Search NASA⌕ Search

SEARCH · Search NASA

Results for “Machine learning (ML)”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,549 records · Page 86

Integration of Decentralized Graph-Based Multi-Agent Reinforcement Learning with Digital Twin for Traffic Signal Optimization

Machine learning (ML) methods, particularly Reinforcement Learning (RL), have gained widespread attention for optimizing traffic signal control in intelligent transportation systems. However, existing ML approaches often exhibit limitations in scalability and adaptability, particularly within large traffic networks. This paper introduces an innovative solution by integrating decentralized graph-based multi-agent reinforcement learning (DGMARL) with a Digital Twin to enhance traffic signal optimization, targeting the reduction of traffic congestion and network-wide fuel consumption associated with vehicle stops and stop delays. In this approach, DGMARL agents are employed to learn traffic state patterns and make informed decisions regarding traffic signal control. The integration with a Digital Twin module further facilitates this process by simulating and replicating the real-time asymmetric traffic behaviors of a complex traffic network. The evaluation of this proposed methodology utilized PTV-Vissim, a traffic simulation software, which also serves as the simulation engine for the Digital Twin. The study focused on the Martin Luther King (MLK) Smart Corridor in Chattanooga, Tennessee, USA, by considering symmetric and asymmetric road layouts and traffic conditions. Comparative analysis against an actuated signal control baseline approach revealed significant improvements. Experiment results demonstrate a remarkable 55.38% reduction in Eco_PI, a developed performance measure capturing the cumulative impact of stops and penalized stop delays on fuel consumption, over a 24 h scenario. In a PM-peak-hour scenario, the average reduction in Eco_PI reached 38.94%, indicating the substantial improvement achieved in optimizing traffic flow and reducing fuel consumption during high-demand periods. These findings underscore the effectiveness of the integrated DGMARL and Digital Twin approach in optimizing traffic signals, contributing to a more sustainable and efficient traffic management system.

42 ENGINEERING↗

Development of Short-Term Forecasting Models Using Plant Asset Data and Feature Selection

Nuclear power plants collect and store large volumes of heterogeneous data from various components and systems. With recent advances in machine learning (ML) techniques, these data can be leveraged to develop diagnostic and short-term forecasting models to better predict future equipment condition. Maintenance operations can then be planned in advance whenever degraded performance is predicted, thus resulting in fewer unplanned outages and the optimization of maintenance activities. This enables lower maintenance costs and improves the overall economics of nuclear power. This paper focuses on developing a short-term forecasting process that leverages a feature selection process to distill large volumes of heterogeneous data and predict specific equipment parameters. A variety of feature selection methods, including Shapley Additive Explanations (SHAP) and variance inflation factor (VIF), were used to select the optimal features as inputs for three ML methods: long short-term memory (LSTM) networks, support vector regression (SVR), and random forest (RF). Each combination of model and input features was used to predict a pump bearing temperature both 1 and 24 hours in advance, based on actual plant system data. The optimal inputs for the LSTM and SVR were selected using the SHAP values, while the optimal input for the RF consisted solely of the response variable itself. Each model produced similar 1-hour-ahead predictions, with root mean square errors (RMSEs) of roughly 0.006. For the 24-hour-ahead predictions, differences could be seen between LSTM, SVR, and RF, as reflected by model performances of 0.036 +- 0.014, 0.0026 +- 0, and 0.063 +- 0.004 RMSE, respectively. As big data and continuous online monitoring become more widely available, the proposed feature selection process can be used for many applications beyond the prediction of process parameters within nuclear infrastructure.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

A Hybrid Deep Learning Approach to Cosmological Constraints from Galaxy Redshift Surveys

