Nutrient over-enrichment of estuarine environments is increasing globally. However, it is difficult to determine the eutrophication trend in estuaries over long periods of time because long-term monitoring records are scarce and do not permit the identification of baseline environmental conditions. In this study, preliminary diatom based transfer functions for the inference of total phosphorus (TP) and total nitrogen (TN) in east-Australian sub-tropical estuaries were developed to address the deficiency in knowledge relating to historical estuary water quality trends. The transfer functions were created from a calibration set consisting of water quality and associated surface sediment diatom assemblage data from fifty-two sub-tropical estuaries in New South Wales and Queensland, Australia. Following data screening processes, Canonical Correspondence Analysis confirmed that TP and TN both explained significant, independent variation in the diatom assemblages. Variance partitioning, however, indicated that the TP was confounded with and may receive some strength from TN. WA and WA-PLS 2 component models for TP that included all calibration set sites yielded statistically weak results based on the jack-knifed r 2 scores $$ \left( {r_{\text{jack}}^{{^{ 2} }} \, = 0.22\;{\text{and}}\;0. 2 2 {\text{ respectively}}} \right) $$ . Removal from the calibration set of 12 sites that had all PO4, NH4, NO2, and NOx concentrations below detection limit resulted in a substantial improvement in WA-PLS 2 component TP model scores $$ \left( {r_{\text{jack}}^{{^{ 2} }} \; = \;\,0.69} \right) $$ , indicating that this model is statistically robust, and thus suitable for down core nutrient reconstructions. Caution, however, is required when developing diatom based inference models in Australian estuaries as nutrient cycling processes may have the potential to influence diatom based transfer functions. The model reported on here provides a foundation for reconstructing nutrient histories in eastern Australian sub-tropical estuaries in the absence of monitoring data.