Assimilation of trace gas retrievals with the Local Ensemble Transform Kalman Filter
Kuhl, David Derieg
MetadataShow full item record
Over the 50 year history of Numerical Weather Prediction (NWP), the focus has been on the modeling and prediction of meteorological parameters such as surface pressure, temperature, wind, and precipitation. However, due to concerns over pollution and to recent advancements in satellite technologies, an increasing number of NWP systems have been upgraded to include capabilities to analyze and predict the concentration of trace gases. This dissertation explores some of the specific issues that have to be addressed for an efficient modeling of the concentration of the trace gases. These issues include modeling the effects of convective mixing on the concentration of the trace gases and the multivariate assimilation of space-based observations of the concentration of the trace gases. In this dissertation, we assimilate observations of the concentration of trace gases with an implementation of the Local Ensemble Transform Kalman Filter (LETKF) data assimilation system on the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) NWP model. We use a modified version of the NCEP GFS model that was operational in 2004 at resolution T62/L28. We modify the model by adding parameterization for the process of convective mixing of the trace gases. We consider two specific trace gases: ozone (O<sub>3</sub>) and carbon monoxide (CO). We incorporate these gases into the model by using 3-dimensional time-dependent O<sub>3</sub> and CO production-loss values from the Real-time Air Quality Modeling System (RAQMS) global chemical model. The O<sub>3</sub> observations we assimilate are from the Solar Backscatter UltraViolet generation 2 (SBUV/2) satellite instrument (version 8) flown on the NOAA 16 and 17 satellites. The CO observations we assimilate are from the Measurements Of Pollution In The Troposphere (MOPITT) instrument (version 3) flown on the NASA TERRA satellite. We also develop a new observation operator for the assimilation of retrievals with the LETKF. We carry out numerical experiment for the period between 000UTC 1 July 2004 to 000UTC 15 August in the summer of 2004. The analysis and forecast impact of the assimilation of trace gas observations on the meteorological fields is assessed by comparing the analyses and forecasts to the high resolution operational NCEP GFS analyses and to radiosonde observations. The analysis and forecast impact on the trace gas fields is assessed by comparing the analyzed and predicted fields to observations collected during the Intercontinental Chemical Transport Experiment (INTEX-A) field mission. The INTEX-A field mission was conducted to characterize composition of pollution over North America, thus providing us with ozonesonde and aircraft based verification data. We find that adding the process of convective mixing to the parameterization package of the model and the assimilation of observations of the trace gases improves the analysis and forecast of the concentration of the trace gases. In particular, our system is more accurate in quantifying the concentration of O<sub>3</sub> in the troposphere than the original NCEP GFS. Also, our system is competitive with the state-of-the-art RAQMS atmospheric chemical model in analyzing the concentration of O<sub>3</sub> and CO throughout the full atmospheric model column. The assimilation of O<sub>3</sub> and CO observations has a mixed impact on the analysis and forecast of the meteorological fields. We find that most of the negative impact on the meteorological fields can be eliminated, without much reduction to the positive impact on the trace gas fields, by inflating the prescribed variance of the trace gas observations. The appendices of this dissertation reproduces two papers on related research. The first paper covers the northward front movement and rising surface temperatures in the great planes. The second paper covers the assessment of predictability with a Local Ensemble Kalman Filter.