pH beräkningar för ytvatten: slumpvisa och systematiska fel av olika pH modeller

While modeled MAGIC ANC usually corresponds reasonably well with measured ANC there were, however, major differences between measured and modeled pH values. Standard error of pH calculated with Excel (the "pH-snurran") and constants from Hruska et al (2003) was around 0.3 pH units but coulde be as high as 0.5 pH units (Köhler 2011, Fölster 2011 and figures in the appendix). Standard deviation of the pH calculated with Excel (the "pH-snurran") is now around 0,05 pH units when using the new model constants. The uncertainty in modeled pH values are strongly influenced by the uncertainty in ANC. Differences in measured or various modeled ANC can easily be 0.005 meq L-1. According to Köhler et al. (2002), such differences may cause pH errors of up to 0.7 pH units. To be able to model the pH without any systematic errors using MAGIC modeled ANC both access to data for NH4, F, and object-specific solubility constants for gibbsite is required. The random error introduced when using the measured ANC for calculations is around pH 0.3 pH units. Assessments of the acidification based on measured ANC and pH differences of only 0.4 pH units are considered to be rather uncertain. The validation or prediction of pH values due to changed ​​liming or during episodes should be done with the model of Köhler and Hruska (2014) and CBALK because the precision is much better. Large systematic differences in pH can be caused by incorrect carbon dioxide pressure. The model may be either run according to Sobek et al. (2003) or with a constant carbon dioxide pressure.