12B.5 Intercomparison of Seven Planetary Layer/Surface Layer Physics Schemes over Complex Terrain for Battlefield Situational Awareness Applications

Thursday, 7 June 2018: 9:15 AM
Colorado B (Grand Hyatt Denver)
Richard S. Penc, U.S. Army Research Laboratory, WSMR, NM; and J. A. Smith, J. W. Raby, R. E. Dumais Jr., B. P. Reen, and L. Dawson
Manuscript (601.8 kB)

The WRF model is capable of producing global and regional scale forecasts, as well as longer term, decadal, climate simulations. Spatially, the model physics are capable of resolving large scale synoptic weather systems, mesoscale phenomena such as severe storms, and sub-km scale phenomena down to the boundary layer eddies through inclusion of a large eddy simulation (LES) model. In addition, the model includes a wide array of physics packages that can be selected by the user, parameterizing sub grid scale processes including solar (SW) and long wave (LW) radiation, cumulus (CU), and shallow cumulus (SCU), cloud microphysics (MP), planetary boundary layer (PBL), surface layer (SL), and integrated land surface model (LSM) options. The WRF stands out among other operational weather forecast models in its flexibility of choice in physical parameterizations.

As a community model, the WRF code and physics is not static. Aside from bug fixes and performance improvements, the modeling system is continually being updated with the latest physical parameterization schemes. This research was performed with WRF v.3.8.1. In this release, there are twelve PBL, eight SL, six LW, seven SW, twenty one cloud MP, eleven CU, and three SCU options selectable by the user. Along with tunable model parameters (diffusion, for example) there are almost limitless combinations of schemes and parameterization schemes to choose from. The present study focuses on the PBL/SL combinations because of the importance of simulating boundary layer processes, which drive atmospheric motions near the surface where people reside, ground forces operate, artillery meteorology impacts are greatest, and where land mobility is impacted. Of the twelve PBL schemes available in WRF, seven are examined for their overall performance in a complex terrain environment. The model domain for these exercises is a Southern California domain, approximately centered near San Diego, California. It was chosen primarily because of the availability of high density surface observations, and variety of terrain and land use. A triple nest configuration (9-3-1km) of WRF v3.8.1 was run for a single day in which conditions were quiescent in order to observe the evolution of the PBL in the absence of rapidly changing synoptic scale weather influences.

This study examines the PBL/SL performance driven by the need for the WRE-N modeling system to be preconfigured to run optimally in non-predisclosed forward deployed locations, primarily for situational awareness. The MYNN, MYJ, YSU, ACM2, QNSE and Shin-Hong PBL/SL options were compared, representing both local and non-local transport physics. While the performance of these schemes does not differ significantly for the daytime convective boundary layer, notable differences exist between the schemes in the case of the nocturnal boundary layer. For the daytime convective boundary layer, no single scheme stands out as the best to use in our complex terrain environment. All configurations show that the model has the greatest difficulty capturing the transition between daytime and nighttime boundary layer. The near surface temperature, PBL structure, and vertical profiles are important for our application, and for the prediction of weather effects on Army operations using weather impacts decision aids (WIDAs). Both the YSU and MYJ schemes, used in the common WRE-N configurations and in operational configurations, appear to perform reasonably well. Because the scale-aware scheme (Shin and Hong, 2013) addresses the scale dependency problem, this parameterization would be preferable for use with domains with a grid spacing < 1km. However, there was no notable improvement when used with our inner 1 km grid spacing. For our domain, the QNSE scheme appears to have performed best during the forecast period. Statistical analysis shows that the PBL scheme accounts for about 3-4% of the total variance, and any physics based ensemble must consider more than solely PBL/SL options. This is confirmed by constructing Talagrand diagrams for the mini-ensemble. Research looking into the relative contributions to model uncertainty due to the other physics schemes, and initialization data are addressed in the Design of Experiments approach described by Smith and Penc (2016), and Smith et al. (2018).


Shin, H., and S. Hong, 2013: Analysis of Resolved and Parameterized Vertical Transports in Convective Boundary Layers at Gray-Zone Resolutions. J. Atmos. Sci., 70, 3248-3261.

Smith, J.A. and R.S. Penc, A Design of Experiments Approach to Evaluating Parameterization Schemes for Numerical Weather Prediction: Problem Definition and Proposed Solution Approach. Joint Statistical Meetings Proceedings, Section on Statistics in Defense and National Security, Conference on Applied Statistics in Defense 2015, 2016: p. 4183-4192.

Smith, J.A., R. Penc, and J.W. Raby, Statistical Design of Experiments in Numerical Weather Prediction: Emerging Results, in 98th Annual AMS Meeting, Joint with 25th Conference on Probability and Statistics. 2018, American Meteorological Society: Austin, TX. Paper 6.1.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner