170 likes | 258 Views
The Center for Atmospheric Chemistry and the Environment. EnKF Assmilation of Chemical Tracer Information in a 2-D Sea Breeze Model . Amy L. Stuart, Altug Aksoy, Fuqing Zhang, and John W. Nielsen-Gammon.
E N D
The Center for Atmospheric Chemistry and the Environment EnKF Assmilation of Chemical Tracer Information in a 2-D Sea Breeze Model Amy L. Stuart, Altug Aksoy, Fuqing Zhang, and John W. Nielsen-Gammon Work sponsored in part by the Texas Environmental Research Consortium and the Texas Commission on Environmental Quality
Research Questions • Does the EnKF perform well in a forced-dissipative dynamical system with no mechanisms for rapid error growth? • Can observations of chemical tracers effectively improve the meteorological and chemical analysis?
Outline • Model description • Ensemble characteristics • Meteorological assimilation • Chemical assimilation
Model Description • 2-D: 500 km x 3 km + sponge layers • Grid spacing: 4 km x 50 m • Prognostic variables: horizontal vorticity, buoyancy, concentration • Sinusoidally-varying buoyancy source over land plus stochastic white noise • Tracer source 28 km inland
EnKF Configuration (1) • Observed variable: Buoyancy or Concentration • Observations: Surface observations on land • Observational error: Standard deviation of 10-3 ms-2 or 10-7 kg/m3 • Observation spacing: 40 km (10 grid points)
EnKF Configuration (2) • Covariance localization: Gaspari and Cohn’s (1999) fifth-order correlation function with 100 grid-point radius of influence • Observation processing: Sequential (Snyder and Zhang 2003) with no correlation between observation errors • Filter: Square-root after Whitaker and Hamill (2002) with no perturbed observations
The sea breeze model: The sea breeze cycle Buoyancy (ms-2) Vorticity (s-1) Sea breze front develops at the coast 123 Hour Forecast (3:00PM Local) Onset of the sea breeze
The sea breeze model: The sea breeze cycle Buoyancy (ms-2) Vorticity (s-1) Vertical gravity waves emanate from the PBL Sea breze front matures and penetrates inland 129 Hour Forecast (9:00PM Local) Peak sea breeze
The sea breeze model: The sea breeze cycle Buoyancy (ms-2) Vorticity (s-1) Sea breze front weakens 135 Hour Forecast (3:00AM Local) Onset of the land breeze Land breeze front develops
The sea breeze model: The sea breeze cycle Buoyancy (ms-2) Vorticity (s-1) 141 Hour Forecast (9:00AM Local) Peak land breeze Land breeze front matures yet is not as strong as the sea breeze front
The sea breeze model: Forecast spread Vorticity Buoyancy • Buoyancy spread dominated by initial error spread; little diurnal variability • Initial vorticity spread advected out of the domain; strong diurnal variability • Buoyancy power spectrum dominated by large-scale initial-condition error • Vorticity power spectrum reflects smaller-scale frontal dynamics and is flatter
The sea breeze model: Perfect-model EnKF Results Vorticity Buoyancy • Buoyancy is the observed variable; its error reduction is more dramatic and faster • Buoyancy error saturates at a magnitude comparable to observational error • Unlike buoyancy, vorticity error and spread exhibit diurnal signal
Mean Predicted Concentrations Peak land T Peak sea breeze • Sea breeze recirculation allows concentrations to build near source Peak land breeze Peak heating 3 km Source 500 km Land Sea
Predicted Concentration Uncertainties Peak land T Peak sea breeze • Ensemble standard deviation has diurnal variability, grows in transition between land to sea breeze Peak land breeze Peak heating
Evolution of Domain Average Errors Error Vorticity (s-1, x103) • EnKF assimilation of concentration observations reduces error in both meteorological variables and concentrations Bouyancy (m/s2) Concentration (kg/m3, x10-7) noon peak sea breeze peak sea breeze peak land breeze
Targeted Single Observation Design • Pre- vs post- network assimilation domain uncertainty norms • Locations of promising adaptive observations are similar before and after regular network assimilation.
Conclusions • EnKF works for sea breeze • Chemical data assimilation improves chemistry and meteorology • Ensemble can predict optimal locations for targeted observations • Next: imperfect model and parameter estimation…