We present a deep machine learning (ML)–based technique for accurately determining σ g and Ω m from mock 3D galaxy surveys. The mock surveys are built from the AbacusCosmos suite of N -body simulations, which comprises 40 cosmological volume simulations spanning a range of cosmological parameter values, and we account for uncertainties in galaxy formation scenarios through the use of generalized halo occupation distributions (HODs). We explore a trio of ML models: a 3D convolutional neural network (CNN), a power spectrum–based fully connected network, and a hybrid approach that merges the two to combine physically motivated summary statistics with flexible CNNs. We describe best practices for training a deep model on a suite of matched-phase simulations, and we test our model on a completely independent sample that uses previously unseen initial conditions, cosmological parameters, and HOD parameters. Despite the fact that the mock observations are quite small (~0.07 h -3 Gpc 3 ) and the training data span a large parameter space (six cosmological and six HOD parameters), the CNN and hybrid CNN can constrain estimates of σ g and Ω m to ~3% and ~4%, respectively.

79 ASTRONOMY AND ASTROPHYSICS↗

Inferring the Redshift of More than 150 GRBs with a Machine-learning Ensemble Model

Gamma-ray bursts (GRBs), due to their high luminosities, are detected up to a redshift of 10, and thus have the potential to be vital cosmological probes of early processes in the Universe. Fulfilling this potential requires a large sample of GRBs with known redshifts, but due to observational limitations, only 11% have known redshifts (z). There have been numerous attempts to estimate redshifts via correlation studies, most of which have led to inaccurate predictions. To overcome this, we estimated GRB redshift via an ensemble-supervised machine-learning (ML) model that uses X-ray afterglows of long-duration GRBs observed by the Neil Gehrels Swift Observatory. The estimated redshifts are strongly correlated (a Pearson coefficient of 0.93) and have an rms error, namely, the square root of the average squared error $\langle$Δz 2 $\rangle$, of 0.46 with the observed redshifts showing the reliability of this method. The addition of GRB afterglow parameters improves the predictions considerably by 63% compared to previous results in peer-reviewed literature. Finally, we use our ML model to infer the redshifts of 154 GRBs, which increase the known redshifts of long GRBs with plateaus by 94%, a significant milestone for enhancing GRB population studies that require large samples with redshift.

79 ASTRONOMY AND ASTROPHYSICS↗

Advanced Offshore Hazard Forecasting to Enable Resilient Offshore Operations

Paper prepared for the Offshore Technology Conference, 2024. Hazards in the offshore environment can imperil successful energy operations, whether those operations are conventional, renewable, or for decarbonization. The expanding accessibility of data science and the advanced applications of machine learning (ML) models creates an opportunity to assess potential hazards and the infrastructure they impact. We present a use case demonstrating the combined application of published ML tools to U.S. federal waters of the Gulf of Mexico, an actively explored region for offshore energy that is affected by variable metocean conditions and geologic processes contributing to potential hazards.

Mark-Moser, Mackenzie K.↗

Performance Analysis of an Optimization Algorithm for Metamaterial Design on the Integrated High-Performance Computing and Quantum Systems

Optimizing metamaterials with complex geometries is a big challenge. Although an active learning algorithm, combining machine learning (ML), quantum computing, and optical simulation, has emerged as an efficient optimization tool, it still faces difficulties in optimizing complex structures that have potentially high performance. In this work, we comprehensively analyze the performance of an optimization algorithm for metamaterial design on the integrated HPC and quantum systems. We demonstrate significant time advantages through message-passing interface (MPI) parallelization on the high-performance computing (HPC) system showing approximately 54% faster ML tasks and 67 times faster optical simulation against serial workloads. Furthermore, we analyze the performance of a quantum algorithm designed for optimization, which runs with various quantum simulators on a local computer or HPC-quantum system. Results showcase ~24 times speedup when executing the optimization algorithm on the HPC-quantum hybrid system. This study paves a way to optimize complex metamaterials using the integrated HPC-quantum system.

