Empirical orthogonal functions (EOFs) of indirect archives of environmental change are increasingly used to identify coherent trends between palaeoclimate records, to separate externally forced patterns from locally driven idiosyncrasies. Lake sediments are particularly suited to such syntheses: they are abundant in most landscapes and record a wide array of information, yet local complexities often conceal or confuse the climate signal recorded at individual sites. Lake sediment parameters usually exhibit non-linear, multivariate and indirect responses to climate, therefore identifying coherent patterns between two or more lake records presents a complex challenge. Ideally, the selection of representative variables should be non-subjective and inclusive of as many different variables as possible, allowing for unexpected correlations between sites. In order to meet such demands, we propose a two-tier ordination procedure whereby site-specific (local) ordinations, obtained using Detrended Correspondence Analysis (DCA), are nested within a second, regional EOF. Using the local DCAs as representative variables allows the retention of a larger fraction of variance from each site, removes any subjectivity from variable selection and retains the potential for observing multiple, coherent signals from within and between each dataset. We explore this potential using four decadally resolved diatom records from volcanic lakes in Western Victoria, Australia. The records span the 1000 years prior to European settlement in CE 1803. Our analyses reveal at least two coherent patterns of ecological change that are manifest in each of the four datasets, patterns which may have been overlooked by a single-variable, empirical orthogonal function approach. This intra-site coherency provides a valuable step towards understanding multi-decadal hydroclimate variability in southeastern Australia.