Emission Measurements of Volatile Organic Compounds with the SOF ...

Emission Measurements of Volatile Organic Compounds with the SOF method in the Rotterdam Harbor 2008

Johan Mellqvist, Jerker Samuelsson, Brian Offerle, Håkan Salberg, John Johansson, and Sini Jaakkola

FluxSense AB, Göteborg, September 2009

FluxSense Report 2009-09-09 Title: Emission Measurements of Volatile Organic Compounds with the SOF method in the Rotterdam Harbor 2008 Authors: Johan Mellqvist, Jerker Samuelsson, Brian Offerle, Håkan Salberg, John Johansson, and Sini Jaakkola FluxSense AB Hörsalsvägen 11 S-412 96 Göteborg Sweden

E-mail: [email protected] [Cover Graphic: SOF measurement in Rotterdam harbor 2008. The path driven with the SOF measuring van displayed with a colored scale ranging from blue low staples, indicating low emission, to red tall staples indicating higher emission. White arrows show wind direction. Aerial photo obtained from Google Earth Pro.]

2

Summary The total emissions of volatile organic compounds (VOCs) from industrial sources in the Rotterdam harbor area has been quantified in a field study during August and September 2008, on behalf of the DCMR (Environmental Protection Agency Rijnmond). The method applied was the Solar Occultation Flux (SOF) method. This is a remote sensing method based on measuring infrared intensity of solar spectra from a mobile platform (car, boat). Utilizing known spectroscopic absorption features of VOCs the total mass of alkanes along the path of the solar light can be retrieved from the infrared solar spectra, and with a good knowledge of the wind profile emission rates can be determined. In addition to the SOF measurements, canister sampling, with subsequent gas chromatography analysis of 32 VOC species, was carried out to obtain an estimate of the relation between alkanes and other groups of VOCs, i.e. aromatics, alkenes and alkynes. To obtain wind information, measurements from several wind masts operated by the DCMR in the Rotterdam area were used together with measurements conducted within the project from a 17 m high mobile mast and balloon launches of GPS sondes. In the field study measurements were conducted on 17 days at six subareas of the Rijnmond harbor area, enclosed by the A4 highway in the east, the A15/N15 highway in the south, the Nieuwe Waterweg in the north and the sea in the west, respectively. The subareas were selected according to geographical and commercial boundaries and differed in terms of extent, type and number of facilities enclosed, but all contained a various mixture of refineries, oiland product storage facilities and petrochemical plants. For the month of September 2008, the measured alkane emission in the studied part of the Rotterdam area was 3998 tons/month (5553 kg/h) with an uncertainty of around 21% and a variability (1σ) of 338 tons. This is 4.4 times (340%) higher than the corresponding monthly VOC emission data of 911 tons/month (1265 kg/h) that is reported to the DCMR using conventional emission estimates. For the individual subareas the discrepancy between measured and reported VOC emissions varies between 2 to 14 times, see table below. Noteworthy is that the measured emissions only include alkanes whereas the reported ones also includes aromatic VOCs and alkenes and that the obtained discrepancies are consistent with SOF studies in other countries. As an upper estimate about one third of the discrepancy in this study could be explained by the fact that upset emissions are included to a larger extent in the SOF measurements than in the reported emissions and to the fact that the latter data corresponds to an annual estimate which is based on calculations that use annual mean values of physical parameters such as wind speed, solar radiation and ambient temperature. The average temperature during the campaign was for instance 8oC above the annual mean while the wind speed was slightly higher, 6.9 m/s instead of 5 m/s. Area 1 2 3 4 5 6 Total

Measured alkane emission (kg/h) 792 704 964 671 768 1654 5553

Reported VOC emission (kg/h) 391 284 119 81 269 121 1265

Discrepancy (Factor) 2.0 2.5 8.1 8.3 2.9 13.7 4.4

3

Table of Contents 1. Introduction .......................................................................................................................... 5 2. Methods ................................................................................................................................. 6 2.1 The SOF method............................................................................................................. 6 2.1.1 General ..................................................................................................................... 6 2.1.2 Details of method..................................................................................................... 7 2.1.3 Flux calculation ..................................................................................................... 11 2.2

Canister sampling for plume speciation............................................................... 13

3. Wind measurements........................................................................................................... 14 4. Results ................................................................................................................................. 18 4.1 SOF results.................................................................................................................... 18 4.1.1 Area 1 ..................................................................................................................... 20 4.1.2 Area 2 ..................................................................................................................... 25 4.1.3 Area 3 ..................................................................................................................... 28 4.1.4 Area 4 ..................................................................................................................... 31 4.1.5 Area 5 ..................................................................................................................... 32 4.1.6 Area 6 ..................................................................................................................... 35 4.2 Canister sampling......................................................................................................... 42 5. Discussion............................................................................................................................ 46 5.1 Measurement uncertainty............................................................................................ 46 5.2 Representativeness of data and comparison to reported values .............................. 47 6. Acknowledgements............................................................................................................. 53 7. References ........................................................................................................................... 53 Appendix I: SOF transects .................................................................................................... 55 Appendix II: Reported emission calculations...................................................................... 61 Appendix III: Canister sampling .......................................................................................... 62

4

1. Introduction A survey of industrial releases of non methane volatile organic compounds (NMVOCs) has been conducted in the harbor area of Rotterdam using an infrared absorption spectroscopy based method, Solar Occultation Flux (SOF). The measurements were conducted over the period August – September 2008 on a total of 17 days. The harbor area was divided into six sub areas, Figure 1.1. For each of the six areas SOF measurements of alkanes were conducted during three or more days with different wind directions in order to quantify the emissions from each area. In addition, canister sampling with subsequent gas chromatography analysis was made to obtain an estimate of the relation between alkanes and other groups of VOCs, i.e. aromatics, alkenes and alkynes. The different areas in this study contains a mixture of refineries, tank storage facilities and chemical industries as given in Table 1.1. Table 1.1 Description of the studied areas. Area Refinery Tank storage 1 X X 2 X X 3 X 4 X X 5 X X 6 X X * Minor

6

5

Chemical industry X X X* X

Ship loading X X X X X X

4 3 2

1

Figure 1.1 Overview of main subareas of the Rotterdam Harbor studied in the project. Colored lines and boxes emerging from the picture correspond to SOF measurement transects.

