We present herein a new basis for measuring river discharge in ungauged catchments. Surrogate runoff (SR) is created using remotely sensed data to compensate for the absence of ground streamflow measurements. Because of their widespread availability, remotely sensed SR products are attractive, with approaches such as satellite-derived measurement-calibration ratio (C/M ratio). However, the use of the C/M ratio suffers from its limited penetration through ground vegetation canopies. While a microwave signal with a longer wavelength has been used to enhance the penetration capability, the coarseness of the spatial resolution of the microwave signal offsets its improvement due to the inherent assumptions in the C/M ratio, i.e., selecting two contrasting pixels (i.e., measurement and calibration) at the same time. To address both issues, this study proposes a new SR formulation using a longer wavelength (L-band microwave) with a better assumption for handling coarse grids, whereby the temporal variability of dryness against the driest state in each grid is used. The performance of the new SR is assessed for 467 Australian Hydrologic Reference Station catchments. Results show considerable improvements in the Pearson linear correlation (R) between the proposed SR and streamflow: 44% of the study areas show R higher than 0.4 with the new approach, whereas only 13% of the study areas show R higher than 0.4 with the currently used alternative (C/M ratio derived from Ka-band microwave). Overall, the resulting SR is dramatically improved by using the newly designed SR approach with the L-band microwave signal.

Robust and reliable projections of future streamflow are essential to create more resilient water resources, and such projections must first be bias corrected. Standard bias correction techniques are applied over calendar-based time windows and leverage statistical relations between observed and simulated data to adjust a given simulated datapoint. Motivated by a desire to connect the statistical process of bias correction to the underlying dynamics in hydrologic models, we introduce a novel windowing technique for projected streamflow wherein data are windowed based on hydrograph-relative time, rather than Julian day. We refer to this method as ‘seasonally anchored’. Four existing bias correction methods, each using both the standard day-of-year and the novel windowing technique, are applied to daily streamflow simulations driven by 10 global climate models across a diverse subset of six watersheds in California to investigate how these methods alter the model climate change signals. Among the methods, only PresRat preserves projected annual streamflow changes, and does so for both windowing techniques. The seasonally anchored window PresRat reduces the ensemble bias by a factor of two compared to quantile mapping (Qmap), cumulative distribution function transform (CDFt), and equidistant quantile matching (EDCDFm) methods. For wet season flows, PresRat with seasonally anchored windowing best preserves the original model change over the entire distribution, particularly at the highest quantiles, and the other three methods show improved performance using the novel windowing method. Concerning temporal shifts in seasonality, PresRat and CDFt preserve the original model signals with both the novel and standard windowing methods.

Increasing wildfire and declining snowpacks in mountain regions threaten water availability. We combine satellite-based fire detection with snow seasonality classifications to examine fire activity in California’s seasonal and ephemeral snow areas. We find a nearly tenfold increase in fire activity during 2020 and 2021 compared to 2001-2019 as measured by satellite data. Accumulation season snow albedo declined 17-77% in two burned sites as measured by in-situ data relative to un-burned conditions, with greater declines associated with increased soil burn severity. By enhancing snowpack susceptibility to melt, decreased snow albedo drove mid-winter melt during a multi-week midwinter dry spell in 2022. Despite similar meteorological conditions in 2013 and 2022, which we link to persistent high pressure weather regimes, minimal melt occurred in 2013. Post-fire differences are confirmed with satellite measurements. Our findings suggest larger areas of California’s snowpack will be increasingly impacted by the compounding effects of dry spells and wildfire.

Hundreds of ancient palaeolake basins have been identified and catalogued on Mars, indicating the distribution and availability of liquid water as well as sites of astrobiological potential. Palaeolakes are widely distributed across the Noachian aged terrains of the southern highlands, but Arabia Terra hosts few documented palaeolakes and even fewer examples of open-basin palaeolakes. Here we present a detailed topographic and geomorphological study of a previously unknown set of seven open-basin palaeolakes adjacent to the planetary dichotomy in western Arabia Terra. High resolution topographic data were used to aid identification and characterisation of palaeolakes within subtle and irregular basins, revealing two palaeolake systems terminating at the dichotomy including a ~160 km chain of six palaeolakes connected by short valley segments. Analysis and correlation of multiple, temporally distinct palaeolake fill levels within each palaeolake basin indicate a complex and prolonged hydrological history during the Noachian. Drainage catchments and collapse features place this system in the context of regional hydrology and the history of the planetary dichotomy, showing evidence for the both groundwater sources and surface accumulation. Furthermore, the arrangement of large palaeolakes fed by far smaller palaeolakes, indicates a consistent flow of water through the system, buffered by reservoirs, rather than a catastrophic overflow of lakes cascading down through the system.

