Assimilating Doppler Wind Lidar in a High-Resolution Atmospheric Model for Airborne Contaminant Applications

Doppler wind lidar (DWL) data provide 3-dimensional realizations of the atmosphere at high spatial and temporal resolutions (e.g., 200-meter horizontal resolution every few minutes).  A DWL, or network of DWLs, provides extremely valuable information that can be assimilated into an atmospheric model to improve the accuracy (or can be used as validation) of the model’s analyses and forecasts.  Ground-based DWL data can resolve complex flow features within the atmospheric boundary layer (ABL) that are unobservable by standard observation networks.  Proper representation of the ABL is crucial for air-quality and airborne hazard prediction since most sources of contaminants and affected populations are located within the ABL.  For example, the boundary layer height, which is a function atmospheric stability, determines how much a contaminant is diluted as it moves downwind. A relatively small error in ABL height can result in large errors in the predicted severity of a given hazardous contaminant. 

    There are a couple fundamental components of atmospheric models and assimilation algorithms that need to be considered before attempting to assimilate high-resolution (i.e., un-thinned) DWL data.  First, the atmospheric model needs to be integrated at a grid resolution comparable to that of the DWL data (e.g., on the order of 100s of meters).  Otherwise, the model will not be capable of resolving flow features that are present in the observed data (i.e., errors of representativeness). Second, to statistically assimilate the DWL data, model error and cross-gridpoint error correlations need to be quantified or estimated.  Statistical algorithms (e.g., minimum variance or maximum likelihood) require model and observation errors to determine the relative weights assigned to either the model values or the DWL values as well as how the DWL information is spread across model grid points.

  This research focuses on the assimilation of DWL data to improve high-resolution atmospheric model accuracy whereas plume transport accuracy will be addressed in future work.  We demonstrate a ground-based lidar assimilation system using the Weather Research and Forecasting (WRF) atmospheric model and the Gridpoint Statistical Interpolation (GSI) assimilation system.  We demonstrate various formulations of model error estimates and analyze the corresponding sensitivities.  We also evaluate the frequency of assimilation cycles and the impact on operational run time.  Although Lidar scans can occur every few minutes, the atmospheric models my not be able to keep up.  We then evaluate how different datasets (i.e., radar, surface, and upper-air data can compliment the DWL inputs by providing other atmospheric parameters (e.g., temperature, pressure and humidity) that help establish physical balance in the model fields.  Finally, we attempt to validate short-duration WRF forecasts using subsequent DWL scans.