Impact of basic soil parameters on pesticide disappearance investigated by multivariate partial least square regression and statistics.

Dissipation time is a key parameter when studying and modeling the environmental fate of pesticides. This study was conducted to characterize the variability of pesticide disappearance in soil and to identify possible controlling parameters related to intrinsic soil properties and microbiology. Multivariate data analysis was used to study spatial variability in three horizons from 24 sandy soil profiles. The time for 50% disappearance (DT(50)) was characterized for two herbicides, metribuzin (MBZ) and MCPA, and methyltriazine amine (MTA; transformation product of metsulfuron-methyl, tribenuron-methyl, thifensulfuron-methyl, and chlorsulfuron). Normal and log-normal distributions were compared for DT(50) and soil properties and descriptive statistics were calculated. Conformity with log-transformed distributions was observed and assuming normality of the DT(50) data would cause 5 to 35% overestimation. Mean DT(50) were: MCPA 9.5, MBZ 168, and MTA 127. Significant effect of soil depth on DT(50) was shown for MCPA and MBZ, with low values in deeper horizons. Simple linear correlation for combinations of MCPA, MTA, pH, and total organic carbon (TOC) was observed. Using partial least squares regression (PLS) 71 to 85% of the total DT(50) variance was explained. A specific predictor variable could not be identified as the controlling components differed within horizons and compounds. For MCPA the overall important predictor variables were microbiology and TOC, whereas for MTA and MBZ it was inorganic variables (Al, Fe, cation exchange capacity, base saturation percent, and pH) and microbiology. The study indicates that PLS generated input data can improve pesticide fate modeling and reduce the uncertainty in dissipation estimation.