Search Results

You are looking at 1 - 10 of 37 items for :

  • Deep learning x
  • Journal of the Atmospheric Sciences x
  • All content x
Clear All
Noah D. Brenowitz, Tom Beucler, Michael Pritchard, and Christopher S. Bretherton

, including a machine-learning parameterization, should capture the dependence of convection to these parameters. One such parameter is the LTS: LTS = θ ⁡ ( 700   hPa ) − SST , where θ is the potential temperature and SST is the sea surface temperature. Low LTS indicates the lower troposphere is conditionally unstable, favoring deep convection. A second controlling parameter is the midtropospheric moisture, defined by Q = ∫ 850 hPa 550 hPa q T   d p g . Cumulus updrafts entrain surrounding air as they

Restricted access
A. K. Showalter

fig. 2 of the Byers and Rodebush reportimply a much deeper on-shore flow than off-shoreflow during a twenty-four hour period. If a value of25 were arbitrarily added to all of the convergenceand divergence values thus displacing the zero lines infig. 2 downward approximately 25 points, the relativedepths of the day and night breeze would appear tobe approximately correct. Most meteorologists arefamiliar with the fact that on-shore breezes arestronger than off-shore land breezes, but my discussion

Full access
Daniel J. Kirshbaum and Dale R. Durran

in the orographic cloud. As discussed by Banta (1990) , one form of moist instability termed “latent” instability is characterized by the existence of convective available potential energy (CAPE) in the orographically modified flow. As with conditionally unstable flow over flat terrain, air parcels lifted to the level of free convection in a latently unstable atmosphere can develop into deep convective storms. This type of instability was present in the Big Thompson flash-flood of 1976

Full access
Hugh Morrison, Marcus van Lier-Walqui, Matthew R. Kumjian, and Olivier P. Prat

-level knowledge and strictly enforce integral constraints, while using Bayesian inference to determine uncertain microphysical process representations. Another approach is to develop full parameterizations using machine learning without specifying any explicit underlying physical framework (e.g., Rasp et al. 2018 ; Gentine et al. 2018 ; O’Gorman and Dwyer 2018 ; Brenowitz and Bretherton 2018 ). For example, this was recently applied using deep neural network learning to parameterize all subgrid

Free access
Grant J. Firl and David A. Randall

possible to guide parameterization development, perhaps large-eddy simulation (LES) could be useful. Indeed, Bogenschutz et al. (2010) used LES data from shallow and deep convective cases to calculate the moments necessary to test many assumed-PDF schemes (represented by the dashed line in Fig. 1 ). This technique eliminates errors in moment generation (such as those associated with the various higher-order closure assumptions) and provides a way to objectively compare existing schemes. In this work

Full access
Harold D. Orville and Lansing J. Sloan

argument of an exponential, which isapproximated by a series expansion, exceeded 1. Usinga one-dimensional parcel model, we checked the saturation technique against an iteration technique for thesaturation temperature and did not find significantdifferences. However, both methods reproduce a pseudoadiabat to within a degree only in the upper levels.Consequently, we have continued to use the Ogurasaturation technique. The deeper the atmosphere thegreater the discrepancy between the numerical technique

Full access
S. Christodoulaki, G. Petihakis, N. Mihalopoulos, K. Tsiaras, G. Triantafyllou, and M. Kanakidou

and is characterized by an excess of evaporation over precipitation and river runoff that creates a complex circulation system. Water with a low N:P ratio is penetrating in the Mediterranean from the Atlantic through the Strait of Gibraltar to compensate for the water deficit caused by evaporation ( Coste et al. 1988 ; Gomez et al. 2000 ). The surface seawater in the Mediterranean is further depleted toward the east ( Azov 1991 ) because of the westward transport of the underlying deep

Full access
Sungsu Park

scheme (i.e., free-tropospheric transport), while nonlocal transport is parameterized by saturated shallow and deep convection schemes. In the alternative process-dependent parameterization, local transport is parameterized by a moist turbulence scheme in CAM5 that simulates both dry and saturated turbulent transport in the entire atmospheric layers with a single set of moist physics formulas. The remaining nonlocal transport is performed by UNICON that consists of subgrid convective updrafts

Full access
Sam F. Iacobellis II and Richard C. J. Somerville

used to model vertical diffusion of heat and moisture in the model troposphere(up to 100 mb). The diffusivity coefficient, Ka, hasbeen set to 100/>2 (m2 s-l) where O is the density inkg m-3. The deep convection parameterization used in thisstudy follows Kuo's (1974) general approach and isbased on the version of the Kuo parameterization dueto Anthes (1977). The parameterization of shallowconvection is taken from Tiedtke et al. (1988). Themodel recognizes two distinct types of convection

Full access
A. Deloncle, R. Berk, F. D’Andrea, and M. Ghil

clustering methodology as Kondrashov et al. (2004) to define weather regimes and the preferred transition paths between them. Statistical learning techniques are then applied to exploit this knowledge for forecasting purposes. The paper is organized as follows. In section 2 , the atmospheric model and the preprocessing performed to obtain the weather regimes and the transition paths are briefly described; some details on the model appear in appendix A . In section 3 , we present the two main

Full access