Statistical modeling of methane production from landfill samples

Abstract

Multiple-regression analysis was conducted to evaluate the simultaneous effects of 10 environmental factors on the rate of methane production (MR) from 38 municipal solid-waste (MSW) samples collected from the Fresh Kills landfill, which is the world's largest landfill. The analyses showed that volatile solids (VS), moisture content (MO), sulfate (SO42-), and the cellulose-to-lignin ratio (CLR) were significantly associated with MR from refuse. The remaining six factors did not show any significant effect on MR in the presence of the four significant factors. With the consideration of all possible linear, square, and cross-product terms of the four significant variables, a second-order statistical model was developed. This model incorporated linear terms of MO, VS, SO42-, and CLR, a square term of VS (VS2), and two cross-product terms, MO x CLR and VS x CLR. This model explained 95.85% of the total variability in MR as indicated by the coefficient of determination (R2 value) and predicted 87% of the observed MR. Furthermore, the t statistics and their P values of least-squares parameter estimates and the coefficients of partial determination (R values) indicated that MO contributed the most (R = 0.7832, t = 7.60, and P = 0.0001), followed by VS, SO42-, VS2, MO x CLR, and VS x CLR in that order, and that CLR contributed the least (R = 0.4050, t = -3.30, and P = 0.0045) to MR. The SO42-, VS2, MO x CLR, and CLR showed an inhibitory effect on MR. The final fitted model captured the trends in the data by explaining vast majority of variation in MR and successfully predicted most of the observed MR. However, more analyses with data from other landfills around the world are needed to develop a generalized model to accurately predict MSW methanogenesis.