5

2. Methods 2.1 The SOF method 2.1.1 General The Solar Occultation Flux (SOF) method is relatively new and was developed from a number of different research projects [Mellqvist, 1995; 1999, Galle, 1999]. The method utilizes the sun as the light source and gas species that absorb in the infrared portion of the solar spectrum are measured from a mobile platform. The method is often used to survey the VOC emissions from individual industrial plants. A typical survey is conducted over 10 measurement days and measurements are conducted both outside and inside the facility, to determine total emissions and to attribute emissions to specific areas within the facility, respectively. The SOF method works well for rapid screening of emissions over an industrial facility and operates in the day under sunny conditions. The SOF measurements are often complemented by mobile extractive FTIR measurements [Galle, 2001; Börjesson, 2009] conducted from the SOF vehicle with a separate instrument, for carrying out cloudy day or night time measurements, typically studies of full-cycle tank- or ship loading emissions. Tracer gas is positioned at the location of the leak and then the ratio of tracer gas and leaking VOC is measured by extracting the gas plume into a gas cell and then analyzing the gas concentrations by infrared spectroscopy. The SOF method has been applied in more than 40 individual plant surveys over the last 7 years and at several larger campaigns both in Europe and the USA (Table 2.1). For Swedish refineries the measurement show that the VOC emissions typically correspond to 0.03-0.09% of their throughput of oil, with more than half of the emissions originating from oil and product storage [Kihlman 2005a, 2005b]. There SOF measurements are the basis for the reported VOC emissions. Table 2.1 Measurement studies of industrial VOC emissions that have been conducted using SOF measurements (Solar Occultation Flux Method). Several reports can be found at www.fluxsense.se Measurements at Individual Plants (10 days survey) Akzo Nobel surface chemistry, Stenungsund, Sweden 2003, 2004, 2005, 2006, 2007, 2008, Borealis Polyethylene, Stenungsund, Sweden 2000, 2006, 2007, 2008, Göteborg oil harbor Göteborg, Sweden 2002, 2003, 2004, 2006, 2007, 2008, 2009 Nynas Petroleum Nynäshamn, Sweden 2005 Nynas Petroleum Göteborg, Sweden 2005, 2006 Preem refinery Göteborg, Sweden 2002, 2003, 2004, 2006, 2007, 2008, 2009 Preem refinery Lysekil, Sweden 2002, 2003, 2004, 2005, 2006, 2007, 2008 Shell refinery Göteborg, Sweden 2002, 2003, 2004, 2006, 2007, 2008, Chemical plant Austria 2008 and 2009 Measurement campaigns at industrial areas Refineries and petrochemical plants Tula, Mexico 2003 Refineries, storage depots and Texas, USA 2006 and 2009 petrochemical plants at the Houston Ship Channel Refineries, storage depots and Normandie, France 2008 petrochemical industries at Le Havre and Gravenchon Refineries, storage depots and Rotterdam, The 2008 petrochemical industries at the Netherlands Rotterdam harbor

6

In various studies it has been found that the measured emissions using SOF are 5-10 times higher than the reported emission estimates based on calculation techniques. For instance in a recent study in Houston it was shown that the industrial releases of alkenes, on average, were 10 times higher than what was reported [Mellqvist, 2009]. These results were supported by airborne measurements [De Gouw, 2009]. For alkanes the discrepancy factor was about 8 [Mellqvist, 2007]. 2.1.2 Details of method The SOF method is based on the recording of broadband infrared spectra of the sun with a Fourier transform infrared spectrometer (FTIR) that is connected to a solar tracker. The latter is a telescope that tracks the sun and reflects the light into the spectrometer independent of its position. From the solar spectra it is possible to retrieve the path-integrated concentration (column, see Eq. 1) in the unit mg/m2 of various species between the sun and the spectrometer. In Figure 2.1 a measurement system is shown built into a van. The system consists of a custom built solar tracker, transfer optics and a Bruker IR-cube FTIR spectrometer with a spectral resolution of 0.5 cm-1, equipped with both an MCT detector and an InSb detector. Optical interference filters are used to optimize the signal to noise ratio (S/N) of the measurements.

Figure 2.1 The SOF system elevated through the roof top of the mobile van. The solar tracker (front left) transmits the solar light into the infrared spectrometer (mid right with a GPS on top) independent of the vehicle’s position.

To obtain the gas emission from a source, the car is driven in such way that the detected solar light cuts through the emission plume. This is illustrated in Figure 2.2. To calculate the gas emission the wind direction and speed is also required and these parameters are usually measured from high masts and towers.

7

Figure 2.2 In the Solar Occultation Flux (SOF) method gases are measured by observing solar light in the infrared portion of the solar spectrum from a moving vehicle. The vehicle transects the entire plume obtaining column totals and emission rates for the gases of interest.

The spectral retrieval is conducted by the custom software (QESOF) [Kihlman, 2005b] in which calibration spectra are fitted to the measured spectra using nonlinear multivariate analysis. Calibration data from the HITRAN database [Rothman, 2005] are used to simulate absorption spectra for atmospheric background species at the actual pressure, temperature and instrumental resolution of the measurements. The same approach is applied for several retrieval codes for high resolution solar spectroscopy [Rinsland, 1991; Griffith, 1996] and QESOF has been tested against these with good results. For the retrieval of alkanes, high resolution spectra of propane, n-butane and n-octane were obtained from the PNL (Pacific Northwest Laboratory) database [Sharpe, 2004] and these are degraded to the spectral resolution of the instrument by convolution with the instrument lineshape. The uncertainty in the absorption strength of the calibration spectra is about 3.5% for all three species. The VOCs leaking from refineries correspond mostly to alkanes (by mass). These compounds are retrieved in the infrared region between 3.3-3.7 μm (2700-3005 cm-1), using the vibration transition in the carbon and hydrogen bond (CH-stretch). The absorption features of the different alkanes are similar and interfere with each other, but since the number of absorbing C-H-bonds is directly related to the molecule mass, the total alkane mass can be retrieved despite the interference. Aromatics and alkenes also have absorption features in the CHstretch region, but mainly below 3.33 μm for the most abundant species. The absorption cross section for alkenes is also far weaker here than in the 10.21-10.58 µm region. The SOF alkane channel retrieval used in the Rotterdam study fits calibration spectra of propane, n-butane and n-octane to the recorded spectra, using a resolution of 8 cm-1. A study

8

of the cross sensitivity of various compounds to this alkane mass retrieval method has been investigated by retrieving spectra of typical plume constituent mixes. The results showed that for alkanes (C2 to C9) the applied SOF alkane channel mass retrieval retrieved the alkane mass within a factor 0.75 to 1.22 for the different compounds (0.991 on average). For nine of the most abundant alkenes the cross sensitivities varies between 0.029 and 0.549, and the corresponding values for aromatic hydrocarbons are 0.000-0.254, see Table 2.2. For the average plume composition of alkanes, alkenes, aromatics and alkynes in the Rotterdam study, the SOF alkane channel retrieved the alkane mass within 3.2%. The retrieval has little sensitivity to methane when operating at a spectral resolution of 8 cm-1, since most of the methane absorption lines are already fully absorbed due to the high levels of CH4 in the atmospheric background (1.8 ppm). Table 2.2 Cross sensitivity for the SOF alkane channel retrieval for the main 32 VOCs. Values correspond to the retrieved alkane mass (e.g. 1 mg m-3 n-pentane is retrieved as 1.013 mg m-3 of alkane, and 1 mg m-3 of toluene is interpreted as 0.092 mg m-3 alkanes). Alkanes

