Tropical cyclones in Earth's atmosphere form due to a combination of warm ocean waters, high atmospheric humidity, persistent deep convection, and preexisting low-level vorticity. Although Saturn's largest moon, Titan, has a global Earth-like tropical climate with a volatile methane cycle and sufficient convective activity to power the energetics of a mature cyclone, no such storm has yet been observed. Previous work has suggested that Titan's polar lakes could be suitable for tropical cyclone development, based on inferences of low-level absolute vorticity and surface heat fluxes. However, key background variables, such as temperature and specific humidity, differ between the atmospheres of Earth and Titan, making direct comparisons difficult. By systematically varying rotation rate, condensable volatility, and equatorial landmass, we reveal how modest adjustments to planetary parameters can restructure the balance between dynamical forcing, thermodynamic potential, and environmental ventilation. The picture that emerges is one in which Earth-like systems naturally support broad regions of cyclogenesis in the deep tropics, whereas Titan-like systems shift environmental favorability to narrow, poleward bands where vorticity, air-sea disequilibrium, and atmospheric moisture align only episodically. These results not only refine our understanding of Titan's potential for tropical cyclone activity but also highlight the sensitivity of cyclogenesis pathways to planetary-scale controls. More broadly, the framework developed here provides a new foundation for exploring cyclone-favorable climates across a wide range of terrestrial worlds, underscoring that the familiar paradigms of Earth's tropical cyclone behavior may represent just one expression within a much wider spectrum of atmospheric possibilities.

Saturn's largest moon, Titan, has an Earth-like volatile cycle, but with methane playing the role of water and surface liquid reservoirs geographically isolated at high latitudes. We recreate Titan's characteristic dry hydroclimate at the equator of an Earth-like climate model without seasons and with water as the condensable by varying a small set of planetary parameters. We use three observationally motivated criteria for Titan-like conditions at the equator: 1) the peak in surface specific humidity is not at the equator, despite it having the warmest annual-mean temperatures; 2) the vertical profile of specific humidity in the equatorial column is nearly constant through the lower troposphere; and 3) the relative humidity near the surface at the equator is significantly lower than saturation (lower than 60%). We find that simply reducing the available water at the equator does not fully reproduce Titan-like conditions. We additionally vary the rotation period and volatility of water to mimic Titan's slower rotation and more abundant methane vapor. Longer rotation periods coupled with a dry equatorial surface meet fewer of the Titan-like criteria than equivalent experiments with shorter rotation periods. Experiments with higher volatility of water meet more criteria than those with lower volatility, with some of those with the highest volatility meeting all three, demonstrating that an Earth-like planet can display Titan-like climatology by changing only a few physical parameters.

This work presents strategies to learn an Energy-Based Model (EBM) according to the desired length of its MCMC sampling trajectories. MCMC trajectories of different lengths correspond to models with different purposes. Our experiments cover three different trajectory magnitudes and learning outcomes: 1) shortrun sampling for image generation; 2) midrun sampling for classifier-agnostic adversarial defense; and 3) longrun sampling for principled modeling of image probability densities. To achieve these outcomes, we introduce three novel methods of MCMC initialization for negative samples used in Maximum Likelihood (ML) learning. With standard network architectures and an unaltered ML objective, our MCMC initialization methods alone enable significant performance gains across the three applications that we investigate. Our results include state-of-the-art FID scores for unnormalized image densities on the CIFAR-10 and ImageNet datasets; state-of-the-art adversarial defense on CIFAR-10 among purification methods and the first EBM defense on ImageNet; and scalable techniques for learning valid probability densities. Code for this project can be found at https://github.com/point0bar1/ebm-life-cycle.