Transformation of rainfall to runoff is a complex hydrological phenomenon involving various interconnected processes. Besides, the distribution of rainfall and basin characteristics are not uniform across time and space leading to a poor understanding of the process. Hydrologists have been using various hydrological models to understand transformation of rainfall into runoff. Conceptual models developed in the 1960s represent various individual components of hydrological cycle via interconnected conceptual elements, thus model various aspects of the hydrological cycle. On the other hand, data-driven models such as Artificial Neural Networks (ANNs) are widely regarded as universal approximators due to their ability to model many complex problems. Very few studies reported the application of a widely used conceptual model, Sacramento Soil Moisture Accounting model (SAC-SMA), in the Indian river basins context. Considering that the hydrological cycle is very complex and may never be fully understood in detail, conceptual models like Sacramento Soil Moisture Accounting model (SAC-SMA) can be integrated with data-driven models which can take care of poorly described and understood aspects of hydrological modelling. In this study, a hybrid rainfall-runoff model was developed and applied over the Godavari river basin in India at multiple spatial scales for capturing the spatial variations in model inputs and catchment charateristics.The hybrid model by virtue of the semi-distributed configuaration and addition of ANN component led to improved simulations of streamflow in comparison to the standalone SAC-SMA model.

A common viewpoint across the Earth science community is that global soil moisture estimates from satellite L-band (1.4 GHz) measurements represent moisture only in the shallow soil layers (0-5 cm) and are of limited value for studying global terrestrial ecosystems because plants use water from deeper rootzones. Here, we argue that such a viewpoint is flawed for two reasons. First, microwave soil emission theory and statistical considerations of vertically correlated soil moisture information together indicate that L-band measurements are typically representative of soil moisture within at least the top 15-25 cm, or 3-5 times deeper than commonly thought. Second, in reviewing isotopic tracer field studies of plant water uptake, we find a global prevalence of vegetation that primarily draws moisture from these upper soil layers. This is especially true for grasslands and croplands covering more than a third of global vegetated surfaces. While shrub and tree species tend to draw deeper soil moisture, these plants often still preferentially or seasonally draw water from the upper soil layers. Therefore, L-band satellite soil moisture estimates are more relevant to global vegetation water uptake than commonly appreciated, and we encourage their application across terrestrial hydrosphere and biosphere studies.

With the intensification of climate change and population growth, human-water relationships (HWR) have changed from the simple utilization of water resources to changing the spatial distributions and distribution proportions of water resources through formulating corresponding policies, such as Chinese ecological civilization policy. However, the impact of the ecological civilization policy on the evolution of HWR is still unclear. Here, taking the 600-year old “Tunpu” area as a typical study area, this research analyses the evolution of HWR over different space and time spans based on the Remote Sensing Hydrological Station (RSHS) technology, an improved water balance formula and the transition theory. The results show that at the village scale, the water cycle structure of a typical village has remained stable, and deforestation has increased the proportion of runoff to precipitation by 10.62%. At the basin scale, due to land use/cover changes and precipitation fluctuations, the trend of the runoff changes from slowly decreasing to accelerated increases, with change rate increasing from -0.073×104 m3·a-1 in the Ming Dynasty (1470-1636) to 30.946×104 m3·a-1 in the China stage (1949-2020). HWR have developed from the initial balanced resource-rich period to the unbalanced extensive-development period and have finally changed into a rebalancing period under the influence of the ecological civilization policy. Four stages of HWR are as follows: predevelopment (1470-1685), take off (1685-1912), acceleration (1912-2000) and rebalancing (2000-2020). This research indicates that the ecological civilization policy can rebuild HWR, and it is expected to provide enlightenment for future construction of the ecological civilization.

