A quantitative reconstruction of recent environmental change at Signy Island, maritime Antarctic, is performed using a model based on lake-sediment characteristics.

20th Century climatic warming is the probable cause for recent deglaciation on Signy Island. Proxies in glacial lake-sediments can be used to model the rates and directions of environmental change. Previous studies have made this link qualitatively. Here, a quantitative link is made between climate, glacial extent and activity and lake-sediment response.

The Island's 17 freshwater lakes display different characteristics within a small geographical area. Multiple cores from each basin yielded 209 surface-sediment (0-1 cm) samples which were analyzed for their lithology, particle-size and mineral magnetism. Ordination and classification identify relationships between the sediment variables and contemporary environmental factors responsible for within- and between-lake variability. Variance partitioning using a 'minimum adequate model' of 14 significant environmental variables provides insights into the balance of limnological and catchment factors affecting overall sediment character and its component fractions. A portion of unexplained variance suggests that other, unmeasured variables also play a role in determining sediment character.

Catchment ice-cover is selected for reconstruction on the basis of its gradient length. An inverse regression model is constructed using four variables: % dry weight, % Loss On Ignition, mean particle-size and HARD%, a measure of antiferromagnetic mineral concentration. A more parsimonious model can also be constructed using the first three selected variables, thus excluding the potential effects of source mineralogical and post-depositional changes in mineral magnetism. Two methods of reconstruction - Partial Least Squares (PLS) regression and analogue matching (Modern Analog Technique) are tested using the four and three variables models with various screened training-sets. Cross-validated prediction errors are ca.13%. Calibration is performed at four sites using optimal models to reconstruct ice-cover and results compare favorably with historical (1968 AD) records. The model is a sensitive proxy for erosion. Reconstructions are compared with temperature records and affirm that climatic warming is the most likely cause for observed ice-retreat. Based on simple measures, the model could see application in similar locations in the Antarctic, Arctic and Alpine regions.