Precipitation forecasting for tailings storage facility water management

Understanding precipitation accumulation patterns is critically important to ensure proper infrastructure design, water management planning and risk management, and reliably inform decision-making for mine management. Statistical analysis of historical data is often used to forecast precipitation, but exclusive use of on-site weather station data is often insufficient because it does not consider regional climatology, and the period of record is usually limited. Further, while event-based risk (such as overtopping potential caused by a 100-year, 24-hour event) can be readily quantified, longer-term patterns are often only vaguely defined and generally not explicitly tied to probabilistic outcomes (for example: What if there is a wet year? What if there are two wet years in a row because of El Niño?). Because of this, the ability to characterize both short- and long-term historic precipitation patterns and incorporate this into probabilistic forecasts using fewer statistical models and maximizing the information gained from existing data is of great interest. This case study generally explains how to overcome this constraint through the development of future daily precipitation sequences with exceedance probabilities for multiyear durations, and how these can be used to support decisions on tailings management. While there are several methodological advantages to this approach, ultimately, these improvements facilitate clear and reliable estimates of long-term, multiyear risks to decision-makers. In this case study, site management needed to know not only what level of event-based precipitation accumulation would cause runoff volumes to encroach on established freeboard requirements in a tailings storage facility — that is, event-based 100-year, 24-hour, 72-hour probable maximum precipitation (PMP) — but also how multiyear precipitation patterns (and their associated probabilities) could create adverse conditions that could impact operations and dam safety. The facility evaluated in this case study has one on-site weather station, with a period of record limited to 12 years. Regionally, there are 17 weather stations near the area of interest that are candidates to be incorporated into model forecast scenarios. After review of the 17 regional weather stations, two regional weather stations near the operating mine were chosen, using criteria such as a long period of record (more than 100 years at each weather station), data quality, reasonable distance from area of interest, and similar climatology. Following the selection of these quality-controlled regional weather stations, daily precipitation levels from the selected stations were aggregated into nonoverlapping n-year duration totals (such as one-year, two-year, up to five-year accumulation horizons). Goodness-of-fit and correlation measures between aggregated regional data and on-site measurements were used to identify the best distribution for developing the exceedance probability (EP) curves. These curves were used to derive a depth-duration-frequency table — Table 1 shows cumulative precipitation scaled to on-site conditions for various durations and EPs. For example, in a five-year duration, the data (more than 200 years of station data) indicate only a 0.5 percent probability of exceeding 1,948 mm of total precipitation. Figure 1: Thirty precipitation patterns to accumulate 1,948 mm of total precipitation for five years and a 0.5% exceedance probability.(Source: Freeport-McMoRan). As a final step, 30 separate independent daily precipitation time series were extracted from the historical data for each EP and future duration based on how closely the cumulative precipitation in the sequence matched the respective EP precipitation total. Although we considered the selection of 30 time series in this example to be sufficient to capture the general variability of accumulating precipitation to a given total, more sequences could be developed, based on the number of station-years available, and in some instances the number of sequences may be limited by the available gauge data periods of record. An illustration of the five-year and 0.5 percent EP precipitation values is shown in Fig. 1. The selection of 30 different patterns provides a robust and data-based method of forecasting to account for variable temporal patterns to accumulate 1,948 mm and serves as a strong foundation for risk management and decision-making. The technical benefits of such an approach are that patterns present in the long-term record near a site of interest can be used, thereby making it no longer necessary to explicitly model the independence and time-series correlations of precipitation patterns (such as Markov chains, independent monthly precipitation probability distributions) that are commonly used in water balance modeling forecasts. From a management perspective, the first benefit of this approach is that site management was able to evaluate scenarios associated with different forecasted periods and their associated EPs, as shown in Table 1. The EP explicitly defines the probability of accumulating a certain amount of precipitation over a given time frame that is based on many station-years of data. A second benefit is that the EP and corresponding duration can be modeled using a variation of temporal patterns (30 sequences) to assess the systematic potential of exceeding a given criterion — for example, freeboard. Finally, this methodology allows the regional data to be used to their utmost and removes a requirement to statistically model climatologic patterns in model inputs. For example, this method intrinsically accounts for month-to-month and year-to-year correlations because of extended dry/wet periods caused by macroscale climate events. These patterns are very difficult to explicitly model, and any month-to-month and year-to-year patterns that emerge from modeling the climate series as independent monthly probabilities would only be random rather than part of a longer-term signal that is present in the regional data. To address the concern of site management regarding the likelihood of exceeding a freeboard limit, seven recurrence probabilities and five-time horizons (35 scenarios), shown in Table 1, were modeled, each using all 30 precipitation scenarios in GoldSim, version 12.1 (Gold Technology Group, Seattle, WA, 2020). This required 1,050 simulations. Furthermore, the simulations included a 72-hour PMP event during the later months of the simulation period. The summary of simulations can be seen in Table 2. To determine freeboard exceedance, there are three categories: (1) Unlikely, in green, (2) Potentially, in yellow and (3) Likely, in red. Unlikely is used when the average of 30 sequences does not exceed the criteria, and fewer than five sequences do exceed the criteria. Potentially is assigned when the average of 30 sequences does not exceed the criteria, but a total of five or more individual sequences exceeds the criteria. Likely represents the condition when the average of the 30 sequences exceeds criteria. Of the 35 scenarios, all but four were determined unlikely to exceed the minimum freeboard criteria of 1.5 m. The remaining four scenarios were estimated to potentially exceed the criteria. For the four potential cases highlighted in yellow, the numbers given represent the number of sequences out of 30 that exceed the criteria: The overall average freeboard for the 30 sequences did not exceed the criteria in these cases. Together, these results show that the current operating practices at the site are adequate to safely manage extreme multiyear precipitation accumulations combined with the 72-hour PMP for all durations at the 2 percent EP and greater. For the 1 percent and 0.5 percent EPs, the results show there is a potential for the freeboard criteria to be exceeded (for example, instead of 1.5 m of freeboard, the facility only has 1.3 m of freeboard) for the one-, three- and four-year durations, but only after a very low probability (1 percent, 0.5 percent) precipitation pattern and a rare extreme storm event: that is, PMP. Regional precipitation analysis is designed for use in generating daily precipitation patterns that are representative of future precipitation levels at a site. Utilizing data from regional weather stations overcomes the limitations of on-site data and quantifies the uncertainty over various durations. More importantly, the results in this case study indicate that the proposed approach reflects actual daily precipitation patterns and can be used to quantify risk going forward. Taken together, these findings significantly improve the capacity to forecast precipitation and can improve water management practices in the planning stages. This enables site managers to make proactive and informed decisions in consideration of precipitation accumulation patterns as well as extreme events, and their combinations.

Precipitation forecasting for tailings storage facility water management

Authored by: Homin Kim and Brant Whitling

To view the past issues of the UCA magazine you must be a member of SME/UCA.

Mining Engineering’s Weekly Rundown: Episode 73

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

Mining Engineering’s Weekly Rundown: Episode 73

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

To view the past issues of the UCA magazine you must be a member of SME/UCA.

Call for nominations: SME, SMEF, AIME and division awards

Inside MINEXCHANGE 2025: Industry innovations, insights and future trends

Related Articles

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

The Art and Science of the Flowsheet

UCA Division officer nominations

Mining Engineering’s Weekly Rundown: Episode 73

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

Mining Engineering’s Weekly Rundown: Episode 73

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

Use Single Sign On