Search NASA⌕ Search

Engineering topics

Martin, Thomas

Publications and source records attributed to Martin, Thomas.

Evidential Deep Learning: Enhancing Predictive Uncertainty Estimation for Earth System Science Applications

Abstract Robust quantification of predictive uncertainty is a critical addition needed for machine learning applied to weather and climate problems to improve the understanding of what is driving prediction sensitivity. Ensembles of machine learning models provide predictive uncertainty estimates in a conceptually simple way but require multiple models for training and prediction, increasing computational cost and latency. Parametric deep learning can estimate uncertainty with one model by predicting the parameters of a probability distribution but does not account for epistemic uncertainty. Evidential deep learning, a technique that extends parametric deep learning to higher-order distributions, can account for both aleatoric and epistemic uncertainties with one model. This study compares the uncertainty derived from evidential neural networks to that obtained from ensembles. Through applications of the classification of winter precipitation type and regression of surface-layer fluxes, we show evidential deep learning models attaining predictive accuracy rivaling standard methods while robustly quantifying both sources of uncertainty. We evaluate the uncertainty in terms of how well the predictions are calibrated and how well the uncertainty correlates with prediction error. Analyses of uncertainty in the context of the inputs reveal sensitivities to underlying meteorological processes, facilitating interpretation of the models. The conceptual simplicity, interpretability, and computational efficiency of evidential neural networks make them highly extensible, offering a promising approach for reliable and practical uncertainty quantification in Earth system science modeling. To encourage broader adoption of evidential deep learning, we have developed a new Python package, Machine Integration and Learning for Earth Systems (MILES) group Generalized Uncertainty for Earth System Science (GUESS) (MILES-GUESS) ( https://github.com/ai2es/miles-guess ), that enables users to train and evaluate both evidential and ensemble deep learning. Significance Statement This study demonstrates a new technique, evidential deep learning, for robust and computationally efficient uncertainty quantification in modeling the Earth system. The method integrates probabilistic principles into deep neural networks, enabling the estimation of both aleatoric uncertainty from noisy data and epistemic uncertainty from model limitations using a single model. Our analyses reveal how decomposing these uncertainties provides valuable insights into reliability, accuracy, and model shortcomings. We show that the approach can rival standard methods in classification and regression tasks within atmospheric science while offering practical advantages such as computational efficiency. With further advances, evidential networks have the potential to enhance risk assessment and decision-making across meteorology by improving uncertainty quantification, a longstanding challenge. This work establishes a strong foundation and motivation for the broader adoption of evidential learning, where properly quantifying uncertainties is critical yet lacking.

Schreck, John S.↗

Increasing the Reproducibility and Replicability of Supervised AI/ML in the Earth Systems Science by Leveraging Social Science Methods

Artificial intelligence (AI) and machine learning (ML) pose a challenge for achieving science that is both reproducible and replicable. The challenge is compounded in supervised models that depend on manually labeled training data, as they introduce additional decision-making and processes that require thorough documentation and reporting. We address these limitations by providing an approach to hand labeling training data for supervised ML that integrates quantitative content analysis (QCA)—a method from social science research. The QCA approach provides a rigorous and well-documented hand labeling procedure to improve the replicability and reproducibility of supervised ML applications in Earth systems science (ESS), as well as the ability to evaluate them. Specifically, the approach requires (a) the articulation and documentation of the exact decision-making process used for assigning hand labels in a “codebook” and (b) an empirical evaluation of the reliability” of the hand labelers. In this paper, we outline the contributions of QCA to the field, along with an overview of the general approach. We then provide a case study to further demonstrate how this framework has and can be applied when developing supervised ML models for applications in ESS. With this approach, we provide an actionable path forward for addressing ethical considerations and goals outlined by recent AGU work on ML ethics in ESS.

58 GEOSCIENCES↗

Play Fairway analysis of geothermal resources across the State of Hawai‘i: 4. Updates with new groundwater chemistry, subsurface stress analysis, and focused geophysical surveys

This paper is the fourth in a series on a play fairway analysis of geothermal resources across the State of Hawai‘i. Here we describe recent exploration activities that include groundwater sampling in ten locations statewide, as well as geophysical surveys on Lana‘i, across the SW rift of Haleakala Volcano (Maui), and surrounding Mauna Kea Volcano (Hawai‘i Island). We derive a first-order method for computing topographic stress using Green’s functions, finding that topography can induce appreciable crustal stress. We develop a new method for incorporating depth information about resistivity, density, and topographic stresses into our previously published equations to calculate resource probability and confidence. We incorporate newly collected and modeled data into our calculations to update statewide maps of probability and confidence. Lana‘i Island and southeast Mauna Kea are identified as the top targets for exploratory drilling. The east rift of Haleakala Volcano, the southern-most region of Mauna Loa’s SW rift (Hawai‘i Island), and central Kauai are identified as targets for geophysical surveying.

