed preserved glacial lake deposits channelized by 4-6 newly identified southward flowing paleo-rivers, potentially joining at the confluence of a greater, previously mapped paleo-river. Interpreted unconformities and depositional shifts in glacial lake sediments added complexities to Glacial Lake Connecticut’s formative chronology and the dynamics of its meltwater sources, drainage, and estuarine transition. 2-D CHIRP sub-bottom profiles proved to be an effective tool for the preliminary identification of submerged and buried features within a complex, geodiverse region of the Long Island Sound.