Kim, Seongmin↗

On the Need to Align Intent and Implementation in Uncertainty Quantification for Machine Learning

Quantifying uncertainties for machine learning (ML) models is a foundational challenge in modern data analysis. This challenge is compounded by at least two key aspects of the field: (a) inconsistent terminology surrounding uncertainty and estimation across disciplines, and (b) the varying technical requirements for establishing trustworthy uncertainties in diverse problem contexts. In this position paper, we aim to clarify the depth of these challenges by identifying these inconsistencies and articulating how different contexts impose distinct epistemic demands. We examine the current landscape of estimation targets (e.g., prediction, inference, simulation-based inference), uncertainty constructs (e.g., frequentist, Bayesian, fiducial), and the approaches used to map between them. Drawing on the literature, we highlight and explain examples of problematic mappings. To help address these issues, we advocate for standards that promote alignment between the \textit{intent} and \textit{implementation} of uncertainty quantification (UQ) approaches. We discuss several axes of trustworthiness that are necessary (if not sufficient) for reliable UQ in ML models, and show how these axes can inform the design and evaluation of uncertainty-aware ML systems. Our practical recommendations focus on scientific ML, offering illustrative cases and use scenarios, particularly in the context of simulation-based inference (SBI).

Trivedi, Shubhendu [MIT] (ORCID:0000000312374301)↗

Fast Machine Learning for Quantum Control of Microwave Qudits on Edge Hardware

Quantum optimal control is a promising approach to improve the accuracy of quantum gates, but it relies on complex algorithms to determine the best control settings. CPU or GPU-based approaches often have delays that are too long to be applied in practice. It is paramount to have systems with extremely low delays to quickly and with high fidelity adjust quantum hardware settings, where fidelity is defined as overlap with a target quantum state. Here, we utilize machine learning (ML) models to determine control-pulse parameters for preparing Selective Number-dependent Arbitrary Phase (SNAP) gates in microwave cavity qudits, which are multi-level quantum systems that serve as elementary computation units for quantum computing. The methodology involves data generation using classical optimization techniques, ML model development, design space exploration, and quantization for hardware implementation. Our results demonstrate the efficacy of the proposed approach, with optimized models achieving low gate trace infidelity near $10^{-3}$ and efficient utilization of programmable logic resources.

Sanders, Flor [Columbia U.]↗

Data from: Coupled machine learning-ecosystem ensemble models substantially improve predictions of nitrous oxide (N 2 O) fluxes from US croplands

Nitrous oxide (N₂O) is a potent and persistent greenhouse gas, with rising atmospheric concentrations driven in part by inefficient use of synthetic nitrogen (N) fertilizers in agriculture. Predicting soil N₂O emissions is challenging due to high spatial and temporal variability arising from complex soil biogeochemical processes. Process-based ecosystem models and standalone machine learning (ML) approaches without extensive site-specific calibration often miss high emission episodes. Here, we show how an Ensemble Modeling System (EMS) based on outputs from an ensemble of ecosystem models coupled to an ensemble of ML models can improve predictions and understanding of N2O fluxes from US cropland. Trained and validated on approximately 12,000 N2O chamber measurements at 17 U.S. Midwest sites (six crops, 35 management practices), the EMS accurately predicted daily fluxes of N2O at both training (R² = 0.84, RMSE = 16.4 g N ha⁻¹ d⁻¹) and held-out testing sites (R² = 0.84, RMSE = 6.2 g N ha⁻¹ d⁻¹). Analyses identified six dominant N₂O drivers: soil organic carbon (SOC), NH₄⁺, NO₃⁻, water-filled pore space (WFPS), soil temperature, and biomass production. Wet, warm soils produced large N₂O peaks only with sufficient SOC and mineral N; in low-SOC soils, fluxes remained low. Incorporating these drivers into process-based models might significantly improve their predictive capacity. The EMS demonstrates a strong potential to predict N₂O fluxes at unseen sites, enabling more reliable regional inventories, improved gap-filling where measurements are sparse, and enhanced understanding of mechanisms to advance targeted mitigation strategies in food, feed, and bioenergy crops.

