Search NASA⌕ Search

SEARCH · Search NASA

Results for “mean squared error”

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 55 records · Page 3

Structure classification and melting temperature prediction in octet AB solids via machine learning

Machine learning methods are being increasingly used in condensed matter physics and materials science to classify crystals structures and predict material properties. However, the reliability of these methods for a given problem, especially when large data sets are unavailable, has not been well studied. By addressing the tasks of classifying crystal structure and predicting melting temperatures of the octet subset of AB solids, we performed such a study and found potential problems with using machine learning methods on relatively small data sets. At the same time, however, we can reaffirm the potential power of such methods for these tasks. In particular, we uncovered an important new material feature, the excess Born effective charge, that significantly increased the accuracy of the predictions for the classification problem we defined. This discovery leads us to propose a new scale for the degree of ionicity and covalency in these solids. More specifically, we partitioned the crystal structures of a set of 75 octet solids into those that are ionic and covalent bonded and thus performed a binary classification task. We found that using the standard indices (r σ ,r π ), suggested by St. John and Bloch several decades ago, enabled an average success in classification of 92%. Using just r σ and the excess Born effective charge ΔZ A of the A atom enabled an average success of 97%, but we also found relatively large variations about these averages that were dependent on how certain machine learning methods were used and for which a standard deviation was not a proper measure of the degree of confidence we can place in either average. Instead, we calculated and report with 95 % confidence that the traditional classification pair predicts an accuracy in the interval [ 89%, 95%] and the accuracy of the new pair lies in the interval [96 %, 99%]. For melting temperature predictions, the size of our data set was 46. We estimate the root-mean-squared error of our resulting model to be 11% of the mean melting temperature of the data, but we note that if the accuracy of this predicted error is itself measured, our estimated fitting error itself has a root-mean-square error of 50%. In short, what we illustrate is that classification and regression predictions can vary significantly, depending on the details of how machine learning methods are applied to small data sets. This variation makes it important, if not essential, to average the predictions and compute confidence intervals about these averages to report results meaningfully. However, when properly used, these statistical methods can advance our understanding and improve predictions of material properties even for small data sets.

36 MATERIALS SCIENCE↗

Probabilistic Day-Ahead Forecasting Using an Analog Ensemble Approach for Wind Farm Grid Services

Wind resource assessment and wind power forecasting are used in research and industry to anticipate future power output at scales ranging from individual wind turbines to entire wind farms. Probabilistic day-ahead wind forecasting is useful for anticipating how a wind farm could potentially participate in the day-ahead market by providing upper and lower bounds for expected power generation, thus informing grid operators of its uncertainty. Understanding this uncertainty is part of a larger project focused on building a platform that combines efforts in weather forecasting, aerodynamic and economic modeling to create maximum value of a wind plant to better provide services to the grid. This effort is also known as the Atmosphere to Electrons to Grid (A2E2G) project. One method for producing a probabilistic forecast is through the analog ensemble approach (Delle Monache et al., 2011). This method leverages historical forecasts and their corresponding observations as a training data set from which future forecasts can be made. For some future forecast, the most similar historical forecasts (analogs) are identified on a regular time basis such as once per a 3-hour window. The most similar analogs, based on a metric such as root mean square error (RMSE), are recorded and their corresponding verifying observations are used as an ensemble member for this future forecast. Prior work in this area demonstrates improvements over raw Numerical Weather Prediction (NWP) forecasts and shows skill similar to techniques such as logistic regression and machine learning (Delle Monache et al., 2013; Alessandrini et al., 2015). Here, we take the High-Resolution Rapid Refresh model (HRRR) day-ahead forecast (0-36 hours) to create a probabilistic day-ahead forecast using an analog ensemble approach. The HRRR has an hourly temporal resolution, with a spatial resolution of 3 km. The 12 UTC HRRR model run is downloaded every day for one year from August 2019 - July 2020, with the first 11 months serving as a bank of analogs from which the forecasting algorithm can create a probabilistic forecast. Once downloaded, the original HRRR forecast is temporally interpolated to 5-minutes, aligning with both the temporal resolution of the observations as well as the timescale relevant for day-ahead power forecasts. The forecast is validated at the M2 tower at the Flatirons Campus of the National Renewable Energy Laboratory (NREL) at a typical wind turbine height of 80 m. Variables such as wind speed, wind direction, and turbulence intensity are incorporated into the probabilistic forecast model and weighted according to their relative importance to the forecast. Based on metrics such as mean bias error (MBE), mean absolute error (MAE), and root mean square error, the analog ensemble forecast outperforms the raw HRRR forecast during the testing period of July 2020. Figure 1 illustrates an example day-ahead forecast compared against the verifying observations. The general variability and ramps are captured throughout the day, with potential to further improve the analog ensemble model through machine learning techniques.