Ethane

Cross sensitivity (alkane mass) 0.753

Alkenes

Cross sensitivity (alkane mass)

ethylene

0.055

Propane

0.985

propylene

0.236

n-butane

0.993

1,3-butadiene

0.029

i-butane

1.079

c-2-butene

0.375

n-pentane

1.013

t-2-butene

0.375

i-pentane

0.965

1-butene

0.388

2m-pentane

0.969

t-2-pentene

0.549

3m-pentane

0.915

1-pentene

0.384

cyclo-hexane

1.217

c-2-pentene

0.425

n-hexane

1.003

Aromatics

n-heptane

1.005

benzene

0

n-octane

1.009

toluene

0.092

i-octane

0.91

m+p-xylene

0.166

n-nonane

1.051

o-xylene

0.186

ethyl-benzene

0.224

0

1,3,5-trimethylbenzene

0.227

0.115

1,2,4-trimethylbenzene

0.254

Alkynes ethyne propyne

In the left part of Figure 2.3 a solar spectra is shown which has been recorded downwind a refinery emitting VOCs. In the right part of the figure is shown the same spectrum ratioed to a "clean air" background spectrum which has been recorded immediately before the plume transect. This transmittance spectrum has been fitted to a mix of calibration spectra by the QESOF spectral retrieval routine.

9

Figure 2.3 Example of a solar spectra in the alkane CH-stretch region, recorded downwind an industrial complex (left). To the right the measured and the fitted mix of calibration spectra are shown (relative to a clean air background reference), obtained using the QESOF spectral retrieval algorithm.

The performance of the SOF technique has been tested by comparing it to other methods and tracer gas release experiments. In one experiment, tracer gas (SF6) was released from a 17 m high mast on a wide parking lot. The emission rate was then quantified by SOF measurements 50-100 m downwind the source, yielding an accuracy of 10 % for these measurements when averaging 5-10 transects [Kihlman, 2005b]. More difficult measurement geometries have also been tested by conducting tracer gas releases of SF6 from the top of crude oil tanks. In this case, for close by measurements in the disturbed wind field at a downwind distance of about 5 tank heights, the overestimation was 30%, applying wind data from a mast at about 1 tank height elevation [Kihlman, 2005a; Samuelsson, 2005]. The SOF method has also been compared against other techniques. In an experiment at the Nynas refinery in Sweden a fan was mounted outside the ventilation pipe, sucking out a controlled VOC flow from the tank. The pipe flow was measured using a so called pitot pipe and the concentration was analyzed by FID (Flame ionization detector) which made it possible to calculate the VOC emission rate of 12 kg/h. In parallel, SOF measurements were carried out at a distance corresponding to a few tank heights, yielding an emission rate of 9 kg/h, hence a 26 % underestimation in this case. Similar measurements from a joint ventilation pipe from several Bitumen cisterns yielded a FID value of 7 kg/h and only 1% higher emission from the SOF measurements [Samuelsson, 2005]. During a campaign in Texas in 2006, the SOF method was used in parallel to airborne measurements of ethene fluxes from a petrochemical industrial area in Mt Belvieu [De Gouw, 2009]. The agreement was here within 50% and in this case most of the uncertainties were in the airborne measurements. The SOF method has not been directly compared to the laser based technique DIAL [Walmsley, 1998] which has commonly been used for VOC measurements. Nevertheless, measurements at the same plant in Sweden (Preem refinery) yield very similar results when measuring at different years. Differences have been seen for bitumen refineries however [Samuelsson, 2005]. All in all the experiment described in this report, is consistent with an uncertainty budget of 20-30%, see section 5.1.

10

2.1.3 Flux calculation The gas flux in the SOF method is obtained first by adding the column measurements conducted across the gas plume as illustrated in Figure 2.4. The integrated mass of the key species across the plume is hence obtained. To obtain the flux this value is then multiplied by the average wind speed of the plume, uaverage.

Figure 2.4 Illustration of the SOF measurement.

The flux calculation is shown in Eq. 1. Here x corresponds to the travel direction, z to the height direction, u’ to the wind speed orthogonal to the travel direction (x), Zsun to the distance to the sun, and Hmix to the maximum height of the plume. For the solar measurements the slant angle of the sun is compensated for, by multiplying the concentration with the cosine factor of the solar zenith angle.

⎛ flux = ∫ ⎜⎜ x1⎝

x2 ⎞ ∫0 conc( z) ⋅ u' ( z) ⋅ dz ⎟⎟⎠dx = u'average x∫1column( x)dx

x 2 Zsun

Eq. 1

Hmix

Where u 'average =

∫ u '⋅dz Hmix

∫ dz

Zsun

, column =

∫ conc( z ) ⋅ dz

o

The wind is not straightforward to obtain since it is usually complex close to the ground and increases with the height. What helps the situation is that SOF measurements only can be done in sunny conditions. This is advantageous since it corresponds to unstable meteorological conditions for which wind gradients are smoothed out by convection. Over relatively flat terrain the mean wind varies less than 20% between 20 and 100 m height using standard calculations of logarithmic wind. This is illustrated for the harbor of Göteborg in Figure 2.5. Here the average daytime wind velocity and wind direction profile for all sunny days during August of 2004 have been simulated [Kihlman, 2005a] using a meteorological flow model denoted TAPM [Hurley, 2005].

11

In addition, for meteorological conditions with considerable convection, the emission plume from an industry mixes rather quickly vertically giving a more or less homogeneous distribution of the pollutant versus height through the mixing layer even 10 km downwind. In addition to the atmospheric mixing, the plumes from process industries exhibit an initial lift since they are usually hotter than the surrounding air. The rapid well-mixed assumption agrees with results from airborne measurements made by NOAA [De Gouw, 2009] and Baylor University [Buhr, 2006] during an air quality campaign in Texas (TexAQS 2006) in which also SOF measurements were conducted [Mellqvist, 2009]. The NOAA measurements indicate that the gas plumes from the measured industries mix evenly from the ground to 1000 m altitude, i.e. throughout the entire mixing layer, within 1000-2000 s transport time downwind the industrial plants. This indicates a vertical mixing speed of the plume between 0.5 to 1 m/s. This is further supported by Doppler LIDAR measurements by NOAA showing typical daytime vertical mixing speeds of ±(0.5-1.5) m/s [Tucker, 2007].

