Professional Scientist
Illinois State Water Survey

INTRODUCTION

The work reported here has two separate focus areas. The focus of our dispersion modeling studies was to improve the description of odor transport to better account for effects of elevated sources, trees surrounding a hog facility, and terrain features such as ravines.

The second focus was the development of methods to calibrate an odor-measurement device developed throughout the SRI and operating procedures that assure the instrument yields reproducible results.

ODOR DISPERSION MODELING

Previous work of the effort on dispersion modeling of hog odor employed a Gaussian plume model to summarize odor transport with respect to distance, prevailing wind direction, and odor emission rate, to determine the frequency the ground-level odor is above the detection threshold at various distances from the source. This information is useful for estimating the impact of a hog facility in producing noticeable odors in the vicinity and in assessing the impacts of odor-control measures. The dispersion model also was extended to estimate the impact of elevated release of odor on ground-level concentrations.

The Gaussian plume model lacks the sophistication to address many questions that arise, such as the effect of vegetation in the vicinity of the facility. Questions that are of interest: To what degree could vegetation barriers reduce odor impact? What would be the optimal orientation of the barriers? Another issue that often arises in connection with odor complaints is the speculation that odor follows terrain features such as a ravines and is able to remain above the detection threshold for considerably longer transport distances under favorable atmospheric conditions.

To address these issues, two new approaches beyond use of a Gaussian dispersion model have been initiated to investigate different aspects of odor transport. First, a “micro” model was developed, by scaling-down a 3-D air-quality model, to resolve concentrations over spatial distances appropriate for the description of odor dispersion. The air-quality model has been extensively used at a horizontal resolution of 10 km and the corresponding

For the odor problem, all chemical reactions are ignored because too little is known about the chemical transformation of odor components in the atmosphere.

Deposition of odor to surfaces is addressed in the micro model, but there is uncertainty concerning what deposition rate to use.

In order to compute the transport, a meteorological field describing the atmospheric motion must be input to the air-quality model. From other studies we have meteorological output from version 5 of the mesoscale meteorological model (MM5), a model widely used in the meteorological community, for summer months 1995. At present the micro model uses the MM5 output with the winds from MM5 extrapolated from the 10-km resolution grid to the100-m grids.

The second approach is to use Fluent, a computational fluid dynamics code, which resolves momentum, energy (temperature), and (optionally) species transport. Use of Fluent involves constructing a detailed (3-dimensional if desired) grid of the area to be modeled, specifying boundary conditions, and running the simulation to obtain convergence of the solution. One use of Fluent in the present context is to derive more precise meteorological fields for input to the micro model. Other applications are to derive meteorological and species dispersion fields in the vicinity of a hog facility to determine effects of tree barriers, elevated emissions, and a ravine. It presently appears that Fluent can be configured to take input from MM5 over the scale of 10 km and produce meteorological and species fields sufficiently detailed in the vicinity of features such as buildings, tree barriers, elevated stacks, and ravines. This is possible because the finite-element grid strategy employed in Fluent allows the grid to be fine in places where more detail is required and coarse at other areas in the model domain.

The practical question: Can the grid and model strategy be laid out with sufficient skill to meet these computational constraints yet not require excessive amounts of computer time? (An account was obtained on an NCSA supercomputer for the SOWM project with a fixed computer time allotment. The NCSA account affords access to Fluent which would otherwise be costly.)

An alternative is to use Fluent in a nested fashion. In the present case this involves running with a 10-km horizontal domain using input from MM5 to obtain meteorological and temperature fields at 100-m horizontal and approximately 1-m vertical resolution. This output could then be used to drive the micro model or to drive a nested Fluent model. The micro model source emissions from a building are specified by setting the concentration in the lowest grid cells (100 m x 100 m x 3 m) to a higher level (corresponding to the estimated building emission rate times the model time step of approximately 1 min) and setting initial concentration in surrounding grids to zero. An elevated stack emission is simulated in the micro-model by specifying an elevated grid (100 m x 100 m x 1m) at an increased concentration.

