Tracking a dynamic odor plume to the odor source in a real world environment using a chemical sensor is not a trivial task. In general, odor distribution in the real world is temporally and spatially fluctuated in a random manner owing to turbulent airflow. Therefore, evaluating a tracking strategy in a real world environment presents a certain difficulty in setting up a consistent environment for repetitive testing. This paper presents a simulation framework for the development and evaluation of a tracking strategy in a consistent virtual environment. A dynamic turbulent environment model with an odor source was modeled using Computational Fluid Dynamics (CFD) software, and the resulting data were used for a plume tracking simulation. This paper addresses the problem of estimating the trend of a fluctuating odor concentration for odor plume tracking. An estimation method for the odor concentration trend based on a regression was compared with the one based on an adaptive threshold. Each method was embedded into a chemotaxis-anemotaxis based plume tracking strategy to localize the source. The performances of the tracking strategy employing these different methods were compared. In addition, data smoothing was also examined to improve the effectiveness of the estimation methods. The investigation shows that the tracking strategy employing the combined smoothing-regression trend estimation has the highest success rate and minimum average time performance compared to the other tested methods. This paper also shows that selection of the key parameter of the method affects the tracking performance.