Long-Term Tracer Study at Los Alamos, New Mexico. Part II: Evaluation and Comparison of Several Methods to Determine Dispersion Coefficients

Abstract

A long-term tracer study was conducted to evaluate dispersion in complex terrain in support of air quality and emergency response initiatives. An array of monitors continuously measured external radiation levels at different directions and distances from a 30-m-high radionuclide emission source. The monitors were located at distances ranging from 550 to 2800 m downwind of the release stack. Winds were measured near the source and 700 m downwind, across a deep canyon. A Gaussian model using simplified and improved Draxler dispersion coefficients was used to estimate concentrations. Daily predicted values closely agree with measured values, while the correlation coefficient nearly equals 0.9. Model results differ somewhat depending on the wind data used, with the near-source wind data providing the best agreement. The routine difference in winds between the source and receptors across the canyon are demonstrated to be important in estimating long-term concentrations in specific sectors downwind. Comparison of predicted and measured values at receptor locations across the canyon with those on a mesa confirms the validity of the widely used reflection factor in the Gaussian model. Comparison of daily predicted with measured values were also made for several methods that estimate dispersion coefficients. The widely used Pasquill-Gifford (P-G) sigma curves give the lowest correlation (r = 0.76) and greatest underprediction (42%) of measured radiation levels at night. The ratio of daily predicted to measured values improves from 0.82 to 0.88 when coefficients from the next highest stability category are used. This case also provides closer agreement of day and night predicted-to-measured ratios of 0.80 and 0.94, respectively. The use of urban coefficients provides too much dispersion. The P-G dispersion coefficients are also representative when σ? is used to determine the stability category. However, the use of σ?-determined P-G dispersion coefficients results in too little dispersion, as daytime predictions overestimate and nighttime predictions underestimate measured levels by 37% and 19%, respectively. The direct use of turbulence is superior to all of the indirect methods studied. Comparisons of predicted with measured 15-min values were made at 15 sites. Predicted values agree with measured values by a factor of 2 only 50% of the time when the wind direction was strictly applied in the model. However, the frequency of agreement within a factor of 2 improved to 80% when the wind direction was allowed 5° of uncertainty. An uncertainty of 20° allows reliable predictions of levels at even the bottom of an adjacent canyon. Peak-to-mean ratios of measured values were also examined at three primary sites during 1986. The ratios increase slightly from the outside edge toward the center of the plume.