Simulations of human behavior in water resources systems are challenged by uncertainty in model structure and parameters. The increasing availability of observations describing these systems provides the opportunity to infer a set of plausible model structures using data-driven approaches. This study develops a three-phase approach to the inference of model structures and parameterizations from data: problem definition, model generation, and model evaluation, illustrated on a case study of land use decisions in the Tulare Basin, California. We encode the generalized decision problem as an arbitrary mapping from a high-dimensional data space to the action of interest and use multi-objective genetic programming to search over a family of functions that perform this mapping for both regression and classification tasks. To facilitate the discovery of models that are both realistic and interpretable, the algorithm selects model structures based on multi-objective optimization of (1) their performance on a training set and (2) complexity, measured by the number of variables, constants, and operations composing the model. After training, optimal model structures are further evaluated according to their ability to generalize to held-out test data and clustered based on their performance, complexity, and generalization properties. Finally, we diagnose the causes of good and bad generalization by performing sensitivity analysis across model inputs and within model clusters. This study serves as a template to inform and automate the problem-dependent task of constructing robust data-driven model structures to describe human behavior in water resources systems.

We surveyed the subsurface structure in eastern Coprates and Capri Chasmata in the equatorial region using high-resolution visible images, digital terrain models, and radar sounding data. We identified subsurface reflectors in four areas of the chasmata. At the stratigraphic exposure on the chasmata walls, the corresponding depth of the reflector is ~60 m. The bulk dielectric constants of the layers above the reflectors are calculated as 3.4-4.0, suggesting a rock-air mixture with ~46.1% porosity, or a rock-air-ice mixture with ~21.2% water-ice fraction. Recent climate models suggest that water-ice is unstable on the surface around the equatorial regions. However, considering the recent high obliquity that occurred ~0.4 Ma and a slow diffusivity of water-ice, the existence of subsurface water-ice deeper than a few meters cannot be ruled out. If water-ice is actually contained in the layer, our results show the maximum volume of putative water-ice in the chasmata is 16.6 km.

As a result of uneven density of data collection, level-2 satellite gravimetry data suffer from global north-south striping. By applying various filtering methods, several studies have addressed the mitigation of the data. However, the studies mainly addressed the issue on a global scale, and the local effects were not considered. On the other hand, water research, especially inland hydrology, usually deals with small-scale fitures such as lakes and watersheds. Therefore, the local data de-striping methods need special attention. This research presents a new analytical method to de-stripe gravimetry data based on the spatial contrast of signals. The approach strikes a balance between de-striping and signal preservation. Using a-priori information obtained from the gravimetry data, the de-striping method first estimates the spatial gradient of the signal and optimizes a Poisson filter based on this information to de-stripe the data. Unlike the other approaches, the optimized filter is dynamic and accounts for temporal variations in the signal contrast, such as seasonality. The proposed approach is applied to ten globally distributed study areas to derive a general scheme. Detailed processes and evaluations are applied to two study areas: the Caspian Sea and the Congo River Basin. Results are visually assessed for spatial fit and for temporal consistency by comparison with results from other filters. The use of a dynamic filter set specified for each region and time point allows us to preserve local hydrologic signals that are susceptible to globally optimized filters. It also allows filter-related errors to be effectively constrained.

We present an initial assessment of using tomography on single-spacecraft images to reconstruct 3D X-ray emissions from the Earth’s magnetosheath. 3D structures in the Earth’s magnetosphere have been studied using superposed epoch techniques with single-point single-spacecraft observations. They have yielded great insights, but some studies are observation starved, particularly for infrequent solar wind conditions. Global imaging data have provided more insight about these structures, but are 2D projections of 3D structures. We explore the use of tomographic reconstruction techniques to understand what can be extracted from global images from a single spacecraft. The Solar wind Magnetosphere Ionosphere Link Explorer (SMILE) mission, due to launch in 2024 on a 3-year mission, will carry a soft X-ray imager which will capture emissions from portions of the magnetosheath and upstream solar wind. We already demonstrated that the 3D shape of the magnetopause and the bow shock can be extracted from such images with suitable assumptions. The next step is to examine whether full 3D reconstructions of the emissions are possible. We explore the limited range of viewing angles, which affect the accuracy of the reconstructions and introduce artifacts in some cases, and the low count-rates in the images which introduce noise in the reconstructions which must be filtered out. Despite these limitations we show that it is possible to reconstruct some aspects of the magnetosheath global morphology using single-spacecraft soft X-ray imaging. Plans for similar missions which overlap with SMILE, open the possibility of multi-spacecraft tomography, to be addressed in a separate paper.