Grid-based spatially distributed hydrological modeling became feasible along with advances in watershed routing scheme, remote sensing technology, and computing resources. Such modeling is expected to be common in routine hydrological analysis and watershed management planning as it can maximize the use of spatial information and provide the detailed picture of transport processes. However, the heavy computational requirement and resulting long running time are still barriers that prevent the spatially detailed modeling practices from being employed widely, particularly in a fine-resolution large-scale study. Parallelizing computational tasks has been successful in mitigating the difficulty. We propose a noble way to improve the simulation efficiency of direct runoff transport processes by carefully grouping watershed areas based on the time-area routing scheme. The proposed parallelization method was applied to simulating the runoff routing processes of three watersheds draining the areas of 3.79 km2, 133.59 km2, and 2,800 km2 respectively at a 30-m resolution. Results demonstrated that the new method could substantially improve the computational efficiency of the time-area routing method with common computing resources. The efficiency of the parallelization scheme was not limited by the hierarchical relationship between upstream and downstream catchments along flow paths, which could be possible with the Lagrangian flow tracking strategy of the time-area routing method.

This mass balance study was intended to provide up-to-date information about the water quality of the headwater streams draining to the Mohican and Walhonding rivers. This data will be used to define target locations for conservation practices, including agricultural and stormwater management practices. During the study, 124 sites were sampled twice in 2021: during spring high-flow conditions (May) and fall low-flow conditions (August).

While more hydrological data is being generated than ever before, the power of modelling this collected information is not fully realized unless it is of high quality, especially considering hydrological data from sensor networks, which is often errant due to the possibility of malfunction or non-conducive environmental conditions. Fluctuations or errors are difficult to predict, identify, and interpret. Manual models of quality assurance are not designed for managing datasets with continuous timeseries or spatially extensive coverage, resulting in time- consuming models that rely on humanmade decision making and lack statistical inference. This research hypothesizes that the stochasticity of rainfall and deterministic properties of flow can be used in concert to create a more characteristic quality assurance model for high-resolution environmental data. An automated implementation of this model is presented herein with the application of two use-cases, which maintains statistical integrity and circumvents biases and potential for user error of manual frameworks.

Not only are reservoir managers and aquatic scientists concerned with the environmental effects of water quality, civil engineers must also consider water quality to comply with regulations in the construction of new reservoirs, or in making structural and operational modifications to existing reservoirs. This study establishes a machine learning approach for predicting Carlson’s Trophic State Index (CTSI), which is a frequently used metric of water quality in reservoirs. Data collected over ten years (1995-2016) from the stations at 20 reservoirs in Taiwan were preprocessed as the input for the modeling system. Four well-known artificial intelligence (AI) techniques, ANN (Artificial Neural Network), SVM (Support Vector Machine), CART (Classification And Regression Technique), and LR (Linear Regression), were used to analyze in baseline and ensemble scenarios. Moreover, one variation of support vector machine was integrated with a metaheuristic optimization algorithm to develop a hybrid AI model. The comprehensive comparison demonstrated that the ensemble ANN model, based on tiering method, is more accurate than the other single, ensemble, and hybrid models. The novelty of this study is providing a new approach of AI models, reducing the complexity of measuring three traditional parameters of CTSI formula, as an alternative to the conventional approach to predicting CTSI. This work contributes to the improvement of water quality management by providing a versatile technique that offers diverse predictive methods to meet the specific requirements of practitioners.

Throughout the last years, there is an increasing interest of the geoscientific community in using terrestrial gravimetry as an integrative and non-invasive method for observing mass change and mass redistribution in the environment due geophysical processes. The nature of the observed processes and the need for nearby data collection often require the gravimeters to be installed at remote field sites. In contrast to classical deployment at permanent observatory sites, this often is a challenge because there are three main requirements to be fulfilled for continuous high-quality operation of the gravimeter: electrical power, stable site conditions, and data connection. Whereas the latter can usually be accomplished by wireless solutions, the second requirement is more demanding as it requires an adequate design of a gravimeter enclosure and a stable pillar, while the first requirement so far has been practically insolvable in the absence of a power line. Here, we present the prototype of a mobile field container for gravity monitoring that fulfils all above requirements: the gPhone-solar-cube. The container consists of a cubic steel container as used by ships and trucks with edge length of about 2 meters. We optimized all components to host a continuously operating gPhoneX. Components include temperature shielding, ventilation, solar panels, power management and monitoring, storage batteries and an integrated backup generator to guarantee self-sufficient power supply, data loggers and wireless data transfer components. Furthermore we developed a new type of gravimeter pillar which is simple to install and to remove, without connection to the container floor to avoid vibration transfer. The pillar is large enough to accommodate two CG-6 field gravimeters, next to the gPhoneX. Other instruments integrated are a complete weather station and a cosmic ray neutron probe. The gPhone-solar-cube has been installed in the Ore mountains, Germany, as a continuously operating gravity reference station for time-lapse field surveys with CG-6 gravimeters to assess water storage changes in the course of heavy precipitation events. After 6 months of field operation, all requirements concerning data transmission, remote access, energy consumption, pillar stability and reliable gravity data were continuously met.

