Fig. 3

Trend shape simulation. A comparison of a observed and b simulated compound concentration trends mapped to a domain and range of 0 to 1. Observed data has been smoothed using cubic regression splines with a maxium of 4 degrees of freedom. The line widths represent estimated standard error ranges from the regression spline model. Simulated trend generation is described in the text and the line widths have been set to a constant 5 % of maximum value. Lines were plotted using a high degree of transparency, with darker areas indicating a higher density of curves passing through them. In b, 100 trends were generated for each of the classifications using Eqs. 1–2 and parameter values listed in Table 1