Diatom-salinity relationships in wetlands: assessing the influence of salinity variability on the development of inference models

Abstract

Diatoms are among the most widely used indicators of human and climate induced wetland salinity history in the world. This is particularly as a result of the development of diatom-based models for inferring past salinity. These models have primarily been developed from relationships between diatoms and salinity measured at the time of sampling or during the preceding year. Although within site variation in salinity has the potential to reduce the efficacy of such models, its influence has been rarely considered. Hence, diatom–conductivity relationships in eight seasonally monitored wetlands have been investigated. In developing a diatom–conductivity transfer function from these sites, we sought to assess the influence of conductivity variation on diatom inference model performance. Our sites were characterised by variability in conductivity that was not correlated to its range and thus were well suited to an investigation of this type. We found, contrary to expectations, that short-term (seasonal) changes in conductivity which were often dramatic did not result in unduly reduced transfer function performance. By contrast, sites that were more variable in the medium term (5–6 years) tended to have larger model errors. In addition, we identified a secondary ecological gradient in the diatom data which could not be related to any measured variable (including pH, turbidity or nutrient concentrations).