Abstract Baseflow is often regarded as the streamflow component derived predominantly from groundwater discharge. The estimation of baseflow is important for water supply, water allocation, investigation of contamination impacts, low flow hydrology and flood hydrology. Baseflow is commonly estimated using graphical methods, recursive digital filters (RDFs), tracer based methods, and conceptual models. Of all of these methods, RDFs are the most commonly used, due to their relatively easy and efficient implementation. This paper presents a generic framework for assessing and improving the performance of RDFs for baseflow estimation for catchments with different characteristics and subject to different hydrological conditions. As part of the framework, a fully integrated surface water/groundwater (SW/GW) model is used to obtain estimates of streamflow and baseflow for catchments with different properties, such as soil types and rainfall patterns. An RDF is then applied to the simulated streamflow to assess how well the baseflow obtained using the filter matches the baseflow obtained using the fully integrated SW/GW model. In order to improve the performance of the filter, the user-defined parameter(s) controlling filter operation can be adjusted in order to obtain the best match between the baseflow obtained using the filter and that obtained using the fully integrated SW/GW model (i.e. through calibration). The proposed framework is tested by applying it to a common SW/GW benchmarking problem, the tilted V-catchment, for a range of soil properties. HydroGeoSphere (HGS) is used to develop the fully integrated SW/GW model and the Lyne and Hollick (LH) filter is used as the RDF. The performance of the LH filter is assessed using the commonly used value of the filter parameter of 0.925, as well as calibrated filter parameter values. The results obtained show that the performance of the LH filter is affected significantly by the saturated hydraulic conductivity (Ks) of the soil and that calibrated LH filter parameter can result in significant improvements in filter performance.