Rivers and streams are an important component of the global carbon budget, emitting CO2 to the atmosphere. However, our ability to accurately predict carbon fluxes from stream systems remains uncertain due to small scales of pCO2 variability within streams (100-102 m), which make monitoring intractable. Here we incorporate CO2 input and output fluxes into a stream network advection-reaction model, representing the first process-based representation of stream CO2 dynamics at watershed scales. This model includes groundwater (GW) CO2 inputs, water column and benthic hyporheic zone (BZ) respiration, downstream advection, and atmospheric exchange. We evaluate this model against existing statistical methods including upscaling techniques and multiple linear regression models through comparisons to high-resolution stream pCO2 data collected across the East River Watershed in the Colorado Rocky Mountains. The stream network model accurately captures topography-driven pCO2 variability and significantly outperforms multiple linear regressions for predicting pCO2. Further, the model provides estimates of CO2 contributions from internal versus external sources and suggests that streams transition from GW- to BZ-dominated sources between 3rd and 4th Strahler orders, with GW and BZ accounting for 53 and 47% of CO2 fluxes from the watershed, respectively. Lastly, stream network model CO2 fluxes are 5-13x times smaller than upscaling technique predictions, largely due to inverse correlations between stream pCO2 and atmosphere exchange velocities. Taken together, the stream network model presented improves our ability to predict and monitor stream CO2 dynamics, and future applications to regional and global scales may result in a significant downward revision of global flux estimates.

Most organic matter delivered to the seafloor is degraded within the bioturbated layer. Theory and empirical evidence have shown that organic matter reactivity relates to the age of a particle. However, due to particle mixing, the age-depth linear relation induced by sediment accretion is obfuscated. Here we combine a Lagrangian particle tracking model that resolves the age distribution of particles in the bioturbated zone and couple it to age-dependent organic matter degradation. Depth profiles for organic carbon concentration, reactivity and degradation rate are presented for sediments receiving low-quality and high-quality organic matter in coastal, continental slope and deep-sea environments. Our results show that a simple first-order kinetics model suffices for well-mixed sediments and systems receiving pre-processed materials. A reactive continuum approach is needed for poorly-mixed sediments receiving high-quality organic matter and for sediments below the bioturbated layer.

We evaluate the mass balance of the Russia High Arctic Archipelago between April 2002 and August 2016 employing independent estimates obtained using time-variable gravity from the NASA/DLR GRACE mission and satellite altimetry data from the NASA ICESat and the ESA CryoSat-2 missions. We present a new algorithm designed to derive ice elevation change maps using altimetry data and we evaluate its performance over the region of interest. Gravimetric and Altimetric observations provide consistent results and show that over the period under analysis, glaciers in the region have lost mass at a rate of 15 +/-7 Gt/yr corresponding to a sea level contribution of 0.039 mm/yr. The mass loss increased after 2010, reaching a maximum rate of 25 +/-7 Gt/yr between 2010 and 2016. The increased mass loss was associated with high thinning rates at low elevations (below 500 m), with marine-terminating glaciers thinning significantly faster than those terminating on land. The mass loss process was associated to a shift in climatic conditions in the region due to enhanced atmospheric and ocean temperatures and decreased sea ice concentrations. These results indicate that glaciers in the region are sensitive to variations of both climatic mass balance and ice discharge.

1. Noble gas tracers can infer the rate of sea ice production in polynyas. 2. Frazil ice in polynyas appears to block air-sea gas exchange mechanisms. 3. The solubility pump is influenced by glacial ice melt and sea ice formation.

