WRF-based Weather Running Estimate – Nowcast Tool for Army Applications
December 13, 2012
- ARL is developing a Weather Running Estimate-Nowcast (WRE-N) tool
- Near-surface meteorological changes on timescales ranging from minutes to a few hours are critical to ground operations
- Goal is to be able to execute the WRE-N in a forward tactical setting on a platform no larger than a small linux cluster and potentially as small as a high-end laptop
The U.S. Army Research Laboratory, Computational and Information Sciences Directorate, Atmospheric Modeling Applications Branch located at White Sands Missile Range, N.M., is developing a Weather Running Estimate-Nowcast (WRE-N) tool.
WRE-N frequently updates the local environmental four-dimensional (x,y,z,t) meteorological grids or "data cubes" for use by the Army and Air Force tactical weather systems and decision aids to improve tactical execution strategy, situational awareness, and decision making related to weather when conducting military operations. Such data cubes often deviate at any given time and location from the coarser resolution operational (and potentially stale) model guidance from a forecast center (i.e., Air Force Weather Agency). For the Army, near-surface meteorological changes on timescales ranging from minutes to a few hours are critical to ground operations.
Nowcasting is a method of diagnosing and forecasting the current and very short-term weather situation by combining existing forecast guidance with observational data, using various methods of analysis/fusion and assimilation, and then extending this meteorological "picture" forward in time to produce a very short-range forecast or prediction. The predictive component can be produced in many ways, such as through cycling (coupling analyses to prognostic numerical weather prediction (NWP) models), extrapolation, statistical methods, or expert system approaches. WRE-N provides a 3-D gridded analysis of the "current" weather (the WRE), and a 3-D short-range gridded prediction counterpart (the Nowcast).
The Weather Research and Forecast (WRF) model, used as the operational forecast model at both the National Centers for Environmental Prediction and the Air Force Weather Agency, is the engine which generates both the WRE and the Nowcast in the current WRE-N.
The technique of "observation nudging" 4-D data assimilation is used by WRF to produce a new WRE through execution of a short preforecast "dat assimilation" period, while the WRF prognostic dynamical equations extend the WRE fields forward several hours to generate a new short-term Nowcast. The combination of the WRE and Nowcast meteorological grids projected over a short time forward from the present, constitute the 4-D data cube.
"WRE-N is designed to provide local corrections to the regional very short-range forecast data cube along with Army-required finer scale domains and resolutions," said Robert Dumais, Jr, project lead at White Sands. "This allows for a computationally fast, mesoscale meteorological analysis, and very short range prediction (Nowcasting) capability specifically for the Distributed Common Ground System-Army (DCGS-A) along with Army/Marine Artillery profiler. The WRE-N would serve as a 'weather' Intel component of DCGS-A. In short, this implementation of WRE-N allows 3-D gridded distributions of temperature, dew-point temperature and horizontal wind vector components to be updated, analyzed and Nowcasted with a rapid refresh rate, making more timely use of local weather observations."
According to Dumais, being able to accurately update high spatial and temporal resolution meteorological information for Army operations, generated by fusing or assimilating current local meteorological observational data nearer to the time they were taken, is often better than relying on results obtained by potentially outdated and previously executed coarser resolution model forecasts alone.
"Although current state-of-the-art numerical weather prediction models offer an unprecedented capability to simulate small-scale mean atmospheric structures, individual deterministic model solutions tend to suffer from day-to-day spatial and temporal errors in amplitude, phase and variability," said Dumais.
He said to correct for these kinds of errors and model biases, it is necessary to assimilate real meteorological observational data through, for example, NWP model forecast cycling. The advanced assimilation methods that have been developed over the past few decades now offer dynamically, physically consistent, and statistically optimal methods of "correcting" NWP model tendencies on the fly.
"If the "current" weather situation is accurately analyzed, it may be possible to reasonably predict very short-term weather situations (Nowcasting)," said Dumais. "Our ability to accomplish this will increase as techniques continue to improve and better methods of assimilating observed cloud-scale fields and dynamically coupling them to NWP model fields are produced."
The WRE-N will provide critical battlefield weather information that can enhance Soldier survivability, force sustainment, situational awareness, and can enhance and extend the operational capabilities of existing and notional Army systems.
The complexity of WRE-N is that for Army needs we are tailoring the WRF model at unprecedented spatial and temporal resolutions, assimilating non-traditional and well as traditional sources of battlefield weather information and establishing "compute-friendly" configurations that would make WRE-N amenable to running on more modest compute platforms, which could conceivably be hosted at a forward echelon (such as in DCGS-A or artillery)," said Dumais.
Dumais said the ultimate goal is to be able to execute the WRE-N in a forward tactical setting on a platform no larger than a small linux cluster and potentially as small as a high-end laptop.