Development and evaluation of a diatom‐conductivity model from lakes in West Greenland

• Published in: Freshwater Biology
• Main Authors: RYVES, D. B., MCGOWAN, S., ANDERSON, N. J.
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 2002
• Subjects: Greenland; Kangerlussuaq; Ice Sheet; Søndre strømfjord
• Online Access: http://dx.doi.org/10.1046/j.1365-2427.2002.00832.x; https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1046%2Fj.1365-2427.2002.00832.x; https://onlinelibrary.wiley.com/doi/pdf/10.1046/j.1365-2427.2002.00832.x

1. The area around Kangerlussuaq (Søndre Strømfjord; West Greenland, 67°N 51°W) contains thousands of lakes ranging from coastal, dilute (conductivity < 30 μS cm –1 ) oligotrophic systems to subsaline (∼4000 μS cm –1 ), closed basin lakes close to the ice sheet margin. In closed basins, salinity (or conductivity) is often a proxy for effective moisture, and thus palaeorecords of lake conductivity can provide valuable palaeoclimatic data. Little or nothing is known about the recent history of these lakes and hence it is difficult to evaluate how they will respond to effects of future changes. 2. Over 100 lakes have been sampled (1996–2000) between the ice sheet and the outer coast for a variety of water chemical and limnological variables. Surface sediments were taken from a subset of 40 lakes and analysed for diatoms. Diatom responses to 28 environmental variables were analysed by multivariate ordination techniques and indicate that the main gradient is highly correlated to conductivity (explaining ∼12% of species variance). Despite the relatively short gradient (24–4072 μS cm –1 ), diatom assemblages exhibit a clear response to conductivity. The most saline lakes do not contain a true saline flora. 3. We developed a range of weighted averaging (WA) and weighted‐averaging partial least squares (WA‐PLS) models from this training data set and found two component WA‐PLS models performed best. The effects of data transformation and omission of dissolution susceptible species ( Diatoma spp.) on model performance were also examined. The error statistics for the preferred WA‐PLS (2) model ( r 2 jack =0.88, root mean square error of prediction (RMSEP)=0.217 log μS cm –1 ) compare well with other published models. 4. A 210 Pb‐dated short core from a meromictic, subsaline lake (Braya Sø; location 67°N, 51°W, max. depth 23 m; conductivity 2600 μS cm –1 ) was analysed for diatoms. Diatom preservation is poor and some taxa (e.g. Diatoma spp.) are badly corroded. Lake water conductivity was reconstructed using WA‐PLS ...