Groundwater resources of the Southern High Plains/Ogallala Aquifer in Texas and New Mexico are being depleted due to groundwater mining for irrigation. Inevitably, resource depletion leads to calls for water conservation in agriculture. Conservation can take two forms: a reduction in irrigation depth, and a reduction in irrigated area, which in economics are termed the “intensive margin” and “extensive margin” of agricultural water use. In the Southern High Plains, we find different effects on water table elevation arising from these two approaches. This research presents a coupled model of landscape and groundwater hydrology. Model results indicate that a 50% reduction in irrigation water application would limit loss in irrigable area to about 1% of existing irrigable land per year. This is approximately half the rate of depletion from a ‘business as usual’ scenario. Relative benefits of each conservation approach varied: areas with a high density of irrigated land experienced greater benefits from a reduction in irrigation depth, whereas reducing irrigated acreage maintained water tables more in areas with a low density of irrigated land. This project demonstrates that strategies for irrigation water management can support conservation goals. However, model results also demonstrated that even a 50% reduction in irrigation water use –which would be politically and economically unfeasible in Texas and New Mexico – would still result in overall depletion of the regional aquifer.

The direct observation of water movement via Electrical Resistivity Imaging (ERI) can leverage the understanding of the processes that lead to the occurrence of variable residence times (RT) within the Critical Zone (CZ). While hydrological processes at natural landscapes are often space and time-variable, quantitatively estimating solute transport with ERI under transient conditions is challenging due to necessary considerations of moisture states and electrical properties of the medium. Here, we introduce the use of Periodic Steady State (PSS) theory applied to electrical resistivity of soils to provide a simple solution to the problems and report a laboratory experiment to test the proposed method. We used a 1 m3 sloping lysimeter to represent the hydrological functioning of natural hillslopes, equipped with electrodes to provide cross-borehole images of bulk soil electrical conductivity and performed a 28-days experiment in which a periodic irrigation was applied. A saline tracer was added to the lysimeter in two irrigation pulses and subsequent pulses were applied until the tracer was flushed out. ERT-surveys and estimates of background soil-water conductivity were used to quantitatively estimate solute breakthrough throughout the different cross-sections. Integrated lysimeter-scale images were superimposed with the water table progression throughout the experiment to leverage the understanding of flow and transport processes responsible for the tracer mobilization. Our study introduces a novel method for laboratory experimentation at mesocosm scales using ERT and provides valuable insight into the role of water table dynamics in mediating the occurrence of variable flow pathways within hillslopes.

The downward flux of organic carbon exported from the surface ocean is of great importance to the Earth’s climate because it represents the major pathway for transporting CO from the surface ocean and atmosphere into the deep ocean and sediments where it can be sequestered for a long time. Here we present global-scale estimates for the export fluxes of total, dissolved, and particulate organic carbon (TOC, DOC, and POC, respectively) constrained by observed thorium-234 (Th) activity and dissolved phosphorus (DIP) concentration in a global inverse biogeochemical model for the cycling of phosphorus and Th. We find that POC export flux is low in the subtropical oceans, indicating that a projected expansion of the subtropical gyres due to global warming will weaken the gravitational biological carbon pump. We also find that DOC export flux is low in the tropical oceans, intermediate in the upwelling Antarctic zone and subtropical south Pacific, and high in the subtropical Atlantic, subtropical north Pacific, and productive subantarctic zone (SAZ). The horizontal distribution of DOC export ratio (F/F) increases from tropical to polar regions, possibly due to the detrainment of DOC rich surface water during mixing events into subsurface waters (increasing the strength of the mixed layer pump poleward due to stronger seasonality). Large contribution to the export flux from DOC implies that the efficiency with which photosynthetically fixed carbon is exported as particles may not be as large as currently assumed by widely used global export algorithms.

