Research Lines

​My research interest is in Artificial Intelligence (AI) applied to earth science. I have been dedicated to study machine learning algorithms and physics-based models integrated to autonomous systems (aerial, surface or aquatic) to develop the next generation of intelligent models and robots. My research has been driven by my own curiosity on how machines can learn to obtain rapid data, infer earth system patterns, and represent complex processes in nature. I aim to innovate the state of the art in contextual modeling to create new methodological frameworks that move the science forward through numerical models and intelligent system that learn from limited data sets under limited time. Previous AI algorithms, based mainly in statistical representations, are limited in reasoning capability, rely in large data sets for prediction, and sometimes, training data creates maladaptation. I work on novel scientific frameworks where physics-based models are merged to machine learning algorithms to provide solutions to earth science problems. My goal is to achieve a deep understanding of processes through context adaptation models capable of perceiving, learning, abstracting and reasoning. These AI models over time can build underlying self-explanations and make decisions that allow them to characterize real-world phenomena and understand why they are making those decisions.

Eddy resolving three-dimensional simulations combined with field and laboratory observations allow elucidation of key elements of fluid dynamics and sediment transport in fluvial environments. In this research, I develop a state-of-the-art Computational Fluid Dynamics (CFD) model for the study of sediment transport and morphodynamic processes in fluvial systems. My primary goal is to achieve a deeper understanding of fluid and sediment motion in river systems, and how these motions are represented at different spatial scales. Thus, the developed multi-physics model is capable of predicting sediment transport and bed evolution

change, while resolving the energetically important eddies in the flow. The model has been validated in a large-scale canyon-bound river and can be used for a range of applications in water resources over a wide range of scales, from laboratory scales to large rivers. This physically-based model is also used as a prediction tool to understand and forecast the fundamental physics of erosion and deposition processes in river networks including different scenarios that will affect the river morphodynamics. Advances in our understanding of the role played by macro-turbulence events in sediment dynamics provide deeper insights into what determines the geomorphologic change in rivers. The results of this research lead to the development of more efficient models that can be used in lower-order models commonly employed to predict bed evolution in loose bed channels and river reaches (e.g., 3-D Reynolds averaged Navier-Stokes or 2-D depth-averaged models). Figure 1 shows the last step of a numerical simulation of the flow and concentration of sediment along the Colorado River in the Eminence Break fan-eddy complex.

I study erosion of river sandbars that examines the impact of dam operation criteria on the bank slope stability leading to mass failure through seepage. This research work is developed by means of full-scale laboratory experiments to simulate the sandbar beaches of the Lower Colorado River Basin in Marble and Grand Canyons. Findings reveal that there is no significant correlation between diurnal stage fluctuation caused by the Glen Dam operation criteria and slope stability. The erosion of sandbars with a slope greater than the equilibrium slope is inevitable and after the bars reach the equilibrium slope, mass failure and seepage erosion cease. These findings are essential for decision-makers and scientists to restrict up ramp and down ramp river stage rates caused by the Glen Canyon Dam operation rule.

​

​

​

​

​

​

​

​

​

​

​

​

​

​