numerical weather prediction↗

Use of multiple gauges and microwave attenuation of precipitation for satellite verification

In this paper both a microwave attenuation measurement along a horizontal line and multiple point gauge measurements are analyzed as possible ground-truth designs to validate satellite precipitation retrieval algorithms at the field of view spatial level (typically about 20 km). The design consists of comparing a sequence of pairs of contemporaneous measurements taken from the ground and from space. The authors examine theoretically the variance of expected differences between the two systems. The line average measurement leads to a smaller mean-square error compared to the case of a single point gauge, since some of the small-scale variability of the rain field is smoothed away by the line integration. The multiple point gauge measurements also give smaller mean-square error than that of a single point gauge. The centroid of the line and point gauge configurations are considered to be located randomly inside the field of view for different overpasses. A space-time spectral formalism is used with a noise-forced diffusive rain field to find the mean-square error. By considering instantaneous ground and satellite measurement pairs over about 50 visits when raining, we can reduce the expected error to approximately 10% of the standard deviation of climatological variability. This is considered to be a useful level of tolerance for identifying biases in the retrieval algorithms. It is found that the multiple point gauges (especially two gauges) are the economical ground-truth design compared to the microwave attenuation based on the mean-square error comparison. The major finding of this study is that a significant improvement over the point gauge is obtained by adding a single additional piece of information; adding more gauges or extending the line of attenuation is not an important improvement.

Ha, Eunho↗

Quantifying Aerosol Direct Effects from Broadband Irradiance and Spectral Aerosol Optical Depth Observations

We outline a methodology using broadband and spectral irradiances to quantify aerosol direct effects on the surface diffuse shortwave (SW) irradiance. Best Estimate Flux data span a 13 year timeframe at the Department of Energy Atmospheric Radiation Measurement Program’s Southern Great Plains (SGP) site. Screened clear-sky irradiances and aerosol optical depth (AOD), for solar zenith angles ≤ 65°, are used to estimate clear-sky diffuse irradiances. We validate against detected clear-sky observations from SGP’s Basic Radiation System (BRS). BRS diffuse irradiances were in accordance with estimates, producing a root-mean-square error and mean bias errors of 4.0 W/m2 and -1.4 W/m2, respectively. Absolute differences show 99% of estimates within ±10 W/m2 (10%) of the mean BRS observations. Clear-sky diffuse estimates are used to derive quantitative estimates of aerosol radiative effects, represented as the aerosol diffuse irradiance (ADI). ADI is the contribution of diffuse SW to global SW, attributable to scattering of atmospheric transmission by natural plus anthropogenic aerosols. Estimated slope for the ADI as a function of AOD indicates an increase of ~22 W/m2 in diffuse SW for every 0.1 increase in AOD. Such significant increases in the diffuse fraction could possibly increase photosynthesis. Annual mean ADI is 28.2 W/m2, and heavy aerosol loading at SGP provides up to a maximum increase of 120 W/m2 in diffuse SW over background conditions. With regard to seasonal variation, the mean diffuse forcings are 17.2, 33.3, 39.0, and 23.6 W/m2 for winter, spring, summer, and fall, respectively.

