Geostatistics as a basis to the CMLS pesticide simulation model with validation in soil columns

Abstract

The use of simulation models is probably the most efficient means for predicting the behavior of pesticides in the soil-plant-water system. The CMLS (Chemical Movement in Layered Soils) simulation model for predicting the fate of pesticides was used for studying the behavior of tebuthiuron, a herbicide used in sugar cane crops, from a sampling grid with 111 sampling points 200 m apart from one another and encompassing three types of soil: Ustic Quartzipsamment, Rhodic Hapludox and Typic Hapludox, all with medium and clay textures. The 373 points assessed by the simulator, generated from samples coming from the original grid and through the geostatistical methods of variography and ordinary kriging, returned the depth values reached by the herbicide after six years of simulation (1989-1995). For the Ustic Quartzipsamment, tebuthiuron, in four simulated points, returned depth values above 43 m and a maximum 50 m, with a certain amount of the product still remaining in the soil that was close to 10% of the original 1.1 kg ha-1 applied. Results from the column assay used for validating the model showed that the model overestimated the depth reached by the herbicide in 6.6% as compared to the column value for the Ustic Quartzipsamment. The depth was underestimated in 4.5% and 20% for the Typic Hapludox and the Rhodic Hapludox, respectively. These data support the adequacy of the model for assessing the fate of tebuthiuron in both Ustic Quartzipsamment and Typic Hapludox.