Dynamic Simulation of a Tailing Storage Facility

The below article was written by Joseph Donnelly (Hatch).

Introduction

Within the mining industry, GoldSim is frequently employed in the simulation of surface impoundments used for long-term storage of mine tailings.  These impoundments, or tailings storage facilities (TSFs), commonly receive waste output from ore processing facilities in slurry form consisting of fine-grained particles and water.  As the slurry is deposited within the TSF, solids settle forming a beach while the water collects down-gradient in a water reclaim pond.  As part of operations, the location of slurry deposition is changed over time in order to control the slope and shape of the tailings beach as it forms.   The variation in slurry outfall locations results in a continually changing geometry of the tailings beach and the associated water reclaim pond.  

The variable geometry of TSFs can be difficult to represent within a dynamic predictive model.  This article demonstrates a methodology that can be used to simulate the dynamic geometry (stage-storage-area) of a typical TSF.  The methodology incorporates an initial geometric configuration based on available survey information imported from a single two-dimensional lookup table within GoldSim.   Further, a simple forecasting method involving offsets to the final survey is provided based on an assumed tailing density and deposition footprint. In summary, the method includes:

• Incorporation of historical bathymetric surveys
• Forecasting capabilities beyond the last bathymetric survey
• Demonstrate the modeling approach using hypothetical data in a GoldSim model.

Incorporating Bathymetric Surveys

Active mining operations typically have multiple surveys of the TSF.  

The survey data can be used to generate simple two-dimensional relationships representing stage-storage-area for the TSF at the date of each survey.  In order to represent these relationships dynamically in GoldSim, we have developed the following approach:

STEP 1 – Interpolate measured survey data so that all stage-storage-area data can be aligned on consistent increments at the date of each survey as shown in Figure 1.  

STEP 2 – Import the stage-storage data using a 2D lookup table in GoldSim (see Figure 2).  The water level within the TSF can be estimated from the dynamic water volume within the TSF and the dynamic stage-storage data. The first argument in the following selector demonstrates the use of a 2D lookup table.

Figure 1 - Spreadsheet image of four interpolated stage-storage curves taken from hypothetical survey data

Figure 2 - Importing and referencing interpolated bathymetry in a GoldSim model

Forecasting TSF Geometry

Detailed deposition planning can be used to generate future stage-storage-area curves for a TSF based on assumed tailing density, beach slopes and accumulated tailing.  When this type of data is available, it can be incorporated into a GoldSim model in the same method previously discussed.  The following method provides a simplified alternative based on an offset from the last available survey data.  The method is outlined in the following steps:

STEP 1 - A dynamic TSF offset is calculated based on the estimated deposition area footprint, tailing density, and accumulated tailing deposition since the time of the last TSF survey. Figure 3 is a screen capture of how this might be accomplished in GoldSim. 

STEP 2 – The TSF offset is added to the last stage-storage area curve to estimate future water elevations using a Selector in GoldSim.  This is demonstrated in the second argument of the selector in Figure 2, above.  Conversely, the TSF offset can be subtracted from future levels to lookup the water storage volume, as demonstrated in the second argument of the selector shown in Figure 4.

Figure 3 - GoldSim model calculation of a TSF offset from the last available survey data.

Figure 4 - Importing and referencing interpolated bathymetry in a GoldSim model.

Validation of Forecasting Geometry

TSF survey data from an anonymous site was used to validate the proposed forecasting methodology.  Nine TSF surveys were provided spanning over a period of 4 years.  Static tailing density and deposition footprint were estimated for the facility and the model was used to forecast eight stage-storage curves offset from the first curve.  Results of the analysis are provide in Figure 5. Graphically, the method tends to predict future water surface elevations given only this minimal input data.  There is some error at lower water storages; however, given that the lower areas are typically pooled with water and difficult to accurately measure, the results appear to represent future survey data.

Figure 5 - Forecasted versus measured TSF stage-storage data over a four year validation period.

Results

The historic and forecasting TSF geometry was used in a hypothetical model to demonstrate these methods of representing TSF geometry.  The conceptual model of the TSF is provided below with reference to Figure 6, with the following components:

• Four historic TSF surveys and measured water levels.
• Tailing deposition with assumed density and deposition footprint.
• Probabilistic simulation of precipitation, snow pack, snow melting, and runoff. 
• Decant pumping from the reclaim pond.
• Evaporation from beach and pond areas. 
• Water lost to entrainment and seepage.

Figure 6 - Conceptual flow model of a hypothetical TSF.

A one year calibration period and three year forecasting period was simulated. Custom statistics were used to illustrate the dam crest height, freeboard, measured water levels and probability based water levels in a single time series plot referenced to Figure 7.

Figure 7 - Probabilistic simulation of a hypothetical TSF indicating decreased PMF storage between dam crest raises and freeboard encroachment for the 1%, 50% and 99% of Monte Carlo results.

Conclusions

Two methods are provided to predict stage-storage data for typical TSF.  The first method can be used to easily import existing TSF survey data into a model for historic simulation and calibration.  An advantage of this method is that new data can be added to the right of the 2D lookup table the model will automatically import the data for model updates.

The second method forecasts the stage-storage data. The method is demonstrated to represent TSF configurations for up to four year periods.  

For more information, please contact the author using the contact information shown below.

Joseph Donnelly, P.E., Specialist Water Resource Engineer │ 4715 Innovation Drive, Suite 110, Fort Collins, CO 80525 │ joseph.donnelly@hatchusa.com │ 970.225.8222