broadband↗

First estimation of high-resolution solar photovoltaic resource maps over China with Fengyun-4A satellite and machine learning

Fengyun-4A (FY-4A), which is the latest-generation Chinese geostationary meteorological satellite, measures solar reflection and thermal emission with high temporal, spatial, and spectral resolutions. It is expected to be highly beneficial for solar resource assessment and forecasting in China. This study is the first to estimate, using FY-4A and a random forest model, the global horizontal irradiance (GHI) at a 4-km–15-min spatio-temporal resolution over China, as a means to arrive at a solar photovoltaic (PV) resource map. In terms of GHI estimates, the root mean square error and mean bias error between hourly measured and retrieved values are 147.02 (35.2%), –5.64 W/m 2 (–1.4%), respectively, whereas the values of daily estimates are 29.20 (18.0%), –2.97 W/m 2 (–1.3%). The retrieval accuracy is found much better for instances with solar zenith angles smaller than 60°. Relatively larger errors are found at locations in the Sichuan Basin and northeastern China, which can be attributed to bright surfaces and/or strong cloud transients. With the retrieved irradiance, PV resource is derived through a physical model chain. The annual mean PV resource map suggests that, over most of the west areas, the annual mean effective irradiance exceeds 1700 kWh/m 2 , with the highest value found in Tibet (around 2000 kWh/m 2 per annum). Eastern China has an annual effective irradiance of only 1300–1500 kWh/m 2 . In conclusion, the region with poorest solar resource is the Sichuan Basin (less than 1100 kWh/m 2 per annum).

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

A semblance measure for model comparison

Algorithmic and computational advances have made it possible that geophysical survey and earth model design can be aided by many systematic trial inverse-modelling runs with synthetic data. Such may, for example, come up in machine-learning approaches. Automated image appraisal pertaining to such applications will involve common statistical tests for goodness-of-data fit as a primary evaluation method. However, solution non-uniqueness may render multiple images equivalent in terms of their data fit, requiring secondary categorizers. A logical choice for classifying synthetic-imaging results quantifies the goodness of model fit where a known reference model replaces the observational input. The task of model intercomparison in terms of measuring the resemblance to the reference model poses challenges to common distance-based metrics like root mean square error and mean absolute error. First, distance-based metrics can introduce spurious contributions when smooth models with fuzzy target contours are to be compared against a sharp reference. Second, large differences due to parameter-estimation overshoots can dominate distance metrics. Here, we propose a remedy that is referred to as semblance and is based on the idea of logistic functions, where a binary-dependent variable adds non-zero or zero accumulation terms for the, respectively, passing or failing of preset target thresholds. This classifying approach is amenable to an objective where model feature recognition is primary. Numerical comparisons to distance-based metrics provide evidence for the advantages of the semblance in view of this objective. Geophysical imaging in conjunction with machine-learning is seen as a benefitting upcoming application area.

58 GEOSCIENCES↗

A novel adaptive discrete cosine transform-domain filter for gap-inpainting of high resolution PET scanners

Purpose: Several positron emission tomography (PET) scanners with special detector block arrangements have been developed in recent years to improve the resolution of PET images. However, the discontinuous detector blocks cause gaps in the sinogram. This study proposes an adaptive discrete cosine transform-based (aDCT) filter for gap-inpainting. Methods: The gap-corrupted sinogram was morphologically closed and subsequently converted to the DCT domain. A certain number of the largest coefficients in the DCT spectrum were identified to determine the low-frequency preservation region. The weighting factors for the remaining coefficients were determined by an exponential weighting function. The aDCT filter was constructed and applied to two digital phantoms and a simulated phantom introduced with various levels of noise. Results: For the Shepp-Logan head phantom, the aDCT filter filled the gaps effectively. For the Jaszczak phantom, no secondary artifacts were induced after aDCT filtering. The percent mean square error and mean structure similarity of the aDCT filter were superior to those of the DCT2 filter at all noise levels. For the simulated striatal dopamine innervation study, the aDCT filter recovered the shape of the striatum and restored the striatum to reference activity ratios to the ideal value. Conclusions: The proposed aDCT filter can recover the missing gap data in the sinogram and improve the image quality and quantitative accuracy of PET images.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Iterative metal artifact reduction for x-ray computed tomography using unmatched projector/backprojector pairs

