499 
500 
reduction of these uncertainties would result in better fitting between modelling and measurements 
both in duration and concentrations of time-series. 
501 Another source of uncertainty may be the exact route of the vessel in relation to the wind 
502 speed, expressed by the angle ® (Figure 3). Small uncertainties in the vessel course may be due to 
503 route adjustments and GPS positioning limitations, which are not reflected in the AIS data. This 
504 will have an impact on how the plume crosses the MS location (Figure 2) and, in turn, this will 
505 affect both the duration and the individual concentrations of peaksum. To demonstrate this, Figure 
506 11 depicts the alteration in concentration time-series duration and shape considering ® deviations 
3507 for Ship 1, as an example. The baseline corresponds to the AIS recordings with an angle wo= 62°, 
508 as also shown in Figure 11a. The angle has been first varied by +7.5°, a range which we expect to 
509 xcorrespond to realistic possible deviations of the actual route. As an extreme case of low 
510 probability, we also demonstrate what deviations of the order of +15° would mean to the measured 
511 signal. Figure 11b shows that the decrease of ® creates a more tightened peaksum with smaller 
>12 duration. On the contrary, increasing this angle causes a curve flattening, resulting to a longer time 
313 duration. The MOP of time-series that correspond to - 7.5° and + 7.5° are equal to 87 ppm and 104 
514 ppm, respectively, with a value of the baseline result at 107 ppm, showing that peaksum results 
515 are not significantly affected within such a range of angle deviations. The exceptional +15° case 
516 leads to a MOP of only 33 ppm, which is clearly different to the baseline result. This sensitivity 
517 analysis shows that the method developed is rather robust in the uncertainty of angle ® for 
518 moderate deviations. However, significant deviations from the vessel known route may indeed 
519 have a strong impact on the plume concentrations modelled for the monitoring station. By 
520  adjusting the parameter ©, it is feasible to improve time-series both in concentrations magnitude 
521 and duration simultaneously 
272