Ground-based monitoring of the distributions and dynamic change in local area of trace gases can play an important role in pridicting the local atmosphere and designing the space-based instruments for sounding atmospheric moleculars. Based on the trace gas inversion algorithm SFIT2, the optimization process is improved and a novel inversion model is developed based on truth region method. A solution to the convergence of algorithm when the a prior i and true atmospheric information have great differences is found, and effective inversion of trace gases is realized when atmospheric parameters are hard to get. The total column amount and the volume mixing ratio of carbon monoxide are retrieved from the ground-based infrared ultraspectral data during September and October in Beijing. The results improve the validity of this modified algorithm and the ability for understanding the trace gases and their effects on environment and climate.