<p>We recreate a Titan-like climate using an Earth-like global climate model (GCM) by varying a small set of planetary parameters. Understanding the range of possible climate states for Earth-like planets is important for interpreting exoplanet observations and Earth’s own climate history. We find that simply reducing the available water at the surface does not fully reproduce Titan-like conditions. This may indicate that there are many possible “in-between” states an Earth-like planet can have that span the gap between the Earth and Titan climate archetypes. We use three observationally motivated criteria to determine Titan-like conditions: 1) the peak in surface specific humidity is not at the equator, despite it having the warmest annual-mean temperatures (Ádámkovics et al. 2016); 2) the vertical profile of specific humidity in the equatorial column is nearly constant through the lower troposphere (Niemann et al. 2005); and 3) the relative humidity near the surface at the equator is significantly lower than saturation (lower than 60%; Niemann et al. 2005; Tokano et al. 2006). We first limit the available water by placing a continental land strip centered on the equator and varying its width. This mimics Titan’s dry tropics and wet poles, and could be similar to past continental arrangements in Earth’s history. Land strips alone allow some experiments to meet two Titan-like criteria, but none show the near-constant vertical profile of specific humidity. We take three of these land strip widths and vary the rotation period, starting with Earth’s rotation and moving towards Titan’s (16 Earth days). Slowing the rotation results in fewer experiments meeting any of the Titan-like criteria due to increased access to oceanic moisture from the widened Hadley Circulation. For the same three land strip widths and using Earth rotation, we vary the volatility of the condensable via a constant multiplied to the saturation vapor pressure. Titan’s condensable, methane, is more volatile under Titan’s surface conditions than water is on Earth, resulting in high specific humidities. By artificially increasing the saturation vapor pressure, we can approximate this effect without changing the properties of the condensable. Experiments with a volatility constant of 2.5 (the maximum used in this work) meet all three Titan-like criteria, demonstrating that an Earth-like planet can display Titan-like climatology by changing only a few physical parameters.</p>

This work proposes a method for using any generator network as the foundation of an Energy-Based Model (EBM). Our formulation posits that observed images are the sum of unobserved latent variables passed through the generator network and a residual random variable that spans the gap between the generator output and the image manifold. One can then define an EBM that includes the generator as part of its forward pass, which we call the Hat EBM. The model can be trained without inferring the latent variables of the observed data or calculating the generator Jacobian determinant. This enables explicit probabilistic modeling of the output distribution of any type of generator network. Experiments show strong performance of the proposed method on (1) unconditional ImageNet synthesis at 128x128 resolution, (2) refining the output of existing generators, and (3) learning EBMs that incorporate non-probabilistic generators. Code and pretrained models to reproduce our results are available at https://github.com/point0bar1/hat-ebm.

The vulnerability of deep networks to adversarial attacks is a central problem for deep learning from the perspective of both cognition and security. The current most successful defense method is to train a classifier using adversarial images created during learning. Another defense approach involves transformation or purification of the original input to remove adversarial signals before the image is classified. We focus on defending naturally-trained classifiers using Markov Chain Monte Carlo (MCMC) sampling with an Energy-Based Model (EBM) for adversarial purification. In contrast to adversarial training, our approach is intended to secure pre-existing and highly vulnerable classifiers. The memoryless behavior of long-run MCMC sampling will eventually remove adversarial signals, while metastable behavior preserves consistent appearance of MCMC samples after many steps to allow accurate long-run prediction. Balancing these factors can lead to effective purification and robust classification. We evaluate adversarial defense with an EBM using the strongest known attacks against purification. Our contributions are 1) an improved method for training EBM's with realistic long-run MCMC samples, 2) an Expectation-Over-Transformation (EOT) defense that resolves theoretical ambiguities for stochastic defenses and from which the EOT attack naturally follows, and 3) state-of-the-art adversarial defense for naturally-trained classifiers and competitive defense compared to adversarially-trained classifiers on Cifar-10, SVHN, and Cifar-100. Code and pre-trained models are available at https://github.com/point0bar1/ebm-defense.

