SEARCH · Search NASA
Results for “Forest Modeling”
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.
Belowground Response to Drought in a Tropical Forest Soil. I. Changes in Microbial Functional Potential and Metabolism
We report that global climate models predict a future of increased severity of drought in many tropical forests. Soil microbes are central to the balance of these systems as sources or sinks of atmospheric carbon (C), yet how they respond metabolically to drought is not well-understood. We simulated drought in the typically aseasonal Luquillo Experimental Forest, Puerto Rico, by intercepting precipitation falling through the forest canopy. This approach reduced soil moisture by 13% and water potential by 0.14 MPa (from -0.2 to -0.34). Previous results from this experiment have demonstrated that the diversity and composition of these soil microbial communities are sensitive to even small changes in soil water. Here, we show prolonged drought significantly alters the functional potential of the community and provokes a clear osmotic stress response, including the production of compatible solutes that increase intracellular C demand. Subsequently, a microbial population emerges with a greater capacity for extracellular enzyme production targeting macromolecular carbon. Significantly, some of these drought-induced functional shifts in the soil microbiota are attenuated by prior exposure to a short-term drought suggesting that acclimation may occur despite a lack of longer-term drought history.
Estimating forest variables from fusion of SAR and TM data and analytical scattering and reflectance models
A method for simultaneous integration of synthetic aperture radar (SAR) and optical remote sensing data in an estimation algorithm is presented which results in estimates of foliage mass over a larger range of values and more accurately than would be possible with either data type alone.
A model for the propagation of uncertainty from continuous estimates of tree cover to categorical forest cover and change
Explore the source record for details and available documents.
Leveraging 35 years of Pinus taeda research in the southeastern US to constrain forest carbon cycle predictions: regional data assimilation using ecosystem experiments
Predicting how forest carbon cycling will change in response to climate change and management depends on the collective knowledge from measurements across environmental gradients, ecosystem manipulations of global change factors, and mathematical models. Formally integrating these sources of knowledge through data assimilation, or model–data fusion, allows the use of past observations to constrain model parameters and estimate prediction uncertainty. Data assimilation (DA) focused on the regional scale has the opportunity to integrate data from both environmental gradients and experimental studies to constrain model parameters. Here, we introduce a hierarchical Bayesian DA approach (Data Assimilation to Predict Productivity for Ecosystems and Regions, DAPPER) that uses observations of carbon stocks, carbon fluxes, water fluxes, and vegetation dynamics from loblolly pine plantation ecosystems across the southeastern US to constrain parameters in a modified version of the Physiological Principles Predicting Growth (3-PG) forest growth model. The observations included major experiments that manipulated atmospheric carbon dioxide (CO 2 ) concentration, water, and nutrients, along with nonexperimental surveys that spanned environmental gradients across an 8.6 × 10 5 km 2 region. We optimized regionally representative posterior distributions for model parameters, which dependably predicted data from plots withheld from the data assimilation. While the mean bias in predictions of nutrient fertilization experiments, irrigation experiments, and CO 2 enrichment experiments was low, future work needs to focus modifications to model structures that decrease the bias in predictions of drought experiments. Predictions of how growth responded to elevated CO 2 strongly depended on whether ecosystem experiments were assimilated and whether the assimilated field plots in the CO 2 study were allowed to have different mortality parameters than the other field plots in the region. We present predictions of stem biomass productivity under elevated CO 2 , decreased precipitation, and increased nutrient availability that include estimates of uncertainty for the southeastern US. Overall, we (1) demonstrated how three decades of research in southeastern US planted pine forests can be used to develop DA techniques that use multiple locations, multiple data streams, and multiple ecosystem experiment types to optimize parameters and (2) developed a tool for the development of future predictions of forest productivity for natural resource managers that leverage a rich dataset of integrated ecosystem observations across a region.
