Sherrie Wang

Minerals play a critical role in the advanced energy technologies necessary for decarbonization, but characterizing mineral deposits hidden underground remains costly and challenging. Inspired by recent progress in generative modeling, we develop a learning method which infers the locations of minerals by masking and infilling geospatial maps of resource availability. We demonstrate this technique using mineral data for the conterminous United States, and train performant models, with the best achieving Dice coefficients of 0.31 ± 0.01 and recalls of 0.22 ± 0.02 on test data at 1×1 sq mi spatial resolution. One major advantage of our approach is that it can easily incorporate auxiliary data sources for prediction which may be more abundant than mineral data. We highlight the capabilities of our model by adding input layers derived from geophysical sources, along with a nation-wide ground survey of soils originally intended for agronomic purposes. We find that employing such auxiliary features can improve inference performance, while also enabling model evaluation in regions with no recorded minerals.

<strong class="journal-contentHeaderColor">Abstract.</strong> <span>We present the first global, data-driven analysis of power plant NO</span><span>₂ </span><span>plume visibility from space. Using TROPOMI observations over 6,000 of the world&rsquo;s highest-emitting power plants and hourly CEMS data for 500 U.S. plants, we develop an automated algorithm that labels plumes and attributes them to their sources with 98 % accuracy. We then train a machine learning model to predict plume detectability from environmental, meteorological, and observational variables (F1 score &gt; 0.65, AUC &gt; 0.8). Out of 25 variables, we find that NO</span><span><em>_x</em> </span><span>emission rate, surface albedo, wind speed, and sensor zenith angle jointly explain much of the detection variability. An hourly NO</span><span>_<em>x</em> </span><span>emission rate of &asymp; 400 kg/h corresponds to a 50 % detection probability on average, but detection rates vary from &lt; 20 % to &gt; 60 % under different combinations of these conditions. These results provide the first empirical quantification of the physical and environmental factors that govern NO</span><span>₂ </span><span>plume visibility in satellite data, establishing a foundation for models to use similar predictors as auxiliary variables when quantifying emission rates from plume appearance.</span>

Near-surface atmospheric conditions can differ sharply over tens to hundreds of meters due to land cover and topography, yet this variability is absent from current weather analyses and forecasts. It is unclear whether such meter-scale variability reflects irreducibly chaotic dynamics or contains a component predictable from surface characteristics and large-scale atmospheric forcing. Here we show that a substantial, physically coherent component of meter-scale near-surface weather is statistically recoverable from existing observations. By conditioning coarse atmospheric state on sparse surface station measurements and high-resolution Earth observation data, we infer spatially continuous fields of near-surface wind, temperature, and humidity at 10 m resolution across the contiguous United States. Relative to ERA5, the inferred fields reduce wind error by 29% and temperature and dewpoint error by 6%, while explaining substantially more spatial variance at fixed time steps. They also exhibit physically interpretable structure, including urban heat islands, evapotranspiration-driven humidity contrasts, and wind speed differences across land cover types. Our findings expand the frontier of weather modeling by demonstrating a computationally feasible approach to continental-scale meter-resolution inference. More broadly, they illustrate how conditioning coarse dynamical models on static fine-scale features can reveal previously unresolved components of the Earth system.

Near-surface atmospheric conditions can differ sharply over tens to hundreds of meters due to land cover and topography, yet this variability is absent from current weather analyses and forecasts. It is unclear whether such meter-scale variability reflects irreducibly chaotic dynamics or contains a component predictable from surface characteristics and large-scale atmospheric forcing. Here we show that a substantial, physically coherent component of meter-scale near-surface weather is statistically recoverable from existing observations. By conditioning coarse atmospheric state on sparse surface station measurements and high-resolution Earth observation data, we infer spatially continuous fields of near-surface wind, temperature, and humidity at 10 m resolution across the contiguous United States. Relative to ERA5, the inferred fields reduce wind error by 29% and temperature and dewpoint error by 6%, while explaining substantially more spatial variance at fixed time steps. They also exhibit physically interpretable structure, including urban heat islands, evapotranspiration-driven humidity contrasts, and wind speed differences across land cover types. Our findings expand the frontier of weather modeling by demonstrating a computationally feasible approach to continental-scale meter-resolution inference. More broadly, they illustrate how conditioning coarse dynamical models on static fine-scale features can reveal previously unresolved components of the Earth system.