300

300

250

250

200

200 Height (m)

Height (m)

In the case of the SOF measurements conducted at the Rotterdam harbor in this study, the measurements were conducted downwind the industries at a typical plume transport time of 40 s to 950 s, which according to the discussion above means that the emission plumes had time to mix up to heights of several hundred meters above the ground, e.g. above the first 50 100 m where the wind is usually disturbed due to various structures. The windprofile at Rotterdam was actually measured with GPS balloon sondes, as described in section 3 and in general the average wind from ground to 200 m altitude was used to derive the emission rates.

150

150

100

100

50

50

0

0.8 1 1.2 1.4 Relative wind velocity to 25 m

0 -5 0 5 10 15 Relative wind direction (degrees) to 25 m

Figure 2.5 Average daytime wind velocity and wind direction profile retrieved by simulation above Gothenburg harbor area averaged over all sunny days during a month with a wind-speed of 3-6 m/s at ground. The error bars indicate standard deviation between daily averages [Kihlman 2005a].

12

2.2

Canister sampling for plume speciation of aromatics, alkanes and alkenes

The SOF method is presently not able to measure aromatic species directly and therefore canister samples are taken at various positions downwind of the specific areas from which aromatic emissions are to be determined. Subsequently the samples are analyzed and masses of alkanes, alkenes and aromatics are obtained. To obtain the aromatic emissions the ratio of aromatics to alkanes are multiplied by the alkane emissions measured by SOF. Table 2.2 presents the 32 VOCs included in the analysis. Air sampling was conducted using evacuated inert (Silonite™) canisters (Entech Instruments Inc., California USA) with an integration time (time to fill) ranging from 10 minutes to 4 hours depending on sampling location (typically 60 minutes). Samples were collected outside the process areas on local roads driving through the plume repeatedly during continuous sampling and generally the SOF-method was used to localize the emission plume. Sample blanks for background concentrations of ambient air were obtained on the same occasions using ten minute integration samples outside the plume. All samples were analyzed using gas chromatography (GC) at an accredited lab (IVL Swedish Environmental Research Institute, Göteborg, Sweden). Aromatic species analyzed were benzene, toluene, ethylbenzene, o-xylene, m+p-xylene, 1,3,5-, and 1,2,4-trimethybenzene (TMB). The detection limits are between 0.02-0.12 μgm-3 for all compounds in Table 2.2 [Potter, 2005].

13

3. Wind measurements Primary wind measurements were made using ground masts at different heights and geographical locations, selected to cover the large area extent of the Rotterdam harbor. The locations of the masts are marked with red flags in Figure 3.1. These ground measurements were complemented with a total number of 18 GPS balloon sondes launched throughout the campaign, launch sites marked by white circles in Figure 3.2. Two typical sonde measurements are shown in Figure 3.2 and 3.4. Noteworthy is the relatively small variation versus height in both wind direction and speed, caused by the sunny weather conditions with strong convection that smoothes out height variations.

Figure 3.1 Wind measurement sites from masts (red flags) and GPS balloon sonde sites (white circles). The mobile wind wagon was operated by FluxSense, whereas Geulhaven and Hoek van Holland were part of the DCMR wind meter network.

The mobile wind wagon was operated with a calibrated Young wind monitor (model 051035), mounted on a telescopic mast with maximum height of 17 m. Data was recorded averaging wind speed and wind direction for 30 s (also storing variability for that time period) to a logger (Campbell Scientific, CR10). The mobile wind wagon was operated by FluxSense, whereas the 15 m Hoek van Holland and the 11 m Geulhaven wind meters were part of the DCMR wind data network. The GPS balloon sondes that were launched were of type Intermet iMet-1, transmitting GPS position, temperature and relative humidity information every second to a ground receiver (403 MHz channel, Intermet iMet-3000).

14

Figure 3.2 Wind speed and wind direction as a function of altitude obtained from a GPS balloon sounding on Sep 9 2008 around noon time. The red line with circles corresponds to wind direction, whereas the black line with dots corresponds to wind speed.

Figure 3.3. Wind speed and wind direction as a function of altitude obtained from a GPS balloon sounding on Sep 11 2008 around noon time. The red line with circles corresponds to wind direction, whereas the black line with dots corresponds to wind speed.

15

To investigate the variations in the wind over height and at different geographical locations, the different mast winds have been compared to the average wind up to 200 m obtained from the GPS sonde measurement. The latter was used as reference wind since it is estimated to be the most probable wind for the gas plumes when measured by the SOF in this study, as discussed in section 2.1. The wind comparison is shown in Figure 3.4 and Figure 3.5. Also shown are winds from the GPS sondes averaged from ground level up to 100 and 500 m. The ground mast wind meter data used in the comparison corresponds to 30 minute averages, distributed in time as ±15 minutes from respectively GPS sonde launch. The wind measurements spread within ± 15 % (±1σ) for wind speed deviation, respectively ±11° (±1σ) in direction deviation compared with the 0-200 m balloon sounding. Table 3.1 shows the mean and standard deviation of the difference between the 200 m average and the other measurements, as relative difference for wind speed and absolute difference for wind direction.

Figure 3.4 Comparison of wind speed measurements versus GPS sondes from 0 – 200 m.s.l. for the period 9 – 20 September 2008. The dotted lines represent ± 38 % which enclose all data. Dash-dot lines correspond to ±15% (±1σ) which enclose 68% of the data. The ground mast wind meter data corresponds to 30 minute averages, distributed in time as ±15 minutes from respectively GPS sonde launch.

16

Figure 3.5 Comparison of wind direction measurements versus GPS sondes from 0 – 200 m.s.l. for the period 9 – 20 September 2008. The dotted lines represent ± 38º which enclose all data. Dash-dot lines correspond to ±11º (±1σ) which enclose 68% of the data. The ground mast wind meter data corresponds to 30 minute averages, distributed in time as ±15 minutes from respectively GPS sonde launch. Table 3.1 Comparison of wind measurements versus GPS sondes from 0 – 200 m.s.l.

GPS sonde 0-100 m GPS sonde 0-500 m Mobile wind wagon 17 m Hoek van Holland 15 m Geulhaven 11 m

Mean Relative Difference (%) from GPS sonde average 0-200m (+/- 1 sd)

Mean Difference (°) from GPS sonde average 0-200m (+/- 1 sd)

-7.1 (+/- 9.3) 3.5 (+/- 12.1) -21.8 (+/- 15.6) -4.5 (+/- 14.4) -3.3 (+/- 20.1)

2.1 (+/- 5.2) -3.5 (+/- 5.5) -2.2 (+/- 11.5) 3.2 (+/- 15.4) -2.9 (+/- 11.9)

As discussed based on plume transport time, plume lift and atmospheric mixing, the 0-200 m wind average is believed to best represent the average plume mass transport. The GPS sonde data has limited time coverage, whereas the ground wind meters represent a continuous data set. Hence the ground data set has been used in the emission calculations for the individual SOF transects, but adjusted to the 0-200 m level using the systematic mean relative differences in Table 3.1.

