Multicomponent reactive transport models for acid mine drainage and radioactive waste disposal

Resumen

Mine activities may pose severe impacts in water resources and ecosystems when pyrite-rich materials are exposed to atmospheric conditions by mine activities and generate acid mine drainage (AMD) waters. AMD is a world-wide environmental issue. Its prediction and modeling is a difficult and costly task. The As Pontes open pit mine in A Coruña is one of the largest coal mines in Spain. Mine and dewatering activities in As Pontes ended in December 2007. The coal mine dumps located outside the pit have a surface area of 11.4 km2. An artificial lake is being formed in the former open pit with runoff waters coming from the outside dumps, the drainage basin of the mine and with pristine waters diverted from the nearby Eume river. As Pontes dumps contain a wide variety of mine soils of various weathering stages and physical and chemical properties. The temperate and wet climate of the study area favor weathering and leaching processes. Pyrite caused soil and water acidification during the construction of the dumps. Acidity promoted also the dissolution of large amounts of aluminum and iron. The runoff waters of the outside dumps are collected into the North and the South perimeter channels. Dump runoff waters were treated in the liquid-effluent treatment TEL plant during mine operation. Since mine closure in December 2007, they are diverted to the open pit lake. Their water quality has improved over the last 20 years due to the remedial actions taken by ENDESA to achieve acceptable environmental conditions. There are seasonal changes in the chemical composition, c(t), which are related to the seasonal variability of the stream flows, Q(t). The appropriate management of As Pontes dump waters and the need to predict the future time evolution of the water quality of the dumps waters require the use of hydrological and hydrochemical models which take into account simultaneously: 1) The hydrology of the dumps; 2) The geochemical processes and reactions taking place in the dumps; and 3) The types of chemical waters resulting from such reactions. Carbon steel and compacted bentonite have been proposed as candidate materials for the overpack and the buffer, respectively, in the multi-barrier system of a deep geological repository for high level radioactive waste (HLW). Corrosion of the carbon steel will affect the lifetime of the overpack, the time evolution of the chemistry of the bentonite porewater and the migration of radionuclides from the canister to the host rock through the bentonite. The assessment of the performance of a highlevel radioactive waste repository requires the use of models for the long-term prediction of radionuclide migration through the natural and engineered barriers. These models must consider simultaneously the flow of groundwater, the transport of chemicals and radionuclides and geochemical processes. Therefore, a reactive transport model (RTM) is needed to assess the combined effect of geochemical and hydrodynamic processes. RTM provides a tool for comprehensive, quantitative and predictive treatment of chemical reactions and mass transfer within a geosystem and allows the quantification of the transport of contaminants through natural and engineered barriers. This dissertation deals with the application of RTM to acid mine drainage and radioactive waste disposal. The reclamation of areas affected by acid mine drainage and the safe disposal of radioactive waste in deep geological repositories are environmental challenges of paramount importance. The geochemical models used for acid mine drainage and radioactive waste disposal share the complexity of the key physical and chemical processes controlling the geochemistry of the system and the relevance of surface chemistry processes. The research group of UDC led by Professor Javier Samper has been working on the development and the application of surface water and groundwater quality numerical models. This dissertation has benefited from the outcomes of such research projects and in turn has contributed to the achievements of such projects. A coupled hydrological and full-mixing geochemical model has been developed for the quantitative prediction of the transient evolution of the stream flows, Q(t), and the water composition, c(t), of the runoff waters of the As Pontes outside dumps. The model takes into account simultaneously: 1) The hydrological water balance of the dumps; 2) The key geochemical reactions taking place in the dumps; and 3) The types of chemical waters resulting from such reactions. The model is based on the assumptions that the chemistry of the runoff water samples is the result of the mixing of the following three chemical end-members: 1) Unaffected natural waters; 2) Acid highly-mineralized waters and 3) Neutralized highly-mineralized waters. The amount of each end-member in a given chemical sample is measured in terms of the chemical mixing fractions (CMF). The chemical compositions of the end-members are estimated from measured chemical data using multivariate statistical and graphic methods. The hydrology of the As Pontes outside dumps is represented with a lumped hydrological model. The model computes daily values of the hydrological components from hydrometeorological data. The total runoff of the dumps is the sum of the following three components: 1) Direct surface runoff having a response time equal to the concentration time, 2) Interflow or subsurface flow which has a response time of some days; and 3) Deep groundwater flow having a response time on the order of months. The fraction of each component in a given day is known as flow mixing fractions (FMF). The hydrological model and the geochemical model are coupled by a linear relationship between flow mixing fractions and chemical mixing fractions. FMF and CMF are related through a hydrology-chemistry matrix A. Its entry, Aij, is equal to the fraction of the j-th chemical component in the i-th hydrological component. The coupled hydrological and geochemical model is solved sequentially. First, the hydrological model is solved with the