Since the 1960s, carbon cycling in the high-latitude northern forest (HLNF) has experienced dramatic changes: most of the forest is greening and net carbon uptake from the atmosphere has increased. During the same time period, the COseasonal cycle amplitude (SCA) has almost doubled. Disentangling complex processes that drive these changes has been challenging. In this study, we substitute spatial sensitivity to temperature for time to quantify the impact of temperature increase on Gross Primary Production (GPP), total ecosystem respiration (TER), the fraction of Photosynthetic Active Radiation (fPAR), and the resulted contribution of these changes in amplifying the COSCA over the HLNF since 1960s. We use the spatial heterogeneity of GPP inferred from solar-induced chlorophyll Fluorescence in combination with net ecosystem exchange (NEE) inferred from column COobservations made between 2015 and 2017 from NASA’s Orbiting Carbon Observatory -2. We find that three quarters of the spatial variations in GPP and in the fPAR absorbed by the HLNF can be explained by the spatial variation in the growing season mean temperature (GSMT). The long term hindcast captures both the magnitude and spatial variability of the trends in observed fPAR. We estimate that between 1960 and 2010, the increase in GSMT enhanced both GPP and the SCA of NEE by ~20%. The calculated enhancement of NEE due to increase in GSMT contributes 56–72% of the trend in the CO SCA at high latitudes, much larger than simulations by most biogeochemical models.

The partial reflection facility of the Polar Geophysical Institute (the Tumanny observatory, 69.0N, 35.7E) has observed behavior of the high-latitude lower ionosphere during the 20 March 2015 total solar eclipse. There were several effects during the eclipse. At the heights of 60–80 km the ionosphere has shown the effect of a “short night”, but at the higher altitudes local enhanced electron concentration had a wave-like form. Data received by the riometer of the Tumanny observatory have also shown wave-like behavior. The behavior can be explained by influence of acoustic-gravity waves which originated after cooling of the atmosphere during the lunar shadow supersonic movement, and transport processes during the eclipse. During the 21 August 2017 solar eclipse there was a substorm at the high latitudes. But after the end of the substorm in the region of the Tumanny observatory the observed amplitudes of the reflected waves had wave effects which could be connected with the coming waves from the region of the eclipse. The wave features were also shown in the behavior of the total electron content (TEC) of the lower ionosphere. During several solar eclipses it was implemented observations of lower ionosphere behavior by the partial reflection facility of the Tumanny observatory. The consideration of the lower ionosphere TEC had revealed common features in the TEC behavior during the eclipses. The photochemical theory of processes in the lower ionosphere is very complicated and up to now it is not completely developed. Therefore introduction of the effective coefficients determining the total speed of several important reactions has been widely adopted when modeling the D-region of the ionosphere. However, experimental opportunities for obtaining effective recombination coefficients are rather limited. One of the methods to estimate effective recombination coefficients uses the phenomenon of a solar eclipse. During solar eclipses at the partial reflection facility of the Tumanny observatory observations were carried out. It gave possibility to obtain the behavior of the electron concentration in time at the selected heights. Using the obtained experimental profiles, the effective recombination coefficients at the D-region heights of the ionosphere have been evaluated.

We use Wasserstein Generative Adversarial Networks to learn and integrate multi-scale features in segmented three-dimensional images of porous materials, enabling the dependable generation of large-scale representations of complex porous media. A Laplacian pyramid generator is introduced which creates pore-space features with a hierarchy of spatial scales. Feature statistics mixing regularization enhances the ability of the model generation to reliably maintain multi-scale pore-space features of images by increasing diversity. The method is tested on a variety of X-ray images of porous rocks. The generated images can be of any size -- cm-scale ten-billion-cell images are generated to demonstrate the power of the approach -- which have two-point correlation functions, porosity, permeability, Euler characteristic, curvature, and specific surface area close to the training datasets. The images demonstrate a considerable improvement over previously- published studies using generative adversarial networks.