Essentials Plasma Factor XIII, a heterodimer of A and B subunits FXIIIA2B2, is a transglutaminase enzyme with a well‐established role in haemostasis. Cells of bone marrow and mesenchymal lineage express the FXIII‐A gene (F13A1) that encodes the cellular form of the transglutaminase, a homodimer of the A subunits, FXIII‐A. FXIII‐A was presumed to function intracellularly, however, several lines of evidence now indicate that FXIII‐A is externalised by an as yet unknown mechanism This review describes the mounting evidence that FXIII‐A is a diverse transglutaminase with many intracellular and extracellular substrates that can participate in an array of biological processes

Geomorphic features typically associated with extreme rainfall events in terrestrial settings, including extensive fluvial features and alluvial fans, have been detected on Titan's surface. Methane flow from precipitation on Titan can transport sediments and potentially erode the icy bedrock, but averaged precipitation rates from prior global-scale modelling are too low by at least an order of magnitude to initiate sediment transport of observed grain sizes at low latitudes. Here, we quantify the regional magnitude, frequency and variability of extreme rainfall events from simulations of present-day Titan, with a general circulation model coupled to a land model partially covered by wetlands reservoirs that can capture Titan's regionally varying hydroclimate. We find that the most extreme storms tend to occur in the mid-latitudes, where observed alluvial fans are most concentrated. Storms capable of sediment transport and erosion occur at all latitudes in our simulations, consistent with the observed global coverage of fluvial features. Our results demonstrate the influential role of extreme precipitation in shaping Titan's surface. We therefore suggest that, similarly to Earth but differently from Mars, active geomorphic work may be ongoing in the present climate on Titan.

We search for meridional variation in the abundance of CH3D relative to CH4 on Titan using near-IR spectra obtained with NIRSPAO at Keck, which have a photon-limited signal-to-noise ratio of similar to 50. Our observations can rule out a larger than 10% variation in the column of CH3D below 50 km. The preferential condensation of the heavy isotopologues will fractionate methane by reducing CH3D in the remaining vapor, and therefore these observations place limits on the amount of condensation that occurs in the troposphere. While previous estimates of CH3D fractionation on Titan have estimated an upper limit of -6 parts per thousand, assuming a solid condensate, we consider more recent laboratory data for the equilibrium fractionation over liquid methane, and use a Rayleigh distillation model to calculate fractionation in an ascending parcel of air that is following a moist adiabat We find that deep, precipitating convection can enhance the fractionation of the remaining methane vapor by -10 to -40 parts per thousand, depending on the final temperature of the rising parcel. By relating fractionation of our reference parcel model to the pressure level where the moist adiabat achieves the required temperature, we argue that the measured methane fractionation constrains the outflow level for a deep convective event. Observations with a factor of at least 4-6 times larger signal-to-noise are required to detect this amount of fractionation, depending on the altitude range over which the outflow from deep convection occurs. (C) 2016 Elsevier Inc. All rights reserved.

The mechanisms behind the occurrence of large cloud outbursts and precipitation on Titan have been disputed. A global- and annual-mean estimate of surface fluxes indicated only 1% of the insolation, or similar to 0.04 W m(-2), is exchanged as sensible and/or latent fluxes. Since these fluxes are responsible for driving atmospheric convection, it has been argued that moist convection should be quite rare and precipitation even rarer, even if evaporation globally dominates the surface-atmosphere energy exchange. In contrast, climate simulations indicate substantial cloud formation and/or precipitation. We argue that the top-of-atmosphere (TOA) radiative imbalance is diagnostic of horizontal heat transport by Titan's atmosphere, and thus constrains the strength of the methane cycle. Simple calculations show the TOA radiative imbalance is similar to 0.5-1 W m(-2) in Titan's equatorial region, which implies 2-3 MW of latitudinal heat transport by the atmosphere. Our simulation of Titan's climate suggests this transport may occur primarily as latent heat, with net evaporation at the equator and net accumulation at higher latitudes. Thus, the methane cycle could be 10-20 times previous estimates. Opposing seasonal transport at solstices, compensation by sensible heat transport, and focusing of precipitation by large-scale dynamics could further enhance the local, instantaneous strength of Titan's methane cycle by a factor of several. A limited supply of surface liquids in regions of large surface radiative imbalance may throttle the methane cycle, and if so, we predict more frequent large storms over the lakes district during Titan's northern summer.