LDI-MVFNet: A Multi-view fusion deep network for leachate distribution imaging

Area of Science:

• Environmental Geophysics
• Hydrogeology
• Solid Waste Management

Background:

• Groundwater pollution from landfill leachate is a global challenge.
• Electrical Resistivity Tomography (ERT) offers non-destructive monitoring but single arrays have limitations in complex media.
• Accurate detection of concealed landfill leakage is crucial for groundwater protection.

Purpose of the Study:

• To develop and validate a novel deep network for multi-view fusion (LDI-MVFNet) to invert subsurface resistivity.
• To infer leachate distribution and dynamics more accurately than traditional ERT methods.
• To improve the monitoring and early warning systems for groundwater pollution.

Main Methods:

• Designed a novel deep network, LDI-MVFNet, for multi-view fusion of ERT data.
• Validated the LDI-MVFNet using synthetic models and a salt tracer experiment.
• Compared LDI-MVFNet inversion results against single-array ERT methods (Dipole-Dipole, Wenner-Schlumberger, Pole-Pole).

Main Results:

• LDI-MVFNet demonstrated superior performance in inverting real resistivity distribution.
• The average RMSE for synthetic models was 0.98, significantly outperforming single arrays (e.g., Dipole-Dipole at 3.86).
• 2D cross-section imaging showed LDI-MVFNet's superiority in noise suppression and inversion accuracy.

Conclusions:

• Multi-view data fusion in ERT significantly enhances the accuracy of subsurface resistivity inversion.
• LDI-MVFNet provides a more reliable method for detecting and mapping groundwater pollution from landfill leachate.
• The findings support the advancement of solid waste management and groundwater protection strategies.