Background: Capillary trapping of gas bubbles and oil blobs within water-saturated media plays an important role for underground gas storage and secondary oil recovery. Wettability and roughness of the surface are elementary properties of a porous medium that determine the trapping efficiency. In previous work [1,2], we demonstrated that glass beads and natural sands display a significant difference (15%) of the trapped gas phase. Here, we carry out a systematic study of the capillary trapping efficiency in dependence of the wettability and surface roughness. Methods: We conducted a series of column experiments to study capillary trapping of gaseous CO2 using both glass beads and natural sands as sediments. Based on the high-resolution non-invasive micro-CT visualization method and subsequent image processing, we quantified capillary trapping efficiency, gas-cluster morphology and gas-cluster size distribution. We used the silanization method for varying degrees of wettability resulting in three different contact angles on microscopic soda lime glass slides: (i) Piranha cleaning (= 7°), (ii) untreated glass (= 30°) and (iii) silanized glass (=100°). Results: We observed that by-pass trapping is the dominant trapping mechanism in glass beads (smooth surfaces). The displacement process is piston-like. For natural sands (rough surface), thick film flow occurs, causing an efficient snap-off trapping mechanism. Our results indicate that the capillary trapping efficiency of natural sands is stronger reduced by a transition from water-wet to CO2-wet 3-phase system (increasing contact angle) when compared to glass beads. [1] H. Geistlinger, I. Ataei-Dadavi, S. Mohammadian, and H.-J. Vogel (2015) The Impact of Pore structure and Surface Roughness on Capillary Trapping for 2D- and 3D-porous media: Comparison with Percolation theory. Special issue: Applications of Percolation theory, Water Resour Res, 51, doi:10.1002/2015WR017852. [2] H. Geistlinger, I. Ataei-Dadavi, and H.-J. Vogel (2016) Impact of Surface Roughness on Capillary Trapping Using 2D-Micromodel Visualization Experiments. Transport in Porous Med., DOI 10.1007/s11242-016-0641-y.

Machine learning has the potential to automate the analysis of vast amounts of raw geophysical data, allowing scientists to monitor changes in key aspects of our climate such as cloud cover in real-time and at fine spatiotemporal scales. However, the lack of large labeled training datasets poses a significant barrier for effectively applying machine learning to these applications. Transfer learning, which involves first pretraining a neural network on an auxiliary “source” dataset and then finetuning on the “target” dataset, has been shown to improve accuracy for machine learning models trained on small datasets. Across prior work on machine learning for geophysical imaging, different choices are made about what data to pretrain on, and the impact of these choices on model performance is unclear. To address this, we systematically explore various settings of transfer learning for cloud classification, cloud segmentation, and aurora classification. We pretrain on different source datasets, including the large ImageNet dataset as well as smaller geophysical datasets that are more similar to the target datasets. We also experiment with multiple transfer learning steps where we pretrain on more than one source dataset. Despite the smaller source datasets’ similarity to the target datasets, we find that pretraining on the large, general-purpose ImageNet dataset yields significantly better results across all of our experiments. Transfer learning is especially effective for smaller target datasets, and in these cases, using multiple source datasets can give a marginal added benefit.

Water returned to the atmosphere as evapotranspiration (ET) is approximately 1.6x greater than global river discharge and has wide-reaching impacts on groundwater and streamflow. In the U.S. Midwest, widespread land conversion from prairie to cropland has altered spatiotemporal patterns of ET, yet there is no consensus on the direction of change in ET or the mechanisms controlling changes. We aimed to harmonize findings about how land use change affects ET in the Midwest. We measured ET at three locations within the Long-Term Agroecosystem Research (LTAR) network along a latitudinal gradient with paired rainfed cropland and prairie sites at each location. At the northern locations, the Upper Mississippi River Basin (UMRB) and Kellogg Biological Station (KBS), the cropland has annual ET that is 84 and 29 mm/year higher, respectively, caused primarily by higher ET, likely from soil evaporation during springtime when agricultural fields are fallow. At the southern location, the Central Mississippi River Basin (CMRB), the prairie has 69 mm/year higher ET, primarily due to a longer growing season. To attribute differences in springtime ET to specific mechanisms, we examine the energy balance using the Two-Resistance Method (TRM). Results from the TRM demonstrate that higher surface conductance in croplands is the primary factor leading to higher springtime ET from croplands, relative to prairies. Results from this study provide critical insight into the impact of land use change on the hydrology of the U.S. Corn Belt by providing a mechanistic understanding of how land use change affects the water budget.

We have used the LOw-Frequency ARray (LOFAR) to image a few lightning flashes during a particularly severe thunderstorm. The images show an exceptional amount of VHF activity at altitudes above 10 km. Much of this is in the form of small-scale discharges occurring seemingly randomly around the centers of active storm cells. Because of their small and incidental structure we refer to these as ‘speckles’. A detailed investigation shows strong evidence that these speckles are indicative of positive leader channels and that they are equivalent to the needle activity seen around positive leader tracks at lower altitudes.