SEARCH · Engineering Papers

Results for “Random forests”

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.

SourceResearch connectionFrom yearThrough year

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

At least 451 records · Page 25

All sources NASA NTRS DOE OSTI

DOE OSTI2023

Improving E3SM Land Model Photosynthesis Parameterization via Satellite SIF, Machine Learning, and Surrogate Modeling

The parameterization of key photosynthesis parameters is one of the key uncertain sources in modeling ecosystem gross primary productivity (GPP). Solar-induced chlorophyll fluorescence (SIF) offers a good proxy for GPP since it marks the actual process of photosynthesis; while machine learning (ML) provides a robust approach to model the GPP-SIF relationship. Here, we trained the boosted regressing tree (BRT) and the Random Forest ML models with Greenhouse Gases Observing Satellite SIF data and in situ GPP observations from 49 eddy covariance towers. These trained ML GPP-SIF models were fed into the Energy Exascale Earth System Model (E3SM) Land Model (ELM) to generate ELM-simulated global SIF estimates, which were then benchmarked against satellite SIF observations with a surrogate modeling approach. Our results indicated good modeling performance of the ML-based GPP-SIF relationship. The ELM model when fed with the ML GPP-SIF models also can well predict the spatial-temporal variations in SIF. We also found high model accuracy for the surrogate modeling. Model parameter sensitivity analysis suggested that the fraction of leaf nitrogen in RuBisCO (flnr) is the most sensitive parameter to the SIF; other sensitive parameters include the Ball-Berry stomatal conductance slope (mbbopt) and the vcmax entropy (vcmaxse). The posterior uncertainty in simulated GPP was greatly reduced after benchmarking, and the model produced improved spatial patterns of mean GPP relative to FLUXCOM GPP. Our integrated approach provides a new avenue for improving land models and using remote-sensing SIF, which can be further improved in the future with more ground- and satellite-based observations.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Machine Learning of Key Variables Impacting Extreme Precipitation in Various Regions of the Contiguous United States

Abstract Amplification in extreme precipitation intensity and frequency can cause severe flooding and impose significant social and economic consequences. Variations in extreme precipitation intensity, frequencies, and return periods can be attributed to many physical variables across spatial and temporal scales. Here we employ ensemble machine learning (ML) methods, namely random forest (RF), eXtreme Gradient Boosting (XGB), and artificial neural networks (ANN), to explore key contributing variables to monthly extreme precipitation intensity and frequency in six regions over the United States. We further establish emulators for return periods. Results show that the ML models for intensity perform better in regions with obvious seasonality (i.e., Northern Great Plains, Southern Great Plains, and West Coast) than the other three regions (Northeast, Southwest, and Rocky Mountains), while for frequency the models perform well for most regions. The Shapley additive explanation is used to help explain the relationships between extreme precipitation characteristics and identify top variables for RF and XGB. We find that latent heat flux, relative humidity, soil moisture, and large‐scale subsidence are key common variables across the regions for both monthly intensity and frequency, and their compound effects are non‐negligible. The developed ML models capture the probability and return period of extreme precipitation well for all regions and may be used for decision making (e.g., infrastructure planning and design).

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Ensemble Estimation of Historical Evapotranspiration for the Conterminous U.S.

Abstract Evapotranspiration (ET) is the largest component of the water budget, accounting for the majority of the water available from precipitation. ET is challenging to quantify because of the uncertainties associated with the many ET equations currently in use, and because observations of ET are uncertain and sparse. In this study, we combine information provided by available ET data and equations to produce a new monthly data set for ET for the conterminous U.S. (CONUS). These maps are produced from 1895 to 2018 at an 800 m spatial scale, marking a finer resolution than currently available products over this time period. In our approach, the relative performance of a suite of ET equations is assessed using water balance, flux tower, and remotely sensed ET estimates. At the observation locations, we use error distributions to quantify relative weights for the equations and use these in a modified Bayesian model averaging weighted ensemble approach. The relative weights are spatially generalized using a random forest regression, which is applied to wall‐to‐wall explanatory variable maps to generate CONUS‐wide relative weight maps and ensemble estimates. We assess the performance of the ensemble using a reserved subset of the observations and compare this performance against other national‐scale map products for historical to modern ET. The ensemble ET maps are shown to provide an improved accuracy over the alternative comparison products. These ET maps could be useful for a variety of hydrologic modeling and assessment applications that benefit from a long record, such as the study of periods of water scarcity through time.

