Prediction of Dimethyl Disulfide Levels from Biosolids Using Statistical Modeling

Two statistical models were used to predict the concentration of dimethyl disulfide (DMDS) released from biosolids produced by an advanced wastewater treatment plant (WWTP) located in Washington, DC, USA. The plant concentrates sludge from primary sedimentation basins in gravity thickeners (GT) and sludge from secondary sedimentation basins in dissolved air flotation (DAF) thickeners. The thickened sludge is pumped into blending tanks and then fed into centrifuges for dewatering. The dewatered sludge is then conditioned with lime before trucking out from the plant. DMDS, along with other volatile sulfur and nitrogen-containing chemicals, is known to contribute to biosolids odors. These models identified oxidation/reduction potential (ORP) values of a GT and DAF, the amount of sludge dewatered by centrifuges, and the blend ratio between GT thickened sludge and DAF thickened sludge in blending tanks as control variables. The accuracy of the developed regression models was evaluated by checking the adjusted R² of the regression as well as the signs of coefficients associated with each variable. In general, both models explained observed DMDS levels in sludge headspace samples. The adjusted R² value of the regression models 1 and 2 were 0.79 and 0.77, respectively. Coefficients for each regression model also had the correct sign. Using the developed models, plant operators can adjust the controllable variables to proactively decrease this odorant. Therefore, these models are a useful tool in biosolids management at WWTPs.

Key Words:

• Dimethyl disulfide
• Biosolid odors
• Statistical modeling.

ACKNOWLEDGMENTS

We would like to thank Professor Eric Seagren at the University of Maryland as well as the reviewer and editor for insightful comments and discussions.

Notes

¹Note that the coefficients for the variables are only valid within the ranges of practical operations. For example, for the variable blend ratio ², the range is (0.5)² to (2.0)².

²These elasticity computations involved multiplying the average value for the variable by 1.1.