Correcting for bias in freshwater total nitrogen concentrations obtained with a modified standard (NS4743) method

Abstract

Concentrations of total nitrogen (TotN) in freshwater monitoring programs were unreliable for a limited period, probably because of inadequate sample digestion when using a modified version of method NS 4743 (“method A”) for TotN analysis. The method resulted in underestimation of TotN, especially for humus-rich surface waters. Where re-analysis of samples with the original NS 4743 method (“method B”) is not possible, TotN may be corrected using a regression model fitted to ca. 1000 freshwater samples with analytical results from both method A (likely resulting in biased TotN concentrations; TotNold) and method B (resulting in presumably correct values; TotNnew). Here we present the dataset, the correction model, and a validation of the model using independent datasets. The regression model used TotNold and TOC and their interaction as predictors, and had an R2 of 0.956 and a root mean squared error (RMSE) of 31.87 μg N/L. The model predicted TotNnew well when tested on independent datasets from lakes and headwater streams. We recommend that the function is used to estimate TotN in long-term monitoring programs of lakes and headwater streams (1-3 order streams) where TotN was analysed with the modified standard method (method A), but not for rivers (>6-order streams) since the difference between TotN from the two methods was not significant for these water types.