17

4. Results The main focus of the project was to obtain reliable statistics of the VOC emissions from the six different sub-areas of the Rotterdam harbor area. Public roads were used to access the emission plume for measurements. For all six areas total emission measurements of alkanes were conducted with the SOF method and canister sampling was performed in order to obtain an estimate of the relation between aromatics, alkenes and alkanes. Canister samples were primarily used for estimating the uncertainty in the SOF alkane mass retrieval, but also to get a rough estimate of the emissions of aromatic and alkene VOC species. In the case of area 1, the total area emission could be further separated to east and west halves. At Area 6, in addition to total emission measurements around the area, measurements have also been made within the area itself during half of a day. The inside measurements represent close-by snapshots of the different facility emissions within that area.

4.1 SOF results Table 4.1 shows the measured emissions and number of measurement days for respectively area, and Figure 4.1 graphically display the data. Table 4.1 Summary of the alkane emissions in Rotterdam Harbor area obtained by the SOF measurements. Area

Alkane emission (kg/h)

1

792 ± 128

2

704 ± 104

3

964± 219

4

671 ± 175

5

768 ± 272

6

1654 ± 334

Total

5553 ± 540

18

1654

768

671 964

792 704 Figure 4.1 Measured alkane emission as kg/h with error bars corresponding to ±21% uncertainty. Colored road segments correspond to SOF measurement paths around the areas.

In the following result sections, the general emission distribution is given as histogram plots, showing the number of measurements that fall within respectively emission size interval. For each area one or more examples of measurements deviating somewhat from the general picture is then presented.

19

4.1.1 Area 1 Area 1 located at the eastern end of the Rotterdam harbor, mainly inhabits a larger refinery, chemical industries, tank storage, truck and boat loading. The wind measured at Geulhaven (35 m height) was used for the flux calculations. Measurements were made mainly in northeast to south-east wind directions. Four of the 29 transects were made with north-west wind direction. The average value of the 29 SOF transects, distributed over 6 different days at area 1, was 792 kg/h (Table 4.2). The frequency distribution of the measurements is shown as a histogram plot in Figure 4.2. The 1σ spread of the data set correspond to 16 % of the average emission, which can be considered as fairly confined, indicating that few upsets occurred during the measurements. The red curve in Figure 4.2 shows a fitted probability distribution to the measured data set. The most probable (peak) value of the fitted probability distribution is believed to correspond to the continuous emissions (819 kg/h), to be used as a basis to estimate an annual emission. This is further discussed in section 5.2. Table 4.2 Summary for SOF measurements at Area 1. Day

Time span

Nr of transects

Emission (kg/h)

Wind (m/s | deg)

080829

165754 -174515

3

785±94

3.8-4.3 | 318-338

080831

145411 -150836

1

613

4.4 | 149

080914

90743 -133929

11

725±107

5.5-9.4 | 54-85

080916

115944 -154343

4

1004±65

4.0-4.4 | 31-50

080917

115629 -122128

2

813±62

4.2-4.6 | 59-76

080918

135538 -153102

8

811±225

4.5-6.7 | 67-88

Average

-

(total 29)

792±128 kg/h

-

20

Figure 4.2 Emission frequency distribution of measurements conducted at Area 1. The blue staples show the distribution of transects over emission size interval, and the red line shows the fitted probability distribution of the data. Average value for the 29 transects was 792 kg/h and most probable value for the fitted distribution of the histogram was 819 kg/h.

21

Area 1 can easily be split in two sub areas, area 1 western part and area 1 eastern part, Figure 4.3. Measurements at Area 1 western part showed an average emission of 290 kg/h, Table 4.3, and at the eastern part 585 kg/h, Table 4.4.

Figure 4.3 Area 1 SOF measurement (total, west and east) on 15th of September. The color index represents the line-integrated vertical total column of alkanes in mg/m2, ranging from 0 (blue) to 625 (red). Each dot corresponds to one measurement, and the dot size is scaled with the total column value. The lines emerging from the dots point up in the wind, and thus towards potential emission sources.

22

Table 4.3 Summary for SOF measurements at Area 1 western part. Day

Time span

Nr of transects Emission (kg/h) Wind (m/s | deg)

080829 080909 080914 080915 080916 080917

172459 -175730 135559 -152321 93822 -134834 105758 -111009 124941 -130044 114659 -115647

3 6 3 2 1 2

223±133 171±34 227±226 586±51 431 102±58

3.9-4.0 | 318-340 6.0-7.0 | 141-149 5.8-9.4 | 59-86 5.2-5.2 | 42-50 3.9 | 42 4.1-4.4 | 38-45

Average

-

(total 17)

290±181

-

Table 4.4 Summary for SOF measurements at Area 1 eastern part. Day

Time span

Nr of transects Emission (kg/h) Wind (m/s | deg)

080829 080914 080915 080916 080917 080918

171817 -175454 93822 -141959 105514 -111009 114234 -115035 114356 -142545 115918 -132714

3 4 3 1 10 9

385±55 782±287 532±151 494 627±239 687±146

3.9-4.0 | 319-341 5.8-9.4 | 59-91 4.9-5.2 | 41-50 3.7 | 62 3.8-4.6 | 38-85 4.7-6.3 | 101-124

Average

-

(total 30)

585±142

-

On two occasions during the measurements at area 1 eastern part there were emissions that might not be considered as normal, possibly due to ongoing cleaning activities according to the proprietor. On the 15th of September at 10:55 an emission of 25.7 kg/min was measured for the eastern part of area 1, Figure 4.4. Clearly the plume deviates from the normal pattern. Integrating the emission from the atypical sector itself (red high columns) an emission of 15.0 kg/min was derived, leaving 641 kg/h for the remaining part, which is well in line with the overall average of 585±142 kg/h for the whole area. Again on the 18th of September at 12:25 there was an emission of 29.0 kg/min measured for the eastern part, where the atypical peak solely gave 19.2 kg/min, Figure 4.5, leaving 592 kg/h for the remaining part. None of these upset emission events were included in the average calculation.

23

Figure 4.4 Upset emission event at area 1 eastern part on the 15th of September, 25.7 kg/min. Wind speed 5 m/s direction 40 degrees north east. The color index represents the line-integrated vertical total column of alkanes in mg/m2.

24

Figure 4.5 Upset emission event at area 1 eastern part on the 17th of September. Wind speed 5.8 m/s direction 117 degrees south east. The color index represents the line-integrated vertical total column of alkanes in mg/m2.

