Jennifer Baus (jab30@po.cwru.edu)

and

Aaron A. Jennings, Professor (aaj2@po.cwru.edu)

Department of Civil Engineering

Case Western Reserve University

Cleveland, OH 44106-7201

Passive soil vapor extraction is a method of soil remediation that uses natural meteorological driving gradients to drive vapor extraction wells. To evaluate the effectiveness of this type of technology, the impact of random meteorological conditions must be simulated. This presentation will discuss efforts made to simulate random atmospheric pressures that may be used as transient boundary conditions on vapor extraction models. In the past, this type of analysis has been done for weather patterns as they vary from month to month or day to day but rarely had it been attempted on an hourly basis. One of the reasons for this is the complex interactions of the seasonal weather patterns and variations as well as diurnal or daily weather patterns. It may also be true that few applications require meteorological variability tuned to this fine time scale. However, if a method can be developed to simulate natural weather patterns on fine time scales, the results can be used to conduct detailed studies of remediation techniques as they are affected by transient weather conditions.

To accomplish the required analysis, several decades of pressure monitoring data was compiled for several Ohio cities in order to determine if the statistical properties of atmospheric pressure recorded at each of these locations (airports) had similar characteristics. Hourly pressure values were analyzed for recorded records of as long as 40 years (e.g. up to 350,400 data points per site). Special analytical techniques had to be devised to handle this volume of data. After extensive analysis it was discovered that there were strong similarities in the statistical responses of all locations analyzed.

Once the statistical properties of the hourly pressure response could be quantified, two methods of generating synthetic data with the appropriate statistical response were evaluated. The first of these uses randomly generated pressure run magnitudes and durations sampled from the governing PDF’s to simulate the desired signal. The second uses a more formal time series analysis based on point-to-point correlations to accomplish the desired result.

The results of the first method have been less than successful. Although the simulated pressure patterns had the correct statistical properties, they tended to stray far from the natural central tendency. The results of the second method appear to be more promising, but this work is still in progress.