L. Kattner et al.: Monitoring compliance with sulfur content regulations of shipping fuel 
10089 
www.atmos-chem-phys.net/15/10087/2015/ 
Atmos. Chem. Phys., 15,10087-10092, 2015 
ble measurement conditions and at the same time provide a 
compact and transportable set-up. Data were stored in an in 
tegrated data logger with the time resolution of 1 min. De 
spite different time resolutions of the instruments, we used 
data normalised to 1 min, which is sufficient for the analy 
sis of emission events with a duration in the order of several 
minutes. 
NO*, SO2 and O3 were measured with instruments from 
the Horiba AP-370 series, which are certified instruments 
according to EU directives (EN14211 for NO*, EN 14212 
for SO2 and EN 14625 for O3) used by German authorities 
for standard air pollution measurements. CO2 was measured 
with a Licor 840A analyser. The O3 measurements were not 
used for this study and are just mentioned for completeness. 
For SO2: the Horiba APSA-370 is based on the UV- 
fluorescence method, using the excitation of SO2 molecules 
by UV light and measuring the fluorescence which is a func 
tion of SO2 concentration. The response time of the instru 
ment is specified to be less than 120 s. Calibration was car 
ried out with a standard gas mixture from Air Liquide with 
a concentration of 99.7 ppb SO2 with an accuracy of 5 %. In 
addition, a daily control was obtained by the measurement 
of zero gas produced with a scrubber, and span gas from an 
internal permeation source with 175 ppb SO2. There is a NO 
cross sensitivity for SO2 which gives an SO2 signal for 0.8 % 
of the NO signal. We have determined this value of 0.8 % via 
a set of six calibration measurements of different NO con 
centrations between 100 and 470 ppb. 
For NO* : the Horiba APNA-370 measures the chemilumi 
nescence of NO molecules reacting with O3. To obtain in 
formation about the NO2 concentration, the device contains 
a deoxidation converter to transfer NO2 molecules to NO. 
The NO2 concentration is calculated by the difference of total 
NO*, representing NO + NO2, and NO without conversion. 
The response time for measurement of both gases is 90s. 
The instrument is calibrated with an Air Liquide standard 
gas mixture with a concentration of 216.0 ppb NO and an ac 
curacy of 5 %. A daily control with scrubber-produced zero 
gas and an NO2 span gas of 105 ppb is also implemented. 
For CO2: the Licor 840A is a non-dispersive infrared gas 
analyser. It has a response time of 1 s and was calibrated 
with two Air Liquide standard gas mixtures with 306.6 and 
990.0ppm CO2 with an accuracy of 2 %. 
The trace gas measurements were complemented with 
measurements of wind, temperature, air pressure and precip 
itation by a compact weather station (Lufft WS600). With 
an AIS (automatic identification system) receiver the infor 
mation transmitted by passing ships was collected, which in 
cludes identification number, name and type of the ship as 
well as position, course and speed. 
3 Data analysis 
To obtain the sulfur content of ship fuel in use, the enhance 
ment of SO2 and CO2 in measurements affected by exhaust 
gases is measured, and the ratio of these SO2 and CO2 peaks 
is used to calculate the fuel sulfur content. The combination 
of the trace gas peak time, the wind direction and the AIS in 
formation enables the identification of the peak-related ship. 
When wind conditions are favourable for measurements, 
the plumes of ships passing the instrument leave a distinc 
tive enhancement of the measured component against back 
ground concentrations. Since this enhancement is most sig 
nificant in NO measurements, and NO is an indicator for 
recent combustion processes, these NO peaks are used to 
identify the time stamp of a ship emission event. For these 
time stamps, peaks in CO2 are then identified, which is more 
complicated because background concentrations are larger 
and more variable due to the surrounding vegetation. Back 
ground signals for each gas are determined via a customised 
running mean filter. The SO2 signals are only analysed for 
those events for which there was a significant CO2 peak and 
a clearly determinable background. For all peaks the indi 
vidual peak area above the background concentration is de 
termined. This accounts for the difference in peak width for 
each gas due to different time resolutions of the respective in 
struments. The peak area value of the SO2 peaks is corrected 
with 0.8 % of the peak area value of the NO peaks to ac 
count for the cross sensitivity. With the assumption that fuel 
contains 87 ± 1.5 % carbon (Cooper et ah, 2003) and 100 % 
of the sulfur and the carbon content of the fuel are emitted 
as SO2 and CO2 respectively, the sulfur fuel content (SFC) 
mass percent can be calculated as follows: 
SFC[%] = 
S[kg] 
fuel [kg] 
SO2 [ppm] • A(S) 
CO2 [ppm] • A(C) 
SO2 [ppb] 
r L -[ • 0.232 [%] 
CO2 ppm 
87 [%] 
(1) 
where A(S) is the atomic weight of sulfur and A(C) the 
atomic weight of carbon. Using this formula, it is relatively 
simple to calculate the sulfur content for each set of peaks. 
For a discussion about the uncertainties of this formula see 
Sect. 3.1. 
The second part of the data analysis is the attribution of the 
identified emission events to individual passing ships. Within 
30 min before each event, which is characterised by the time 
the emissions arrive at the instruments, the AIS data are anal 
ysed for ship positions close to the measurement site. In com 
bination with wind information, this yields the identification 
of the individual ships which have caused the emission in 
most cases. The time the plume travels from being emitted 
to being analysed is about 2 to 10 min, depending on wind 
speed and direction. However, there are events in which there