Purpose: Metal artifact reduction (MAR) is a major problem and a challenging issue in x-ray computed tomography (CT) examinations. Iterative reconstruction from sinograms unaffected by metals shows promising potential in detail recovery. This reconstruction has been the subject of much research in recent years. However, conventional iterative reconstruction methods easily introduce new artifacts around metal implants because of incomplete data reconstruction and inconsistencies in practical data acquisition. Hence, this work aims at developing a method to suppress newly introduced artifacts and improve the image quality around metal implants for the iterative MAR scheme. Methods: The proposed method consists of two steps based on the general iterative MAR framework. An uncorrected image is initially reconstructed, and the corresponding metal trace is obtained. The iterative reconstruction method is then used to reconstruct images from the unaffected sinogram. In the reconstruction step of this work, an iterative strategy utilizing unmatched projector/backprojector pairs is used. A ramp filter is introduced into the back-projection procedure to restrain the inconsistency components in low frequencies and generate more reliable images of the regions around metals. Furthermore, a constrained total variation (TV) minimization model is also incorporated to enhance efficiency. The proposed strategy is implemented based on an iterative FBP and an alternating direction minimization (ADM) scheme, respectively. The developed algorithms are referred to as “iFBP-TV” and “TV-FADM,” respectively. Two projection-completion-based MAR methods and three iterative MAR methods are performed simultaneously for comparison. Results: The proposed method performs reasonably on both simulation and real CT-scanned datasets. This approach could reduce streak metal artifacts effectively and avoid the mentioned effects in the vicinity of the metals. The improvements are evaluated by inspecting regions of interest and by comparing the root-mean-square errors, normalized mean absolute distance, and universal quality index metrics of the images. Both iFBP-TV and TV-FADM methods outperform other counterparts in all cases. Unlike the conventional iterative methods, the proposed strategy utilizing unmatched projector/backprojector pairs shows excellent performance in detail preservation and prevention of the introduction of new artifacts. Conclusions: Qualitative and quantitative evaluations of experimental results indicate that the developed method outperforms classical MAR algorithms in suppressing streak artifacts and preserving the edge structural information of the object. In particular, structures lying close to metals can be gradually recovered because of the reduction of artifacts caused by inconsistency effects.

60 APPLIED LIFE SCIENCES↗

A Comparison of Infectious Disease Forecasting Methods across Locations, Diseases, and Time

Accurate infectious disease forecasting can inform efforts to prevent outbreaks and mitigate adverse impacts. This study compares the performance of statistical, machine learning (ML), and deep learning (DL) approaches in forecasting infectious disease incidences across different countries and time intervals. We forecasted three diverse diseases: campylobacteriosis, typhoid, and Q-fever, using a wide variety of features (n = 46) from public datasets, e.g., landscape, climate, and socioeconomic factors. We compared autoregressive statistical models to two tree-based ML models (extreme gradient boosted trees [XGB] and random forest [RF]) and two DL models (multi-layer perceptron and encoder–decoder model). The disease models were trained on data from seven different countries at the region-level between 2009–2017. Forecasting performance of all models was assessed using mean absolute error, root mean square error, and Poisson deviance across Australia, Israel, and the United States for the months of January through August of 2018. The overall model results were compared across diseases as well as various data splits, including country, regions with highest and lowest cases, and the forecasted months out (i.e., nowcasting, short-term, and long-term forecasting). Overall, the XGB models performed the best for all diseases and, in general, tree-based ML models performed the best when looking at data splits. There were a few instances where the statistical or DL models had minutely smaller error metrics for specific subsets of typhoid, which is a disease with very low case counts. Feature importance per disease was measured by using four tree-based ML models (i.e., XGB and RF with and without region name as a feature). The most important feature groups included previous case counts, region name, population counts and density, mortality causes of neonatal to under 5 years of age, sanitation factors, and elevation. This study demonstrates the power of ML approaches to incorporate a wide range of factors to forecast various diseases, regardless of location, more accurately than traditional statistical approaches.

