A mechanistic model to predict soil thickness in a valley area of Rio Grande do Sul, Brazil

Soil thickness is an important soil characteristic changing over space and time. In this study, we used a mechanistic soil landscape models to predict soil thickness and show it under development over time. The study was conducted in an 8,118 ha area in Vale dos Vinhedos, Rio Grande do Sul State, Brazil. Different soil production functions (SPF) combined with a landscape evolution model (LEM) were explored. The SPF calculated the soil production rates and LEM calculated erosion and deposition patterns. We evaluated two types of model. Model 1 was used to predict the current soil thickness. The model equals the erosion estimations (by a LEM) to the soil production rate (by a SPF). Three types of SPF were tested, based on a spatial variation of soil moisture. A steady-state condition was assumed, considering soil production rates similar to erosion rates. The model simulated erosion events to 1 year, using a Digital Elevation Model (DEM). A soil survey with observed soil thickness was used to validate the different models. Model 2 used the soil thickness estimation from Model 1 to simulate the soil thickness changes up to 100 kyr, considering the balance between soil production rate and soil eroded or deposited. The soil thickness changes were evaluated in different landscape positions. In Model 1, the linear correlation between observed and predicted soil thickness varied between 0.25 and 0.49, with higher linear correlation in models using soil moisture data. The RMSE under different models varied between 34 cm and 37 cm. Overall, soil depth was more accurately predicted in the upland areas than in the valley bottom areas. Model 2 suggested that the soil thickness variation largely depended on the landscape position. The average soil thickness changed from initial 67 cm (0 Kyr) to 103 cm (100 kyr).