Environmental Systems: Complexity and Uncertainty
A vital system, the environment is also one of the most challenging to simulate. Better predictions of environmental impacts and how to mitigate them are among society’s most pressing future needs. Meeting them is all the more challenging because environmental systems entail a complex, dynamic interplay among many
- chemical, and
- biological processes.
Moreover, their material properties are spatially heterogeneous, meaning they are unevenly distributed in various concentrations in the physical world. Reliable environmental data, particularly the sub-surface variety, is key to prediction, but it is scarce and notoriously difficult to acquire. Hence, environmental system simulations are inherently uncertain and require real-time calibration and statistical concepts to support risk assessment and decision making to make them useful for environmental engineering and society.
An Integrative, Interactive Approach
In a single integrative, transferable and efficient computational approach, we combine
- simulations under uncertainty,
- data assimilation,
- and risk assessment.
This approach provides an interactive framework for different application tasks, such as calibration, prediction, design and control under uncertainty.
As a demonstration project to illustrate the progress we have made toward interactive environmental engineering, we have developed a simulation of carbon dioxide (CO2) injection into deep geologic formations to mitigate global warming.
Assessing the potentials and risks of this technology as an interim solution requires predicting the multi-scale, multi-physical processes involved,
- quantifying uncertainties,
- designing, controlling and visualizing robust injection strategies, and
- communicating the results at the science-policy interface.
Read more below about the four key elements that make up our vision for developing interactive environmental engineering:
A highly complex challenge
Carbon capture and storage (CCS) has been proposed as an interim technology for reducing CO2 emissions. The storage and subsurface flow of CO2 after injection is governed by complex and non-linear multi-physics and multi-scale laws:
- unstable multiphase flow
- phase transitions (fluid-gas-super-critical)
- dissolution into water
- rock deformation
- temperature effects
CCS requires highly-developed simulation tools
- efficient and adaptive discretization schemes and
- numerical algorithms for multi-physical and multi-scale processes.
The resulting codes are scalable to future high-performance computing architectures and equipped with total error control for accurate yet robust simulation.
Blind perfection carries a high price
Current numerical simulation models for the multi-phase flow and transport processes are not up to the demands of repeated simulations required for uncertainty quantification, optimization, control and risk assessment. Even a single simulation run may require parallel high-performance computing. Reduced models can drastically speed up
- control, and
- risk analysis.
Despite this reduction, they retain all relevant aspects of the original model with only a small, controlled error.
Response surface approach
We compress the reaction of complex simulations to parameter changes into a simple, yet smart polynomial response surface using so-called arbitrary polynomial chaos expansion. All repetitive follow-up tasks evaluate the response surface in the blink of an eye.
Inconsiderate confidence in simulation is dangerous
To assess the potential of CCS as interim solution to global warming, ideally, we should know all possible reactions of the environment to CCS. Uncertain parameter values can change simulation outcomes by factors of 10, 100 or more. Therefore, our ability to quantify its uncertainties and risks play a key role and may even be more important than detailed numerical perfection of simulation codes.
Knowing the unknown
Our reduced models allow applying accurate and holistic methods for uncertainty quantification and using them in probabilistic risk assessment. This way, we can predict the likely extent of the benefits and harms we inflict on the environment and then make them inputs to decision making.
Even hightech can fail
CCS should be implemented in a manner that confines even improbable dangers below acceptable risk thresholds. For example, one could choose the injection rate and depth to avoid geomechanical cracking at a 99% safety level. The probability of failure and negative impacts will depend on the interplay between
- controllable engineering aspects and
- the system’s uncertainty aspects.
Optimizing with rational safety margins
Robust design directly impacts uncertainties and failure probabilities. It is a way of minimizing operating risks. Our proposed integrative approach simply builds design variables into the response surface. This assures optimization while guaranteeing that known risks remain below a level designated as acceptable. Our optimized monitoring strategies help to identify and reduce residual risks.
Research on CCS is inseparable from these social dimensions:
- consumer behaviour and population growth
- technological development and economic/ ecological aspects
- sustainability and risk acceptance
- opinion making and legislation
Without adequate visualization and communication, simulation results, because they can be misinterpreted in dangerous fashion, can easily become counterproductive if they become political footballs.
Bridge to decision makers
Simulations should not be understood as “truth-generating machines.” Decision makers are prone to underestimating the role of uncertainties in environmental simulations. Hence, SimTech studies the process of communicating simulations to policy makers.
Our visualization methods facilitate exploring possible CCS scenarios graphically and interactively to include their environmental impacts and uncertainties.