Soil baseline is regarded as an important indicator in the assessment of grassland ecological and environmental damage. How to determine the soil baseline, find the suitable method, and determine the damage degree of soil have become the key point of grassland protection and ecological assessment. In this study, we selected three methods to determine the soil nutrients baseline (total organic matter (TOM), N, P, and K) in Shengli mining area of Xilin Gol Grassland and Baorixile mining area of Hulun Buir Grassland in northern China based on practical and accurate criteria. The soil nutrients' baselines were analyzed by historical data method, reference site method, and the statistical method. The results showed relatively small difference between the soil nutrient level of Shengli mining area and its soil baseline of reference site method. The content of K was disturbed by mining but its average had not decreased significantly compared with the baseline values, so mining activities had no direct impact on the soil nutrient level of Shengli mining area. The average content of TOM in the soil of Baorixile mining area was obviously lower than the baselines of the three methods. The average contents of N and K were similar to the baseline of reference site method, and the average content of P was in the range of three baselines. As a result, mining had a direct impact on the content of TOM in the soil of Baorixile mining area, but no direct impact on the content of N, P, and K. All three methods could be used to determine the grassland soil baseline, but the results were different, and each method had its own advantages and disadvantages. Historical data method reflected the actual historical state of the grassland in the study area, and could be used as a comparative standard for the assessment and identification of ecological environment damage, but it lacked consideration of natural evolution and other factors. Reference site method was most effective when areas with the same natural conditions and ecological function were available, which could be very hard to find in reality. The calculation process of statistical method was prompt and convenient, and its utilization of data was quite high, whereas it was not suitable for the grassland area with high degree of human interference. The most appropriate method for baseline determination may depend on the availability of sample data and evaluation scale. Reference site method would be preferred, while historical data method and statistical method can be the substitutes. It is recommended to use multiple methods together to verify the result.