Because the ability of selected indicators in assessing ecological risk at different temporal scales is not the same, it is necessary to clear the definite comparability of such indicators at temporal scale to explore a new method for dynamic assessing the ecological risk. In this case, five mining cities in Liaoning Province were selected as the study area, with the application of learning vector quantization (LVQ) neural network, the significance of the indicators for the ecological risk assessment was quantitatively analyzed to clarify their characteristics at temporal scale. The expression with two-dimension (long-term and short-term) of temporal scale was put forward as a new method to assess the ecological risk for mining cities. The results showed that the amount of industrial SO2 removed per output value, the amount of industrial dust removed per output value, coverage rate of urban green space, precipitation, coordination degree among subsystems, percentage of mining practitioners, and current year investment on pollution abatement projects were effective at long-term temporal scale, while the other indicators acted at short-term temporal scale. With the combination of long-term and short-term temporal scales, the dynamic assessment of ecological risk for mining cities could be expressed on two-dimension of temporal scale. It was found that Fuxin City got the highest ecological risk in current status, with the risk increasing most in Fushun City at the short-term temporal scale as well as in Chaoyang City at the long-term temporal scale. The method adopted in this study might act as a significant guidance in dynamic controlling and integrative management of ecological risk for mining cities.