A testate amoebae transfer function from Sphagnum-dominated peatlands in the Lesser Khingan Mountains, NE China

We present a testate amoebae training set for building a paleohydrology transfer function. Ninety-one samples were collected from three Sphagnum peatlands in the Lesser Khingan Mountains, NE China. Redundancy analysis revealed that depth to the water table (DWT) and moisture content (% water) are the primary factors that control testate amoebae assemblages. Transfer functions for prediction of these two environmental variables were developed. The root mean square error (RMSEP) for DWT and moisture content were 6.74 cm and 1.49 %, respectively, assessed with "leave-one-out" cross validation. We applied a more robust cross validation method for clustered structure data, "leave-one-site-out," and the RMSEP of the best performance model increased to 6.90 cm and 1.67 %, but all models still had predictive power. The effect of uneven sampling was tested using new statistical approaches. Greater numbers of samples in the middle range of the gradient yielded smaller RMSEP values than did samples from the extreme wet and dry ends of the spectrum, where there were fewer samples. Our results indicate this training set is a potentially important tool for paleoenvironmental reconstruction in the Lesser Khingan Mountains, NE China. It will contribute to understanding climate change, particularly past monsoon activity, in this region