Sam J. Silva’s research while affiliated with University of Southern California and other places

Neural network emulators have become an invaluable tool for climate prediction tasks but do not have an inherent ability to produce equitable predictions (e.g., predictions which are equally accurate across different regions or groups of people). This motivates the need for explicit internal representations of fairness. To that end, we draw on methods for enforcing physical constraints in emulators and propose a custom loss function which punishes predictions of unequal quality across any prespecified regions or category, here defined using Human Development Index. This loss function weighs a standard error metric against another which captures inequity between groups, allowing us to adjust the priority of each. Our results show that emulators trained with our loss function provide more equitable predictions. We empirically demonstrate that an appropriate selection of an equity priority can minimize loss of performance, mitigating the tradeoff between accuracy and equity.

Chemical cycling drives the production and loss of many important atmospheric constituents. The speed of atmospheric chemical cycling is a particularly valuable indicator for characterizing and measuring the effects of such cycles on oxidant chemistry, air quality, and climate. Here, we apply graph theoretical methods to explicitly quantify and analyze the characteristic timescales of gas‐phase chemical cycles in the troposphere and stratosphere, as simulated by the GEOS‐Chem chemical mechanism. We identify all two‐, three‐, and four‐reaction cycles in the mechanism and calculate a characteristic timescale for each individual cycle. We find that the speed of chemical cycling varies by orders of magnitude at any given location but tends to be faster in urban‐ and biogenically‐dominated chemical regions, and slower during the night. We further quantify the fraction of cycling that contains a rate‐determining step, and explicitly demonstrate the large potential for mechanisms to recycle oxidants like OH.

Background Time-stratified case-crossover (CC) and Poisson time series (TS) are two popular methods for relating acute health outcomes to time-varying ubiquitous environmental exposures. Our aim is to compare the performance of these methods in estimating associations with rare, extreme heat exposures and mortality—an increasingly relevant exposure in our changing climate. Methods Daily mortality data were simulated in various scenarios similar to observed Los Angeles County data from 2014 to 2019 (N = 367,712 deaths). We treated observed temperature as either a continuous or dichotomized variable and controlled for day of week and a smooth function of time. Five temperature dichotomization cutoffs between the 80th and 99th percentile were chosen to investigate the effects of extreme heat events. In each of 10,000 simulations, the CC and several TS models with varying degrees of freedom for time were fit to the data. We reported bias, variance, and relative efficiency (ratio of variance for a “reference” TS method to variance of another method) of temperature association estimates. Results CC estimates had larger uncertainty than TS methods, with the relative efficiency of CC ranging from 91% under the 80th percentile cutoff to 80% under the 99th percentile cutoff. As previously reported, methods best capturing data-generating time trends generally had the least bias. Additionally, TS estimates for observed Los Angeles data were larger with less uncertainty. Conclusions We provided new evidence that, compared with TS, CC has increasingly poor efficiency for rarer exposures in ecological study settings with shared, regional exposures, regardless of underlying time trends. Analysts should consider these results when applying either TS or CC methods.

Computational models of atmospheric composition are not always physically consistent. For example, not all models respect fundamental conservation laws such as conservation of atoms in an interconnected chemical system. In well performing models, these unphysical deviations are often ignored because they are frequently minor, and thus only need a small nudge to perfectly conserve mass. Here we introduce a method that anchors a prediction from any numerical model to physically consistent hard constraints, nudging concentrations to the nearest solution that respects the conservation laws. This closed-form model-agnostic correction uses a single matrix operation to minimally perturb the predicted concentrations to ensure that atoms are conserved to machine precision. To demonstrate this approach, we train a gradient boosting decision tree ensemble to emulate a small reference model of ozone photochemistry and test the effect of the correction on accurate but nonconservative predictions. The nudging approach minimally perturbs the already well-predicted results for most species, but decreases the accuracy of important oxidants, including radicals. We develop a weighted extension of this nudging approach that considers the uncertainty and magnitude of each species in the correction. This species-level weighting approach is essential to accurately predict important low concentration species such as radicals. We find that applying the species-weighted correction slightly improves overall accuracy by nudging unphysical predictions to a more likely mass-conserving solution.