The traditional ocean color remote sensing usually focuses on using optical inversion models to estimate the properties of in-water components from the above-surface spectra, so we call it the spectrum-concentration (SC) scheme. Unlike the SC scheme, this study proposed a new research scheme, distribution-distribution (DD) scheme, which uses statistical inference models to estimate the possibility distribution of these in-water components, based on the possibility distribution of the observed spectra. The DD scheme has the advantages that (1) it can rapidly give the key and overview information of the interest water, instead of using the SC scheme to compute each image pixel, (2) it can assist the SC scheme to improve their models and parameters, and (3) it can provide more valuable information for better understanding and indicating the features and dynamics of aquatic environment. In this study, based on Landsat-8 images, we analyzed the spectral possibility distributions (SPD) of 688 global water and found many of them were normal, lognormal, and exponential distributions, but with diverse patterns in distribution parameters such as the mean, standard deviation, skewness and kurtosis. Furthermore, we used Monte-Carlo and Hydrolight simulations to study the theoretical and statistical connections between the possibility distributions of in-water components and SPDs. The simulation results were basically consistent with the observations on the real water. Then by using the simulation and field measured data, we proposed a bootstrap-based DD scheme and developed some simple statistical inference models to estimate the distribution parameters of yellow substance in lakes. Since DD scheme is still on its early stage, we also suggested some potential and useful topics for the future work.

One third of all coastlines worldwide consist of permafrost. Many of these permafrost coasts are presently exposed to greater environmental forcing as a consequence of climate change, such as a lengthening of the open water season, intensified storms, and higher water and air temperatures. As a result, increasing erosion rates are currently reported from various sites across the Arctic. It is crucial to synthetize these data on Arctic shoreline dynamics in order to improve our understanding on present coastal dynamics on the pan-Arctic scale. A first synthesis product was released in form of the Arctic Coastal Dynamics databse in 2012, which included data published until 2009 (Lantuit et al., 2012). Since then, numerous publications and data products were published on short and long term changes of Arctic coasts across a wide range of study sites. We made an extensive literature review of publications released within the last 10 years and updated the shoreline change data section in the Arctic Coastal Dynamics database. While in 2009 for one percent of the Arctic shoreline data on coastal dynamics was available, the addition of new data leads to a broader data coverage, which is mainly the effect of the greater availability of remotely sensed products for analyses conducted in these remote regions. Further, the additional data allow us to update the current mean rate of Arctic shoreline change.

Extreme floods which overwhelm the capacity of a system of flood control dams may result in overtopping one or more of those structures. Traditional US Army Corps of Engineers analysis of hydrologic hazards isolates the study area to a single dam. However, in watersheds where flood hazard is managed by several dams, the estimate for the annual probability of overtopping a dam may be influenced by the operation of one or more other dams in that system. Evaluation and prioritization of modifications for dam safety in a portfolio of structures requires a sound estimate of overtopping probability for every structure. In an effort to properly characterize the hydrologic hazard for five dams in the Trinity River Basin above Dallas, Texas, synthetic weather generation coupled with hydrologic and reservoir models is applied to extend the stage-frequency curve for each dam beyond the observed record. The synthetic weather model is comprised of processes which typify floods most likely to result in overtopping the study dams: 1) continuous, local-scale precipitation and temperature sampling to characterize antecedent hydrologic conditions, 2) intermittent (inhomogenous Poisson), synoptic-scale precipitation sampling based on regional precipitation-frequency analysis to generate hazardous floods, 3) k-nearest-neighbor resampling of precipitation and temperature spatiotemporal patterns and 4) temporal disaggregation of daily precipitation to hourly using correlated Brownian processes. Interrelations between local-scale precipitation, synoptic-scale precipitation and temperature are preserved using a Gaussian copula. Natural variability in annual maximum reservoir stage is described using a stratified sampling scheme used to disproportionately represent extreme floods in a fixed sample of 1,000 events, resulting in fewer model events required to span the probability space from 0.5 to 10-8 annual exceedance probability. Knowledge uncertainty in model components is estimated using a parametric bootstrap, resulting in multiple realizations of synthetic weather. Each weather realization of 1,000 events generated using varying parameters is routed using hydrologic and reservoir models for the system which produce a posterior distribution of annual overtopping probability for each structure.

