Impact of Atmospheric Data Assimilation on the Prediction of Shallow and Deep Convective Clouds near the SGP site during HI-SCALE

Poster PDF

Description

The Holistic Interactions of Shallow Clouds, Aerosols, and Land-Ecosystems (HI-SCALE) field campaign was conducted in Oklahoma near the ARM Southern Great Plains (SGP) site during the spring and late summer in 2016. It was designed to provide set of atmospheric measurements useful for investigation on coupled processes which govern the life cycle of shallow clouds, which have not been represented adequately in existing models. Atmospheric models use various simplified parameterizations for boundary layer, cumulus clouds, and microphysics processes; therefore, model errors are inevitable and will accumulate as the model integrates. Forecast errors also arise from uncertainties in the initial conditions. Data assimilation is a practical technique that can mitigate the problem of uncertain initial conditions by optimally combining information of model predictions with observations. In this study, the Community Gridpoint Statistical Interpolation (GSI) system is used to assimilate atmospheric data in coupling with the Weather Research and Forecasting (WRF) model. In addition to operational surface, upper air, and satellite data, the meteorological measurements collected during HI-SCALE and at the ARM SGP site are assimilated to optimize the regional analysis. The impact of data assimilation on simulated shallow and deep convective clouds is quantified. A three-dimensional variational assimilation (3DVar) scheme implemented in GSI serves as the core of assimilation with background error covariance computed using forecasts from the NCEP’s NAM model. For all assimilation experiments, a 12-hour spin-up forecast is conducted within four nested WRF domains before the assimilation, helping generate fine-scale feature through nonlinear interactions. Cycling of 3DVar data assimilation then proceeds at 6-hour intervals for one day. Each of the five analyses can be used as initial conditions for a forecast at particular time. The 3DVar experiments are compared with forecasts without data assimilation and evaluated quantitatively with observed meteorological data. Experiments are performed with and without ARM data, to assess the value of the SGP megasite and HI-SCALE measurements. Not surprisingly, data assimilation has an improvement on both the analyses and predictions. In the future, we anticipate data assimilation as a tool to reduce the uncertainties associated with initial conditions, so that we can focus on issues associated with model parameterizations.