Looking at water modeling for NorthMet

By Matt Ehling

Editor’s note – Since 2013, PRM has obtained thousands of pages of documents from state and federal regulatory agencies related to mining in northern Minnesota.

Last month, the environmental review of PolyMet’s proposed NorthMet mining project reached another benchmark. The Minnesota Department of Natural Resources (DNR) announced the completion of the third draft of NorthMet’s Supplemental Draft Environmental Impact Statement (SDEIS). The SEDIS document covers key aspects of NorthMet’s proposed operations, environmental impacts, and mitigation procedures. Completion of the document is necessary before PolyMet is able to apply for any mining permits.

NorthMet background
The copper-sulfide mining that PolyMet proposes to undertake differs significantly from the “ferrous” mining that has occurred at the state’s taconite mines over the past century. Unlike ferrous mining, waste rock from copper-sulfide mining acidifies once exposed to oxygen, and can then contaminate adjacent waters.

On June 26th, DNR Commissioner Tom Landwehr told MPR News that the SEDIS shows that “PolyMet has a plan” for addressing this issue, and for meeting state and federal environmental standards. In public statements and regulatory filings, PolyMet has expressed confidence in the use of a “reverse osmosis” mitigation process for treating contaminated water that will accrue at the mine site. The latest version of the SDEIS notes that such treatment would be required “indefinitely.”

As with previous drafts, the newest version of the SDEIS relies on computer modeling to provide predictions about future environmental impacts, including the prospects for seepage from NorthMet’s waste rock containment basin. These include computer models commissioned by PolyMet – such as its proprietary “GoldSim” water modeling software.

Constructing assumptions about the environmental impact of mining operations is a complex task – particularly given the extended period over which environmental mitigation will be required for NorthMet. Thus, the role of computerized modeling like GoldSim has played a central role in the debates over the environmental footprint of the project.

The importance placed on these computer models is reflected in internal correspondence among the regulatory agencies tasked with overseeing the NorthMet project. This correspondence (obtained by PRM in 2014) reveals pointed disagreements about the sufficiency of the NorthMet computer models. Given the age of the documents (most date from 2011 to 2013) some of the disagreements may tend toward the historical. However, it is worth noting that concerns about GoldSim and other models remain in the Environmental Protection Agency (EPA)’s most recent set of public comments on the SDEIS. Likewise, comments submitted by tribal authorities contain similar concerns, indicating that opinion on NorthMet’s water impacts is far from settled.

What is GoldSim?
A 2012 summary of GoldSim prepared by the Environmental Resources Management (ERM) consulting group describes the software as a “proprietary but publicly available dynamic system model” programmed to predict NorthMet’s effects on the quality of surface and groundwater. The summary notes that two GoldSim models were developed for the NorthMet SDEIS – one for the mine site, and one for the plant site and tailings basin.

According to the ERM document, both models incorporate water flow and other “hydrologic” processes, as well as chemical processes such as leaching from mine tailings. The models are described as “probalistic,” in that uncertainty in the parameters that describe the release and transport of chemicals is used to estimate uncertainty in the predictions made by the resulting models.

A 2011-era memo also notes that PolyMet sought to use separate MODFLOW software to model groundwater flow and seepage rates from the NorthMet tailings basin. Water flow from the basin has been an area of significant interest during the environmental review process, given the possibility that contaminated water could escape from its containment area in the mine site, and seep into aquifers or waterways.

Early questions about modeling
Records obtained by PRM show discussion and disagreement about the project’s computer modeling, as agency staff and contractors weighed the appropriateness of design assumptions and inputs.

Documents show that the DNR and other “co-lead” agencies worked with ERM to evaluate and review aspects of the NorthMet process. In November of 2011, David Blaha of ERM posed questions to DNR staff about the parameters for GoldSim. In e-mail correspondence, Blaha registered concerns that certain model parameters were trending toward “the conservative side.” Blaha further noted that “we don’t want conservatism built into the input assumptions,” as such assumptions might bias the model output. Instead, Blaha advocated for a set of “reasonable input assumptions that are not skewed one way or the other” to assure appropriate model results.

Documents indicate that agency discussions led to several modifications in which the GoldSim model was tweaked and its parameters were adjusted. However, criticism of the modeling continued.

In a March, 2012 letter to co-lead personnel, John Coleman of the Great Lakes Indian Fish and Wildlife Commission (GLIFWC) claimed that computer models for the NorthMet mine site had been calibrated to “under-represent true baseflow.” According to Coleman, the use of such parameters would generate models that showed the groundwater system “moving an unrealistically small quantity of water,” thus underestimating potential environmental impacts. Coleman ended his letter by noting that “modeling efforts that are based on faulty initial assumptions and not on field observations will not be able to reasonably predict impacts.”

That same month, Coleman also raised concerns about a lack of “sensitivity analysis” in NorthMet water modeling. Such analysis examines how uncertainty in input parameters affects the resulting uncertainty of model outputs. In a February, 2012 e-mail, Coleman stated that “sensitivity analysis is integral to model development” and that it should be applied to both the GoldSim and MODFLOW models alike.