agricultural sciences↗

Interpreting machine learning prediction of fire emissions and comparison with FireMIP process-based models

Annual burned areas in the United States have increased 2-fold during the past decades. With more large fires resulting in more emissions of fine particulate matter, an accurate prediction of fire emissions is critical for quantifying the impacts of fires on air quality, human health, and climate. This study aims to construct a machine learning (ML) model with game-theory interpretation to predict monthly fire emissions over the contiguous US (CONUS) and to understand the controlling factors of fire emissions. The optimized ML model is used to diagnose the process-based models in the Fire Modeling Intercomparison Project (FireMIP) to inform future development. Results show promising performance for the ML model, Community Land Model (CLM), and Joint UK Land Environment Simulator-Interactive Fire And Emission Algorithm For Natural Environments (JULES-INFERNO) in reproducing the spatial distributions, seasonality, and interannual variability of fire emissions over the CONUS. Regional analysis shows that only the ML model and CLM simulate the realistic interannual variability of fire emissions for most of the subregions (r >0.95 for ML and r =0.14~0.70 for CLM), except for Mediterranean California, where all the models perform poorly (r =0.74 for ML and r <0.30 for the FireMIP models). Regarding seasonality, most models capture the peak emission in July over the western US. However, all models except for the ML model fail to reproduce the bimodal peaks in July and October over Mediterranean California, which may be explained by the smaller wind speeds of the atmospheric forcing data during Santa Ana wind events and limitations in model parameterizations for capturing the effects of Santa Ana winds on fire activity. Furthermore, most models struggle to capture the spring peak in emissions in the southeastern US, probably due to underrepresentation of human effects and the influences of winter dryness on fires in the models. As for extreme events, both the ML model and CLM successfully reproduce the frequency map of extreme emission occurrence but overestimate the number of months with extremely large fire emissions. Comparing the fire PM 2.5 emissions from the ML model with process-based fire models highlights their strengths and uncertainties for regional analysis and prediction and provides useful insights into future directions for model improvements.

54 ENVIRONMENTAL SCIENCES↗

New insights on the prevalence of drizzle in marine stratocumulus clouds based on a machine learning algorithm applied to radar Doppler spectra

Abstract. The detection of the early growth of drizzle particles in marine stratocumulus clouds is important for studying the transition from cloud water to rainwater. Radar reflectivity is commonly used to detect drizzle; however, its utility is limited to larger drizzle particles. Alternatively, radar Doppler spectrum skewness has proven to be a more sensitive quantity for the detection of drizzle embryos. Here, a machine learning (ML)-based technique that uses radar reflectivity and skewness for detecting small drizzle particles is presented. Aircraft in situ measurements are used to develop and validate the ML algorithm. The drizzle detection algorithm is applied to three Atmospheric Radiation Measurement (ARM) observational campaigns to investigate the drizzle occurrence in marine boundary layer clouds. It is found that drizzle is far more ubiquitous than previously thought; the traditional radar-reflectivity-based approach significantly underestimates the drizzle occurrence, especially in thin clouds with liquid water paths lower than 50 g m−2. Furthermore, the drizzle occurrence in marine boundary layer clouds differs among the three ARM campaigns, indicating that the drizzle formation, which is controlled by the microphysical process, is regime dependent. A complete understanding of the drizzle distribution climatology in marine stratocumulus clouds calls for more observational campaigns and continuing investigations.