59 BASIC BIOLOGICAL SCIENCES↗

Machine Learning-Based Process Control for Injection Molding of Recycled Polypropylene

The increased interest in artificial intelligence in manufacturing has driven the adoption of machine learning to optimize processes and improve efficiency. A key challenge in injection molding is the variability of recycled materials, which affects part quality and processing stability. This study presents a novel closed-loop process control approach for injection molding, leveraging machine learning to adaptively predict processing inputs and quality outcomes. The methodology was tested on five blends of recycled polypropylene (rPP), using artificial neural networks (ANNs), linear regression, and polynomial regression to model the relationships between material properties and process parameters. The dataset was split 80/20 into training and testing sets. The ANN model was implemented using TensorFlow and Keras, with six hidden layers of 32 neurons per layer, ReLU activation, and an Adam optimizer. Empirical tuning and early stopping were used to optimize performance and prevent overfitting. Predictions were evaluated based on mean absolute error (MAE), mean squared error (MSE), and percentage error. The results showed that yield stress, ultimate elongation, and part weight were accurately predicted within a 5% error for linear and polynomial regression models and within a 10% error for the ANN. However, modulus predictions were less reliable, with errors of ~11% for ANN and linear regression and ~40% for polynomial regression, reflecting the inherent variability of this property in rPP blends. Predictions of processing inputs had errors ranging from 3% to 25%, depending on the model and response variable. No single modeling approach was consistently superior across all responses, highlighting the complexity of the relationship between material properties, process parameters, and quality metrics. Overall, the work demonstrates that closed-loop process control, powered by machine learning, can effectively predict key quality parameters in injection molding of recycled materials. The proposed approach can improve process stability and material utilization, facilitating increased adoption of sustainable materials.

Krantz, Joshua↗

An LES-based airborne Doppler lidar simulator and its application to wind profiling in inhomogeneous flow conditions

Wind profiling by Doppler lidar is common practice and highly useful in a wide range of applications. Airborne Doppler lidar can provide additional insights relative to ground-based systems by allowing for spatially distributed and targeted measurements. Providing a link between theory and measurement, a first large eddy simulation (LES)-based airborne Doppler lidar simulator (ADLS) has been developed. Simulated measurements are conducted based on LES wind fields, considering the coordinate and geometric transformations applicable to real-world measurements. The ADLS provides added value as the input truth used to create the measurements is known exactly, which is nearly impossible in real-world situations. Thus, valuable insight can be gained into measurement system characteristics as well as retrieval strategies. As an example application, airborne Doppler lidar wind profiling is investigated using the ADLS. For commonly used airborne velocity azimuth display (AVAD) techniques, flow homogeneity is assumed throughout the retrieval volume, a condition which is violated in turbulent boundary layer flow. Assuming an ideal measurement system, the ADLS allows to isolate and evaluate the error in wind profiling which occurs due to the violation of the flow homogeneity assumption. Overall, the ADLS demonstrates that wind profiling is possible in turbulent wind field conditions with reasonable errors (root mean squared error of 0.36 m s –1 for wind speed when using a commonly used system setup and retrieval strategy for the conditions investigated). Nevertheless, flow inhomogeneity, e.g., due to boundary layer turbulence, can cause an important contribution to wind profiling error and is non-negligible. Results suggest that airborne Doppler lidar wind profiling at low wind speeds (<5 m s –1 ) can be biased, if conducted in regions of inhomogeneous flow conditions.

