Robust Quantification Research Articles

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 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 uncertainty with one model. This study compares the uncertainty derived from evidential neural networks to that obtained from ensembles. Through applications of 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. In order to encourage broader adoption of evidential deep learning, we have developed a new Python package, MILES-GUESS (https://github.com/ai2es/miles-guess), that enables users to train and evaluate both evidential and ensemble deep learning.

BS-clock, advancing epigenetic age prediction with high-resolution DNA methylation bisulfite sequencing data.

DNA methylation patterns provide precise and accurate estimates of biological age due to their robustness and predictable changes associated with aging processes. Although several methylation aging clocks have been developed in recent years, they are primarily designed for DNA methylation array data, which has limited CpG coverage and detection sensitivity compared to bisulfite sequencing data. Here, we present BS-clock, a novel DNA methylation clock for human aging based on bisulfite sequencing data. Using BS-seq data from 529 samples retrieved from four tissues, our BS-clock achieves higher correlations with chronological age in multiple tissue types compared to existing array-based clocks. Our study revealed age-dependent aging rates across different age stages and disease conditions, and overall low cross-tissue prediction capability by applying the model trained on one tissue type to others. In summary, BS-clock overcomes limitations of array-based techniques, offering genome-wide CpG site coverage and more robust and accurate aging quantification. This research paves the way for advanced epigenetic studies of aging and holds promise for developing targeted interventions to promote healthy aging. All analysis codes for reproducing the results of the study are publicly available at https://github.com/hucongcong97/BS-clock. Supplementary data are available at Bioinformatics online.

Variational inference for acceleration of SN Ia photometric distance estimation with BayeSN

Abstract Type Ia supernovae (SNe Ia) are standarizable candles whose observed light curves can be used to infer their distances, which can in turn be used in cosmological analyses. As the quantity of observed SNe Ia grows with current and upcoming surveys, increasingly scalable analyses are necessary to take full advantage of these new datasets for precise estimation of cosmological parameters. Bayesian inference methods enable fitting SN Ia light curves with robust uncertainty quantification, but traditional posterior sampling using Markov Chain Monte Carlo (MCMC) is computationally expensive. We present an implementation of variational inference (VI) to accelerate the fitting of SN Ia light curves using the BayeSN hierarchical Bayesian model for time-varying SN Ia spectral energy distributions (SEDs). We demonstrate and evaluate its performance on both simulated light curves and data from the Foundation Supernova Survey with two different forms of surrogate posterior – a multivariate normal and a custom multivariate zero-lower-truncated normal distribution – and compare them with the Laplace Approximation and full MCMC analysis. To validate of our variational approximation, we calculate the pareto-smoothed importance sampling (PSIS) diagnostic, and perform variational simulation-based calibration (VSBC). The VI approximation achieves similar results to MCMC but with an order-of-magnitude speedup for the inference of the photometric distance moduli. Overall, we show that VI is a promising method for scalable parameter inference that enables analysis of larger datasets for precision cosmology.

Functional quantile principal component analysis.

This paper introduces functional quantile principal component analysis (FQPCA), a dimensionality reduction technique that extends the concept of functional principal components analysis (FPCA) to the examination of participant-specific quantiles curves. Our approach borrows strength across participants to estimate patterns in quantiles, and uses participant-level data to estimate loadings on those patterns. As a result, FQPCA is able to capture shifts in the scale and distribution of data that affect participant-level quantile curves, and is also a robust methodology suitable for dealing with outliers, heteroscedastic data or skewed data. The need for such methodology is exemplified by physical activity data collected using wearable devices. Participants often differ in the timing and intensity of physical activity behaviors, and capturing information beyond the participant-level expected value curves produced by FPCA is necessary for a robust quantification of diurnal patterns of activity. We illustrate our methods using accelerometer data from the National Health and Nutrition Examination Survey, and produce participant-level 10%, 50%, and 90% quantile curves over 24 h of activity. The proposed methodology is supported by simulation results, and is available as an R package.

Quantifying the impact of COVID-19 restrictions on air pollution in Ahvaz: a comparative dual-approach assessment of observed against baseline and forecasted criteria air pollutants.

