The parameterization of moist convection contributes to uncertainty in climate modeling and numerical weather prediction. Machine learning (ML) can be used to learn new parameterizations directly from high-resolution model output, but it remains poorly understood how such parameterizations behave when fully coupled in a general circulation model (GCM) and whether they are useful for simulations of climate change or extreme events. Here we focus on these issues using idealized tests in which an ML-based parameterization is trained on output from a conventional parameterization and its performance is assessed in simulations with a GCM. We use an ensemble of decision trees (random forest) as the ML algorithm, and this has the advantage that it automatically ensures conservation of energy and nonnegativity of surface precipitation. The GCM with the ML convective parameterization runs stably and accurately captures important climate statistics including precipitation extremes without the need for special training on extremes. Climate change between a control climate and a warm climate is not captured if the ML parameterization is only trained on the control climate, but it is captured if the training includes samples from both climates. Remarkably, climate change is also captured when training only on the warm climate, and this is because the extratropics of the warm climate provides training samples for the tropics of the control climate. In addition to being potentially useful for the simulation of climate, we show that ML parameterizations can be interrogated to provide diagnostics of the interaction between convection and the large-scale environment. Plain Language SummarySmall-scale features such as clouds are typically represented in climate models by simplified physical models, and these simplified models introduce errors and uncertainties. A promising alternative approach is to use machine learning to train a statistical model to represent small-scale processes based on output from expensive physics-based models that better represent the small-scale processes. Here we use idealized tests to explore the implications of incorporating a machine-learning model of atmospheric convection in a climate model. We find that such an approach can give accurate simulations of mean climate and heavy rainfall events. The machine-learning model does not work well for global warming if it is only trained on the current climate. However, it does work well for global warming if trained on both the current and warmer climates, and it works surprisingly well if only trained on the warmer climate. We also show that the machine-learning model can be used to better understand the underlying physical processes.
We report on an all-sky search with the LIGO detectors for periodic gravitational waves in the frequency range 50 -1000 Hz and with the frequency's time derivative in the range ÿ1 10 ÿ8 Hz s ÿ1 to zero. Data from the fourth LIGO science run (S4) have been used in this search. Three different semicoherent methods of transforming and summing strain power from short Fourier transforms (SFTs) of the calibrated data have been used. The first, known as StackSlide, averages normalized power from each SFT. A ''weighted Hough'' scheme is also developed and used, which also allows for a multiinterferometer search. The third method, known as PowerFlux, is a variant of the StackSlide method in which the power is weighted before summing. In both the weighted Hough and PowerFlux methods, the weights are chosen according to the noise and detector antenna-pattern to maximize the signal-to-noise ratio. The respective advantages and disadvantages of these methods are discussed. Observing no evidence of periodic gravitational radiation, we report upper limits; we interpret these as limits on this radiation from isolated rotating neutron stars. The best population-based upper limit with 95% confidence on the gravitational-wave strain amplitude, found for simulated sources distributed isotropically across the sky and with isotropically distributed spin axes, is 4:28 10 ÿ24 (near 140 Hz). Strict upper limits are also obtained for small patches on the sky for best-case and worst-case inclinations of the spin axes.
We carry out two searches for periodic gravitational waves using the most sensitive few hours of data from the second LIGO science run. Both searches exploit fully coherent matched filtering and cover wide areas of parameter space, an innovation over previous analyses which requires considerable algorithm development and computational power. The first search is targeted at isolated, previously unknown neutron stars, covers the entire sky in the frequency band 160 -728.8 Hz, and assumes a frequency derivative of less than 4 10 ÿ10 Hz=s. The second search targets the accreting neutron star in the lowmass x-ray binary Scorpius X-1 and covers the frequency bands 464-484 Hz and 604-624 Hz as well as the two relevant binary orbit parameters. Because of the high computational cost of these searches we limit the analyses to the most sensitive 10 hours and 6 hours of data, respectively. Given the limited sensitivity and duration of the analyzed data set, we do not attempt deep follow-up studies. Rather we concentrate on demonstrating the data analysis method on a real data set and present our results as upper limits over large volumes of the parameter space. In order to achieve this, we look for coincidences in parameter space between the Livingston and Hanford 4-km interferometers. For isolated neutron stars our 95% confidence level upper limits on the gravitational wave strain amplitude range from 6:6 10 ÿ23 to 1 10 ÿ21 across the frequency band; for Scorpius X-1 they range from 1:7 10 ÿ22 to 1:3 10 ÿ21 across the two 20-Hz frequency bands. The upper limits presented in this paper are the first broadband wide parameter space upper limits on periodic gravitational waves from coherent search techniques. The methods developed here lay the foundations for upcoming hierarchical searches of more sensitive data which may detect astrophysical signals.
We analyzed the available LIGO data coincident with GRB 070201, a short-duration, hard-spectrum gamma-ray burst (GRB) whose electromagnetically determined sky position is coincident with the spiral arms of the Andromeda galaxy (M31). Possible progenitors of such short, hard GRBs include mergers of neutron stars or a neutron star and a black hole, or soft gamma-ray repeater (SGR) flares. These events can be accompanied by gravitational-wave emission. No plausible gravitational-wave candidates were found within a 180 s long window around the time of GRB 070201. This result implies that a compact binary progenitor of GRB 070201, with masses in the range 1 Msun < m1 < 3 Msun and 1 Msun < m2 < 40 Msun, located in M31 is excluded at >99% confidence. If the GRB 070201 progenitor was not in M31, then we can exclude a binary neutron star merger progenitor with distance D < 3.5 Mpc, assuming random inclination, at 90% confidence. The result also implies that an unmodeled gravitational-wave burst from GRB 070201 most probably emitted less than 4.4*10^(-4) Msun c^2 (7.9*10^50 ergs) in any 100 ms long period within the signal region if the source was in M31 and radiated isotropically at the same frequency as LIGO’s peak sensitivity (f about 150 Hz). This upper limit does not exclude current models of SGRs at the M31 distance
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.