Fluent can be used as an emissions model to specify a building source as emanating from the sides of a building or an elevated source from a stack. The simulation could proceed to arrive at concentrations averaged over the micro model grid as a better approximation of a building or elevated source for the micro model. Simulations of the effect of trees or a ravine cannot be easily performed with the micro model. If the Fluent simulation over the entire 10-km domain cannot adequately resolve flow features over trees or a ravine in a single simulation, a smaller (~1-km) Fluent model can be nested to take boundary flow from the 10-km Fluent simulation in a two-step process.

Odor Measurement

The odor-measurement effort has shown promise as a tool for routine odor measurement. Odor is measured by determining the response of alkaline aerosol particles (typically NaOH) to the acids in swine odor. The method of detection that has shown most promise relies on operation of a series flow arrangement of two differential mobility analyzers (DMA1 and DMA2), each of which pass a narrow size range of aerosol particles depending on the voltage setting. (The device is named “nanonose” because its operation depends on nanometer-sized aerosol particles.) A successful procedure has been to set DMA1 to pass, for example, 25-nanometer (nm) diameter particles (actually a narrow size distribution centered at 25 nm), react the particles with the odor sample, heat the sample to volatilize reaction products, and set DMA2 to pass 15-nm particles. The increase in the number of 15-nm particles, produced by volatilizing reaction products from the 25-nm particles, is the raw signal that correlates with odor concentration.

From past experimental results, there is no question that the system works. It gives a response to hog odor at levels comparable to the human detection threshold. (It also responds in the same way to organic acid vapors.) Odor measurement by dilution olfactometry determines the odor-detection threshold, the odor level below which it cannot be detected by repeated measurements with a trained odor panel. The minimum requirement for the nanonose system is to determine a reproducible response to a given level of hog odor. Ideally that odor level would be the detection threshold. If we know the nanonose response for the detection threshold, determined by calibrations with olfactometry, then a sample of the synthetic mixture at an unknown concentration could be repeatedly measured and diluted until the nanonose response matches that of the threshold. The nanonose concentration of the sample could then be reported in terms of multiples of the detection threshold as is the practice in dilution olfactometry. (Likely the nanonose system would be calibrated so the odor concentration could be determined from a single measurement.) Now suppose similar measurements are made to determine the variation of the nanonose response to threshold levels of hog odor from different sources. If it can be established that variation in the detection threshold concentration found with nanonose does not vary substantially from that found by olfactometry, then the nanonose system can be regarded as a stand-alone alternative to olfactometry for measurement of hog odor. If there turns out to be considerable variation in the nanonose threshold reading for different hog odors, then the system would be less independent. This could happen, for example, if the fraction of organic acids in hog odor varies considerably among facilities. In such case the nanonose would need to be correlated with olfactometry for each facility and periodically when sampling from that given facility.