In response to the COVID-19 pandemic, the Iranian government swiftly implemented immediate and decisive measures to control the spread of the infection. This study aims to demonstrate the impact of restriction measure on air pollution, also to highlight the potential variability in results that can arias from different methodological approach. A comprehensive dual-approach assessment was conducted to evaluate the effect of the lockdown measures on criteria air pollutants. Firstly, a traditional approach compared air quality during the pandemic period with baseline conditions from 2013 to 2019. Secondly, observed air pollution values during different periods with varying restrictions in 2020 were compared with expected values. This comprehensive analysis allows for a robust comparison and quantification of the impact of different lockdown measures in Ahvaz. The study revealed significant changes in air pollutant concentrations in Ahvaz during 2020, with variations observed across different pollutants. Notable reductions were observed in O3 levels, particularly in November (-54.44% compared to the baseline) and December (-63.58% compared to expected values). Decreases in CO levels were observed in multiple months, while substantial reductions in PM10 and PM2.5 were observed during various periods. Inconsistencies in the magnitudes and directions of changes were found when comparing baseline and forecasted values. The overall stringency index showed an inverse association with changes in O3, NO2, and CO, with international travel controls and restrictions on internal movement having significant impacts. This study provides valuable insights into the impact of COVID-19 lockdown measures on air pollution in Ahvaz, Iran, using a comprehensive dual-approach assessment. The findings highlight the effectiveness of these measures in reducing specific criteria air pollutants and emphasize the importance of implementing appropriate strategies for air quality management during similar public health emergencies.

Deep Spectral Library of Mice Retina for Myopia Research: Proteomics Dataset generated by SWATH and DIA-NN

The retina plays a crucial role in processing and decoding visual information, both in normal development and during myopia progression. Recent advancements have introduced a library-independent approach for data-independent acquisition (DIA) analyses. This study demonstrates deep proteome identification and quantification in individual mice retinas during myopia development, with an average of 6,263 ± 86 unique protein groups. We anticipate that the use of a predicted retinal-specific spectral library combined with the robust quantification achieved within this dataset will contribute to a better understanding of the proteome complexity. Furthermore, a comprehensive mice retinal-specific spectral library was generated, encompassing a total identification of 9,401 protein groups, 70,041 peptides, 95,339 precursors, and 761,868 transitions acquired using SWATH-MS acquisition on a ZenoTOF 7600 mass spectrometer. This dataset surpasses the spectral library generated through high-pH reversed-phase fractionation by data-dependent acquisition (DDA). The data is available via ProteomeXchange with the identifier PXD046983. It will also serve as an indispensable reference for investigations in myopia research and other retinal or neurological diseases.

Deep Conformal Supervision: Leveraging Intermediate Features for Robust Uncertainty Quantification.

Trustworthiness is crucial for artificial intelligence (AI) models in clinical settings, and a fundamental aspect of trustworthy AI is uncertainty quantification (UQ). Conformal prediction as a robust uncertainty quantification (UQ) framework has been receiving increasing attention as a valuable tool in improving model trustworthiness. An area of active research is the method of non-conformity score calculation for conformal prediction. We propose deep conformal supervision (DCS), which leverages the intermediate outputs of deep supervision for non-conformity score calculation, via weighted averaging based on the inverse of mean calibration error for each stage. We benchmarked our method on two publicly available datasets focused on medical image classification: a pneumonia chest radiography dataset and a preprocessed version of the 2019 RSNA Intracranial Hemorrhage dataset. Our method achieved mean coverage errors of 16e-4 (CI: 1e-4, 41e-4) and 5e-4 (CI: 1e-4, 10e-4) compared to baseline mean coverage errors of 28e-4 (CI: 2e-4, 64e-4) and 21e-4 (CI: 8e-4, 3e-4) on the two datasets, respectively (p < 0.001 on both datasets). Based on our findings, the baseline results of conformal prediction already exhibit small coverage errors. However, our method shows a significant improvement on coverage error, particularly noticeable in scenarios involving smaller datasets or when considering smaller acceptable error levels, which are crucial in developing UQ frameworks for healthcare AI applications.