17 WIND ENERGY↗

Large Contribution of Meteorological Factors to Inter-Decadal Changes in Regional Aerosol Optical Depth

Aerosol optical depth (AOD) has become a crucial metric for assessing global climate change. Although global and regional AOD trends have been studied extensively, it remains unclear what factors are driving the inter-decadal variations in regional AOD and how to quantify the relative contribution of each dominant factor. This study used a long-term (1980–2016) aerosol dataset from the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) reanalysis, along with two satellite-based AOD datasets (MODIS/Terra and MISR) from 2001 to 2016, to investigate the long-term trends in global and regional aerosol loading. Statistical models based on emission factors and meteorological parameters were developed to identify the main factors driving the inter-decadal changes of regional AOD and to quantify their contribution. Evaluation of the MERRA-2 AOD with the ground-based measurements of AERONET indicated significant spatial agreement on the global scale (r= 0.85, root-mean-square error = 0.12, mean fractional error = 38.7 %, fractional gross error = 9.86 % and index of agreement = 0.94). However, when AOD observations from the China Aerosol Remote Sensing Network (CARSNET) were employed for independent verification, the results showed that MERRA-2 AODs generally underestimated CARSNET AODs in China (relative mean bias = 0.72 and fractional gross error =−34.3 %). In general, MERRA-2 was able to quantitatively reproduce the annual and seasonal AOD trends on both regional and global scales, as observed by MODIS/Terra, although some differences were found when compared to MISR. Over the 37-year period in this study, significant decreasing trends were observed over Europe and the eastern United States. In contrast, eastern China and southern Asia showed AOD increases, but the increasing trend of the former reversed sharply in the most recent decade. The statistical analyses suggested that the meteorological parameters explained a larger proportion of the AOD variability (20.4 %–72.8 %) over almost all regions of interest (ROIs) during 1980–2014 when compared with emission factors (0 %–56 %). Further analysis also showed that SO2 was the dominant emission factor, explaining 12.7 %–32.6 % of the variation in AOD over anthropogenic-aerosol-dominant regions, while black carbon or organic carbon was the leading factor over the biomass-burning-dominant (BBD) regions, contributing 24.0 %–27.7 % of the variation. Additionally, wind speed was found to be the leading meteorological parameter, explaining 11.8 %–30.3 % of the variance over the mineral-dust-dominant regions, while ambient humidity (including soil moisture and relative humidity) was the top meteorological parameter over the BBD regions, accounting for 11.7 %–35.5 % of the variation. The results of this study indicate that the variation in meteorological parameters is a key factor in determining the inter-decadal change in regional AOD.

Huizheng Che↗

Evaluation of High Mountain Asia-Land Data Assimilation System (version 1) from 2003 to 2016: 2. The impact of assimilating satellite-based snow cover and freeze/thaw observations into a land surface model

This second paper of the two-part series focuses on demonstrating the impact of assimilating satellite-based snow cover and freeze/thaw observations into the hyper-resolution, offline terrestrial modeling system used for the High Mountain Asia (HMA) region from 2003 to 2016. To this end, this study systematically evaluates a total of six sets of 0.01° (∼1 km) model simulations forced by different precipitation forcings, with and without the dual assimilation scheme enabled, at point-scale, basin-scale, and domain-scale. The key variables of interest include surface net shortwave radiation, surface net longwave radiation, skin temperature, near-surface soil temperature, snow depth, snow water equivalent (SWE), and total runoff. First, the point-scale assessment is mainly conducted via evaluating against ground-based measurements. In general, the assimilation enabled estimates are better than no-assimilation counterparts. Second, the basin-scale runoff assessment demonstrates that across three snow-dominated basins, the assimilation enabled experiment yields systematic improvements in all goodness-of-fit statistics through mitigating the negative effects brought by the fixed long-term precipitation correction factors. For example, when forced by the bias-corrected precipitation, the assimilation-enabled experiment improves the bias by 69%, the root-mean-squared error by 30%, and the unbiased root-mean-squared error by 18% (relative to the no-assimilation counterpart). Finally, the domainscale assessment is conducted via evaluating against satellite-based SWE and skin temperature products. Both sets of domain-scale analysis further corroborate the findings in the point-scale evaluations. Overall, this study suggests the benefits of the proposed multi-variate assimilation system in improving the cryospherichydrological process within a land surface model for use in HMA.

