Title of Document : STATISTICAL ODOR PREDICTION MODELS FOR SUPPORTING BIOSOLIDS ODOR MANAGEMENT

Title of Document: STATISTICAL ODOR PREDICTION MODELS FOR SUPPORTING BIOSOLIDS ODOR MANAGEMENT Sirapong Vilalai, Doctor of Philosophy, 2008 Directed By: Associate Professor Steven A Gabriel Department of Civil and Environmental Engineering Biosolids are being beneficially recycled for agricultural purpose. Often, however, biosolids odors diminish marketability of biosolids, bring community opposition or, in the worst case, cause to ban the biosolids land application program. This dissertation is aiming to develop practical biosolids odor prediction models that can be applied for biosolids management on daily basis using the existing data available at the wastewater treatment plant and at the application sites as explanatory variables. Therefore, biosolids producer can use the plant odor predicting models to early detect and notify the hauling contractor when malodorous biosolids are anticipated. With the field odor models, malodorous products can be allocated accordingly to the appropriate sites in preventing the odor complaints from the communities. First, biosolids odors prediction models at wastewater treatment plant were developed using linear regression analysis and categorical data analysis. Biosolids odor was predicted in terms of detection threshold (DT) concentration and class of biosolids odor (odorous or non-odorous). Variables influencing biosolids odor levels at the plant were the percent solids and temperature of biosolids, percentage of the gravity thickener solids (GT) in the blend tank, pH of the GT solids, concentration of the return activated sludge (RAS) at the secondary process, and number of centrifuges running. Second, simulation and sensitivity analysis were conducted on the selected biosolids odor prediction model when uncertainty in the input variables was considered. Two variables (i.e., the number of centrifuges running and the percentage of GT solids in the blend tank) were identified as decision variable that could reduce the probability of producing odorous biosolids. Last, a biosolids odors prediction model for use at field site was developed using ordered logit model. Various variables at the field site (i.e. weather conditions, odor measurement time of the day, wind condition, temperature, and inspector odor sensitivity) were included in the analysis. Finally, variables relating to field odor levels were the biosolids odor levels (detection threshold) at the plant, temperature at the reuse site, and wind conditions. STATISTICAL ODOR PREDICTION MODELS FOR SUPPORTING BIOSOLIDS ODOR MANAGEMENT