Impacts of Degradation on Water, Energy, and Carbon Cycling of the Amazon Tropical Forests
Selective logging, fragmentation, and understory fires directly degrade forest structure and composition. However, studies addressing the effects of forest degradation on carbon, water, and energy cycles are scarce. Here, we integrate field observations and high-resolution remote sensing from airborne lidar to provide realistic initial conditions to the Ecosystem Demography Model (ED-2.2) and investigate how disturbances from forest degradation affect gross primary production (GPP), evapotranspiration (ET), and sensible heat flux (H). We used forest structural information retrieved from airborne lidar samples (13,500 ha) and calibrated with 817 inventory plots (0.25 ha) across precipitation and degradation gradients in the eastern Amazon as initial conditions to ED-2.2 model. Our results show that the magnitude and seasonality of fluxes were modulated by changes in forest structure caused by degradation. During the dry season and under typical conditions, severely degraded forests (biomass loss ≥66%) experienced water stress with declines in ET (up to 34%) and GPP (up to 35%) and increases of H (up to 43%) and daily mean ground temperatures (up to 6.5°C) relative to intact forests. In contrast, the relative impact of forest degradation on energy, water, and carbon cycles markedly diminishes under extreme, multiyear droughts, as a consequence of severe stress experienced by intact forests. Our results highlight that the water and energy cycles in the Amazon are driven by not only climate and deforestation but also the past disturbance and changes of forest structure from degradation, suggesting a much broader influence of human land use activities on the tropical ecosystems.
NASA 1990 Multisensor Airborne Campaigns (MACs) for ecosystem and watershed studies
The Multisensor Airborne Campaign (MAC) focus within NASA's former Land Processes research program was conceived to achieve the following objectives: to acquire relatively complete, multisensor data sets for well-studied field sites, to add a strong remote sensing science component to ecology-, hydrology-, and geology-oriented field projects, to create a research environment that promotes strong interactions among scientists within the program, and to more efficiently utilize and compete for the NASA fleet of remote sensing aircraft. Four new MAC's were conducted in 1990: the Oregon Transect Ecosystem Research (OTTER) project along an east-west transect through central Oregon, the Forest Ecosystem Dynamics (FED) project at the Northern Experimental Forest in Howland, Maine, the MACHYDRO project in the Mahantango Creek watershed in central Pennsylvania, and the Walnut Gulch project near Tombstone, Arizona. The OTTER project is testing a model that estimates the major fluxes of carbon, nitrogen, and water through temperate coniferous forest ecosystems. The focus in the project is on short time-scale (days-year) variations in ecosystem function. The FED project is concerned with modeling vegetation changes of forest ecosystems using remotely sensed observations to extract biophysical properties of forest canopies. The focus in this project is on long time-scale (decades to millenia) changes in ecosystem structure. The MACHYDRO project is studying the role of soil moisture and its regulating effects on hydrologic processes. The focus of the study is to delineate soil moisture differences within a basin and their changes with respect to evapotranspiration, rainfall, and streamflow. The Walnut Gulch project is focused on the effects of soil moisture in the energy and water balance of arid and semiarid ecosystems and their feedbacks to the atmosphere via thermal forcing.
Cross-domain digital twin architecture for predictive maintenance via machine learning and Large Language Models
This research introduces a comprehensive framework for creating and deploying a digital twin platform for continuous monitoring and predictive maintenance within industrial settings. Through utilizing advanced technologies, including Unreal Engine 5, Unity 3D, the Message Queue Telemetry Transport protocol, Random Forest machine learning algorithms, and Large Language Models (LLMs), we establish a platform that digitally reproduces physical equipment and translates digital controls into real-world actions. This facilitates preventive maintenance approaches and improves operational effectiveness. The digital twin platform gathers sensor data from operational equipment, analyzes it using machine learning, and delivers practical insights to prevent potential malfunctions and enhance equipment performance. Furthermore, the incorporation of a web portal enables efficient monitoring and access to historical data, educational materials, and equipment status information. Preliminary findings indicate that digital twins can transform industrial equipment management and maintenance methodologies.
Predicting Air Traffic Management Initiatives Using Supervised Learning
Terminal Traffic Management Initiatives (TMIs) such as Ground Stops (GS) and Ground Delay Programs (GDP) are implemented to manage excess demand or lowered capacity at an airport. Air Traffic Flow Management (TFM) specialists identify situations such as aviation constraints, current and forecasted weather conditions, airport demand and capacity, and initiate TMIs for safe and orderly movement of air traffic. In this paper, we outline supervised learning techniques that can be used to predict and recommend TMIs at an airport based on current weather and airport conditions. Our research involves building classic Machine Learning (ML) models such as Logistic Regression, K-Nearest Neighbor, Random Forest and XGBoost, as well as Long short-term memory (LSTM) networks. We trained the models on 3-year historical data (weather, airport demand, capacity and TMIs) from Newark (EWR) airport which was selected based on its higher TMI implementation rates and varied weather conditions. Although Random Forest and XGBoost algorithms are able to predict if a TMI is needed or not, they have difficulty in predicting specific program type. For this purpose, we found that LSTM time-series forecasting models performed better as they also learn from past TMI program type sequences. This study also lays down the foundation for advanced modeling techniques and architectures to predict TMIs in advance for future periods. The ability to predict TMIs in advance will be highly beneficial to the traffic controllers and managers as this will help them to prepare for and manage TMIs more efficiently.