4.1.2 Area 2 In Area 2 various refinery-, chemical industries and tank storage facilities are located. To calculate the flux from the area, wind from the Geulhaven wind mast was used. The average value of the 18 SOF transects, distributed over 3 different days at area 2, was 704 kg/h (Table 4.5). Figure 4.6 shows the emission frequency distribution of transects conducted at area 2. The 1σ spread of the data set corresponds to 28 % of the average emission. As seen in Figure 4.6 there is a tail of measured emissions on the upper side of the average, suggesting occasional upsets from the normal emission case. The peak of the fitted probability distribution, interpreted as the most probable value for the continuous emission, lies below the average emission – 497 kg/h compared to 704 kg/h. Table 4.5 Summary for SOF measurements at Area 2. Day

Time span

Nr of transects

Emission (kg/h)

Wind (m/s | deg)

080830 080909 080911

173730 -174524 101929 -164835 100608 -154222

1 11 6

823 657±312 631±154

7.6 | 90 4.2-6.4 | 142-152 4.8-6.3 | 142-151

(total 18)

704±104 kg/h

-

Average

25

Figure 4.6 Emission frequency distribution of measurements conducted at Area 2. The blue staples show the distribution of transects over emission size interval, and the red line shows the fitted probability distribution of the data. The average value for the 18 transects is 704 kg/h and most probable value for the fitted distribution of the histogram is 497 kg/h.

Figure 4.7 shows one transect with higher emission than the normal that was seen during the measuring campaign. Many of the transects at area 2 show comparably higher emissions from that particular area. Due to the limitation by access only to public roads, the plume originating from the very east part of area 2 was hard to fully capture. A plausible explanation for the localized emission here could be ship loading or tank emissions.

26

Figure 4.7 High emission 9th of September at 10:05 gave 22.6 kg/min. Wind speed 4.7 m/s direction 150 degrees south east. The color index represents the line-integrated vertical total column of alkanes in mg/m2. The short gaps in the plume cross section correspond to short term cloud passages blocking the sun.

27

4.1.3 Area 3 Area 3 mainly contains tank storage and some minor chemical industries. In total 20 SOF transects were made for area 3. To calculate the flux from the measurements wind from the mobile wind mast was used, located on the land strip just to the north. On average Area 3 was measured to have an emission of 964 kg/h, see Table 4.6. Figure 4.8 shows the emission frequency distribution and the fitted probability distribution respectively, for transects conducted at area 3. No distinct tail of upper end emissions can be seen here, and the peak of the fitted probability distribution (1000 kg/h), interpreted as the most probable value for the continuous emission, is almost collocated with the average. Table 4.6 Summary for SOF measurements at Area 3. Day

Time span

080825 080911 080914 080915 080918 080920

151954 -153108 163926 -171740 182332 -183106 134247 -154102 174834 -181508 110043 -143227

Average -

Nr of transects Emission (kg/h) Wind (m/s | deg) 1 2 1 4 2 10 (total 20)

1131 879±87 707 1140±394 746±102 1183±295

8.6-8.6 | 241-241 4.5-4.5 | 144-154 5.4 | 72 4.9-6.3 | 73-104 3.8-4.2 | 82-96 2.8-5.4 | 62-90

964±219 kg/h

-

Figure 4.8 Emission frequency distribution of measurements conducted at Area 3. The blue staples show the distribution of transects over emission size interval, and the red line shows the fitted probability distribution of the data. The average value for the 20 transects is 964 kg/h and most probable value for the fitted distribution of the histogram is 1000 kg/h.

28

On two occasions apparent outlier values have been measured. The 6th of September at 16:22 35 kg/minute was measured at area 3 and the emission originates from a ship loading nafta without VRU (vapor recovery unit), and on the 15th of September at 14:40 29 kg/minute was measured at area 3, probably also due to ship loading activity. Measurements elsewhere have occasionally shown large emissions for ship loading dependent on the previous product in the hold and whether the holds were ventilated prior to loading, respectively dependent on the usage and/or performance of a VRU. Figure 4.9 and 4.10 shows the emission measurements.

Figure 4.9 The 6th of September at 16:22 with emission 35 kg/minute (nafta loading without VRU, vapor recovery unit, connected). Wind speed 9.2 m/s direction 190 degrees south. The color index represents the lineintegrated vertical total column of alkanes in mg/m2.

29

Figure 4.10 The 15th of September at 14:40 29 kg/minute high emission probably due to ship loading activity. Wind speed 5.7 m/s direction 80 degrees east. The color index represents the line-integrated vertical total column of alkanes in mg/m2. The short gaps in the plume cross section correspond to short term cloud passages blocking the sun.

30

4.1.4 Area 4 This area encompasses a refinery and tank storage facilities. To calculate the flux from the measurements wind from the mobile wind mast was used. On average as measured in 18 transects on 5 different days, Area 4 emitted 671 kg/h, see Table 4.7. Figure 4.11 shows emission frequency distribution of measurements conducted at area 4. The 1σ spread of the data set corresponds to 26 % of the average emission. As seen in Figure 4.11 there is a tail of emissions on the upper side of the average, suggesting occasional upsets from the normal emission case. Similarly to the case for area 2, the peak of the fitted probability distribution for area 4, interpreted as the most probable value for the continuous emission, lies below the average emission – 484 kg/h compared to 671 kg/h. Table 4.7 Summary for SOF measurements at Area 4. Day

Time span

080825 080830 080901 080914 080920

102640 -152020 132058 -132523 153514 -153951 162949 -182432 105448 -142348

Average -

Nr of transects Emission (kg/h) Wind (m/s | deg) 9 1 1 3 4 (total 18)

645±284 710 691 849±137 458±105

4.4-8.2 | 228-244 4.2 | 97 7.6 | 244 5.0-5.4 | 67-81 2.5-4.8 | 65-75

671±175 kg/h

-

Figure 4.11 Emission frequency distribution of measurements conducted at Area 4. The blue staples show the distribution of transects over emission size interval, and the red line shows the fitted probability distribution of the data. The average value for the 18 transects is 671 kg/h and most probable value for the fitted distribution of the histogram is 484 kg/h.

31

On the 14th of September at 17:20 an upset emission of more than 166 kg/minute (Figure 4.12) was measured from area 4. The origin of the upset is unknown.

Figure 4.12 The 14th of September at 17:20, an instantaneous emission of more than 166 kg/minute was measured. Wind speed 4.6 m/s and wind direction 75 degrees east. The color index represents the line-integrated vertical total column of alkanes in mg/m2.