B-146 Unlocking the potential of large-cohort proteomics studies with the Orbitrap Astral mass platform

Abstract Background Large-cohort proteomics analysis using mass spectrometry is a powerful approach to discover and validate new biomarkers. In combination with clinical data and computational analysis, large-cohort proteomics study brings opportunities in improving early diagnosis, refining patient stratification, and predicting/monitoring treatment response. Yet, to achieve meaningful biological insights in large-cohort studies, robust, reproducible, and comprehensive proteome profiling in a high-throughput manner still remains challenging. Here, we use a high-resolution accurate mass (HRAM) Oribtrap Astral platform to enable high-quality and robust protein quantification across thousands of LC-MS/MS analyses. With a throughput of 100 samples/day, we can reproducibly profile ∼9000 proteins from human cell line and ∼800 proteins from undepleted plasma across multiple instruments and more than 10 consecutive days in a 24/7 operation mode. Methods Experiments were performed on multiple Orbitrap Astral mass spectrometers with and without FAIMS Pro interface. Chromatographic separations were performed using a trap/elute method on Vanquish Neo system at a 11-minute gradient and a flow rate of 1ul/min, resulting in a throughput of 100 samples/day. Undepleted plasma digest was analyzed in a 24/7 mode for &amp;gt; 1000 injections on each LC-MS/MS setup. In addition, HeLa digest as quality control (QC) was analyzed every 12hours. The Oribtrap Astral platform was operated in data-independent acquisition (DIA) mode using a full scan with m/z 380-980, and DIA MS/MS scans with an isolation window of 2Th. Resulting DIA raw files were automatically transferred and processed using Chimerys in a beta version of Proteome Discoverer software. Results Multiple LC-MS/MS systems were operated in DIA mode either with or without FAIMS Pro device in a 24/7 operation mode at a throughput of 100SPD. Undepleted plasma digest was analyzed with &amp;gt;1000 injections on each LC-MS/MS setup. To monitor the baseline performance, HeLa digest served as QC were analyzed periodically every 12 hours with 3 technical replicates each time. To effectively manage and analyze these thousands of data files generated, we developed an automated data transfer and analysis pipeline. Automatically, the resulting DIA raw data files were immediately transferred to a server, then processed by state-of-the-art intelligent search algorithm Chimerys. Benefiting from the ultra-high scan speed of up to 200Hz on the Orbitrap Astral mass spectrometer, a much narrower isolation window width of 2Th was applied in the DIA method, comparing to 10-20Th on the classic DIA method. As a result, ∼9000 proteins from HeLa digest and ∼800 proteins from undepleted plasma digest were identified within only a 11-minutes gradient, respectively. More than 80% of the plasma proteins were reproducibly identified and quantified from all the runs on each LC-MS/MS setup, indicating a great reproducibility from run-to-run longitudinally. Stable and robust peptide quantitation was observed by extracting peptides with high, medium, and low abundant across the runs. Importantly, QC showed no performance degradation throughout the entire study, indicating high robustness of the entire LC-MS/MS setup. Conclusions Orbitrap Astral mass platform can comprehensively analyze the proteome of &amp;gt;1000s of sample robustly and reproducibly in a high-throughput manner, addressing the needs in large-cohort studies.

A reassessment of spinal cord pathology in severe infantile spinal muscular atrophy: Reassessment of spinal cord pathology.

Spinal muscular atrophy (SMA) is a life-limiting paediatric motor neuron disease characterised by lower motor neuron loss, skeletal muscle atrophy and respiratory failure, if untreated. Revolutionary treatments now extend patient survival. However, a limited understanding of the foundational neuropathology challenges the evaluation of therapeutic success. As opportunities to study treatment-naïve tissue decrease, we have characterised spinal cord pathology in severe infantile SMA using gold-standard techniques, providing a baseline to measure treatment success and therapeutic limitations. Detailed histological analysis, stereology and transmission electron microscopy were applied to post-mortem spinal cord from severe infantile SMA patients to estimate neuron number at the end of life; characterise the morphology of ventral horn, lateral horn and Clarke's column neuron populations; assess cross-sectional spinal cord area; and observe myelinated white matter tracts in the clinically relevant thoracic spinal cord. Ventral horn neuron loss was substantial in all patients, even the youngest cases. The remaining ventral horn neurons were small with abnormal, occasionally chromatolytic morphology, indicating cellular damage. In addition to ventral horn pathology, Clarke's column sensory-associated neurons displayed morphological features of cellular injury, in contrast to the preserved sympathetic lateral horn neurons. Cellular changes were associated with aberrant development of grey and white matter structures that affected the overall dimensions of the spinal cord. We provide robust quantification of the neuronal deficit found at the end of life in SMA spinal cord. We question long-accepted dogmas of SMA pathogenesis and shed new light on SMA neuropathology out with the ventral horn, which must be considered in future therapeutic design.