Atmospheric chemical reactions play an important role in air quality and climate change. While the structure and dynamics of individual chemical reactions are fairly well understood, the emergent properties of the entire atmospheric chemical system, which can involve many different species that participate in many different reactions, are not well described. In this work, we leverage graph-theoretic techniques to characterize patterns of interaction (“motifs”) in three different representations of gas-phase atmospheric chemistry, termed “chemical mechanisms.” These widely used mechanisms, the master chemical mechanism, the GEOS-Chem mechanism, and the Super-Fast mechanism, vary dramatically in scale and application, but they all generally aim to simulate the abundance and variability of chemical species in the atmosphere. This motif analysis quantifies the fundamental patterns of interaction within the mechanisms, which are directly related to their construction. For example, the gas-phase chemistry in the very small Super-Fast mechanism is entirely composed of bimolecular reactions, and its motif distribution matches that of an individual bimolecular reaction well. The larger and more complex mechanisms show emergent motif distributions that differ strongly from any specific reaction type, consistent with their complexity. The proposed motif analysis demonstrates that while these mechanisms all have a similar design goal, their higher-order structure of interactions differs strongly and thus provides a novel set of tools for exploring differences across chemical mechanisms.

A substantial portion of tropospheric O3 dry deposition occurs after diffusion of O3 through plant stomata. Simulating stomatal uptake of O3 in 3D atmospheric chemistry models is important in the face of increasing drought induced declines in stomatal conductance and enhanced ambient O3. Here, we present a comparison of the stomatal component of O3 dry deposition (egs) from chemical transport models and estimates of egs from observed CO2, latent heat, and O3 flux. The dry deposition schemes were configured as single-point models forced with data collected at flux towers. We conducted sensitivity analyses to study the impact of model parameters that control stomatal moisture stress on modeled egs. Examining six sites around the northern hemisphere, we find that the seasonality of observed flux-based egs agrees with the seasonality of simulated egs at times during the growing season with disagreements occurring during the later part of the growing season at some sites. We find that modeled water stress effects are too strong in a temperate-boreal transition forest. Some single-point models overestimate summertime egs in a seasonally water-limited Mediterranean shrubland. At all sites examined, modeled egs was sensitive to parameters that control the vapor pressure deficit stress. At specific sites that experienced substantial declines in soil moisture, the simulation of egs was highly sensitive to parameters that control the soil moisture stress. The findings demonstrate the challenges in accurately representing the effects of moisture stress on the stomatal sink of O3 during observed increases in dryness due to ecosystem specific plant-resource interactions.

Computational models of atmospheric composition are not always physically consistent. For example, not all models respect fundamental conservation laws such as conservation of atoms in an interconnected chemical system. In well performing models, these nonphysical deviations are often ignored because they are frequently minor, and thus only need a small nudge to perfectly conserve mass. Here we introduce a method that anchors a prediction from any numerical model to physically consistent hard constraints, nudging concentrations to the nearest solution that respects the conservation laws. This closed-form model-agnostic correction uses a single matrix operation to minimally perturb the predicted concentrations to ensure that atoms are conserved to machine precision. To demonstrate this approach, we train a gradient boosting decision tree ensemble to emulate a small reference model of ozone photochemistry and test the effect of the correction on accurate but non-conservative predictions. The nudging approach minimally perturbs the already well-predicted results for most species, but decreases the accuracy of important oxidants, including radicals. We develop a weighted extension of this nudging approach that considers the uncertainty and magnitude of each species in the correction. This species-level weighting approach is essential to accurately predict important low concentration species such as radicals. We find that applying the uncertainty-weighted correction to the nonphysical predictions slightly improves overall accuracy, by nudging the predictions to a more likely mass-conserving solution.