4.1.5 Area 5 Within area 5, tank storage and a chemical industry are the main plant types. In total 27 transects were made for area 5, distributed over six days. To calculate the flux wind from the Hoek van Holland wind mast was used. On average Area 5 emitted 768 kg/h, see Table 4.9. Figure 4.13 shows a histogram over the measured emissions by size interval, along with the fitted probability distribution. The most probable value for the distribution of 784 kg/h, believed to best represent the continuous emission part, is collocated with the average measured emission.

32

Table 4.9 Summary for SOF measurements Area 5. Day

Time span

Nr of transects

Emission (kg/h)

Wind (m/s | deg)

080823 080825 080831 080901 080906 080920

125418 -145047 102102 -151655 173448 -174215 161000 -163404 155104 -155515 104632 -141238

7 10 1 3 1 5

1231±278 819±201 749 799±240 573 437±370

7.7-9.7 | 335-343 7.1-11.3 | 236-256 5.6 | 199 10.5-11.2 | 248-250 10.0 | 232 2.7-5.6 | 64-88

Average

-

(total 27)

768±272 kg/h

-

Figure 4.13 Emission frequency distribution of measurements conducted at Area 5. The blue staples show the distribution of transects over emission size interval, and the red line shows the fitted probability distribution of the data. The average value for the 27 transects is 768 kg/h and most probable value for the fitted distribution of the histogram is 784 kg/h.

Figure 4.14 shows an example of a SOF measurement by the harbor side of area 5 on the 23rd August 2008 13:10. This transect was immediately followed by a transect across the plume from the whole site, all in all 10 minutes for conducting both transects. The port side transect showed an alkane emission of 590 kg/h, whereas the plume from the whole area shortly after showed an emission of 1220 kg/h, leaving 630 kg/h for the part in between. This is well in line with the overall average of 768±272 kg/h. The observed upset explain to some extent the large variability seen in the emissions from this area. Naturally, more measurements and close-by on site measurements are needed to better understand the true origin of the variability.

33

Figure 4.14 SOF measurement at area 5, harbor side, on the 23rd August 2008 13:10, showing an alkane emission of 590 kg/h. Wind 5.5 m/s from the northwest. A transect across the plume from the whole area shortly after (all in all 10 minutes for both transects) showed an emission of 1220 kg/h, leaving 630 kg/h for the part in between. This is well in line with the overall average of 768±272 kg/h.

34

4.1.6 Area 6 Area 6 at the western end of the Rotterdam harbor includes a larger refinery, tank storage and boat loading activities. Flux calculations were made using wind measurements from the Hoek van Holland mast, close to the north. The average value for the emission measurements at area 6 was 1654 kg/h, Table 4.10. Comparing the average with the peak value (1480 kg/h) of the fitted probability distribution in Figure 4.15, one can see that the emissions tend to tail upwards. As discussed, the peak value of the fitted distribution is interpreted to better represent the continuous emission on a time scale beyond the monthly. Table 4.10 Summary for SOF measurements at Area 6. Day

Time span

080823 080825 080901 080910

112420 -154723 113810 -155057 131433 -162539 131635 -132300

Nr of transects Emission (kg/h) Wind (m/s | deg) 6 8 8 1

1855±539 1594±306 1342±173 1592

7.2-10.2 | 329-341 8.9-11.6 | 236-240 10.2-11.2 | 248-252 7.9 | 233

Average

-

(total 23)

1654±334 kg/h

-

Discussing the influence of different plant parts on the emissions, the tank park, water treatment facility and/or ship loading area seem to be strong contributors as seen in Figure 4.16. The 23rd of August, the emissions were above average, with many of the plume transects pointing out increased emissions in a specific area, as seen in Figure 4.16. Figure 4.17 shows another example from the 19th of September with a localized emission of 160 kg/h, obtained from measurements inside area 6. Figure 4.17-4.20 shows some more measurements which should be regarded as close-by snapshots from different facilities within the area. By putting more effort and time into such near-field measurements, one can examine the performance of individual tanks with respect to sealing, filled height, pumping rate etc., and the importance of certain site activities for the overall emission.

35

Figure 4.15 Emission frequency distribution of measurements conducted at Area 6. The blue staples show the distribution of transects over emission size interval, and the red line shows the fitted probability distribution of the data. The average value for the 23 transects is 1654 kg/h and most probable value for the fitted distribution of the histogram is 1480 kg/h.

36

Figure 4.16 A measurement transect across area 6 the 23rd of August at 11:50 with emission 44 kg/min. Wind speed 10.2 m/s direction 337 degrees north-west. The color index represents the line-integrated vertical total column of alkanes in mg/m2.

37

Figure 4.17 Localized emission of 160 kg/h measured 19th of September inside area 6. Wind speed 6.1 m/s and wind direction 19 degrees north. The color index represents the line-integrated vertical total column of alkanes in mg/m2.

38

Figure 4.18 Example of measurement of tank storage and waste water treatment area the 19th of September with emission 854 kg/h. Wind speed 5.8 m/s, wind direction 23 degrees north. The color index represents the lineintegrated vertical total column of alkanes in mg/m2.

39

Figure 4.19 Example of measurement of the process area the 19th of September with emission 341 kg/h. Wind speed 5.3 m/s, wind direction 68 degrees east. The color index represents the line-integrated vertical total column of alkanes in mg/m2.

40

Figure 4.20 Example of measurement of tank storage area the 19th of September with emission 294 kg/h. Wind speed 5.2 m/s, wind direction 23 degrees south north. The color index represents the line-integrated vertical total column of alkanes in mg/m2.

41

4.2 Canister sampling Canister gas samples were collected at five measurement areas between the 17th and 19th of September 2008, to support the SOF measurements and provide added speciation of the VOCs in the emission plumes. A total of 14 canisters, including background samples were collected, and analyzed for 32 different VOCs. The canister sample locations for the SOF measurement areas are shown in appendix III and Table 4.12. All samples were taken from approximately 2 m height, with an integration time of about 1 hour, and give an indication of the VOC contents in the ground level plume from the emission areas. On average the plume sample canisters contained a VOC concentration of 139 μgm-3 (range 20-256 μgm-3), in excess above background concentration. The plume composition as split into alkanes, alkenes and aromatics for the different canister samples are shown in Figure 4.21. To obtain a rough estimate of the aromatic emissions the ratio of aromatics to alkanes were multiplied by the alkane emissions measured by SOF for each area or part of the area. It should be noted that the canister samples were taken mainly to study the cross interference of the VOC plume mixture on the SOF alkane mass retrieval, rather than to obtain a solid estimate of the aromatic VOC emission. More canister samples would be needed to acquire a better statistically constrained aromatic emission estimate. In the original project outline, 4 main areas were identified for measurements. In the end, these were split into 6 areas, since the SOF data set was solid enough to allow for it. Hence, canister data can only be presented for areas 1-4 and 6. Table 4.11 Calculated aromatic emissions by area. (FS correspond to canister field sample as specified in appendix III). Area

1 2 3 4 5 6

Field Sample

Alkane emission (kg/h)

Aromatic/ alkane (mass ratio)

Derived Aromatic Emission (kg/h)

FS 5+11 FS 4 FS 12 FS 2, FS 13 n.m. FS 1,FS 6,FS 9

792 704 964 671 768 1654

0.11 1.03 0.08 0.33 (0.29-0.36) n.m. 0.24 (0.17-0.29)

87 725 77

5553

-

Total (n.m., not measured)

221 (194-241) n.m. 396 (281-479) 1506 (1364-1609)

In general the aromatic to alkane mass content in emission plumes from oil- and refinery plant complexes are measured in the range 5-30 %, which were also the case for areas 1 and 3-6 in the Rotterdam harbor. Field sample 4 from area 2 showed a high aromatic content with an aromatic/alkane mass ratio of 1.03.