Enhanced water management systems depend on accurate estimation of hydraulic properties of subsurface formations. This is while hydraulic conductivity of geologic formations could vary significantly. Herein, we studied an intensively managed area located in the Upper Sangamon Watershed in Central Illinois, U.S.A., and generated 2D maps of hydraulic conductivity over a large-scale region with quantified uncertainties in different depth layers. In doing so, we made use of low cost, small-scale measurements obtained from the Electrical Earth Resistivity together with more accurate, more expensive pumping tests in a calibration framework based on Kriging. We offered a cost-effective approach to reliably characterize the hydraulic conductivity properties in under-sampled sites and can be particularly used in obtaining large-scale parameter maps for a region using small-scale measurements in an efficient way. This work also includes optimal sensor placement, where the best locations for future data collection are selected by considering the current confidence levels estimated by the Kriging model, which is related to the expected value of information from future sensor data. Our approach is based on the Bayesian experimental design, which selects the best locations, out of a set of candidate locations, based on the value of information that each location is expected to offer.

Turbulence generated by aquatic vegetation plays a vital role in the interfacial transfer process at the air-water interface and sediment-water interface (AWI and SWI), impacting the dissolved oxygen (DO) level, a key indicator of water quality for aquatic ecosystems. We investigated the influence of vegetation, under different submergence ratios and plant densities, on the interfacial gas transfer mechanisms. We conducted laboratory experiments in a unidirectional recirculating flume with simulated rigid vegetation on a sediment bed. Two-dimensional planar Particle Image Velocimetry (2D-PIV) was used to characterize the mean flow field and turbulent quantities. Gas transfer rates at the AWI were determined by monitoring the DO concentration during the re-aeration process in water. SWI interfacial transfer fluxes were estimated by measuring the DO concentration difference between the near-surface and near-bed values. Compared to previous observations on a smooth bed without sediment, the presence of sediment enhances the bottom roughness, which generates stronger bed-shear turbulence. The experimental result shows that turbulence generated from the bed does not affect the surface transfer process directly. However, the near-bed suspended sediment provides a negative buoyancy term that reduces the transfer efficiency according to the predictions by a modified Surface Renewal model for vegetated flows. The measured interfacial transfer fluxes across the SWI show a clear dependence on the within-canopy flow velocity, indicating that bed shear turbulence and within-canopy turbulence are critical indicators of transfer efficiency at SWI in vegetated flows. A new Reynolds number dependence model using near-bed turbulent kinetic energy as an indicator is proposed to provide a universal prediction for the interfacial flux across the SWI in flows with aquatic vegetation. Our study provides critical insight for future studies on water quality management and ecosystem restoration in natural water environments such as lakes, rivers, and wetlands.

Stream water temperature is considered a “master variable” in environmental processes and human activities. Existing process-based models have difficulties with defining true equation parameters, and sometimes simplifications like assuming constant values influence the accuracy of results. Machine learning models are a highly successful tool for simulating stream temperature, but it is challenging to learn about processes and dynamics from their success. Here we integrate process-based modeling (SNTEMP model) and machine learning by building on a recently developed framework for parameter learning. With this framework, we used a deep neural network to map raw information (like catchment attributes and meteorological forcings) to parameters, and then inspected and fed the results into SNTEMP equations which we implemented in a deep learning platform. We trained the deep neural network across many basins in the conterminous United States in order to maximize the capturing of physical relationships and avoid overfitting. The presented framework has the ability of providing dynamic parameters based on the response of basins to meteorological conditions. The goal of this framework is to minimize the differences between stream temperature observations and SNTEMP outputs in the new platform. Parameter learning allows us to learn model parameters on large scales, providing benefits in efficiency, performance, and generalizability through applying global constraints. This method has also been shown to provide more physically-sensible parameters due to applying a global constraint. This model improves our understanding of how to parameterize the physical processes related to water temperature.

