Observing System Simulation Experiments (OSSEs) are designed to evaluate the impact of observing system data in numerical weather analysis and prediction. The OSSE procedures are particularly helpful for assessing the data impact for those observing systems which are not yet operational. For satellite observing systems, OSSEs provide a way to evaluate the impact of an instrument while it is still in the design or pre-flight stage.
While OSSEs vary in complexity and the manner in which they are carried out, there exist fundamental procedures common to most of these types of experiments. Usually a forecast model is used first to generate a "truth" or "nature" data set. To generate this data, the forecast model is usually run at the greatest resolution possible and with the most detailed suite of physical parameterizations, under the assumption that the greatest degree of complexity is required to represent nature.
Next, simulated observations are extracted from these "nature" model fields, taking into account the various characteristics of the observing system being simulated. For the case of satellite data, forward models of radiative transfer must be used to translate forecast model variables (atmospheric temperature and moisture profiles, etc.) into the radiances which would be observed at the top of the atmosphere by the particular instrument. A "retrieval" of the desired atmospheric quantities is performed by either a physical or statistical retrieval algorithm using these radiances.
The original "nature" or "truth" model forecast fields are degraded by adding errors to make them representative of the quality of actual atmospheric information which would be available at a data time. The synthetic retrieval data are then assimilated into the "truth+error" set of atmospheric fields. Ideally, if the retrieval data were of perfect accuracy and the assimilation methods were perfectly executed, this procedure would return the "truth+error" atmospheric fields back to the "truth". In practice, the retrieval data has imperfect coverage, inherent errors and the assimilation methods are imperfect, so only a portion of the errors are corrected by the assimilation. The results of the forecast model runs using the "truth", "truth+error" and "truth+error+data assimilation" are subsequently used to evaluate the impact of the observing system and/or assimilation techniques on numerical prediction.
Recent OSSE work at the CIMSS has been directed towards evaluation of data from the Advanced Microwave Sounding Unit (AMSU) and High-resolution Infrared Radiometer Sounder (HIRS) instrument combination which will be flown on NOAA polar orbiting satellites beginning with NOAA-K in late 1995 or 1996. The assessment procedure includes the evaluation of synthetic satellite imagery as a visualization and diagnostic tool (see Diak et al., 1992). The CIMSS Subsynoptic Scale Model (SSM) is used to produce the "truth" model data fields, including three-dimensional fields of cloud liquid water, while the forward radiative transfer models and retrieval methodology described in Eyre (1990) produce the synthetic radiance data from the model fields and retrieve the atmospheric profiles from these data.
The latest OSSE work was conducted by CIMSS scientists Xiaohua Wu, George Diak, and Kit Hayden (see Wu et al., 1995). This work investigates the impact of satellite-observed water vapor and cloud liquid water used to initialize a primitive equation model on short-range precipitation forecasts. Synthetic radiances for the AMSU-HIRS observing system are generated from a high resolution control run made by the SSM used at the CIMSS. The OSSE procedures used to conduct these experiments are similar to those described in the overview section above and were performed for 12-13 March 1991. On this date, a major cyclone was located over the central United States and produced abundant precipitation in the midwestern and eastern portion of the country (the contours are isolines of SSM predicted cloud liquid water; note that the model is able to accurately position the clouds accompanying the weather system).
In addition to the 12-hour control forecast used to generate the "truth" atmosphere, five additional model forecasts were generated by varying the initialization scheme and the types of satellite-observed data assimilated by the initialization. Two of the model forecasts did not use satellite-observed data but differed because one run used an adiabatic initialization scheme while the other was diabatic. A third model run incorporated satellite-observed temperature and humidity profiles, but no cloud liquid water, into a adiabatic initialization. The last two model forecasts incorporated satellite-observed temperature, humidity, and cloud liquid water profiles into the two initialization schemes.
The assimilation of satellite-observed moisture and cloud water together with a diabatic initialization significantly alleviates the model precipitation "spinup" problem. This is demonstrated by the accumulated hourly precipitation and the areal precipitation coverage generated for a subset of the forecast domain. The combination of assimilated moisture and cloud water profiles retrieved from the AMSU-HIRS and a diabatic initialization produces the most accurate accumulated hourly precipitation rates at the third hour of the forecast period and is the quickest to produce a precipitation total that is 70% of the "truth".
Diak, G.R., D. Kim, M.S. Whipple, X. Wu, 1992: Preparing for the AMSU. Bull. Amer. Meteor. Soc., 73, 1971-1984.
Eyre, J.R., 1990: The information content of data from operational satellite sounding systems: A simulation study. Quart. J. Roy. Meteor. Soc., 116,401-434.
Wu, X., G.R. Diak, C.M. Hayden and J.A. Young, 1995: Short-range Precipitation Forecasts Using Assimilation of Simulated Satellite Water Vapor Profiles and Column Cloud Liquid Water Amounts. Mon. Wea. Rev., 123 , 347-365.