Documents indicate that Coleman was first notified about the issue by EPA personnel, who had contacted co-lead staff once they discovered that sensitivity analysis was not being utilized. “We are dissatisfied that we discovered this information blindly,” EPA’s Mike Sedlaceck wrote on January 31, 2012. His e-mail to co-lead personnel further stated that “EPA staff disagree with the conclusion that the sensitivity analysis should not be performed.”

Subsequent agency documents (circa 2013) indicate that sensitivity analysis was later requested to help measure “interaction effects” occurring in the water models, as well as to help refine the modeling of surface water impacts and other environmental outcomes.

Additional review of GoldSim
Throughout 2012 and 2013, regulatory agency staff continued to focus on GoldSim and other modeling efforts. In the interest of testing model reliability, co-lead agencies assembled a quality assurance review team to examine the GoldSim software. The team included agency personnel and outside vendors such as ERM.

A May, 2012 memo from DNR vendor Interralogic to ERM’s David Blaha highlighted one evaluation of the GoldSim plant site model. “Overall,” the memo stated, “we did not identify any major issues that would cast doubt on the validity of the model with regard to input. We did however identify a number of secondary issues.” For example, the memo pointed to inconsistencies in how certain numerical values were represented, and noted that statistical summaries were used instead of actual values in some cases.

Documents indicate that by 2013, the quality assurance team had reached a level of agreement about GoldSim modeling. A technical memo from early 2013 stated that the team’s review of GoldSim provided “good evidence” that the model had “appropriate and mathematically correct algorithms” for estimating water quality impacts.

However, the memo also noted that the computer models for both the mine and plant sites were “very complex.” The authors acknowledged that GoldSim contained thousands of lines of custom programming, and that it was “not feasible to validate and check every line of code” in the model. Instead, the memo stated that the review team used a “combination of independent calculations and professional judgment” to identify potential problems with the code.

In the end, it was the team’s opinion that the GoldSim model had “acceptable reliability” for use in assessing NorhtMet’s environmental impacts.

However, a DNR e-mail from roughly a month later noted possible issues with the degree to which GoldSim captured and rendered existing site conditions. A March 28, 2013 e-mail from the DNR’s Bill Johnson to co-lead personnel discussed memos from Barr Engineering that summarized the sensitivity analysis used for the GoldSim mine site model. In his e-mail, Johnson stated that the Barr memos indicated that “the model results are fairly insensitive to most Mine-site related input parameters and are more sensitive to the background flow and water quality” in the nearby Partridge River.

EPA weighs in
In April of 2013, the EPA weighed in on the work that had been conducted on GoldSim, including the quality assurance process. In a lengthy letter to the DNR and other co-lead agencies, the EPA’s Alan Walts set out his feedback on GoldSim. Walts noted that EPA was satisfied that sensitivity analysis had been conducted, but also expressed the agency’s belief that the entire GoldSim program should have been reviewed and evaluated, rather than just select components. The letter also criticized the decision not to use all the sample data that had been collected at the project site. Walts noted that the decision raised questions about whether the samples that were used were truly representative of conditions at the site.

Further disagreements
Correspondence from mid-2013 revealed continuing disagreements about the efficacy of GoldSim. In an e-mail from June of 2013, the DNR’s Bill Johnson stated that the co-lead agencies would not be requesting any additional “water team” review of the GoldSim model at that point, since the co-leads believed that GoldSim has been adequately documented for use in the SDEIS.

On July 2, 2013, John Coleman of GLIFWC e-mailed Johnson and other co-lead personnel to express his ongoing concerns about GoldSim. Coleman’s
noted that “GoldSim inaccurately predicts existing conditions” and was unlikely to predict future project conditions. Such a model, Coleman wrote, “would not be “adequate to ensure the protection of water resources,” and he asked for the model to be recalibrated.

EPA registers concerns in SEDIS comments
Questions about NorthMet modeling continued to persist, even after the release of the draft SEDIS in early 2014. In its March, 2014 public comments on the SDEIS, EPA continued to disagree with certain assumptions made in the construction of GoldSim modeling.

For instance, EPA’s comments noted that “there is insufficient detail” to explain why “outlier” data were excluded from consideration in the GoldSim model. Likewise, EPA expressed concern that “mercury was not included in the GoldSim model” and that it was “unclear how mercury evaluation criteria” would be determined without including them in the software modeling.

The EPA’s comments also pointed out that MODFLOW program outcomes assumed that no waste water would seep from the soil berm on the east side of NorthMet’s tailings basin. EPA’s comments urged that this assumption be reconsidered, and that GoldSim models be re-calibrated to reflect it. Such comments are reflected in section eight of the newest draft of the SDEIS, which deals with “major differences of opinion” between the co-lead agencies, and other entities such as the EPA and tribal authorities.

PRM offers the above as an overview of modeling-related correspondence from our mining document collection. The full document sets are available for review on the PRM web site. Newer correspondence (including the newest SDEIS draft) will also be available soon.