Yuan Xue↗

Evaluation of High Mountain Asia-Land Data Assimilation System (version 1) from 2003 to 2016: 2. The impact of assimilating satellite-based snow cover and freeze/thaw observations into a land surface model

This second paper of the two-part series focuses on demonstrating the impact of assimilating satellite-based snow cover and freeze/thaw observations into the hyper-resolution, offline terrestrial modeling system used for the High Mountain Asia (HMA) region from 2003 to 2016. To this end, this study systematically evaluates a total of six sets of 0.01° (∼1 km) model simulations forced by different precipitation forcings, with and without the dual assimilation scheme enabled, at point-scale, basin-scale, and domain-scale. The key variables of interest include surface net shortwave radiation, surface net longwave radiation, skin temperature, near-surface soil temperature, snow depth, snow water equivalent (SWE), and total runoff. First, the point-scale assessment is mainly conducted via evaluating against ground-based measurements. In general, the assimilation enabled estimates are better than no-assimilation counterparts. Second, the basin-scale runoff assessment demonstrates that across three snow-dominated basins, the assimilation enabled experiment yields systematic improvements in all goodness-of-fit statistics through mitigating the negative effects brought by the fixed long-term precipitation correction factors. For example, when forced by the bias-corrected precipitation, the assimilation-enabled experiment improves the bias by 69%, the root-mean-squared error by 30%, and the unbiased root-mean-squared error by 18% (relative to the no-assimilation counterpart). Finally, the domain-scale assessment is conducted via evaluating against satellite-based SWE and skin temperature products. Both sets of domain-scale analysis further corroborate the findings in the point-scale evaluations. Overall, this study suggests the benefits of the proposed multi-variate assimilation system in improving the cryospheric-hydrological process within a land surface model for use in HMA.

Yuan Xue↗

Performance Metrics for the Assessment of Satellite Data Products: An Ocean Color Case Study

Performance assessment of ocean color satellite data has generally relied on statistical metrics chosen for their common usage and the rationale for selecting certain metrics is infrequently explained. Commonly reported statistics based on mean squared errors, such as the coefficient of determination (r2), root mean square error, and regression slopes, are most appropriate for Gaussian distributions without outliers and, therefore, are often not ideal for ocean color algorithm performance assessment, which is often limited by sample availability. In contrast, metrics based on simple deviations, such as bias and mean absolute error, as well as pair-wise comparisons, often provide more robust and straightforward quantities for evaluating ocean color algorithms with non-Gaussian distributions and outliers. This study uses a SeaWiFS chlorophyll-a validation data set to demonstrate a framework for satellite data product assessment and recommends a multimetric and user-dependent approach that can be applied within science, modeling, and resource management communities.

remote sensing↗

Reducing Southern Ocean Shortwave Radiation Errors in the ERA5 Reanalysis with Machine Learning and 25 Years of Surface Observations