Reference data, predictors, and probability grids for forest degradation classes in three sites in the Brazilian Amazon
Forest degradation by fires and selective logging is widespread in the Amazon region. We implemented a gradient boosted classification modeling framework to classify intact, logged, and burned forests at three Amazonian sites: Feliz Natal Municipality and Xingu Indigenous Territory in Mato Grosso State, and Saracá-Taquera National Forest in Pará State. We used forest degradation history from Landsat time-series as reference data and textural metrics derived from PlanetScope images as predictors. Textural metrics were computed using the Gray-Level Co-Occurrence Matrix (GLCM) textural technique. Included in the attached zip file are ten files: - a shapefile containing the reference data (fire and selective logging polygons and year of event) for each site; - a multiband tif file containing the 8 GLCM metrics used as predictors (Mean, Variance, Homogeneity, Contrast, Dissimilarity, Entropy, Angular Second Moment, Correlation) at the original PlanetScope resolution (3.125m) for each site; - a multiband tif file containing the 72 aggregated GLCM metrics used as predictors (Mean, Variance, Homogeneity, Contrast, Dissimilarity, Entropy, Angular Second Moment, and Correlation aggregated using the mean, first quartile, third quartile, maximum, median, minimum, root mean square, standard deviation, and skewness statistics) at 562m resolution for each site; - a multiband tif file containing the 3 probability grids for either intact, logged, or burned forests at the aggregation resolution (562m) for each site.
Western Montana Ecological Forecasting: Modeling Habitat Suitability of Mustelid Species to Guide Detection Dog Surveys for Contaminants Monitoring, via Collected Scats in River Systems of Western Montana
Environmental contaminants are becoming increasingly prevalent in riverine ecosystems. The status of contaminants in western Montana’s relatively pristine river systems is largely unknown. Monitoring for heavy metals, brominated flame-retardants (BFRs), and pharmaceuticals is important due to their negative effects on ecosystems. Exposure to these contaminants can have significant endocrine, neurological, and reproductive effects. Contaminants easily travel up the food chain and bioaccumulate in apex predators. As predators with a largely aquatic diet, American mink (Mustela vison) and North American river otter (Lontra canadensis) serve as reliable indicator species of environmental health and the status of contaminants. Analysis of scat from these species is a noninvasive method to measure contaminant levels, and detection dogs from Working Dogs for Conservation (WD4C) have been used to locate these scat samples. To aid in the search of these samples, habitat suitability models were created for mink and otter for the years 2013-2020 and projected to 2040 using the random forest algorithm in the Software for Assisted Habitat Modeling (SAHM). Predictor variable data were acquired from Landsat 8 Operational Land Imager (OLI), Terra Moderate Resolution Imaging Spectroradiometer (MODIS), Global Precipitation Measurement Integrated Multi-satellite Retrievals for GPM (GPM IMERG), Shuttle Radar Topography Mission (SRTM), and Soil Moisture Active Passive (SMAP). Within these models, the most important variable for mink and otter habitat was distance to river. Suitable habitat also corresponded with emergent herbaceous land cover and deeper river locations. These habitat suitability models will inform sampling site section for further contaminant analysis.