Environmental Sciences & Ecology↗

DOE OSTI2024

Global-Scale Convergence Obscures Inconsistencies in Soil Carbon Change Predicted by Earth System Models

Soil carbon (C) responses to environmental change represent a major source of uncertainty in the global C cycle. Feedbacks between soil C stocks and climate drivers could impact atmospheric CO 2 levels, further altering the climate. Here, we assessed the reliability of Earth system model (ESM) predictions of soil C change using the Coupled Model Intercomparison Project phases 5 and 6 (CMIP5 and CMIP6). ESMs predicted global soil C gains under the high emission scenario, with soils taking up 43.9 Pg (95% CI: 9.2–78.5 Pg) C on average during the 21st century. The variation in global soil C change declined significantly from CMIP5 (with average of 48.4 Pg [95% CI: 2.0–94.9 Pg] C) to CMIP6 models (with average of 39.3 Pg [95% CI: 23.9–54.7 Pg] C). For some models, a small C increase in all biomes contributed to this convergence. For other models, offsetting responses between cold and warm biomes contributed to convergence. Although soil C predictions appeared to converge in CMIP6, the dominant processes driving soil C change at global or biome scales differed among models and in many cases between earlier and later versions of the same model. Random Forest models, for soil carbon dynamics, accounted for more than 63% variation of the global soil C change predicted by CMIP5 ESMs, but only 36% for CMIP6 models. Although most CMIP6 models apparently agree on increased soil C storage during the 21st century, this consensus obscures substantial model disagreement on the mechanisms underlying soil C response, calling into question the reliability of model predictions.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

How Well do Earth System Models Capture Apparent Relationships Between Phytoplankton Biomass and Environmental Variables?

Abstract As phytoplankton form the base of the marine food web, understanding the controls on their abundance is fundamental to understanding marine ecology and its sensitivity to global climate change. While many Earth System Models (ESMs) predict phytoplankton biomass, it is unclear whether they properly capture the mechanistic relationships that control this quantity in the real ocean. We used Random Forest analysis to analyze the output of 13 ESMs as well as two observational data sets. The target variable was phytoplankton carbon and the predictors included environmental parameters known to influence phytoplankton, including nutrients, light, mixed layer depth, salinity, temperature, and upwelling. We examined the following: (a) What fractions of variability in ESMs and observations can be linked to the large‐scale environmental variables simulated by ESMs? (b) What are the dominant predictors and relationships affecting phytoplankton biomass? (c) How well do ESMs simulate phytoplankton carbon and do they simulate the relationships we see in observations? About 88%–96% of the variability in observational data sets and greater than 98% in the ESMs was accounted for by environmental variables known to influence phytoplankton biomass. The dominant predictors in the observational data sets were shortwave radiation and dissolved iron, with temperature and ammonium also relatively important. All the ESMs show that shortwave radiation is the most important variable and most of them predict the right sign of sensitivity to most variables. However, the models predict that biomass reaches maximum levels at unrealistically low levels of iron and unrealistically high levels of light.

Environmental Sciences & Ecology↗

DOE OSTI2024

Estimates of Lake Nitrogen, Phosphorus, and Chlorophyll‐ a Concentrations to Characterize Harmful Algal Bloom Risk Across the United States

Abstract Excess nutrient pollution contributes to the formation of harmful algal blooms (HABs) that compromise fisheries and recreation and that can directly endanger human and animal health via cyanotoxins. Efforts to quantify the occurrence, drivers, and severity of HABs across large areas is difficult due to the resource intensive nature of field monitoring of lake nutrient and chlorophyll‐aconcentrations. To better characterize how nutrients interact with other environmental factors to produce algal blooms in freshwater systems, we used spatially explicit and temporally matched climate, landscape, in‐lake characteristic, and nutrient inventory data sets to predict nutrients and chlorophyll‐aacross the conterminous US (CONUS). Using a nested modeling approach, three random forest (RF) models were trained to explain the spatiotemporal variation in total nitrogen (TN), total phosphorus (TP), and chlorophyll‐aconcentrations across US EPA's National Lakes Assessment (n = 2,062). Concentrations of TN and TP were the most important predictors and, with other variables, the RF model accounted for 68% of variation in chlorophyll‐a. We then used these RF models to extrapolate lake TN and TP predictions to lakes without nutrient observations and predict chlorophyll‐afor ∼112,000 lakes across the CONUS. Risk for high chlorophyll‐aconcentrations is highest in the agriculturally dominated Midwest, but other areas of risk emerge in nutrient pollution hot spots across the country. These catchment and lake‐specific results can help managers identify potential nutrient pollution and chlorophyll‐ahot spots that may fuel blooms, prioritize at‐risk lakes for additional monitoring, and optimize management to protect human health and other environmental end goals.