492P Mass spectrometry as a technique for robust quantification of titin and other large muscle disease-associated proteins

492P Mass spectrometry as a technique for robust quantification of titin and other large muscle disease-associated proteins

Machine Learning-Driven RUL Prediction and Uncertainty Quantification for Ball Screw Drives in a Cloud-Ready Maintenance Framework

In today's rapidly evolving industrial landscape, efficient predictive maintenance solutions are essential for minimizing downtime and enhancing productivity. This research introduces an adaptive cloud-based model pipeline for predicting the Remaining Useful Life (RUL) of machine components, specifically ball screws. The pipeline integrates local preprocessing, edge computing, and cloud-based adaptive model training, ensuring data privacy and reducing data transmission volumes. The system classifies wear states using various machine learning mod-els and predicts RUL through regression analysis, incorporating uncertainty quantification for robust maintenance scheduling. The experimental setup includes accelerated degradation of ball screws, with data collected via a three-dimensional accelerometer. Feature extraction and data augmentation techniques are employed to enhance prediction accuracy. Random Forest and Gradient Boosting models demonstrate superior performance, with Random Forest selected for its robustness and uncertainty quantification capabilities. Empirical results indicate high prediction accuracy, with Random Forest achieving up to 91% accuracy in Phase 2. This cloud-ready predictive maintenance framework leverages scalable cloud infrastructure for efficient data processing and real-time updates, offering a practical solution for industrial applications. The proposed approach significantly advances the adoption of digital business models within the manufacturing industry, providing a reliable and efficient tool for predictive maintenance.

Scalable spatiotemporal prediction with Bayesian neural fields

Spatiotemporal datasets, which consist of spatially-referenced time series, are ubiquitous in diverse applications, such as air pollution monitoring, disease tracking, and cloud-demand forecasting. As the scale of modern datasets increases, there is a growing need for statistical methods that are flexible enough to capture complex spatiotemporal dynamics and scalable enough to handle many observations. This article introduces the Bayesian Neural Field (BayesNF), a domain-general statistical model that infers rich spatiotemporal probability distributions for data-analysis tasks including forecasting, interpolation, and variography. BayesNF integrates a deep neural network architecture for high-capacity function estimation with hierarchical Bayesian inference for robust predictive uncertainty quantification. Evaluations against prominent baselines show that BayesNF delivers improvements on prediction problems from climate and public health data containing tens to hundreds of thousands of measurements. Accompanying the paper is an open-source software package (https://github.com/google/bayesnf) that runs on GPU and TPU accelerators through the Jax machine learning platform.

Combining Data Independent Acquisition With Spike-In SILAC (DIA-SiS) Improves Proteome Coverage and Quantification

Data-independent acquisition (DIA) is increasingly preferred over data-dependent acquisition due to its higher throughput and fewer missing values. Whereas data-dependent acquisition often uses stable isotope labeling to improve quantification, DIA mostly relies on label-free approaches. Efforts to integrate DIA with isotope labeling include chemical methods like mass differential tags for relative and absolute quantification and dimethyl labeling, which, while effective, complicate sample preparation. Stable isotope labeling by amino acids in cell culture (SILAC) achieves high labeling efficiency through the metabolic incorporation of heavy labels into proteins invivo. However, the need for metabolic incorporation limits the direct use in clinical scenarios and certain high-throughput experiments. Spike-in SILAC (SiS) methods use an externally generated heavy sample as an internal reference, enabling SILAC-based quantification even for samples that cannot be directly labeled. Here, we combine DIA-SiS, leveraging the robust quantification of SILAC without the complexities associated with chemical labeling. We developed DIA-SiS and rigorously assessed its performance with mixed-species benchmark samples on bulk and single cell-like amount level. We demonstrate that DIA-SiS substantially improves proteome coverage and quantification compared to label-free approaches and reduces incorrectly quantified proteins. Additionally, DIA-SiS proves effective in analyzing proteins in low-input formalin-fixed paraffin-embedded tissue sections. DIA-SiS combines the precision of stable isotope-based quantification with the simplicity of label-free sample preparation, facilitating simple, accurate, and comprehensive proteome profiling.