Work Completed

• The micro model has been developed from a larger-scale air-quality model and tests on the model performance conducted. Fluent has been successfully run to show that meteorological fields of 100-m resolution can be produced from boundary conditions specified over a 10-km domain with moderate amounts of computer time. Therefore Fluent can be used to derive a detailed meteorological field for the micro model from MM5 output. Fluent simulations including inclusion of species transport also have been conducted over a more limited domain of approximately 1-km with geometric details of a specific building. For these problems also, the Fluent model also converges easily.
• The odor-measurement research effort has been focused on developing a calibration method and operating procedure for routine odor measurement. An attempt was made early on to use the system to measure odor samples taken from the HENCO farm in collaboration with an ongoing field research project where samples are collected and analyzed by olfactometry. We were unable to adequately compare results with olfactometry for co-located samples. Since that time major changes, including a re-working of the flow system, have been implemented. An improved method of generating the alkaline aerosol was introduced, resulting in more stable aerosol concentrations and the needed capability to produce much higher concentrations. Finally, a strategy of continually heating the sample inlet was found to greatly reduce the carryover between successive samples, so the instrument response quickly returns to background level following the sample response.
• Following the attempt to compare nanonose results with olfactometry using the HENCO samples, it was concluded that Tedlar bags drastically affect the odor sample. Recent results with the improved nanonose setup confirm that result. Specifically, when acetic-acid vapor is injected into a Tedlar bag containing nitrogen as carrier gas, the nanonose measurement of bag contents show little or no response. When the same procedure is followed except using a Teflon bag, the acetic-acid response is at the expected value as determined by pre-calibration with acetic-acid vapor from a glass jar. It appears that organic acids when injected into a Tedlar bag attach to the walls and are thus removed from the bag air. Likewise, a hog-odor sample collected in a Tedlar bag loses organic acids. This is consistent with HENCO measurements showing samples collected at 4 pm and analyzed at 10 pm gave much higher nanonose readings than the same samples analyzed at 10 am the following day.
• We have obtained reproducible calibration with the nanonose system using acetic acid. The response is stable and can be reproduced from day to day. Optimum settings have been determined for aerosol-generation rate, voltages of the mobility analyzers, temperatures of the heated tubes through which the particles pass, and inlet heating. A method has been developed to confidently assure that the system response is not due to water vapor and that the system measures the same odor response regardless of sample relative humidity. Hog-odor samples obtained by placing a few millimeters of slurry inside a Teflon bag give a strong response equal to the equilibrium concentration of acetic-acid vapor diluted several hundred times. Unfortunately, we did not acquire the Teflon bags until after the olfactometry support to the SRI had ended, so we were unable to compare the nanonose instrument to olfactometry with Teflon bags.
• A persistent feature of the nanonose response is that a hog-odor mixture compared to a single acid is more complex. The final measurement procedure was to inject approximately 10 ml of an odor sample into the nanonose system, and the response forms a peak evolving in approximately 30 seconds. Acetic acid forms a well-defined single peak. A hog-odor sample forms an envelope of multiple peaks. Based on the observed behavior of the system, a plausible hypothesis is that the complex hog-odor sample experiences limited chromatographic separation, so the different acids evolve at different rates. This observation could be exploited by constructing a column to increase the chromatographic separation. Based on retention time, then, the identity of the main peaks could be identified and the system could reveal the relative amounts of the organic acids in a hog-odor sample.

Implications

• Results of simulations with the micro model for sources elevated up to 10-m show that beyond approximately 1 km from the source, elevation has little effect on ground-level odor concentration. In a case where it is desirable to reduce odor level at a mile or so from the source, it is unlikely that releasing the odor from an elevated stack would be effective. As a practical matter, an elevated source would be more effective at distances out to several hundred meters from the source.
• The problem with Tedlar bags removing organic acids from the bag was a serious hindrance in pushing the nanonose measurement technique to field application. The problem also has serious implications because of the widespread use of this method for odor quantification. Unless hog-odor samples in Tedlar bags are analyzed within a few hours after being collected, the organic-acid contribution to the odor will be greatly reduced. By mass, organic acids make up over 90 % of constituents of hog odor. The common observation that odor stays on clothes and hair after only a brief exposure to a hog barn atmosphere underlines the fact that odor attaches to surfaces. The same odor constituents that attach to clothes are likely to be the constituents that attach to Tedlar bags as molecular polarity is a key underlying property. If so, this implies that an odor sample belatedly analyzed from a Tedlar bag could be apparently deficient in commonly recognized hog-odor components, and the process of measurement of odor from Tedlar bags may drastically under represent odor concentration.
• The nanonose measurement instrument is successful from the standpoints of calibration with a laboratory standard and reproducibility of results. The method still must be compared with olfactometry, using Teflon sample bags, to establish whether it can be quantitatively correlated with olfactometry. Using off-the-shelf components, the cost of the device would approach $75 thousand. The device is much more difficult to operate than is an olfactometer. However, a single analysis can be made in 1 min after the instrument is set up. Potentially hundreds of samples could be done during a day compared to only a dozen or so with olfactometry.

*Prepared for presentation at University of Illinois Pork Industry Conference, Champaign, IL, December 11-12, 2003.

Resolution of the micro model is 100 m. In the vertical dimension the thickness of the lowest layer was scaled from 75 m to 1 m. The micro model describes pollutant transport and chemical transformation given the meteorological conditions.