Canada's per capita solid waste disposal rates are among the highest in the world. Landfill gas generation requires more accurate modelling in order to properly compare future emission mitigation or energy production projects. The EPA software LandGEM was selected for its common use in the literature. Two alternative methods to increase accuracy in methane and carbon dioxide estimates were studied. Real-time methane collection data from a municipal landfill in Regina's cold, semi-arid climate were consolidated to fit a linear-interpolated form of LandGEM. LandGEM defaults were found invalid for this landfill due to significant overestimation (76.5% error). Seasonal variations in gas collection were explored, and found that optimal seasonal k and Lo collection parameters had 7.3% error compared to field data, compared to 15.5% error using optimal annual parameters. The optimal kwinter was 0.0082 year-1 and the ksummer was 0.0095 year-1 (14.7% difference). Three pseudo-second order iterative methods were used to fit the model estimates to the daily data, and they were evaluated using RSS and literature values. Optimized parameters were applied to a simple study using LFGcost-Web. The default parameters overestimated the net present worth by 57-107% for three of the four projects. LandGEM assumes that carbon dioxide estimates are a function of methane, and that the two gases make up nearly 100% of gas content. This can lead to oversights in collection system design. A total of 25 cases (five formulas, five approaches) were compared for carbon dioxide collection at four western Canadian landfills. The existing Default with Traces approach overestimated production in 3 of the 4 sites, resulting in the highest RSS. LandGEM's governing formula yielded the most accurate results under this approach (mean RSS increased by 7.0 to 49.9% using other equations). Optimization resulted in the most accurate results for all formulas and approaches, and had the greatest reduction in RSS over the default approach (73.0 to 98.0%). The 1.4 ratio approach yielded the second most accurate results (mean RSS reduction of 66.5%). The annual k formula calculated k’s via two empirical formulas (based on precipitation), and yielded the lowest accuracy in 12 of 20 approaches.