Mapping the Extent of Mangrove Ecosystem Degradation by Integrating an Ecological Conceptual Model with Satellite Data
Anthropogenic and natural disturbances can cause degradation of ecosystems, reducing their capacity to sustain biodiversity and provide ecosystem services. Understanding the extent of ecosystem degradation is critical for estimating risks to ecosystems, yet there are few existing methods to map degradation at the ecosystem scale and none using freely available satellite data for mangrove ecosystems. In this study, we developed a quantitative classification model of mangrove ecosystem degradation using freely available earth observation data. Crucially, a conceptual model of mangrove ecosystem degradation was established to identify suitable remote sensing variables that support the quantitative classification model, bridging the gap between satellite-derived variables and ecosystem degradation with explicit ecological links. We applied our degradation model to two case-studies, the mangroves of Rakhine State, Myanmar, which are severely threatened by anthropogenic disturbances, and Shark River within the Everglades National Park, USA, which is periodically disturbed by severe tropical storms. Our model suggested that 40% (597 km2) of the extent of mangroves in Rakhine showed evidence of degradation. In the Everglades, the model suggested that the extent of degraded mangrove forest increased from 5.1% to 97.4% following the Category 4 Hurricane Irma in 2017. Quantitative accuracy assessments indicated the model achieved overall accuracies of 77.6% and 79.1% for the Rakhine and the Everglades, respectively. We highlight that using an ecological conceptual model as the basis for building quantitative classification models to estimate the extent of ecosystem degradation ensures the ecological relevance of the classification models. Our developed method enables researchers to move beyond only mapping ecosystem distribution to condition and degradation as well. These results can help support ecosystem risk assessments, natural capital accounting, and restoration planning and provide quantitative estimates of ecosystem degradation for new global biodiversity targets.
Directional reflectance distributions of a hardwood and pine forest canopy
A hand-held radiometer with AVHRR bands 1 and 2 was used to measure the directional reflectance distributions for both a hardwood and a pine forest canopy from a helicopter platform; canopy characteristics were also measured on the ground. The reflectance distributions obtained are compared with the scattering behavior of agricultural and natural grassland canopies. In addition, the Kimes (1983) three-dimensional radiative transfer model is used to document the unique radiant transfers that occur in forest canopies in virtue of their geometric structure. Both the measurements and the model calculations show that dense forest canopy scattering is similar to that for crops and grasslands. Attention is given to the effects of sparse forest canopies.
Modeling the Effects of Global Change on Ecosystem Processes in a Tropical Rainforest
Ongoing land-use change and climate change in wet tropical forests can potentially drive shifts in tree species composition, representing a change in individual species within a functional group, tropical evergreen trees. The impacts on the global carbon cycle are potentially large, but unclear. We explored the differential effects of species within this functional group, in comparison with the effects of climate change, using the Century model as a research tool. Simulating effects of individual tree species on biome-level biogeochemical cycles constituted a novel application for Century. Background and Objectives: A unique, long-term, replicated field experiment containing five evergreen tree species in monodominant stands under similar environmental conditions in a Costa Rican wet forest provided data for model evaluation. Our objectives were to gain insights about this forest’s biogeochemical cycles and effects of tree species within this functional group, in comparison with climate change. Materials and Methods: We calibrated Century, using long-term meteorological, soil, and plant data from the field-based experiment. In modeling experiments, we evaluated effects on forest biogeochemistry of eight plant traits that were both observed and modeled. Climate-change simulation experiments represented two climate-change aspects observed in this region. Results: Model calibration revealed that unmodeled soil processes would be required to sustain observed P budgets. In species-traits experiments, three separate plant traits (leaf death rate, leaf C:N, and allocation to fine roots) resulted in modeled biomass C stock changes of >50%, compared with a maximum 21% change in the climate-change experiments. Conclusions: Modeled ecosystem properties and processes in Century were sensitive to changes in plant traits and nutrient limitations to productivity. Realistic model output was attainable for some species, but unusual plant traits thwarted predictions for one species. Including more plant traits and soil processes could increase realism, but less-complex models provide an accessible means for exploring plant-soil-atmosphere interactions.
