10090 
L. Kattner et al.: Monitoring compliance with sulfur content regulations of shipping fuel 
Atmos. Chem. Phys., 15,10087-10092, 2015 
www.atmos-chem-phys.net/15/10087/2015/ 
are two or more ships too close to each other, or where no 
AIS signal was received, such that no single ship can be as 
sociated to the signal. These events are excluded from the 
data set. 
3.1 Uncertainties 
There are several aspects that influence the accuracy of the 
calculated values of the sulfur content for each ship. The SFC 
formula Eq. (1) assumes a 100% conversion from sulfur to 
SO2 during combustion, which is only true for an idealised 
combustion process. There is a range of uncertainty with re 
spect to the amounts of sulfur oxidised and released as par 
ticles. Studies found that there could be an underestimation 
of the sulfur fuel content between 1 and 19 % from assuming 
complete conversion (Schlager et al., 2006; Agrawal et al., 
2008; Moldanova et al., 2009; Balzani Loov et al., 2014). 
The uncertainty or sum of systematic and random error of 
our measurements is determined from a combination of the 
calibration uncertainty and the uncertainty resulting from the 
signal to noise ratio (SNR). CO2 values with a SNR of less 
than 5 are excluded from the data, which leads to an upper 
limit uncertainty of 20 %. However, the majority of CO2 val 
ues have an uncertainty of around 10 %. For SO2 we do not 
exclude data with a low SNR because these are the zero sul 
fur content cases. The SNR of SO2 data for a sulfur content 
of around 0.1 % is 10 or better, with a decrease for lower 
sulfur content values. For an SNR below 5 we consider the 
SO2 signal to be zero. This is only important for the January 
2015 data, since the measured SO2 concentrations in 2015 
are much lower than for the 2014 data. This is shown in Fig. 2 
as a comparison between one week in December 2014 and 
one week in January 2015 with similar weather conditions. 
While no reduction in NO values can be observed, there is a 
large reduction in SO2 values, as expected. 
All uncertainties added up with the root of sum of squares 
method; this gives us an uncertainty range for the sulfur con 
tent calculations of 15-30 %. 4 
4 Results 
Using the method described above we were able to identify 
824 ship plumes of 474 individual ships within the months 
of September, November and December 2014. Unfortunately 
no data are available in October due to instrumentation prob 
lems. This data set is the so-called pre-regulation-change set, 
where the regulatory sulfur fuel content allowed for the ships 
of is 1.0%. The January 2015 data set consists of 589 ship 
plumes of 374 individual ships, which since 1 January 2015 
have to comply with the new 0.1 % rule. As shown in Fig. 3, 
the difference between these two data sets is remarkably ob 
vious. 
In the pre-regulation-change data set, 99.6 % of all ships 
complied with the 1 % sulfur limit with respect to the mea- 
surement uncertainty. This is better than previously pub 
lished compliance rates of 85 % of 174 ship plumes (Beecken 
et al., 2014), although it should be noted that this study did 
not describe the uncertainty considerations and was mea 
sured by aeroplane on the open sea. The latter may imply that 
compliance might not be so high when no direct control is 
possible. Compliance rates at other locations for land-based 
measurements show values of 90 % of 255 ship plumes and 
97 % of 211 ship plumes (Beecken et al., 2015). However, 
a study of Diesch et al., 2013, that describes measurements 
with a mobile laboratory along the Elbe River near our mea 
surement site, found a compliance of nearly 100% for 139 
ship plumes. This could possibly be credited to the special 
location of Hamburg harbour where ships have to go up the 
Elbe for more than 100 km. 
In accordance with the practice in use that fuel samples 
analysed in laboratories are considered as exceeding the 
0.1% sulfur limit in a legally binding way above the value of 
0.149 %, we suggest using a corresponding value of 0.15 % 
as a limit value for discussing the compliance of the ships 
in our January 2015 data set. This is in consistence with 
the formerly stated measurement uncertainties. In Fig. 4, a 
more detailed graph of the January 2015 data is shown. The 
red line shows the 0.1% limit with the shaded area, indicat 
ing a conservative 30 % measurement uncertainty. The blue 
line indicates the suggested 0.15 % limit for compliance dis 
cussion. Of all the ships measured in January, 95.4% were 
complying with the new regulation. There are preliminary re 
sults for first SFC measurements in January 2015 presented 
in Beecken, 2015, which are comparable with our measure 
ments, although with slightly higher uncertainty and lower 
compliance rates. 
The lengths of the ships in 50 m size steps are colour- 
coded in Figs. 3 and 4. Even before the regulation change, 
ships smaller than 100 m did not use fuel with sulfur val 
ues higher than 0.2 %, most likely because their engines can 
not process such fuels or because storage capacity for two 
different kinds of fuels is not available. After the regulation 
change, those smaller ships still do not use fuels that reach up 
to the 0.1% limit allowed. If one considers only those ships 
longer than 100 m that could choose which fuel to use and 
had to change their way of operation, the compliance drops 
to 93 %. 
The number of ships that can be detected for compliance 
depends strongly on the wind conditions. Assuming the aver 
age number of calls in Hamburg harbour according to Ham 
burg port statistics of 800 ships per month means that 1600 
emission events happen at our measurement station of ships 
on their way in and out of the harbour. For months with 
good wind conditions like December 2014 and January 2015, 
we can detect about 30-40 % of those events, for months 
with unfavourable wind conditions, like November 2014, this 
value drops to less than 10 %.