54 ENVIRONMENTAL SCIENCES↗

Climate-driven deterioration of future ozone pollution in Asia predicted by machine learning with multi-source data

Ozone (O 3 ) is a secondary pollutant in the atmosphere formed by photochemical reactions that endangers human health and ecosystems. O 3 has aggravated in Asia in recent decades and will vary in the future. In this study, to quantify the impacts of future climate change on O 3 pollution, near-surface O 3 concentrations over Asia in 2020–2100 are projected using a machine learning (ML) method along with multi-source data. The ML model is trained with combined O 3 data from a global atmospheric chemical transport model and real-time observations. The ML model is then used to estimate future O 3 with meteorological fields from multi-model simulations under various climate scenarios. The near-surface O 3 concentrations are projected to increase by 5 %–20 % over South China, Southeast Asia, and South India and less than 10 % over North China and the Gangetic Plains under the high-forcing scenarios in the last decade of 21st century, compared to the first decade of 2020–2100. The O 3 increases are primarily owing to the favorable meteorological conditions for O 3 photochemical formation in most Asian regions. We also find that the summertime O 3 pollution over eastern China will expand from North China to South China and extend into the cold season in a warmer future. Our results demonstrate the important role of a climate change penalty on Asian O 3 in the future, which provides implications for environmental and climate strategies of adaptation and mitigation.

54 ENVIRONMENTAL SCIENCES↗

Machine-learning-based investigation of the variables affecting summertime lightning occurrence over the Southern Great Plains

Lightning is affected by many factors, many of which are not routinely measured, well understood, or accounted for in physical models. Several commonly used machine learning (ML) models have been applied to analyze the relationship between Atmospheric Radiation Measurement (ARM) data and lightning data from the Earth Networks Total Lightning Network (ENTLN) in order to identify important variables affecting lightning occurrence in the vicinity of the Southern Great Plains (SGP) ARM site during the summer months (June, July, August and September) of 2012 to 2020. Testing various ML models, we found that the random forest model is the best predictor among common classifiers. When convective clouds were detected, it predicts lightning occurrence with an accuracy of 76.9 % and an area under the curve (AUC) of 0.850. Using this model, we further ranked the variables in terms of their effectiveness in nowcasting lightning and identified geometric cloud thickness, rain rate and convective available potential energy (CAPE) as the most effective predictors. The contrast in meteorological variables between no-lightning and frequent-lightning periods was examined for hours with CAPE values conducive to thunderstorm formation. Besides the variables considered for the ML models, surface variables and mid-altitude variables (e.g., equivalent potential temperature and minimum equivalent potential temperature, respectively) have statistically significant contrasts between no-lightning and frequent-lightning hours. For example, the minimum equivalent potential temperature from 700 to 500 hPa is significantly lower during frequent-lightning hours compared with no-lightning hours. Finally, a notable positive relationship between the intracloud (IC) flash fraction and the square root of CAPE ($\sqrt{CAPE}$) was found, suggesting that stronger updrafts increase the height of the electrification zone, resulting in fewer flashes reaching the surface and consequently a greater IC flash fraction.

54 ENVIRONMENTAL SCIENCES↗

Thermodynamics-guided machine learning model for predicting convective boundary layer height and its multi-site applicability

Accurate estimation of convective boundary layer height (CBLH) is vital for weather, climate, and air quality modeling. Machine learning (ML) shows promise in CBLH prediction, but input parameter selection often lacks physical grounding, limiting generalizability. This study introduces a novel ML framework for CBLH prediction, integrating thermodynamic constraints and the diurnal CBLH cycle as an implicit physical guide. Boundary layer growth is modeled as driven by surface heat fluxes and atmospheric heat absorption represented with the low tropospheric stability, using the diurnal cycle as input and output. TPOT and AutoKeras are employed to select optimal models, validated against Doppler lidar-derived CBLH data, achieving an R 2 of 0.84 across untrained years. Comparisons of eddy covariance (ECOR) and energy balance Bowen ratio (EBBR) flux measurements show the same prediction capability. Models trained on the ARM SGP C1 site with ECOR data and tested at E37 and E39 yield R 2 values of 0.79 and 0.81, respectively, demonstrating their adaptability. The ML model trained with all sites' data slightly enhances the performance compared with ML models trained over single-site data. The interquartile range for predicted CBLH is consistently narrower than that for DL-derived CBLH, reflecting lower variability in predicted CBLH compared to DL-derived CBLH, which is influenced by additional factors, which are not well represented with the model inputs. The model's generalizability across multiple sites at the ARM SGP site demonstrates its potential for transfer to greater distances, offering a scalable approach for enhancing boundary layer parameterization in atmospheric models.

