Background/Objectives: Predictive modeling is an inherently imprecise but unescapable tool used throughout the environmental industry. In situ thermal remediation providers depend upon predictive modeling to size their systems, select vapor and liquid treatment options, determine power demand and use, and estimate cleanup times. This paper will explore how site data quality and quantity affect modeling accuracy by examining some of the key algorithms that form the core of the predictive models used by Cascade Thermal.

Approach/Activities: Cascade Thermal’s in situ thermal remediation model evaluates all three of the thermal technologies, Thermal Conduction Heating (TCH), Electrical Resistance Heating (ERH), and Steam Enhanced Extraction (SEE). The building blocks of this predictive model include sub-models that simulate the energy balance, water balance and mass balance of a given treatment volume during the heating process. The modeling approach is iterative: beginning with a black box model in which the sub-models output estimated clean-up time based on generalized site characteristics and ending with a complex multi-layered transient box model that runs each of the sub-models for discrete intervals within the treatment zone. In the multi-layered model, the sub-models are transient, accounting for changes in the volumetric heat capacity, energy flux and mass removal rates according to the changes in saturation, thermal conductivity, contaminant concentration and hydrolysis reactions in each time-step. However, as is true of any model, the output of these simulations is only as valuable as the site data input.

Results/Lessons Learned: Not only do less than ideal site data sets negatively affect the reliability of a projective model, but empirical data about the contaminants and other media present in the treatment volume is often lacking. While standard equations for determining the heat capacity of pure water and sand exist, no such equations are available for multi-contaminant solutions or glacial till. This forces the model to approximate according to governing assumptions Cascade Thermal has developed from experience at hundreds of sites. As more and more site results are added to the model, these approximations become better and better. This is particularly true for estimates of cleanup time.

This paper will open up the Cascade predictive model to allow a view of where it relies upon well established equations and where it uses approximation algorithms. Attention will be paid to those factors, in site data and estimating algorithms, that most influence the modeled values for treatment volume heating rates, contaminant reduction rates, and site cleanup times.