Upscaling reactive transport in heterogeneous porous media
The biggest problem with existing (upscaled) non-Fickian transport models is that they require a tracer test to determine model parameters. Even when those parameters are identified, they are not guaranteed to be transferable and may not even be representative of different scales at the same site. I have been working on developing approaches that formally upscale transport processes by taking into account the small-scale spatial-statistics of the hydraulic conductivity field including specifically the marginal distribution of hydraulic conductivity and its spatial statistics.
This concept is merged with realistic assumptions about data support and availability to produce a generalized transport model that is conditioned to hydrologic observations, without requiring a tracer to constrain the model. The immediate advantage of the resulting framework is faster estimates of upscaled transport behaviors compared to an ensemble of distributed model realizations, and we expect the two approaches will have comparable uncertainty. The correlation model also accounts for changes in the proximity of particles, which provides a more accurate understanding of mixing and reactions, especially when nonlinearities are involved. The expected significance of this project is that it will improve the accuracy of upscaled estimates of transport and reactivity across diverse hydrogeologic settings and this has proven elusive to the greater community. I expect that the improved estimates of solute transport afforded by this project can be applied to a wide range of contamination problems, ultimately reducing both the time and cost to remediate subsurface contaminants.
Publications:
A. Massoudieh and M. Dentz, (2020), Upscaling non-linear reactive transport in correlated velocity fields, Advances in Water Resources, 143, 103680, https://doi.org/10.1016/j.advwatres.2020.103680.
Massoudieh A., M. Dentz, J. Alikhani, (2017), A spatial Markov model for the evolution of the joint distribution of groundwater age, arrival time, and velocity in heterogeneous media, Water Resources Research, 53, pp 5495–5515, doi:10.1002/2017WR020578.
Optimal control in engineered water systems using AI and machine learning
Reinforcement Learning (RL) is a branch of machine learning that has shown a great deal of promise in a wide range of applications including computer games, and robotics. The goal of RL is to discover optimal control strategies composed of actions at any given time that result in the maximization of a reward function representing the long-term costs and benefits of the system. There are potential applications of RL in optimal control of a wide range of water problems including but not limited to determining ecological flows in a water body, wastewater treatment processes, and integrated sustainable stormwater management. Even though RL has been gaining traction in many areas of Engineering, the applications of RL in the area of water engineering has not been explored enough. In the past couple of years, I have been developing tools aiming at utilizing RL in finding optimal operation strategies for 1) biological wastewater treatment and 2) smart and integrated BMPs for stormwater management. For example, due to the deep uncertainty regarding the inputs and the processes in wastewater treatment, determining the optimal operation for example the rates of aeration, the addition of external carbon source for denitrification, rates of return, and waste flows are challenging.
Publications:
K. N. Ngo, Winckel, T. V., Massoudieh, A., Wett, B., Al-Omari A., Murthy, S., Tak ́acs, I., and Hayd́ee De Clippeleir, (2020), Towards more predictive clarification models via experimental determination of flocculent settling coefficient value, Water Research, 116294, 2020.
Le, T., Peng, B., Su, C., Massoudieh, A. Torrents, A., Al‐Omari, A., Murthy, S., Wett, B., Chandran, K., DeBarbadillo, C., Bott, C., De Clippeleir, H., (2019), Nitrate residual as a key parameter to efficiently control partial denitrification coupling with anammox, Water Environ Res, 91: 1455-1465. doi:10.1002/wer.1140.
Le, T., Peng, B., Su, C., Massoudieh, A., Torrents, A., Al‐Omari, A., Murthy, S., Wett, B., Chandran, K., DeBarbadillo, C. and Bott, C., (2018), Impact of carbon source and COD/N on the concurrent operation of partial denitrification and anammox. Water Environment Research, 91(3)- pp 185-197.
Developing agile tools for predicting the behavior of natural and engineered systems
Tools developed for modeling environmental systems in general and water systems in particular often focus on a single domain with a uniform set of governing equations applied to the entire domain. However, many problems involving natural or engineered water systems involving flow, mass transfer, aquatic ecology, particularly when natural systems are affected by engineering infrastructure, require considering interactions between multiple domains, each governed by different sets of processes, models, and governing equations. The application of disjoint domain-specific models limits our ability to take the interactions between these different domains into account. These points highlight the importance of developing modeling frameworks that provide maximum agility, flexibility, and transparency to adapt to a wide range of target problems. One way to achieve this is to avoid hard-coding of the model governing equations to the greatest extent possible and allow users to easily inspect and alter the governing equations used to describe various model components. One area of my research is developing agile and extensible modeling tools for predicting flow, transport, and biogeochemical processes in systems including multiple natural or engineered domains. I have developed GIFMod an agile tool for modeling hydraulics and water quality processes in urban stormwater systems. Recently, OpenHydroQual is a plug-in-based and extensible framework for developing a wide range of water systems.
Some instructional videos on OpenHydroQual can be found here: https://www.youtube.com/channel/UCJ3b55fpMnAam3B5besdHmg