Risk assessment of mining projects is a requirement in the mineral industries. In this process, many risk variables are time-dependent, and the only available data are historical time series. Moreover, in the case of a multivariate scenario, conventional forecasting methods fail to capture conditional dependency across the variables, which is important when there is an underlying causal relationship that needs to be modeled for accurate project evaluation. Thus, we investigated the use of copulas to capture the conditional distribution of the factors involved in a mine risk assessment study. We employed a multivariate copula-based time-series approach to model several uncertain variables. The Autoregressive Fractionally Integrated Moving Average - Generalized Autoregressive Conditional Heteroscedasticity (ARFIMA-GARCH) model was used for the conditional mean and copulas were used to model the error distribution, thus capturing the collective variation and dependence pattern across the variables. The method was implemented to model gold prices, copper prices, and the 10-year US Treasury bond yields and to determine the project’s net present value and probability of being economically feasible. The proposed approach can be used for cases where simulation of multivariate time-series is conducted.

采矿项目的风险评估是采矿业的一项要求。在此过程中，许多风险变量与时间有关，唯一可用的数据是历史时间序列。此外，在多变量情况下，常规的预测方法无法捕获变量之间的条件依存关系，当需要为精确的项目评估建模需要潜在的因果关系时，这一点很重要。因此，我们调查了使用copulas来捕获矿山风险评估研究中所涉及因素的条件分布。我们采用了基于多变量copula的时间序列方法来对几个不确定变量进行建模。将自回归分数积分移动平均值-广义自回归条件异方差模型（ARFIMA-GARCH）用于条件均值，并使用copulas对误差分布进行建模，从而捕获变量之间的集体变化和依存关系。实施该方法的目的是对黄金价格，铜价和10年期美国国债收益率进行建模，并确定项目的净现值和在经济上可行的可能性。所提出的方法可以用于进行多元时间序列模拟的情况。以及10年期美国国债收益率，并确定该项目的净现值和在经济上可行的可能性。所提出的方法可以用于进行多元时间序列模拟的情况。以及10年期美国国债收益率，并确定该项目的净现值和在经济上可行的可能性。所提出的方法可以用于进行多元时间序列模拟的情况。