Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher.
Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?
Some links on this page may take you to nonfederal websites. Their policies may differ from this site.

Free, publiclyaccessible full text available January 1, 2024

Accurate predictions of water temperature are the foundation for many decisions and regulations, with direct impacts on water quality, fishery yields, and power production. Building accurate broadscale models for lake temperature prediction remains challenging in practice due to the variability in the data distribution across different lake systems monitored by static and timeseries data. In this paper, to tackle the above challenges, we propose a novel machine learning based approach for integrating static and timeseries data in deep recurrent models, which we call InvertibilityAwareLong ShortTerm Memory(IALSTM), and demonstrate its effectiveness in predicting lake temperature. Our proposed method integrates components of the Invertible Network and LSTM to better predict temperature profiles (forward modeling) and infer the static features (i.e., inverse modeling) that can eventually enhance the prediction when static variables are missing. We evaluate our method on predicting the temperature profile of 450 lakes in the Midwestern U.S. and report a relative improvement of 4\% to capture data heterogeneity and simultaneously outperform baseline predictions by 12\% when static features are unavailable.more » « less

Abstract Most Upper Jurassic studies of astronomical forcing have focused on deeper‐water sections which are relatively continuous. An Upper Jurassic (Kimmeridgian) section on the greenhouse Adriatic Carbonate Platform, Croatia, was studied to determine if astronomical forcing can be recognized in a 5.8 ± 0.1 Myr duration, disconformity‐prone shallow platform succession. The succession consists of metre‐scale subtidal parasequences intermixed with peritidal parasequences, and intermittent subaerial breccias at sequence boundaries. Ages were constrained by biostratigraphy and δ^{13}C chemostratigraphy, and most sequence boundaries appear to match those of the coastal onlap curve of Haq (2018). Logged sections were converted into depth–rank time series and parasequence–thickness time series. Accumulation rates were statistically evaluated for the rank series against an astronomical‐forcing model, and compared with long‐term accumulation rates (thickness divided by time). The statistical rates were used to select the
ca 100 kyr eccentricity cycle to tune the series. Spectral analysis showed peaks atca 400 kyr (superbundles) andca 100 kyr (bundles), along with obliquity (38 kyr and 27 kyr) and precessional (18−22 kyr) cycles (parasequences). The Kimmeridgian sequences areca 400 kyr,ca 800 kyr andca 1.1 Myr duration. Sequence scale (0.4 to 1.2 Myr) stratigraphic completeness based on statistical accumulation rates versus long‐term rates isca 60%. This study estimatesca 1 Myr missing time in parasequences stacked into superbundles and 1.6 Myr in four major sequence boundaries. Given that the Kimmeridgian was the hottest time of the Middle and Late Jurassic, aquifer eustasy may have influenced the timing of sequence boundaries, although documented late Kimmeridgian cooling could have triggered a glacio‐eustatic component. 
Demeniconi, Carlotta ; Davidson, Ian (Ed.)This paper proposes a physicsguided machine learning approach that combines machine learning models and physicsbased models to improve the prediction of water flow and temperature in river networks. We first build a recurrent graph network model to capture the interactions among multiple segments in the river network. Then we transfer knowledge from physicsbased models to guide the learning of the machine learning model. We also propose a new loss function that balances the performance over different river segments. We demonstrate the effectiveness of the proposed method in predicting temperature and streamflow in a subset of the Delaware River Basin. In particular, the proposed method has brought a 33%/14% accuracy improvement over the stateoftheart physicsbased model and 24%/14% over traditional machine learning models (e.g., LSTM) in temperature/streamflow prediction using very sparse (0.1%) training data. The proposed method has also been shown to produce better performance when generalized to different seasons or river segments with different streamflow ranges.more » « less

Abstract The detection of GW170817, the first neutron starneutron star merger observed by Advanced LIGO and Virgo, and its following analyses represent the first contributions of gravitational wave data to understanding dense matter. Parameterizing the high density section of the equation of state of both neutron stars through spectral decomposition, and imposing a lower limit on the maximum mass value, led to an estimate of the stars’ radii of
km and km (Abbottet al 2018Phys. Rev. Lett .121 161101). These values do not, however, take into account any uncertainty owed to the choice of the crust lowdensity equation of state, which was fixed to reproduce the SLy equation of state model (Douchin and Haensel 2001Astron. Astrophys .380 151). We here reanalyze GW170817 data and establish that different crust models do not strongly impact the mass or tidal deformability of a neutron star—it is impossible to distinguish between lowdensity models with gravitational wave analysis. However, the crust does have an effect on inferred radius. We predict the systematic error due to this effect using neutron star structure equations, and compare the prediction to results from full parameter estimation runs. For GW170817, this systematic error affects the radius estimate by 0.3 km, approximately of the neutron stars’ radii. 
This paper proposes a physicsguided recurrent neural network model (PGRNN) that combines RNNs and physicsbased models to leverage their complementary strengths and improve the modeling of physical processes. Specifically, we show that a PGRNN can improve prediction accuracy over that of physical models, while generating outputs consistent with physical laws, and achieving good generalizability. Standard RNNs, even when producing superior prediction accuracy, often produce physically inconsistent results and lack generalizability. We further enhance this approach by using a pretraining method that leverages the simulated data from a physicsbased model to address the scarcity of observed data. Although we present and evaluate this methodology in the context of modeling the dynamics of temperature in lakes, it is applicable more widely to a range of scientific and engineering disciplines where mechanistic (also known as processbased) models are used, e.g., power engineering, climate science, materials science, computational chemistry, and biomedicine.more » « less

We construct a notion of teleparallelization for Newton–Cartan theory, and show that the teleparallel equivalent of this theory is Newtonian gravity; furthermore, we show that this result is consistent with teleparallelization in general relativity, and can be obtained by nullreducing the teleparallel equivalent of a fivedimensional gravitational wave solution. This work thus strengthens substantially the connections between four theories: Newton–Cartan theory, Newtonian gravitation theory, general relativity, and teleparallel gravity.more » « less