Larsen, S.E., Windolf, J., Tornbjerg, H., Hoffmann, C.C., Søndergaard, M. & Blicher-Mathiesen. 2018. Genopretning af fejlbehæftede kvælstof- og fosforanalyser. Ferskvand. Aarhus Universitet, DCE – Nationalt Center for Miljø og Energi, 72 s. - Teknisk rapport fra DCE - Nationalt Center for Miljø og Energi nr. 110

http://dce2.au.dk/pub/TR110.pdf

In early 2017, the Danish Environmental Protection Agency (DEPA) discovered that the laboratory they had contracted for analyzing water samples from rivers, lakes and agricultural catchments had used an unauthorized method for analyzing total nitrogen (TN) and total phosphorus (TP) in 2016 and the beginning of 2017. The unauthorized method was an online (UV-radiation) method rather than the correct method – offline (or autoclave). The laboratory had informed the DEPA that the incorrect method had introduced a bias and that the concentrations of TN and TP would be underestimated compared to results from analyses using the correct method. The problem with the online method is that it insufficiently destroys the organic fraction of the TN and TP.

In June 2017, the DEPA contacted DCE to set up a program for the remaining part of 2017 with the aim of trying to “reconstruct” correct results of TN and TP for 2016 and the first part of 2017. This program contained approximately 50 sampling sites in rivers with monthly or bi-monthly sampling as well as sampling in a number of lakes. Each water sample was analyzed with the online and offline method and for rivers resulted in a dataset of nearly 400 pairs of online and offline concentrations for TN and 300 for TP, for lakes nearly 90 pairs of online and offline concentrations.

The datasets were analyzed using a number of different statistical methods, depending on the nature of the individual dataset.

Upon initiating the correction program, the DEPA informed DCE that the unauthorized method also had been used from 2010-14 – but by a different contracting laboratory.

The DEPA asked DCE to undertake a number of tasks regarding samples from 2016 and 2017:

1. Based on the dataset generated from rivers, DCE was asked to establish a model with the purpose of correcting each individual result of TN. If it was assessed that it was not possible to correct each individual result, DCE was asked to look at other possibilities and/or scales to correct the TN load calculation.

*DCE recommendation: *

*It is recommended to correct each individual result of TN in rivers* *with the formula given (formula 1). The corrected results can be used for a TN load calculation. This recommendation is also valid for results from samples taken in drainage systems.*

2. Based on the dataset generated from rivers, DCE was asked to establish a model with the purpose of correcting each individual result of TP. If it was assessed that it was not possible to correct each individual result, DCE was asked to look at other possibilities and/or scales to correct the TP load calculation.

*DCE recommendation: *

*It is not recommended to use corrected individual results of TP in rivers.*

*Instead, it is recommended to use the formula found (formula 2) to correct TP concentrations for the calculation of a national TP load or aggregated to major catchments, i.e. different HELCOM areas. *

3. For lakes, DCE was asked to assess whether it was possible to correct TN and TP for each individual lake. If it was assessed that it was not possible to correct the results for each individual lake, DCE was asked to look at other possibilities and/or scales to correct the TP and TN concentrations.

*DCE recommendation:*

*Based on the statistical analysis, it is not recommended – for the time being – to correct individual results of TP and TN from lakes.*

*It should be investigated closer whether a correction can be made by including other supplementary tests and using supplementary parameters, such as suspended matter.*

The DEPA also asked DCE to assess and describe what bias was introduced to the calculated yearly loads of TN and TP by using the online method in 2010-14. This assessment can be made by using results from tests of the method and intercalibrations between laboratories and in the development in the ratio between inorganic and total content of nutrients. DCE was asked to assess whether a correction for 2010-14 could be estimated to increase the certainty of the calculated total loads.

*A preliminary calculation of the load of TN at the measuring points for 2016 shows that the TN load will increase 7% by using the established formula.*

*A first assessment of the loads and concentration data for TN, nitrate and organic N for 2010-14 compared to the previous period shows that, for a great number of measuring points, we see a drop in organic N concentration, which will result in total N consisting almost entirely of nitrate. The same pattern has been seen in 2016/17, where the online method was used.*

*Based on this, DCE assesses that during the period 2010-14 a method with insufficient destruction of the organic part of TN has also been used. When using the correction formula for 2016/17 on data from 2010-14, this shows that the load cannot be corrected consistently. There is an indication that by using the formula for TN found for 2016/17 on data from 2010-14, there will be an overcorrection of the concentrations and further an overcorrection of the calculated loads.*

*For TP, the preliminary analyses indicate that the bias was smaller in 2010-14 compared to 2016/17.*

*Before a correction of TN and TP concentrations for the period 2010-14 can be carried out, the laboratory methods used should be described in more detail. Furthermore, it will be necessary to carry out a similar test of methods as in 2017, but this time with the laboratory that was contracted in the period 2010-14.*

*Based on these preconditions, a correction of the data from 2010-14 will eliminate the bias introduced by the incorrect analysis method for TN. By using a model for the correction, a model uncertainty will be added to the uncertainty of the analysis method, but it is assessed that the uncertainty caused by the model is smaller than the uncertainty of the analysis. DCE therefore assesses that a correction of the TN data from 2010-14 will improve the calculation of the TN load without introducing significant further uncertainty.*

*Concerning TP, the situation in 2010-14 is not clear, as the preliminary analysis indicates that the bias in 2010-14 was smaller than in 2016/17.*