Earth system models struggle to simulate clouds and their radiative effects over the Southern Ocean, partly due to a lack of measurements and targeted cloud microphysics knowledge. We have evaluated biases of downwelling shortwave radiation in the ERA5 climate reanalysis using 25 years (1995–2019) of summertime surface measurements, collected on the Research and Supply Vessel (RSV) Aurora Australis, the Research Vessel (R/V) Investigator, and at Macquarie Island. During October–March daylight hours, the ERA5 simulation of SW down exhibited large errors (mean bias = 54 W m -2 , mean absolute error = 82 W m -2 , root-mean-square error = 132 W m -2 , and R 2 = 0.71). To determine whether we could improve these statistics, we bypassed ERA5’s radiative transfer model for SW down with machine learning–based models using a number of ERA5’s gridscale meteorological variables as predictors. These models were trained and tested with the surface measurements of SW down using a 10-fold shuffle split. An extreme gradient boosting (XGBoost) and a random forest–based model setup had the best performance relative to ERA5, both with a near complete reduction of the mean bias error, a decrease in the mean absolute error and root-mean-square error by 25% ± 3%, and an increase in the R 2 value of 5% ± 1% over the 10 splits. Large improvements occurred at higher latitudes and cyclone cold sectors, where ERA5 performed most poorly. We further interpret our methods using Shapley additive explanations. Our results indicate that data-driven techniques could have an important role in simulating surface radiation fluxes and in improving reanalysis products.

54 ENVIRONMENTAL SCIENCES↗

A study of image quality for radar image processing

Methods developed for image quality metrics are reviewed with focus on basic interpretation or recognition elements including: tone or color; shape; pattern; size; shadow; texture; site; association or context; and resolution. Seven metrics are believed to show promise as a way of characterizing the quality of an image: (1) the dynamic range of intensities in the displayed image; (2) the system signal-to-noise ratio; (3) the system spatial bandwidth or bandpass; (4) the system resolution or acutance; (5) the normalized-mean-square-error as a measure of geometric fidelity; (6) the perceptual mean square error; and (7) the radar threshold quality factor. Selective levels of degradation are being applied to simulated synthetic radar images to test the validity of these metrics.

King, R. W.↗

Mechanistic Modeling of TEG Dehydrator Emissions in Oil and Gas Industry

This work presents a mechanistic modeling approach for simulating methane emissions from triethylene glycol (TEG) dehydrators used in oil & gas (O&G) operations. The model was developed as a modular component of the Mechanistic Air Emissions Simulator (MAES) tool, incorporating species-specific absorption and emission dynamics through two-level, second-order polynomial regression (PR) models trained on ProMax simulation data: (1) species-level regression models that track the transfer rates of individual gas species within the dehydrator unit streams, and (2) outlet flow stream regression models that predict the fraction of inlet gas distributed among the outlet streams of the dehydrator unit. These behaviors were characterized over a range of glycol circulation ratios, wet gas pressures, and temperatures. The model was validated using root mean square error (RMSE) analysis. The species-level PR achieved low root mean square error (RMSE) values (<0.03) for light hydrocarbon species across all dehydrator components, ranging from 0.0009 for methane to 0.029 for normal pentane. Similarly, the outlet-level PR yielded RMSE values below 0.002 for the dry gas fraction, 0.001 for the flash tank fraction, and 0.002 for the still vent fraction, demonstrating strong agreement between predicted and reference ProMax values. When deployed at field facilities, the model significantly improved MAES-simulated dehydrator emissions, revealing that gas-assisted glycol pump emissions are the dominant contributors to both dehydrator-level and site-level methane emissions under uncontrolled conditions. Further analysis of the 154 dehydrator units reported by operators under the AMI 2024 project showed that 54 units (31%) used gas-driven glycol pumps, of which 6 units (11%) operated with uncontrolled flash tanks, and 22 units (40.7%) were identified as potentially oversized. Of the six dehydrator units with uncontrolled gas-assisted pumps, pump emissions accounted for 90.25% of total dehydrator emissions and 63.10% of total site-level emissions. These findings highlight substantial opportunities for emissions mitigation through equipment upgrades.

MAES↗