Environmental Sciences & Ecology↗

DOE OSTI2025

Temperature and Water Levels Collectively Regulate Methane Emissions From Subtropical Freshwater Wetlands

Abstract Wetlands are the largest and most climate‐sensitive natural sources of methane. Accurately estimating wetland methane emissions involves reconciling inversion (“top‐down”) and process‐based (“bottom‐up”) models within the global methane budget. However, estimates from these two model types are inherently interdependent and often reveal substantial discrepancies. To enhance the reliability of both approaches, we need a comprehensive understanding of wetland methane emissions and an independent high‐resolution long‐term flux data set. Here, we employed a data‐driven random forest approach to identify key variables influencing methane emissions from subtropical freshwater wetlands in the Southeastern United States. The model‐estimated monthly mean methane fluxes fit well with measured methane fluxes ( R 2 = 0.67) at four representative FLUXNET‐CH4 wetland sites across the region. Variable importance analysis highlighted the sensitivity of subtropical freshwater wetland methane emissions to variations in both temperature and water levels. High temperatures facilitate methanogenesis by enhancing microbial activities, while elevated water levels maintain anaerobic conditions necessary for methane production. Notably, the response of methane emissions to water level fluctuations is contingent on temperature conditions, and vice versa. Moreover, we constructed the first high‐spatial‐resolution (∼1 km × 1 km) and long‐term (1982–2010) gridded regional wetland methane flux product for the Southeastern United States, estimating annual methane emissions from subtropical freshwater wetlands in the region at 4.93 ± 0.11 Tg CH 4 yr −1 for 1982–2010. This new benchmark product holds promise for validating and parameterizing uncertain wetland methane emission processes in bottom‐up models and provides improved prior information for top‐down models.

He, Keqi [Earth and Climate Sciences Nicholas Scho↗

DOE OSTI2024

Use of Satellite, Surface Observations and Numerical Weather Prediction Model Data to Improve Cloud Base Height and Cloud Base Vertical Velocity Estimation

Cloud base height (CBH) and cloud base vertical velocity (CBVV) are important variables that impact the overall climate in a region as they influence the formulation, longevity, and evolution of clouds. Retrieval of both parameters have long used ground instrumentation (e.g., Doppler lidar (DL), ground base radar); however, retrieving CBH from satellites is particularly challenging given that space-based instruments only observe cloud tops. In this manuscript, CBH is retrieved using a multi-linear regression equation, while CBVV used a random forests model. Both retrievals combine satellite and numerical weather prediction data. The satellite data used are the Visible Infrared Imaging Radiometer Suite imagery, while measurements of CBH and CBVV include DL and radiosonde data at the Southern Great Plains (SGP) Atmospheric Radiation Measurement observatory. Data from 83 summer days (May-August) in 2018–2021 featuring cumulus clouds forced by solar heating were examined and used to train the models, with years 2022–2023 used for validation. Various spatial domains were defined with one large (2.4° longitude by 2.0° latitude) SGP domain being split into smaller sections (smallest being 0.99° and 0.61° longitude and latitude respectably). CBH and CBVV values obtained from the DL as compared to the models show root mean square errors between 150 and 200 m, with CBVV values between 0.45 and 1 ms -1 . Finally, it was found that the CBH formulation performs well over all domains, while the CBVV retrievals become less accurate due to more turbulence being introduced into the observations as the number of DL stations decreases in the smaller domains.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Climate, Hydrology, and Nutrients Control the Seasonality of Si Concentrations in Rivers

