Surface-sediment and epilithic diatom pH calibration sets for remote European mountain lakes (AL:PE Project) and their comparison with the Surface Waters Acidification Programme (SWAP) calibration set

A modern diatom-pH calibration data-set consisting of surface-sediment diatom assemblages from 118 lakes and 530 taxa is presented. The AL:PE data-set is from high-altitude or high-latitude lakes in the Alps, Norway, Svalbard, Kola Peninsula, UK, Slovenia, Slovakia, Poland, Portugal, and Spain (pH range = 4.5-8.0; DOC range = 0.2-3.2 mg l(-1)). In addition, 92 epilithon samples from 22 high-altitude or high-latitude lakes comprise an AL:PE epilithon diatom-pH data-set. Weighted averaging partial least squares regression is used to develop pH-inference models. The AL:PE data-set has a root-mean-square-error of prediction (RMSEP) of 0.33 and a maximum bias of 0.36 pH units and r(2) of 0.82, as assessed by leave-one-out cross-validation. The epilithon data-set has, after data-screening and the deletion of one very obvious outlier, a RMSEP of 0.23 and a maximum bias of 0.18 pH units and r(2) of 0.88. The 167 sample SWAP diatom-pH data-set from lowland or upland lakes in the UK, Norway, and Sweden has a RMSEP of 0.29 and a maximum bias of 0.23 pH units and r(2) of 0.86. The pH optima, as estimated by weighted averaging and Gaussian regression, are compared for the three data-sets (AL:PE, SWAP, AL:PE epilithon). There is a good correspondence between the AL:PE and the AL:PE epilithon optima, but a consistent bias between the AL:PE and SWAP optima, with the SWAP optima being lower than the AL:PE estimates. The predictive performances of the AL:PE and SWAP calibration data-sets are compared using independent test samples and six core sequences, all from high-altitude lakes, one in south-east Siberia and five in eastern Scotland. The results show the importance of using the AL:PE data-set for inferring lake-water pH from diatom assemblages in high-altitude or high latitude lakes with low DOC concentrations.