Northern Rockies Ecological Conservation: Leveraging Earth Observations to Monitor and Predict Populations of Federally Threatened Whitebark Pine (Pinus albicaulis) across the Intermountain West
Whitebark pine (WBP; Pinus albicaulis) is an ecologically important species in North America. As a federally listed threatened species, an understanding of WBP habitat, distribution, and health is important for the natural resource managers of the National Park Service, United States Forest Service, Bureau of Land Management, Fish and Wildlife Service, and non-profit organizations such as the Whitebark Pine Ecosystem Foundation. Previous attempts to develop models of WBP habitat suitability and distribution lack confidence in their validity and integrity for these organizations. The updated models of habitat suitability and distribution developed by this study would provide managers with a capability to be employed in the conservation and future research direction for WBP. Thus, we developed a habitat suitability model of WBP at a high spatial resolution (Landsat 9 Operational Land Image-2, National Land Cover Database, NASA Shuttle Radar Topography Mission; 30m pixels) using a generalized logistic regression with an area under the curve value of 0.754. We extracted spectral reflectance signatures from overlapped ground sample points and Sentinel-2 Multispectral Instrument. The spectral signature analysis indicates WBP is separable from other tree species. We also utilized a visual validation approach and random forest (RF) modeling to separate WBP from limber pine. Through visual validation the RF classifier successfully identified 8out of 10 WBP trees gathered through ground truth points. Additionally, we achieved an overall accuracy of 91%in our confusion matrix for the distribution model using a dependent validation approach. The derived products from this study allow project partners to assess current suitable habitat and apparent health status in areas of identified WBP occurrence, providing data to aid future research regarding WBP health.
Critical Load Exceedances for North America and Europe using an Ensemble of Models and an Investigation of Causes for Environmental Impact Estimate Variability: An AQMEII4 Study
Exceedances of critical loads for deposition of sulphur (S) and nitrogen (N) to different ecosystems were estimated using European and North American ensembles of air quality models, under Phase 4 of the Air Quality Model Evaluation International Initiative (AQMEII4), to identify where risk of ecosystem harm is expected to occur based on model deposition estimates. The ensembles were driven by common emissions and lateral boundary condition inputs. Model output was regridded to common North American and Europe 0.125° resolution domains, which were then used to calculate critical load exceedances. New, targeted deposition diagnostics implemented in AQMEII4 allowed an unprecedented level of post-simulation analysis to be carried out and facilitated the identification of specific causes of model-to-model variability in critical load exceedance estimates. New datasets for North American critical loads for acidity for forest soil water and aquatic ecosystems were combined with the ensemble deposition predictions to show a substantial decrease in the area and number of locations in exceedance between 2010 and 2016 (forest soils: 13.2 % to 6.1 %; aquatic ecosystems: 21.2 % to 11.4 %). All models agreed in the direction of the ensemble exceedance change between 2010 and 2016. The North American ensemble also predicted a decrease in both severity and total area in exceedance between the years 2010 and 2016 for eutrophication-impacted ecosystems in the USA (sensitive epiphytic lichen: 81.5 % to 75.8 %). The exceedances for herbaceous community richness also decreased between 2010 and 2016, from 13.9 % to 3.9 %. The uncertainty associated with the North American eutrophication results is high; there were sharp differences between the models in both predictions of total N deposition and the change in N deposition, and hence in the predicted eutrophication exceedances between the two years. The European ensemble was used to predict relatively static exceedances of critical loads with respect to acidification (4.48 % to 4.32 % from 2009 to 2010) while eutrophication exceedance increased slightly (60.2 % to 62.2 %). While most models showed the same changes in critical load exceedances as the ensemble between the two years, the spatial extent and magnitude of exceedances varied significantly between the models. The reasons for this variation were examined in detail by first ranking the relative contribution of different sources of sulphur and nitrogen deposition in terms of deposited mass and model-to-model variability in that deposited mass, followed by their analysis using AQMEII4 diagnostics, along with evaluation of the most recent literature. All models in both the North American and European ensembles had net annual negative biases with respect to observed wet deposition of sulphate, nitrate and ammonium. Diagnostics and recent literature suggest that this bias may stem from insufficient cloud scavenging of aerosols and gases, and may be improved through the incorporation of multiphase hydrometeor scavenging within the modelling frameworks. The inability of North American models to predict the timing of the seasonal peak in wet ammonium ion deposition (observed maximum was in April, while all models predicted a June maximum) may also relate to the need for multiphase hydrometeor scavenging (absence of snow scavenging in all models employed here). High variability in the relative importance of particulate sulphate, nitrate and ammonium deposition fluxes between models was linked to the use of updated particle dry deposition parameterizations in some models. However, recent literature and further development of some of the models within the ensemble suggests these particulate biases may also be ameliorated via the incorporation of multiphase hydrometeor scavenging. Annual sulphur and nitrogen deposition prediction variability was linked to SO 2 and HNO 3 dry deposition parameterizations, and diagnostic analysis showed that the cuticle and soil deposition pathways dominate the deposition mass flux of these species. Further work improving parameterizations for these deposition pathways should reduce variability in model acidifying gas deposition estimates. The absence of base cation chemistry in some models was shown to be a major factor in positive biases in fine mode particulate ammonium and particle nitrate concentrations. Models employing ammonia bidirectional fluxes had both the largest and the smallest magnitude biases, depending on the model and bidirectional flux algorithm employed. A careful analysis of bidirectional flux models suggests that those with poor NH 3 performance may underestimate the extent of NH 3 emissions fluxes from forested areas. Based on these results, an increased process-research focus is therefore recommended for the following model processes and on observations which may assist in model evaluation and improvement: multiphase hydrometeor scavenging combined with updated particle dry deposition, cuticle and soil deposition pathway algorithms for acidifying gases, base cation chemistry and emissions, and NH 3 bidirectional fluxes. Comparisons with satellite observations suggest that oceanic NH 3 emissions sources should be included in regional chemical transport models. The choice of land use database employed within any given model was shown to significantly influence deposition totals in several instances, and employing a common land use database across chemical transport models and critical load calculations is recommended for future work.