Rain gardens are green stormwater infrastructure that are designed to leverage natural processes to mitigate the impacts of urban stormwater through capturing, infiltrating, and filtering run off. Overtime these systems have the potential to buildup fines and nutrients, impacting their sustainable function. A rain garden’s performance depends on its ability to infiltrate runoff which can be reduced by clogging. Another concern is the potential transport of contaminants from rain gardens to groundwater through deep drainage. This study analyses the spatial and temporal distribution of fines and nutrients in three rain gardens through comprehensive field tests, laboratory testing, and computation analysis. Geomorphic studies were performed by integrating the digital elevation models, derived from Lidar surveys, with the FastMech solver within International River Interface Cooperative (iRIC) software, to model shear stress distribution and sediment transport relative to spatial observations of soil texture and nutrient concentrations within the rain garden. The soil properties were also used in creating models of water infiltration and nutrient sorption using Hydrus 1D. Results show that shear stresses in localized sections of each rain garden can be correlated with fines and nutrient distributions, allowing for prioritizing locations for maintenance. To conclude, LiDAR scans, flow and shear stress models, infiltration and nutrient transport models, field and laboratory soil tests can help us understand the surface dynamics and soil attributes, and gradually gain insight into the GSI performance with time.

Worldwide coastal land-margins are prone to many flood hazards such as astronomical tides, tropical cyclones, sea-level rise, and extreme precipitation events. Compound flood events, in which two or more flooding mechanisms occur simultaneously or in close succession (Santiago-Collazo et al., 2019, https://doi.org/10.1016/j.envsoft. 2019.06.002), can exacerbate the inundation impacts due to the highly non-linear interaction of coastal and hydrologic processes. Furthermore, sea-level rise will increase the hazard at low-gradient coastal land-margins when assessing future projections due to its non-linear nuance on the compound flood (Santiago-Collazo et al., 2021, https://doi.org/10.3389/fclim.2021.684035). Therefore, there is an urgent need to develop new technologies capable of comprehensively studying compound flood events and identifying hotspots prone to these inundations. This research aims to develop a technique capable of defining and classifying coastal land- margins based on physically-based criteria due to surface flow hydrodynamics. A one-dimensional (1-D) hydrodynamic model was used to quantify the hydrodynamic response of thousands of different combinations of input parameters (e.g., astronomical tides, storm surge, precipitation, and landscape) that define a coastal land-margin. This 1-D fully-coupled model, based on the shallow water equations, was applied at a national spatial scale, considering several coastal watersheds within the Gulf of Mexico and the US East coast. One of the main goals of this tool is to identify coastal land-margins vulnerable to compound flood hazards over broad spatial scales (e.g., national or global scale). Findings suggest that low-gradient (e.g., slopes less than 0.01 m km-1) coastal land-margins are more susceptible to compound flood impacts than ones with a steeper gradient under most flooding scenarios. Future research will focus on applying this tool on a worldwide basis to test its capabilities at low-resolution, scarce data regions. A worldwide classification of coastal land-margins may help authorities, policy-makers, and professionals converge on better coastal resilience measures, such as comprehensive compound flood analysis to delineate accurate compound flood hazard maps.Full online poster version at agu2021fallmeeting-agu.ipostersessions.com/Default.aspx?s=FA-1F-20-67-21-4E-E7-69-9F-89-1E-33-BB-3D-2D-40

Some of the Earth system data products such as those from NASA airborne and field investigations (a.k.a. campaigns), are highly heterogeneous and cross-disciplinary, making the data extremely challenging to manage. For example, airborne and field campaign measurements tend to be sporadic over a period of time, with large gaps. Data products generated are of various processing levels and utilized for a wide range of inter- and cross-disciplinary research and applications. Data and derived products have been historically stored in a variety of domain-specific standard (and some non-standard) formats and in various locations such as NASA Distributed Active Archive Centers (DAACs), NASA airborne science facilities, field archives, or even individual scientists’ computer hard drives. As a result, airborne and field campaign data products have often been managed and represented differently, making it onerous for data users to find, access, and utilize campaign data. Some difficulties in discovering and accessing the campaign data originate from the incomplete data product and contextual metadata that may contain details relevant to the campaign (e.g. campaign acronym and instrument deployment locations), but tend to lack other significant information needed to understand conditions surrounding the data. Such details can be burdensome to locate after the conclusion of a campaign. Utilizing consistent terminology, essential for improved discovery and reuse, is also challenging due to the variety of involved disciplines. To help address the aforementioned challenges faced by many repositories and data managers handling airborne and field data, this presentation will describe stewardship practices developed by the Airborne Data Management Group (ADMG) within the Interagency Implementation and Advanced Concepts Team (IMPACT) under the NASA’s Earth Science Data systems (ESDS) Program.