15 GEOTHERMAL ENERGY↗

Conference Proceedings

History and current progress in improving the fuel efficiency of ground based turbine engines for electrical power generation, with details on the manner in which a vane constructed from both monolithic ceramics, and fiber reinforced ceramic matrix composites, can save parasitic cooling air over metals currently employed. An initial design was used to determine the effects on gas turbine simple cycle, with a one point efficiency gain and two percent fuel consumption reduction.

20 FOSSIL-FUELED POWER PLANTS↗

High-Dose-Rate Interstitial Brachytherapy as Monotherapy for Clinically Localized Prostate Cancer: Treatment Evolution and Mature Results

Purpose: To report the clinical outcome of high-dose-rate (HDR) interstitial (IRT) brachytherapy (BRT) as sole treatment (monotherapy) for clinically localized prostate cancer. Methods and Materials: Between January 2002 and December 2009, 718 consecutive patients with clinically localized prostate cancer were treated with transrectal ultrasound (TRUS)-guided HDR monotherapy. Three treatment protocols were applied; 141 patients received 38.0 Gy using one implant in 4 fractions of 9.5 Gy with computed tomography-based treatment planning; 351 patients received 38.0 Gy in 4 fractions of 9.5 Gy, using 2 implants (2 weeks apart) and intraoperative TRUS real-time treatment planning; and 226 patients received 34.5 Gy, using 3 single-fraction implants of 11.5 Gy (3 weeks apart) and intraoperative TRUS real-time treatment planning. Biochemical failure was defined according to the Phoenix consensus, and toxicity was evaluated using Common Toxicity Criteria for Adverse Events version 3. Results: The median follow-up time was 52.8 months. The 36-, 60-, and 96-month biochemical control and metastasis-free survival rates for the entire cohort were 97%, 94%, and 90% and 99%, 98%, and 97%, respectively. Toxicity was scored per event, with 5.4% acute grade 3 genitourinary and 0.2% acute grade 3 gastrointestinal toxicity. Late grade 3 genitourinary and gastrointestinal toxicities were 3.5% and 1.6%, respectively. Two patients developed grade 4 incontinence. No other instance of grade 4 or greater acute or late toxicity was reported. Conclusion: Our results confirm IRT-HDR-BRT is safe and effective as monotherapy for clinically localized prostate cancer.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

GPS-Based Navigation and Orbit Determination for the AMSAT Phase 3D Satellite

This paper summarizes the results of processing GPS data from the AMSAT Phase 3D (AP3) satellite for real-time navigation and post-processed orbit determination experiments. AP3 was launched into a geostationary transfer orbit (GTO) on November 16, 2000 from Kourou, French Guiana, and then was maneuvered into its HEO over the next several months. It carries two Trimble TANS Vector GPS receivers for signal reception at apogee and at perigee. Its spin stabilization mode currently makes it favorable to track GPS satellites from the backside of the constellation while at perigee, and to track GPS satellites from below while at perigee. To date, the experiment has demonstrated that it is feasible to use GPS for navigation and orbit determination in HEO, which will be of great benefit to planned and proposed missions that will utilize such orbits for science observations. It has also shown that there are many important operational considerations to take into account. For example, GPS signals can be tracked above the constellation at altitudes as high as 58000 km, but sufficient amplification of those weak signals is needed. Moreover, GPS receivers can track up to 4 GPS satellites at perigee while moving as fast as 9.8 km/sec, but unless the receiver can maintain lock on the signals long enough, point solutions will be difficult to generate. The spin stabilization of AP3, for example, appears to cause signal levels to fluctuate as other antennas on the satellite block the signals. As a result, its TANS Vectors have been unable to lock on to the GPS signals long enough to down load the broadcast ephemeris and then generate position and velocity solutions. AP3 is currently in its eclipse season, and thus most of the spacecraft subsystems have been powered off. In Spring 2002, they will again be powered up and AP3 will be placed into a three-axis stabilization mode. This will significantly enhance the likelihood that point solutions can be generated, and perhaps more important, that the receiver clock can be synchronized to GPS time. This is extremely important for real-time and post-processed orbit determination, where removal of receiver clock bias from the data time tags is needed, for time-tagging of science observations. Current analysis suggests that the inability to generate point solutions has allowed the TANS Vector clock bias to drift freely, being perhaps as large as 5-7 seconds by October, 2001, thus causing up to 50 km of along-track orbit error. The data collected in May, 2002 while in three-axis stabilized mode should provide a significant improvement in the orbit determination results.

Davis, George↗