An efficient space-fractional dispersion approximation for stream solute transport modeling

Chaopeng Shen, Mantha S. PhanikumarAdvances 2009 · DOI

TL;DR

This paper introduces an efficient numerical method for modeling solute transport in streams, particularly addressing how solutes spread out (dispersion). They propose a new approach that accurately captures non-Fickian dispersion observed in natural streams, offering a better balance between computational accuracy and speed than existing methods. This is crucial for improving predictions of contaminant transport and water quality in rivers.

What Problem Does This Paper Address?

Traditional models of solute transport in rivers, like the transient storage (TS) model, often rely on the classical advection-dispersion equation (ADE). However, field observations show that solute plumes in natural streams spread out faster than predicted by the ADE (non-Fickian behavior), exhibiting heavier tails and faster initial breakthroughs. This discrepancy arises because the ADE assumes a simple, local diffusion process, while real streams involve complex, long-range correlations in solute movement. Space-fractional dispersion equations offer a more promising framework to capture these non-Fickian phenomena. However, existing numerical methods for these equations can be computationally expensive or lack accuracy, hindering their practical application for parameter estimation in complex river systems. This paper aims to bridge this gap by developing and evaluating an efficient and accurate numerical scheme for space-fractional dispersion within a TS model.

Methods

The core of this paper lies in developing and comparing numerical methods for solving the space-fractional dispersion equation. The authors explore three main approaches for approximating the fractional derivative:

  • Grünwald-Letnikov (GL) Approximation:
    • Fully-Implicit (GLFI): A scheme where all terms in the fractional dispersion equation are treated implicitly, leading to a super-triangular system that is computationally intensive.
  • Three-Point Implicit (GL3I): A semi-implicit scheme that treats only a few neighboring points implicitly, resulting in a tri-diagonal system solvable by the Thomas algorithm. However, this scheme exhibits a stringent time-step restriction and can lead to significant errors.
  • Caputo Definition-Based Approximation:
    • Three-Point Implicit (CP3I): This scheme is derived from a finite-volume approach and uses the Caputo definition of the fractional derivative. It also results in a tri-diagonal system and is found to be more stable and accurate than the GL3I scheme.

To further enhance accuracy, the authors employ Richardson Extrapolation (RE) with the GL3I and CP3I schemes, significantly improving their spatial accuracy. For the advection component of the transport equation, they utilize a fourth-order accurate compact scheme, which is known for its good performance in simulating advection-dominated flows. The overall approach is an operator-splitting technique, where advection, fractional dispersion, and any reaction terms are solved sequentially.

Datasets

The paper uses two types of datasets for evaluation:

  1. Analytical Solutions: The authors compare their numerical schemes against exact analytical solutions for the pure fractional diffusion equation and the fractional advection-dispersion equation (FADE) with zero advection.
  2. Field Tracer Data: The proposed Fractional-in-Space Transient Storage (FSTS) model is applied to tracer test data from the Grand River in Michigan. This dataset involves Rhodamine-WT (RWT) tracer measurements at four downstream sites over a 40 km stretch.

Key Insights and Ideas

  • Non-Fickian Dispersion is Real and Important: The paper strongly reinforces that natural streams exhibit dispersion that deviates from the classical Fickian model, and accurately capturing this is key to realistic solute transport modeling.
  • Caputo vs. Grünwald-Letnikov: The Caputo-based approximation (CP3I) generally offers a better balance of accuracy and computational efficiency compared to the Grünwald-Letnikov approximations (GLFI, GL3I), especially when dealing with the fractional tail terms.
  • Richardson Extrapolation is Crucial: For semi-implicit schemes like GL3I and CP3I, Richardson extrapolation is vital for achieving high spatial accuracy and reducing errors, particularly for smaller fractional orders.
  • GL3I's Stringent Time-Step Restriction: The GL3I scheme, while seemingly computationally cheap due to its tri-diagonal system, requires extremely small time steps to maintain acceptable accuracy, making it impractical for many real-world scenarios.
  • FSTS Model's Parsimony: Incorporating fractional dispersion into the TS model (FSTS) allows for a more parsimonious model, meaning fewer parameters might be needed to achieve a good fit to tracer data, especially when compared to standard TS models with a = 2.

Key Results

  • Accuracy: The CP3I-RE scheme, in combination with the compact advection scheme, demonstrated excellent agreement with analytical solutions for both pure fractional diffusion and FADE. It consistently produced more accurate results than the GL3I-RE scheme, especially for higher fractional orders.
  • Computational Efficiency: The CP3I-RE scheme offered a significant advantage in computational time compared to the GLFI-RE scheme, which involves solving a super-triangular system. While GL3I was faster in raw computation, its accuracy limitations at larger time steps made it less viable.
  • Grand River Application: The FSTS model, using the CP3I-RE scheme, successfully described the Rhodamine-WT tracer data from the Grand River. The model achieved a good fit with a parsimonious set of parameters, suggesting that fractional dispersion can better capture the extended residence times observed in natural streams.
  • Order of Accuracy: The CP3I-RE scheme achieved approximately 1.2-order accuracy in space, a notable improvement over the first-order accuracy of the base semi-implicit schemes.

Limitations and Future Work

  • Limitations of GL3I: The paper highlights the severe time-step restriction of the GL3I scheme, making it impractical for many applications where solution quality is paramount.
  • Extension to 2D: The current work focuses on one-dimensional transport. Extending the developed schemes to two-dimensional problems, especially with varying fractional orders in different directions (multiscaling FADE), is noted as a challenge.
  • Parameter Estimation: While the CP3I-RE scheme is deemed suitable for parameter estimation, further exploration of its performance in complex inverse modeling scenarios would be beneficial.
  • Non-Conservative Solutes: The current model is primarily for conservative solutes. Incorporating reaction terms for non-conservative transport would be a valuable extension.
  • Further Investigation of GL3I Behavior: The paper notes the unusual error behavior of the GL3I scheme as step size decreases and suggests this warrants further investigation.

Summary updated

BibTeX
@inproceedings{shen2009efficient,
  author = {Chaopeng Shen and Mantha S. Phanikumar},
  title = {{An efficient space-fractional dispersion approximation for stream solute transport modeling}},
  booktitle = {Advances},
  year = {2009},
  doi = {10.1016/j.advwatres.2009.07.001},
}
Figure