Search for: All records

Physics-guided machine learning (PGML) has become a prevalent approach in studying scientific systems due to its ability to integrate scientific theories for enhancing machine learning (ML) models. However, most PGML approaches are tailored to isolated and relatively simple tasks, which lim- its their applicability to complex systems involving multiple interacting processes and numerous influencing features. In this paper, we propose a Physics-Guided Foundation Model (PGFM) that combines pre-trained ML models and physics- based models and leverages their complementary strengths to improve the modeling of multiple coupled processes. To effectively conduct pre-training, we construct a simulated en- vironmental system that encompasses a wide range of influ- encing features and various simulated variables generated by physics-based models. The model is pre-trained in this sys- tem to adaptively select important feature interactions guided by multi-task objectives. We then fine-tune the model for each specific task using true observations, while maintaining con- sistency with established physical theories, such as the prin- ciples of mass and energy conservation. We demonstrate the effectiveness of this methodology in modeling water temper- ature and dissolved oxygen dynamics in real-world lakes. The proposed PGFM is also broadly applicable to a range of sci- entific fields where physics-based models are being used. 

Sea salt aerosols contribute significantly to the mass loading of ambient aerosol, which may serve as cloud condensation nuclei and can contribute to light scattering in the atmosphere. Two major chemical components commonly found in sea salts are ammonium sulfate (AS) and sodium chloride (NaCl). It has been shown that alkylamines, derivatives of ammonia, can react with ammonium salts in the particle-phase to displace ammonia and likely change the particle properties. This study investigated the effects of atmospheric alkylamines on the composition and properties of sea salt aerosols using a chemical system of methylamine (MA, as a proxy of alkylamines), AS and NaCl (as a proxy of sea salt aerosol). The concentrations of ammonia and MA in aqueous/gas phases at the thermodynamic equilibrium were determined using the Extended Aerosols and Inorganics Model (E-AIM) under varying initial inputs, along with the deliquescence relative humidity (DRH) and the corresponding particle water content. Our findings indicated a notable negative relationship between MA concentration and the DRH for both AS and NaCl while the effect of MA on NaCl is smaller than that on AS. The salt of MA in the particle phase may absorb water vapor and may lead to the displacement reaction between AS and NaCl due to the low solubility of sodium sulfate. The acidity in the particle phase also played a significant role in affecting the DRH of sea salt aerosols. Since both sea salt aerosol and alkylamines are emitted into the atmosphere from the ocean in large quantities, our study suggested the potential impact of alkylamines on the environment and the climate via the modification of sea salt aerosol properties. 

Recent research in atmospheric chemistry suggested that gaseous amines may rapidly react with the acidic components in the aerosol to be incorporated in the particle phase. However, laboratory experiments suggested that these heterogeneous processes may be sensitive to the reaction conditions, such as relative humidity (RH), the initial aerosol acidity and the initial concentration of gaseous ammonia which is ubiquitous in the atmosphere. We studied the heterogenous reactions between several amines and ammonium sulfate using a series of thermodynamic simulations under varying initial conditions, including RH, particle-phase acidity and gaseous amine and ammonia concentrations. Several distinctively different trends in the particle-phase ammonium, amines and water content were observed, depending significantly on the particle-phase acidity and the initial amine to ammonia mole ratio. One notable observation was that alkylamines may facilitate the water uptake of ammonium sulfate even in the presence of 1000 times more ammonia gas. Such change in aerosol water content may alter the surface tension, uptake coefficient and could formation properties of aerosol and influence the radiative forcing of the particles. 

Data transformations (e.g. rotations, reflections,and cropping) play an important role in self supervised learning. Typically, images are transformed into different views, and neural networks trained on tasks involving these views produce useful feature representations for downstream tasks, including anomaly detection. However, for anomaly detection beyond image data, it is often unclear which transformations to use. Here we present a simple end-to-end procedure for anomaly detection with learnable transformations. The key idea is to embed the transformed data into a semantic space such that the transformed data still resemble their untransformed form, while different transformations are easily distinguishable. Extensive experiments on time series show that our proposed method outperforms existing approaches in the one-vs.-rest setting and is competitive in the more challenging n-vs.-rest anomaly detection task. On medical and cyber-security tabular data, our method learns domain-specific transformations and detects anomalies more accurately than previous work.