This study delves into the relationship between lightning frequency and radar echo intensity across three distinct geomorphological regions in Sichuan Province (Sichuan Basin, Western Sichuan Plateau, and Panxi Region) based on lightning and radar echo data from 47 typical thunderstorm events occurring from June to August in 2020 and 2021. Employing boxplot analysis and logarithmic fitting methods, the study selects the maximum echo value within 0.03°×0.03° grid points from SWAN radar echo mosaics as the fitting factor for radar echoes and establishes a specific time window to accumulate lightning frequencies, thereby constructing a comprehensive sample dataset. The results reveal that the median and mean values of lightning frequency and radar echo intensity are relatively close in all three regions, with a generally symmetric data distribution. As lightning frequency increases, echo values tend to converge with reduced dispersion. Despite setting a minimum echo intensity threshold, the span of echo intensities corresponding to lightning frequencies in the Sichuan Basin remains relatively wide. A logarithmic distribution function is adopted to fit lightning frequency and radar echo, yielding satisfactory results in the Sichuan Basin and Western Sichuan Plateau, particularly in the Sichuan Basin where the goodness-of-fit for both the median and mean echo values versus lightning frequency reaches 0.909, indicating a strong correlation and high-precision fit. Conversely, the fit in the Panxi Region is less satisfactory, suggesting a weaker correlation between radar echo values and lightning frequencies in this area. The study further observes that as lightning frequency increases, the median and mean fitting curves of radar echo values in all three regions initially rise rapidly before levelling off, albeit with regional variations. These findings not only uncover the intricate relationship between lightning frequency and radar echo intensity but also provide a scientific basis for regional lightning warning and monitoring. From the perspective of individual case validation of fitted echoes, the fitting results in these three regions exhibit a tendency towards underestimation, likely due to the relatively low grid-point lightning density frequency used in calculations to reduce dispersion and ensure data stability. Nevertheless, the proxy echoes still accurately reflect high-value centre information consistent with observations, which holds significant value for thunderstorm forecasting and warning as well as the application of lightning data. In the future, further validation through more individual cases can be conducted to optimise the fitting algorithm, formula, and method, thereby enhancing the fitting results. Additionally, the fitting formula can be incorporated into model experiments to explore its application effects in different topographical and geomorphological regions.