Simulation of Three Untarped 1,3-Dichloropropene Flux Studies Using HYDRUS
Abstract
Post-application 1,3-dichloropropene (1,3-D) volatilization data from three untarped field studies were simulated using HYDRUS vadose zone models. The volatilization fluxes in the original studies were estimated using the aerodynamic (AD) method. HYDRUS simulated fluxes were calculated using measured or independently estimated input data with the exception that the soil bulk degradation coefficient was an adjustable parameter. The best-fit soil degradation half-lives were similar across all three studies, ranging from 5.1 d to 5.8 d, even though soil types were quite different. The timing and magnitude of HYDRUS-simulated 1,3-D flux densities were within the range of uncertainty of the AD-estimated flux densities in one of the three studies. The peak AD-estimated flux density in the second study occurred several days before the peak HYDRUS modeled flux density, and the cumulative modeled flux was similarly delayed. In contrast, the peak AD-estimated flux and cumulative fluxes in the third study were much later than their corresponding HYDRUS-estimates. Several potential reasons for the deviations between AD-estimated and HYDRUS-modeled flux were identified, including:
• Uncertainty in the AD-estimated fluxes.
• Uncertainty in soil effective vapor phase diffusion coefficients. These coefficients are calculated internally by the HYDRUS based on a user-specified tortuosity model. Those tortuosity models are subject to prediction error, particularly in the wet soil range.
• Inadequate characterization of soil physical properties due to inadequate pre-application soil sampling, leading to inaccurate model parameterization or initial conditions.
• Extensive spatial variability in soil properties.
• Sampling error in field measured concentrations (contributing to flux estimation error).
In future studies, some of these problems can be minimized by increased sampling to better characterize soil properties and water content. However, other problems, such as those related to extreme spatial variability, may require alternate modeling approaches such as probabilistic modeling. Across all three studies, peak modeled fluxes were 71% of the AD-estimated fluxes (range 53 – 81%). However, the AD-estimates themselves are subject to substantial error. Additional work to understand the magnitude and sources of flux estimation error would be useful.