[Objective] This work was to examine several key issues of using the ¹³⁷Cs method to estimate soil erosion rates in order to improve and standardize the method.[Methods] Based on the comprehensive review and synthesis of a large body of published literature and extensive long-term research experience, several key issues in the application of the ¹³⁷Cs method were generalized.[Results] The core assumption that spatial distribution of ¹³⁷Cs inventories is uniform is invalid, and that the ¹³⁷Cs method cannot be used to estimate quantitative soil erosion rates using a single soil core sample. The author emphasized that there existed a random component in ¹³⁷Cs spatial variation, and showed that this spatial random component was the largest uncertainty contributor of the ¹³⁷Cs method using the previous results from a sensitivity and uncertainty analysis.[Conclusion] The erosion estimation error resulting from spatial random variation can be filtered out by using a mean ¹³⁷Cs inventory of multiple independent soil cores or samples. The spatial random component can be removed using appropriate sampling designs for taking independent samples based on geo-statistical properties. Although the ¹³⁷Cs erosion models have been widely used in the literature, none of them has been vigorously tested and validated due to the lack of measured long-term soil loss data, and most models still remain theoretical. Because extremely large differences exist between model estimates, model validation and evaluation are utterly important for successful application and acceptance of the ¹³⁷Cs method.