Accurate estimation of surface soil moisture (SM) in terrestrial ecosystems is essential for understanding hydroclimate dynamics. The L-band Soil Moisture Active Passive (SMAP) mission provides 9-km global daily surface SM by using a microwave radiative transfer model (RTM)-based algorithm. However, the accuracy of SMAP SM is limited in regions with dense vegetation cover and complex surface conditions, due to the empirical parameterization and oversimplified radiative transfer processes. To overcome the limitations, we developed a Process-Guided Machine Learning (PGML) framework to integrate RTM theories and deep learning to predict global daily surface 9-km SM from April 2015 to June 2025. Informed by domain knowledge, we developed the PGML model structure using RTM and hydrological theories, designed a Kling-Gupta efficiency-based cost function, pretrained it with RTM simulations, and fine-tuned it with in-situ measurements. The independent validation shows that PGML SM has strong agreement with in-situ measurements (R = 0.868 and unbiased RMSE = 0.054 m3/m3). This study highlights the potential of PGML to enhance the accuracy of satellite SM, thereby supporting improved water resources and ecosystem management.

ERA5 and MADIS (ground stations) curated weather data for CONUS (2020-2023) This repo contains curated gridded (ERA5) and station (MADIS) weather data for CONUS for the years 2020-2023. The following data is available:- Shapefile of CONUS.- Shapefile containing the location and number of observations (2020-2023) of the MADIS stations- Processed hourly averaged [MADIS](https://madis.ncep.noaa.gov/) data for CONUS (2020-2023)- [ERA5](https://confluence.ecmwf.int/display/CKB/ERA5%3A+data+documentation) re-analysis data for CONUS (2020-2023), gridded For MADIS and ERA5, the following variables are available:- u and v component of wind vector at 10 meters above ground- temperature at 2 meters above ground- dewpoint at 2 meters above ground

ERA5 and MADIS (ground stations) curated weather data for CONUS (2020-2023) This repo contains curated gridded (ERA5) and station (MADIS) weather data for CONUS for the years 2020-2023. The following data is available:- Shapefile of CONUS.- Shapefile containing the location and number of observations (2020-2023) of the MADIS stations- Processed hourly averaged [MADIS](https://madis.ncep.noaa.gov/) data for CONUS (2020-2023)- [ERA5](https://confluence.ecmwf.int/display/CKB/ERA5%3A+data+documentation) re-analysis data for CONUS (2020-2023), gridded For MADIS and ERA5, the following variables are available:- u and v component of wind vector at 10 meters above ground- temperature at 2 meters above ground- dewpoint at 2 meters above ground

Abstract. We present the first global, data-driven analysis of power plant NO2 plume visibility from space. Using TROPOMI observations over 6,000 of the world’s highest-emitting power plants and hourly CEMS data for 500 U.S. plants, we develop an automated algorithm that labels plumes and attributes them to their sources with 98 % accuracy. We then train a machine learning model to predict plume detectability from environmental, meteorological, and observational variables (F1 score &gt; 0.65, AUC &gt; 0.8). Out of 25 variables, we find that NOx emission rate, surface albedo, wind speed, and sensor zenith angle jointly explain much of the detection variability. An hourly NOx emission rate of ≈ 400 kg/h corresponds to a 50 % detection probability on average, but detection rates vary from &lt; 20 % to &gt; 60 % under different combinations of these conditions. These results provide the first empirical quantification of the physical and environmental factors that govern NO2 plume visibility in satellite data, establishing a foundation for models to use similar predictors as auxiliary variables when quantifying emission rates from plume appearance.

Accurate estimation of surface soil moisture (SM) in terrestrial ecosystems is essential for understanding hydroclimate dynamics. The L-band Soil Moisture Active Passive (SMAP) mission provides 9-km global daily surface SM by using a microwave radiative transfer model (RTM)-based algorithm. However, the accuracy of SMAP SM is limited in regions with dense vegetation cover and complex surface conditions, due to the empirical parameterization and oversimplified radiative transfer processes. To overcome the limitations, we developed a Process-Guided Machine Learning (PGML) framework to integrate RTM theories and deep learning to predict global daily surface 9-km SM from April 2015 to June 2025. Informed by domain knowledge, we developed the PGML model structure using RTM and hydrological theories, designed a Kling-Gupta efficiency-based cost function, pretrained it with RTM simulations, and fine-tuned it with in-situ measurements. The independent validation shows that PGML SM has strong agreement with in-situ measurements (R = 0.868 and unbiased RMSE = 0.054 m 3 /m 3 ). This study highlights the potential of PGML to enhance the accuracy of satellite SM, thereby supporting improved water resources and ecosystem management.