Chu, Yufei [Stony Brook Univ., NY (United States)]↗

Synthesis of ARM User Facility Surface Rainfall Datasets to Construct a Best Estimate Value Added Product (PrecipBE)

Surface precipitation measurements are essential for Earth system model (ESM) evaluation and understanding cloud processes. An ever-growing need for robust, temporally evolving, and easy-to-use statistical datasets provides motivation for a baseline ground-based precipitation properties data product. The U.S. Department of Energy Atmospheric Radiation Measurement (ARM) user facility operates an extensive suite of precipitation instruments with various sensitivities and operating mechanisms, which render the decision of which instrument to use based on one or more fixed thresholds challenging and prone to errors and bias. Using a long-term instrument inter-comparison from a unique per-precipitation event perspective, rather than instantaneous sample comparison, we demonstrate that ARM rainfall-measuring instruments are generally consistent with each other at the statistical level. Inter-instrument deviations at the single event level can be large, especially for specific rainfall event properties such as maximum precipitation rates. A machine-learning (ML) analysis using a random forest regressor indicates that in some cases, depending on instrument, local site climatology, and/or specific deployment configuration, certain atmospheric state variables influence the measured quantities in an unpredictable manner. Thus, a-priori weighting of different instruments does not necessarily lead to more accurate and less biased synthesis of instrument data. These results motivate the design of the ARM precipitation best-estimate (PrecipBE) value-added product, which incorporates all valid precipitation data while considering data quality and other instrument limitations. PrecipBE consists of time series and tabular statistics datasets in an easy-to-use and insightful per-precipitation event format. It provides a large set of precipitation event properties supplemented with ancillary data from ARM datasets that correspond to the detected precipitation events. We describe the PrecipBE algorithm and demonstrate its use via the examination of a single-day output as well as a long-term trend analysis of precipitation events at the ARM Southern Great Plains (SGP) site, covering more than 30 years of data. The trend analysis tentatively suggests a long-term temporal tendency for mainly shorter and less intense precipitation events at the SGP site, but a long-term increase in annual rainfall by more than 36 mm (5 %) per decade. This rainfall trend is catalyzed primarily by more extreme event properties of relatively rare, intense precipitation events, with event total and 1 min maximum precipitation rate at a 1 year timeframe increasing up to 5 mm and 9 mm h −1 (several percent) per decade, respectively. While the currently available PrecipBE datasets (at https://adc.arm.gov/discovery/, last access: 8 December 2025) cover rainfall from multiple ARM deployments up to March 2025, PrecipBE is planned to be expanded to include solid-phase precipitation and will soon become an operational product with a several-day lag from real-time. We invite the ARM user community to leverage this new product and welcome user feedback to further enhance the dataset.

Silber, Israel [Pacific Northwest National Laborat↗

Predicting carbon dioxide and energy fluxes across global FLUXNET sites with regression algorithms

