<p>In karst regions, the complexity of aquifer heterogeneity arises from hierarchical interactions of multi-scale dissolution features, including fractures, fissures, and conduits. This complexity causes traditional methods (Theis type-curve matching, straight-line approximation and water-level recovery analyses) to yield inaccurate hydrodynamic parameters when aquifer properties are ignored. This study proposes a new parameter estimation method. The basic procedure of this method is as follows: (1) compile existing field data; (2) identify the aquifer types by comparing the field data to diagnostic curves (log–log drawdown and log-derivative curves); (3) interpret the aquifer properties; then (4) select an appropriate analytical well-test model to estimate parameters. Applied to a test site, the method revealed coexisting confined and unconfined aquifers. The range of transmissivity (15.4–554 m²/d) and storage coefficient (1.0 × 10⁻⁷ to 12.3 × 10⁻⁶) values confirmed strong heterogeneity. To further verify the reliability of the new method, a comparative analysis was conducted on the calculation results of the traditional and new methods. A <i>t</i>-test comparison showed statistically significant differences between the new and traditional methods (<i>t</i> = −6.12 vs. critical <i>t</i> = −2.145 at <i>α</i> = 0.05). Additionally, the robustness analysis demonstrated that the new method has a significantly lower root mean square error (RMSE) and narrower confidence intervals than the traditional method. These results collectively prove the superior reliability of the new methodology. In conclusion, for pumping tests in complex karst settings, comprehensive aquifer characterization followed by model-specific parameter estimation provides more accurate outcomes.</p>