Probe vehicle techniques have been widely applied to the traffic performance studies on freeways and arterials. Today they also show great potential in vehicular energy and emission evaluation as of the development of the microscopic emission model such as Motor Vehicle Emission Simulator (MOVES) and Comprehensive Modal Emissions Model (CMEMS). In this paper, a comparison is conducted among five probe vehicle sampling strategies to find the optimal one to implement future field test in Los Angeles, CA. To measure the performance of each sampling strategy, the vehicle-specific power (VSP) distribution is computed for both sample and population, and the Bhattacharyya distance is introduced to quantify the similarity between two distributions. A dynamic emission estimation model is also proposed in this paper, which estimates the short-term (e.g., 60s) emissions using probes. The methods are calibrated and validated by NGSIM US-101 and Lankershim data. Based on the analysis and numerical results, the authors provide some suggestions to the future field test: 15 probe vehicles, 1 minute as the time interval between vehicles, and 2 minute as the time interval to estimate the dynamic emission performance. This study also provides instructions to other future probe vehicle tests.