code VISUAL-BALAN which provides the daily values of the flow mixing fractions. Then, the chemical compositions of the end members and the chemical mixing fractions are determined using a multivariate statistical method. The entries of the hydrology-chemistry matrix A are estimated with a least squares method which optimizes the consistency between the chemical and hydrological mixing fractions. Finally, the chemical composition of a given sample is computed with a geochemical model which is solved with CORE2DV4 (Samper et al., 2003). This model takes into account the mixing of the chemical end-members as well as a set of chemical reactions such as aqueous complexation, acid-base, redox, dissolution-ex-solution of gases, dissolution-precipitation of Fe and Al minerals and proton surface complexation. The coupled hydrological and geochemical model has been constructed using measured data from sampling point 16B located in the South channel. The compositions of the end-members derived from the statistical method do not lead to a good fit of the measured data. The calibration of the composition of the end-members improves the fit of al chemical variables except for pH data which can only be fit by considering Fe(OH)3(s) and gibbsite dissolution/precipitation. The geochemical model constructed for point 16B was tested with data from Canal 2, an internal sampling point in the dumps having a water quality worse than that of 16B. The model reproduces all chemical data of Canal 2 properly. The model calibrated for point 16B was tested also with data from sampling point 18 which collects the runoff waters of both South and North channels as well as some mine surface runoff. Measured pH data in point 18 were more acid than those of point 16B during mine operation due to the contribution of the acid mine drainage and the fact that the point 18 receives the runoff of the Calvo Sotelo dumps which are known to contain extremely acid-producing areas. After mine closure in December 2007, measured data in sampling points 16B and 18 are similar, and therefore, the pH calculated with the geochemical model of point 16B can reproduce the measured pH at sampling point 18. A reactive transport model has been developed also for the titration experiments of several As Pontes mine water samples including: 1) A surface runoff sample from the West mine and 2) A sample from the Calvo Sotelo dumps. Such titration experiments were performed by Moreira (2010) as part of his Ph D Dissertation and have been modeled here by using a model similar to that of Uhlmann et al. (2004) which has been extended and improved to better reproduce the titration curves of As Pontes samples. Titration curves and their geochemical models are useful to identify and quantify the main buffering mechanisms. Titration curves have been simulated with CORE2DV4. They show the following seven buffering mechanisms: 1) Free hydrogen and hydrogen sulfate ions for pH <3; 2) Schwertmannite precipitation for 3 < pH < 3.4; 3) Precipitation of aluminium minerals for 3.4 < pH < 6; 4) Transformation of ferric minerals from schwertmannite to Fe(OH)3(s) and of aluminium minerals from basaluminite to gibbsite for pH = 6; 5) Surface complexation of cations for 6 < pH < 9; 6) Mn mineral and aqueous complexation of Al(OH)4- for 9 < pH < 10.8; and 7) Aqueous complexation of Fe(OH)4- for pH >10.8. The base neutralization capacity (BNC) of the samples is controlled by the first four buffers and can be calculated approximately as BNC=3·(CFe+CAl)+0.05·Csulfate, (BNC and dissolved concentrations in mol/L). The values of BNC are 9.25·10-3 mol/L for the sample of the West mine and 1.28·10-2 mol/L for the sample from the Calvo Sotelo dumps. Current performance assessment models (PAM) for radionuclide migration through the near field of a high-level radioactive waste (HLW) repository usually rely on simplifying assumptions such as the use of the Kd approach for nuclide sorption. Testing the validity of this assumption has been limited by the lack of: 1) Data on nuclide cation exchange; and 2) Computer codes which could solve for the migration, sorption and precipitation of radionuclides simultaneously with the geochemical evolution of the near field of a HLW repository. Laboratory experiments performed in recent years have provided substantial data and understanding on the mechanisms of nuclide sorption. On the other hand, multicomponent RTM have been developed which can handle radionuclide migration, sorption, and precipitation simultaneous with the multicomponent geochemical evolution of the near field. The validity of the Kd approach for nuclide sorption in the 0.75 m thick compacted bentonite barrier of a spent-fuel repository in granite has been evaluated and tested. Such testing has been performed by comparing the results of a typical PAM with those obtained with a multicomponent reactive transport model which incorporates a mechanistic thermodynamic sorption model. Models have been performed for Cs+. Cs+ sorption onto bentonite occurs mainly via cation exchange. Model results show that the apparent Kd of Cs+, Kda, increases with time due to the decrease of the ionic strength, I, of the bentonite porewater caused by the out-diffusion of the aqueous species from the bentonite into the granite. The simulated values of Kda are related to I through the following equation: A constant-Kd model fails to reproduce the release rate of Cs+ from the near field computed with the RTM. A variable-Kd model which incorporates the dependence of Kda on I reproduces adequately the Cs+ release rate, thus providing a surrogate for the constant-Kd model. The results of the sensitivity runs to model parameters and boundary conditions show that the water flux at the bentonite-granite interface affects strongly the Kda through changes in I while the effective diffusion coefficient of the bentonite plays a minor role on Kda. The increase of cation exchange capacity leads to the increase of Kda, but it does not affect the time evolution of I. Competing cations such as Ni2+ and iron corrosion products decrease slightly the Kda of Cs+ by competing for exchange sites and by increasing the I.