Abstract The seasonal behavior of fluvial dissolved silica (DSi) concentrations, termedDSi regime, mediates the timing of DSi delivery to downstream waters and thus governs river biogeochemical function and aquatic community condition. Previous work identified five distinct DSi regimes across rivers spanning the Northern Hemisphere, with many rivers exhibiting multiple DSi regimes over time. Several potential drivers of DSi regime behavior have been identified at small scales, including climate, land cover, and lithology, and yet the large‐scale spatiotemporal controls on DSi regimes have not been identified. We evaluate the role of environmental variables on the behavior of DSi regimes in nearly 200 rivers across the Northern Hemisphere using random forest models. Our models aim to elucidate the controls that give rise to (a) average DSi regime behavior, (b) interannual variability in DSi regime behavior (i.e., Annual DSi regime), and (c) controls on DSi regime shape (i.e., minimum and maximum DSi concentrations). Average DSi regime behavior across the period of record was classified accurately 59% of the time, whereas Annual DSi regime behavior was classified accurately 80% of the time. Climate and primary productivity variables were important in predicting Average DSi regime behavior, whereas climate and hydrologic variables were important in predicting Annual DSi regime behavior. Median nitrogen and phosphorus concentrations were important drivers of minimum and maximum DSi concentrations, indicating that these macronutrients may be important for seasonal DSi drawdown and rebound. Our findings demonstrate that fluctuations in climate, hydrology, and nutrient availability of rivers shape the temporal availability of fluvial DSi.

Environmental Sciences & Ecology↗

DOE OSTI2024

Empowering Machine Learning Forecasting of Labquake Using Event‐Based Features and Clustering Characteristics

Abstract Following recent advances of machine learning (ML), we present a novel approach to extract spatiotemporal seismo‐mechanical features from Acoustic Emission (AE) catalogs to empower ML‐based forecasting. The AE data were recorded during laboratory stick‐slip experiments on granite samples cut by rough faults. Based on the features computed for a past time window, a random forest (RF) classifier is used to forecast the occurrence of a large magnitude event ( M AE > 3.5) in the next time window. Event‐based features allow us to associate informative time‐space characteristics to each feature and nearest‐neighbor clustering analysis enables us to separate background and clustered seismicity and train individual models. The results show that the separation of AEs enhances the forecasting accuracy from 73.2% for the entire catalog up to 82.1% and 89.0% if background and clustered events are used separately. The presented new approach may be upscaled for applications to forecast tectonic earthquakes.

Karimpouli, Sadegh↗

DOE OSTI2024

Using Machine Learning to Predict Cloud Turbulent Entrainment–Mixing Processes

Different turbulent entrainment–mixing mechanisms between clouds and environment are essential to cloud–related processes; however, accurate representation of entrainment–mixing in weather/climate models still poses a challenge. This study exploits the use of machine learning (ML) to address this challenge. Four ML (Light Gradient Boosting Machine [LGB], eXtreme Gradient Boosting, Random Forest, and Support Vector Regression) are examined and compared. It is found that LGB performs best, and thus is selected to understand the impact of entrainment–mixing on microphysics using simulation data from Explicit Mixing Parcel Model. Compared with traditional parameterizations, the trained LGB provides more accurate microphysical properties (number concentration and cloud droplet spectral dispersion). The partial dependences of predicted microphysics on features exhibit a strong alignment with physical mechanisms and expectations, as determined by the interpreting method, thus overcoming the limitations of the “black box” scheme. The underlying mechanisms are that the smaller number concentration and larger spectral dispersion correspond to more inhomogeneous entrainment–mixing. Specifically, number concentration after entrainment–mixing is positively correlated with adiabatic number concentration and liquid water content affected by entrainment–mixing, and inversely correlated with adiabatic volume mean radius. Spectral dispersion after entrainment–mixing is negatively correlated with liquid water content affected by entrainment–mixing, turbulent dissipation rate and relative humidity of entrained air. Sensitivity analysis further suggests that number concentration is mainly determined by cloud microphysical properties whereas spectral dispersion is influenced by both cloud microphysical properties and environmental variables. The results indicate that the LGB scheme has the potential to enhance the representation of entrainment–mixing in weather/climate models.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2026

A ModEx Framework for Watershed Subsurface Investigation With Limited Geophysical Data Using Machine Learning and Hydrologic Modeling

