A technique to use observed climatological data to improve the quality of simulated climatological data produced by a numerical mesoscale model

The ability to simulate historical and climate conditions by using a numerical model has been developed. This technique has been given the name CLImate statistics by a dynamical MODel (CLIMOD). The evaluation of CLIMOD results has shown considerable success, but there is always some model bias introduced into the simulated climatologies. One technique for removing the model bias is to use climatologies developed from observed data to determine the model bias. From the identified bias a correction factor is calculated and employed through a weighted interpolation technique to remove the bias.

Based upon the results of the CLIMOD research, it was deemed necessary to develop a post-simulation method for blending observed climate data with the model generated climate data to try to reduce the model bias introduced into the simulated climatologies. The objective of the post-simulation blending technique is to create a single database that will contain most or all the attributes of the observational statistics where they are available and blend smoothly to a model-generated database in other areas. At the same time, the statistics in regions without observed data are specified primarily from the model-simulated data but with a reduction in model bias. The use of observed climate data may appear to be somewhat redundant since the model has the ability to assimilate all of the available observed data into the simulations. Thus, the model-generated database will already be a type of blend between the observational data and the model physics. However, the model is capable of simulating only a portion of the variance of each parameter because it cannot simulate atmospheric features that are below the resolvable scale of the model grid. Using a grid mesh finer than 10 km could eliminate this problem. However, that is often not currently computationally feasible when modeling a large region. Sub-grid scale parameterization schemes attempt to account for some of these features but they do not model all of the variability found at small scales. This is particularly true for a variable such as precipitation, which exhibits a great amount of small-scale variability. Importantly, many extreme events (e.g., isolated heavy rains associated with thunderstorms) are associated with small-scale variability.

Results from Saint Louis University's evaluation determined there are four distinct methods to blend or composite the observed data. All are based fundamentally on some form of distant weighting from the observed site. The difference in the methods is in the manner that the autocorrelation function relating the distance and the weight is determined. Analysis of model results indicates that the model statistics typically differ by only a small amount so that the blending should only make small adjustments. Quantifying the differences in blending algorithms is an ongoing part of this research.