Proteome-Scale Tissue Mapping Using Mass Spectrometry Based on Label-Free and Multiplexed Workflows

Multiplexed bimolecular profiling of tissue microenvironment, or spatial omics, can provide deep insight into cellular compositions and interactions in healthy and diseased tissues. Proteome-scale tissue mapping, which aims to unbiasedly visualize all the proteins in a whole tissue section or region of interest, has attracted significant interest because it holds great potential to directly reveal diagnostic biomarkers and therapeutic targets. While many approaches are available, however, proteome mapping still exhibits significant technical challenges in both protein coverage and analytical throughput. Since many of these existing challenges are associated with mass spectrometry-based protein identification and quantification, we performed a detailed benchmarking study of three protein quantification methods for spatial proteome mapping, including label-free, TMT-MS2, and TMT-MS3. Our study indicates label-free method provided the deepest coverages of ∼3500 proteins at a spatial resolution of 50 μm and the highest quantification dynamic range, while TMT-MS2 method holds great benefit in mapping throughput at >125 pixels per day. The evaluation also indicates both label-free and TMT-MS2 provide robust protein quantifications in identifying differentially abundant proteins and spatially co-variable clusters. In the study of pancreatic islet microenvironment, we demonstrated deep proteome mapping not only enables the identification of protein markers specific to different cell types, but more importantly, it also reveals unknown or hidden protein patterns by spatial co-expression analysis.

A Benchmark Study of DFT-Computed p-Block Element Lewis Pair Formation Enthalpies against Experimental Calorimetric Data.

The quantification of Lewis acidity is of fundamental and applied importance in chemistry. While the computed fluoride ion affinity (FIA) is the most widely accepted thermodynamic metric, only sparse experimental values exist. Accordingly, a benchmark of methods for computing Lewis pair formation enthalpies, also with a broader set of Lewis bases against experimental data, is missing. Herein, we evaluate different density functionals against a set of 112 experimentally determined Lewis acid/base binding enthalpies and gauge influences such as solvation correction in structure optimization. From that, we can recommend r2SCAN-3c for robust quantification of this omnipresent interaction.

Solid-phase extraction coupled to automated centrifugal microfluidics SERS: Improving quantification of therapeutic drugs in human serum

Surface-enhanced Raman spectroscopy (SERS) is a powerful method in analytical chemistry, but its application in real-life medical settings has been limited due to technical challenges. In this work, we introduce an innovative approach that is meant to advance the automation of microfluidics SERS to improve reproducibility and label-free quantification of two widely used therapeutic drugs, methotrexate (MTX) and lamotrigine (LTG), in human serum. Our methodology involves a miniaturized solid-phase extraction (μ-SPE) method coupled to a centrifugal microfluidics disc with incorporated SERS substrates (CD-SERS). The CD-SERS platform enables simultaneous controlled sample wetting and accurate SERS mapping. Together with the assay we implemented a machine learning method based on Partial Least Squares Regression (PLSR) for robust data analysis and drug quantification. The results indicate that combining μ-SPE with CD-SERS (μ-SPE to CD-SERS) led to a substantial improvement in the signal-to-noise ratio compared to combining CD-SERS with ultrafiltration or protein precipitation. The PLSR model enabled us to obtain the limit of detection and quantification for MTX as 2.90 and 8.92 μM, respectively, and for LTG as 10.76 and 32.29 μM. We also validated our μ-SPE to CD-SERS method for MTX against HPLC and immunoassay (p-value <0.05), using patient samples undergoing MTX therapy. In addition, we achieved a satisfactory recovery rate (80%) for LTG when quantifying it in patient samples. Our results show the potential of this newly developed approach as a strategy for therapeutic drugs in point-of-care clinical settings and highlight the benefits of automating label-free SERS assays.