Abstract Subsurface heterogeneity influences watershed hydrology strongly but remains difficult to characterize at catchment scales with sparse and costly field data. Geophysical surveys such as electromagnetic induction (EMI) provide local spatial subsurface images yet scaling them to watershed scales and converting EMI‐derived resistivity into hydraulic properties remains a challenge. We present a Model–Experiment (ModEx) framework that integrates limited EMI data with machine learning (ML) and hydrologic modeling to improve process representation and guide field investigations. Sparse EMI surveys were scaled to the catchment scale using a Random Forest model, and the resulting resistivity fields were combined with nearby borehole constraints to parameterize a hydrologic model. The EMI‐informed hydrological simulations improved predictions of streamflow sustained by subsurface flow and shallow saturation patterns. By combining EMI data and ML with hydrologic modeling, the ModEx framework guides future subsurface surveys, providing a transferable and efficient strategy for data–model integration across diverse watersheds. Plain Language Summary Mapping the underground network of soil and rock that controls water is essential for predicting floods and droughts, but seeing underground is difficult and expensive. We cannot drill everywhere, so scientists use geophysical tools to scan broad areas. There are two key challenges: these geophysical scans are often sparse across the whole watershed, and the geophysical data is hard to translate into water‐related properties. We used artificial intelligence to solve these problems. We taught a computer to find patterns linking the limited geophysical data to the land surface properties. This allowed it to fill in the gaps and create a complete, useful subsurface map for the entire watershed. This new map improves hydrologic simulations, leading to more accurate predictions of water movement in the watershed. It also helps scientists build better models with less data and generates a priority map showing where to measure next, making future investigations more efficient. Key Points Limited EMI scaled with ML improves catchment‐scale subsurface parameterization for hydrologic models The framework integrates hydrologic modeling with limited geophysical data to support subsurface investigation design ModEx framework offers a transferable data–model integration strategy that quantifies and reduces uncertainty guiding watershed studies

Chen, Hang↗

DOE OSTI2026

Integrating Maximum Entropy Production Theory and Machine Learning to Improve Global Evapotranspiration Modeling

Accurate estimation of terrestrial evapotranspiration (ET) is vital for understanding global water and energy cycles. However, current global ET estimations are not well constrained. This study introduces an integrated framework combining the Maximum Entropy Production (MEP) theory with Random Forest (RF) model to improve global ET estimation. Specifically, in contrast to direct ET estimation by the RF model, the integrated framework (MEP‐RF) trains to predict error of MEP‐simulated ET. MEP‐RF outperforms RF in spatiotemporal extrapolation. Attribution analysis with in situ observations reveals that the inputs of MEP are the most critical variables for the ET process, including net radiation, vegetated area, soil moisture, and surface temperature. We further drive MEP‐RF with global reanalysis and satellite data sets of these four inputs, yielding a global mean terrestrial ET of 548 mm/year, with 77% attributed to transpiration. The global ET increased at a rate of 0.85 mm/year per year during 2003–2021, primarily due to vegetation greening rather than rising temperature, while decreasing soil moisture led to decreasing regional ET. The integrated framework provides a novel approach for the estimation of global ET without the need for hard‐to‐obtain and thus uncertain inputs, such as wind speed, surface roughness, aerodynamic and canopy stomatal resistance. Therefore, MEP‐RF offers an independent method on existing global ET products. It represents a promising physically based approach that can be incorporated into Earth System Models to enhance water and energy cycle simulations.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2021

Global variation in the fraction of leaf nitrogen allocated to photosynthesis

Plants invest a considerable amount of leaf nitrogen in the photosynthetic enzyme ribulose-1,5-bisphosphate carboxylase-oxygenase (RuBisCO), forming a strong coupling of nitrogen and photosynthetic capacity. Variability in the nitrogen-photosynthesis relationship indicates different nitrogen use strategies of plants (i.e., the fraction nitrogen allocated to RuBisCO; fLNR), however, the reason for this remains unclear as widely different nitrogen use strategies are adopted in photosynthesis models. Here, we use a comprehensive database of in situ observations, a remote sensing product of leaf chlorophyll and ancillary climate and soil data, to examine the global distribution in fLNR using a random forest model. We find global fLNR is 18.2 ± 6.2%, with its variation largely driven by negative dependence on leaf mass per area and positive dependence on leaf phosphorus. Some climate and soil factors (i.e., light, atmospheric dryness, soil pH, and sand) have considerable positive influences on fLNR regionally. This study provides insight into the nitrogen-photosynthesis relationship of plants globally and an improved understanding of the global distribution of photosynthetic potential.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Field experiments show no consistent reductions in soil microbial carbon in response to warming