Floods are the most frequent, costliest natural disasters having devastating consequences on people, infrastructure, and the ecosystem. The accurate and rapid mapping of the flooded areas becomes more crucial when floods strike densely populated cities. During flood events near real-time satellite imagery has been proven to be an efficient management tool for disaster management authorities. However one of the challenges is accurate classification and segmentation of flooded water and permanent water. Binary segmentation using the threshold split-based method is commonly used in this regard, however, the generalization ability of this method is limited due to the effects of backscatter, geographical area, and time of image collection. Recent advancements in deep learning algorithms for image segmentation has demonstrated the excellent potential of Convolutional Neural Networks(CNN) for improving flood detection, although there have been limited studies in this domain due to the lack of large scale labeled flood event dataset. In this project, we present a U-net based deep learning approach by leveraging publicly available Sentinel-1 dataset provided jointly by NASA Interagency Implementation and Advanced Concepts Team and IEEE GRSS Earth Science Informatics Technical Committee. Dataset is composed of 66,810 tiles of 256×256 pixels, distributed respectively across the training, validation and test sets and cover flood events from Nebraska, North Alabama, Bangladesh, Red River North and Florence. Specifically we proposed an Unet architecture based convolutional neural network (CNN) with a backbone of EfficientNetb7, trained against the dataset. We then evaluated the performance of the model with multiple training, testing and validation. Two evaluation methods - Intersection over Union (IOU) and F-Score are adopted to evaluate the model performance. During testing, the model achieved the meanIOU score of 75.06% and F-Score of 74.98%. We hope to further improve the performance of the network by performing hyper-parameter tuning and to develop a model which can be used for near-real-time flood mapping.

Dissolved oxygen (DO) sustains aquatic life and is an essential water quality measure. Our capabilities of forecasting DO levels, however, remain elusive. Unlike the increasingly intensive earth surface and hydroclimatic data, water quality data often have large temporal gaps and sparse areal coverage. Here we ask the question: can a Long Short-Term Memory (LSTM) deep learning model learn the spatio-temporal dynamics of stream DO from intensive hydroclimatic and sparse DO observations at the continental scale? That is, can the model harvest the power of big hydroclimatic data and use them for water quality forecasting? Here we used data from CAMELS-chem, a new dataset that includes sparse DO concentrations from 236 minimally-disturbed watersheds. The trained model can generally learn the theory of DO solubility under specific temperature, pressure, and salinity conditions. It captures the bulk variability and seasonality of DO and exhibits the potential of forecasting water quality in ungauged basins without training data. It however often misses concentration peaks and troughs where DO level depends on complex biogeochemical processes. The model surprisingly does not perform better where data are more intensive. It performs better in basins with low streamflow variations, low DO variability, high runoff-ratio (> 0.45), and precipitation peaks in winter. This work suggests that more frequent data collection in anticipated DO peak and trough conditions are essential to help overcome the issue of sparse data, an outstanding challenge in the water quality community.

The North American Monsoon occurs July-September bringing significant rainfall to Colorado River headwater basins. This rain may buffer streamflow deficiencies caused by reductions in snow accumulation. Using a data-modeling framework, we explore the importance of monsoon rain in streamflow generation over historic conditions in an alpine basin. Annually, monsoon rain contributes 18{plus minus}7% water inputs, generates 10{plus minus}6% streamflow and increases water yield 3{plus minus}2% the following year. The bulk of rain supports evapotranspiration in lower subalpine forests. However, rains have the potential to produce appreciable streamflow at higher elevations where soil storage, forest cover and aridity are low; and rebounds late season streamflow 64{plus minus}13% from simulated reductions in snowpack as a function of monsoon strength. Interannual variability in monsoon efficiency to generate streamflow declines with low snowpack and high aridity, implying the ability of monsoons to replenish streamflow in a warmer future with less snow accumulation will diminish.