Machine Learning for Improving Surface-Layer-Flux Estimates
Abstract Flows in the atmospheric boundary layer are turbulent, characterized by a large Reynolds number, the existence of a roughness sublayer and the absence of a well-defined viscous layer. Exchanges with the surface are therefore dominated by turbulent fluxes. In numerical models for atmospheric flows, turbulent fluxes must be specified at the surface; however, surface fluxes are not known a priori and therefore must be parametrized. Atmospheric flow models, including global circulation, limited area models, and large-eddy simulation, employ Monin–Obukhov similarity theory (MOST) to parametrize surface fluxes. The MOST approach is a semi-empirical formulation that accounts for atmospheric stability effects through universal stability functions. The stability functions are determined based on limited observations using simple regression as a function of the non-dimensional stability parameter representing a ratio of distance from the surface and the Obukhov length scale (Obukhov in Trudy Inst Theor Geofiz AN SSSR 1:95–115, 1946), $$z/L$$ z / L . However, simple regression cannot capture the relationship between governing parameters and surface-layer structure under the wide range of conditions to which MOST is commonly applied. We therefore develop, train, and test two machine-learning models, an artificial neural network (ANN) and random forest (RF), to estimate surface fluxes of momentum, sensible heat, and moisture based on surface and near-surface observations. To train and test these machine-learning algorithms, we use several years of observations from the Cabauw mast in the Netherlands and from the National Oceanic and Atmospheric Administration’s Field Research Division tower in Idaho. The RF and ANN models outperform MOST. Even when we train the RF and ANN on one set of data and apply them to the second set, they provide more accurate estimates of all of the fluxes compared to MOST. Estimates of sensible heat and moisture fluxes are significantly improved, and model interpretability techniques highlight the logical physical relationships we expect in surface-layer processes.
Predictive models of long COVID
Background: The cause and symptoms of long COVID are poorly understood. It is challenging to predict whether a given COVID-19 patient will develop long COVID in the future. Methods: We used electronic health record (EHR) data from the National COVID Cohort Collaborative to predict the incidence of long COVID. We trained two machine learning (ML) models — logistic regression (LR) and random forest (RF). Features used to train predictors included symptoms and drugs ordered during acute infection, measures of COVID-19 treatment, pre-COVID comorbidities, and demographic information. We assigned the ‘long COVID’ label to patients diagnosed with the U09.9 ICD10-CM code. The cohorts included patients with (a) EHRs reported from data partners using U09.9 ICD10-CM code and (b) at least one EHR in each feature category. We analysed three cohorts: all patients (n = 2,190,579; diagnosed with long COVID = 17,036), inpatients (149,319; 3,295), and outpatients (2,041,260; 13,741). Findings: LR and RF models yielded median AUROC of 0.76 and 0.75, respectively. Ablation study revealed that drugs had the highest influence on the prediction task. The SHAP method identified age, gender, cough, fatigue, albuterol, obesity, diabetes, and chronic lung disease as explanatory features. Models trained on data from one N3C partner and tested on data from the other partners had average AUROC of 0.75. Interpretation: ML-based classification using EHR information from the acute infection period is effective in predicting long COVID. SHAP methods identified important features for prediction. Cross-site analysis demonstrated the generalizability of the proposed methodology.