Soil microbes play an essential role in maintaining soil functions and services, but the dynamics of soil microbial biomass carbon (MBC) under global climate change remain unclear. Recently, Patoine et al. combined a global MBC data set with Random Forest modeling and reported that global MBC decreased over 1992–2013, mainly driven by increasing temperatures. Contrarily, using MBC field observations from soil warming manipulation experiments and in-situ long-term measurements across the globe, we found that MBC showed no significant changes under soil warming. Our findings indicate that soil MBC is unlikely to have decreased significantly due to the global warming of 0.28 °C during 1992–2013, and that further mechanistic studies are needed to understand potential changes in MBC under climate change.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Strontium isoscape of sub-Saharan Africa allows tracing origins of victims of the transatlantic slave trade

Abstract Strontium isotope ( 87 Sr/ 86 Sr) analysis with reference to strontium isotope landscapes (Sr isoscapes) allows reconstructing mobility and migration in archaeology, ecology, and forensics. However, despite the vast potential of research involving 87 Sr/ 86 Sr analysis particularly in Africa, Sr isoscapes remain unavailable for the largest parts of the continent. Here, we measure the 87 Sr/ 86 Sr ratios in 778 environmental samples from 24 African countries and combine this data with published data to model a bioavailable Sr isoscape for sub-Saharan Africa using random forest regression. We demonstrate the efficacy of this Sr isoscape, in combination with other lines of evidence, to trace the African roots of individuals from historic slavery contexts, particularly those with highly radiogenic 87 Sr/ 86 Sr ratios uncommon in the African Diaspora. Our study provides an extensive African 87 Sr/ 86 Sr dataset which includes scientifically marginalized regions of Africa, with significant implications for the archaeology of the transatlantic slave trade, wildlife ecology, conservation, and forensics.

Science & Technology - Other Topics↗

DOE OSTI2020

Benchmarking materials property prediction methods: the Matbench test set and Automatminer reference algorithm

Abstract We present a benchmark test suite and an automated machine learning procedure for evaluating supervised machine learning (ML) models for predicting properties of inorganic bulk materials. The test suite, Matbench, is a set of 13 ML tasks that range in size from 312 to 132k samples and contain data from 10 density functional theory-derived and experimental sources. Tasks include predicting optical, thermal, electronic, thermodynamic, tensile, and elastic properties given a material’s composition and/or crystal structure. The reference algorithm, Automatminer, is a highly-extensible, fully automated ML pipeline for predicting materials properties from materials primitives (such as composition and crystal structure) without user intervention or hyperparameter tuning. We test Automatminer on the Matbench test suite and compare its predictive power with state-of-the-art crystal graph neural networks and a traditional descriptor-based Random Forest model. We find Automatminer achieves the best performance on 8 of 13 tasks in the benchmark. We also show our test suite is capable of exposing predictive advantages of each algorithm—namely, that crystal graph methods appear to outperform traditional machine learning methods given ~10 4 or greater data points. We encourage evaluating materials ML algorithms on the Matbench benchmark and comparing them against the latest version of Automatminer.

36 MATERIALS SCIENCE↗

DOE OSTI2021

A study of real-world micrograph data quality and machine learning model robustness

Abstract Machine-learning (ML) techniques hold the potential of enabling efficient quantitative micrograph analysis, but the robustness of ML models with respect to real-world micrograph quality variations has not been carefully evaluated. We collected thousands of scanning electron microscopy (SEM) micrographs for molecular solid materials, in which image pixel intensities vary due to both the microstructure content and microscope instrument conditions. We then built ML models to predict the ultimate compressive strength (UCS) of consolidated molecular solids, by encoding micrographs with different image feature descriptors and training a random forest regressor, and by training an end-to-end deep-learning (DL) model. Results show that instrument-induced pixel intensity signals can affect ML model predictions in a consistently negative way. As a remedy, we explored intensity normalization techniques. It is seen that intensity normalization helps to improve micrograph data quality and ML model robustness, but microscope-induced intensity variations can be difficult to eliminate.

36 MATERIALS SCIENCE↗