UC Riverside

Contaminants in water at the soil surface can be rapidly transported to streams or locations of surface water storage by runoff. This has been reported to be the primary transport route for contaminants on sloping fields and hillslopes. Generally, contaminants are transported faster and in larger amounts with runoff, and much slower and in smaller amounts with infiltration. Runoff transports are strongly affected by surface topography, soil properties, vegetation, and weather. The risk of transport is also highly dependent on the contaminant load at the soil surface, interactions with soil particles, and transformations. An understanding of and the ability to predict processes that influence the transport and fate of contaminants in runoff water are therefore needed to assess and mitigate risks of contamination of surface water supplies on human health.

Physically-based, spatially-distributed models have the potential to be an efficient tool to examine and optimize the removal of contaminants from agricultural runoff through land-use changes and best management practices. In this research, the existing subsurface version of HYDRUS-1D was adapted to simulate uniform or physical nonequilibrium flow and reactive solute transport processes during runoff at the soil surface. The numerical results obtained by the new model produced an excellent agreement with an analytical solution for the kinematic wave equation. Additional model tests further demonstrated the applicability of the adapted model to simulate the transport and fate of many different solutes (non-adsorbing tracers, nutrients, pesticides, microbes, and sediments) that undergo equilibrium and/or kinetic sorption and desorption, and first- or zero-order reactions.

Along with PBMs, data-driven models are becoming increasingly popular for describing the behavior of hydrological and water resources systems since these models can be used to complement or even replace physically based-models when there is a lack of required data. Here we propose a new data-driven model as an alternative to a physically-based overland flow and transport model. Several machine learning techniques including Linear Regression (LR), k-Nearest Neighbor regression (kNN), Support Vector Machine with linear (SVM-L) and non-linear (SVM-NL) kernels, and Deep Neural Networks (DNN) (Neural Networks with multiple hidden layers) were explored to find input - output functional relations. The results indicated that the Deep Neural Network (DNN) model with two hidden layers performed the best among selected data-driven models. This DNN model accurately predicted runoff water quantity over a wide range in parameters. It also predicted well runoff water quality for near-equilibrium solute transport over a wide range in parameters.