Spatio-temporal fields of land–atmosphere fluxes derived from data-driven models can complement simulations by process-based land surface models. While a number of strategies for empirical models with eddy-covariance flux data have been applied, a systematic intercomparison of these methods has been missing so far. In this study, we performed a cross-validation experiment for predicting carbon dioxide, latent heat, sensible heat and net radiation fluxes across different ecosystem types with 11 machine learning (ML) methods from four different classes (kernel methods, neural networks, tree methods, and regression splines). We applied two complementary setups: (1) 8-day average fluxes based on remotely sensed data and (2) daily mean fluxes based on meteorological data and a mean seasonal cycle of remotely sensed variables. The patterns of predictions from different ML and experimental setups were highly consistent. There were systematic differences in performance among the fluxes, with the following ascending order: net ecosystem exchange ( R 2 < 0.5), ecosystem respiration ( R 2 > 0.6), gross primary production ( R 2 > 0.7), latent heat ( R 2 > 0.7), sensible heat ( R 2 > 0.7), and net radiation ( R 2 > 0.8). The ML methods predicted the across-site variability and the mean seasonal cycle of the observed fluxes very well ( R 2 > 0.7), while the 8-day deviations from the mean seasonal cycle were not well predicted ( R 2 < 0.5). Fluxes were better predicted at forested and temperate climate sites than at sites in extreme climates or less represented by training data (e.g., the tropics). Finally, the evaluated large ensemble of ML-based models will be the basis of new global flux products.

54 ENVIRONMENTAL SCIENCES↗

A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations

Photosynthesis plays an important role in carbon, nitrogen, and water cycles. Ecosystem models for photosynthesis are characterized by many parameters that are obtained from limited in situ measurements and applied to the same plant types. Previous site-by-site calibration approaches could not leverage big data and faced issues like overfitting or parameter non-uniqueness. Here we developed an end-to-end programmatically differentiable (meaning gradients of outputs to variables used in the model can be obtained efficiently and accurately) version of the photosynthesis process representation within the Functionally Assembled Terrestrial Ecosystem Simulator (FATES) model. As a genre of physics-informed machine learning (ML), differentiable models couple physics-based formulations to neural networks (NNs) that learn parameterizations (and potentially processes) from observations, here photosynthesis rates. We first demonstrated that the framework was able to correctly recover multiple assumed parameter values concurrently using synthetic training data. Then, using a real-world dataset consisting of many different plant functional types (PFTs), we learned parameters that performed substantially better and greatly reduced biases compared to literature values. Further, the framework allowed us to gain insights at a large scale. Our results showed that the carboxylation rate at 25 °C (V c,max25 ) was more impactful than a factor representing water limitation, although tuning both was helpful in addressing biases with the default values. This framework could potentially enable substantial improvement in our capability to learn parameters and reduce biases for ecosystem modeling at large scales.

54 ENVIRONMENTAL SCIENCES↗

Using automated machine learning for the upscaling of gross primary productivity

Estimating gross primary productivity (GPP) over space and time is fundamental for understanding the response of the terrestrial biosphere to climate change. Eddy covariance flux towers provide in situ estimates of GPP at the ecosystem scale, but their sparse geographical distribution limits larger-scale inference. Machine learning (ML) techniques have been used to address this problem by extrapolating local GPP measurements over space using satellite remote sensing data. However, the accuracy of the regression model can be affected by uncertainties introduced by model selection, parameterization, and choice of explanatory features, among others. Recent advances in automated ML (AutoML) provide a novel automated way to select and synthesize different ML models. In this work, we explore the potential of AutoML by training three major AutoML frameworks on eddy covariance measurements of GPP at 243 globally distributed sites. We compared their ability to predict GPP and its spatial and temporal variability based on different sets of remote sensing explanatory variables. Explanatory variables from only Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance data and photosynthetically active radiation explained over 70 % of the monthly variability in GPP, while satellite-derived proxies for canopy structure, photosynthetic activity, environmental stressors, and meteorological variables from reanalysis (ERA5-Land) further improved the frameworks' predictive ability. We found that the AutoML framework Auto-sklearn consistently outperformed other AutoML frameworks as well as a classical random forest regressor in predicting GPP but with small performance differences, reaching an r 2 of up to 0.75. We deployed the best-performing framework to generate global wall-to-wall maps highlighting GPP patterns in good agreement with satellite-derived reference data. This research benchmarks the application of AutoML in GPP estimation and assesses its potential and limitations in quantifying global photosynthetic activity.

54 ENVIRONMENTAL SCIENCES↗