Value assessment of carbon sequestration in generalized black soil distribution area of Amur River Basin, China

ABSTRACT In recent years, the understanding and resolution of imbalances in ecosystem carbon sequestration services (ECSS) have increasingly relied on robust data quantification and value-accounting mechanisms. Based on the interpretation of primary data, including surface vegetation, 30 cm soil cover, and hydrological information of the Amur River Basin, a carbon sequestration evaluation framework centered on the black soil belt of the river basin (BSBB) was constructed. The results showed that: (1) The annual average carbon sequestration of the BSBB from 2000 to 2019 is estimated to be 12.41 Gt C, with an expected economic value of approximately 607.61 billion yuan (China Yuan, CNY); (2) Daxinganling (DXAL, 3.26 Gt), Heihe (HH, 2.72 Gt) and Mudanjiang (MDJ, 1.81 Gt) were the three sub-regions with the highest annual carbon sequestration service, respectively, and the overall carbon density center has shifted from the south to the north within a decade; (3) The results from the geographic detector and partial correlation analysis showed that the forest area (q = 0.97, r = 0.4) and non-cultivated black soil area (q = 0.96, r = 0.97) were both clear driving factors and strong driving relationships. An increase in the forest area of the BSBB is conducive to upgrading the ECSS. This study helps promote the implementation of a carbon sink balance strategy on a global scale, provides theoretical support for future black soil resource management based on trade-offs and synergies, and provides a dual perspective for balancing ecological environmental protection with social and economic development.

Mechanisms of near-wellbore fracture growth considering the presence of cement sheath microcracks and their implications on wellbore stability

Placement of intended perforations is usually extensive enough into the reservoir to minimize any near wellbore influence on the expected fracture growth. However, some unplanned fractures do not necessarily originate from the intended shot holes. A sufficient volume of published work exists on the processes of crack growth from the near wellbore regions such as from the cement sheath or from fractures observed on the formation wall surrounding the wellbore. A robust quantification of wellbore stability issues should not detach such fracture development, which can potentially result into issues such as cement sheath damage and unnecessary formation damage in the near-wellbore regions. Such fractures may become fluid injection points during stimulation related wellbore pressurization, leading to pronounced development. In such scenarios, fracture development may become a nuisance if it jeopardises the intended structural support and zonal isolation integrity of the cement sheath around the wellbore, leading to wellbore failure, and a potential abandonment. The pressure transmission across this structural seal affects the zonal isolation integrity and its intended structural support leading to wellbore instability. Understanding crack growth behaviour in such cement-rock structural framework is still illusional yet it poses the wellbore to a risky state of instability. Our study provides insight into mechanical and elastic behavior of a tightly bonded cement-rock framework with presence of initial cracks. It also provides insight into how the acting far-field stresses, applied wellbore pressures, crack length, and orientation angle influence the crack growth morphologies including planar fractures, curved and cyclic fractures. The study is important not only for a proper selection of the cement slurry but also for the definition of allowable pressures during the wellbore lifespan. Inappropriately selected parameters might lead to the loss of wellbore integrity to a point as early as the start of a well’s life.

Deep learning for detecting and characterizing oil and gas well pads in satellite imagery

Methane emissions from the oil and gas sector are a large contributor to climate change. Robust emission quantification and source attribution are needed for mitigating methane emissions, requiring a transparent, comprehensive, and accurate geospatial database of oil and gas infrastructure. Realizing such a database is hindered by data gaps nationally and globally. To fill these gaps, we present a deep learning approach on freely available, high-resolution satellite imagery for automatically mapping well pads and storage tanks. We validate the results in the Permian and Denver-Julesburg basins, two high-producing basins in the United States. Our approach achieves high performance on expert-curated datasets of well pads (Precision = 0.955, Recall = 0.904) and storage tanks (Precision = 0.962, Recall = 0.968). When deployed across the entire basins, the approach captures a majority of well pads in existing datasets (79.5%) and detects a substantial number (>70,000) of well pads not present in those datasets. Furthermore, we detect storage tanks (>169,000) on well pads, which were not mapped in existing datasets. We identify remaining challenges with the approach, which, when solved, should enable a globally scalable and public framework for mapping well pads, storage tanks, and other oil and gas infrastructure.

A semi-parametric maximum-likelihood analysis of measurement error in population size estimation

Abstract This work addresses the challenge of measurement errors in capture–recapture (CR) studies with covariates. These errors can introduce bias and undermine inference quality. To address this issue, we introduce a nonparametric measurement error model tailored to the ‘repeated counts’ setting, employing EM-type algorithms for parameter estimation. We use the Horvitz–Thompson estimator for population size estimates. Rigorous simulations, covering varying degrees of measurement error reliability, confirm our approach’s effectiveness. Applied to benchmark datasets, it consistently provides more accurate point estimates and robust uncertainty quantification, enhancing the reliability of CR analyses.