42

100% 80% 60% 40% 20% 0% Alkanes Alkenes Aromatics

FS1

FS 2

FS 4

FS 5

FS 6

FS 8

FS9 FS 11 FS 12 FS13 FS 14

82%

86%

49%

89%

75%

67%

77%

55%

90%

72%

61%

4%

2%

1%

3%

3%

2%

3%

25%

2%

2%

17%

14%

12%

50%

9%

22%

31%

20%

21%

7%

26%

22%

Figure 4.21 Relative VOC mass content in canister samples in the Rotterdam harbor area. (FS correspond to canister field sample as specified in appendix III).

The canister data was used to examine the plume composition interference error in the SOF alkane mass retrieval. For the different plume compositions encountered at the different measurement areas, composite infrared spectra were made from the PNL database, with the corresponding mixture according to the canister plume samples. The retrieved SOF alkane mass in those spectra was then compared to the true alkane mass, as given in Table 4.12. In the average plume composition the SOF alkane mass retrieval overestimated the alkane mass in the plume with 3.2%. For the different measurement areas the SOF alkane mass retrieval were in the range 1.3-8.5% larger than the true alkane mass, with the highest deviation in area 2 having a high aromatic content of the total VOC mass (50.5 % aromatic compared to 48.9 % alkane mass fraction). Table 4.12. Plume composition and SOF alkane mass retrieval interference error. (FS correspond to canister field sample as specified in appendix III). Measurement area

Canister sample

SOF alkane mass retrieval / true alkane mass in composite spectra

1 2 3 4 6 Average plume

FS5+11 FS4 FS12 FS2 FS1 Average(FS1+2+ 4+5+11+12)

1.026 1.085 1.021 1.013 1.049 1.032

Plume composition, mass fraction (%)

alkane 85.3 48.9 90 85.8 81.9 80.3

alkene 4.6 0.7 2.1 1.8 3.9 2.9

aromatic 9.5 50.5 7.4 12.3 13.7 16.4

alkyne 0.6 0 0.6 0.1 0.4 0.4

The canister samples were in general collected at or after sunset, to allow for better downmixing of elevated sources to the ground, and to obtain higher concentrations in the samples which leads to improved GC analysis. It should however be noted that there is always a risk that elevated sources or sources with a strong plume rise might not be proportionally mixed into the plume at ground level, if the plume is sampled close by the site.

43

The cross-interference of various compounds (Table 2.2) on the SOF alkane mass retrieval was determined using the average plume composition in Table 4.12 as a base case. For each compound specified in Table 2.2, an amount of the compound equivalent to 1 % of the overall VOC mass in the average plume spectrum was added to the base spectrum which was then evaluated. The response on the retrieved alkane mass was compared with the added VOC mass (any type of VOC) to give the cross sensitivity for respectively compound. Figure 4.22 shows a plume speciation sampling for area 3 with field sample canister FS12 (appendix III). The blue line corresponds to the integrated path where the canister was continuously sampled, repeatedly back and forth for 60 minutes. SOF measurements indicated the plume location. The black arrows show the variability in the wind direction during sampling and the mean wind speed was 3.5 m/s from the east-north-east. Figure 4.23 shows the result after GC lab analysis with proportions between alkanes, alkenes and aromatics and the speciation within each VOC group.

Figure 4.22 Example of a canister sampling on area 3, with mean wind speed 3.5 m/s from the east-north-east. Arrows indicate the variability of the wind direction during sampling. Plume location was determined with SOF to guide the integration distance of the sample (repeatedly cross plume integration back and forth along the blue line).

44

Figure 4.23 Example of the result from a canister sampling at area 3. The graph at the top left summarizes the distribution of sampled VOC mass into alkanes, alkenes and aromatics. Top right the relative distribution among the alkanes are shown, whereas lower left and lower right show the corresponding for alkenes and aromatics respectively.

45

5. Discussion 5.1 Measurement uncertainty For the SOF flux measurement average emission values, the largest uncertainty originates from the uncertainty in the wind field, specifically in attributing the wind speed correctly for the average plume mass transport. In the Rotterdam survey, wind measurements, with GPS sondes and various wind masts, were set up to serve a large area encompassing several industrial complexes, as shown in Figure 3.1, 3.4 and 3.5. From the wind speed data in Figure 3.4, the 1σ spread relative to the 0-200 m GPS sonde wind is 15%. This spread is attributed as the uncertainty in using the 0-200 m wind for the average plume mass transport, which is the one that we have been using as the most likely plume wind speed during the campaign. Including various ground wind meters and 0-100 respectively 0-500 m averages for the GPS sondes, we assume that the spread in wind measurements cover all variations in the wind due to height and spatial location. The wind direction is also associated with an uncertainty, and this error actually depends on how orthogonal to the plume the transect is being conducted. When the plume transect is done orthogonal to the wind direction an uncertainty in the wind direction does not affect the flux uncertainty to a large extent, but the effect becomes more severe at more oblique angles. An 11º wind direction uncertainty (1σ) was estimated for the Rotterdam measurements, which would imply a 1.9% flux uncertainty at an orthogonal plume transect, and 12 % on average for the transect angle sector 0-45º. The absorption line parameters of the retrieved compounds are well established in published databases, stating an uncertainty of 3.5% for the spectroscopic part [Sharpe, 2004]. To this a retrieval uncertainty is added (6.6%). This is the combined effect of instrumentation and retrieval stability on the retrieved total columns during the course of a plume transect. Also affecting uncertainty is error in the SOF alkane mass retrieval due to interference of various compounds in the plume mixture (3.2 %). The composite flux measurement uncertainty for the SOF measurements, obtained as the square root sum of the quadratic errors for the parameters described above, is around 21% for the retrieved alkane mass, see Table 5.1. Table 5.1 Uncertainty estimation of the flux measurements (the variability of the sources not taken into account). Wind Speed

Wind Direct

Spectroscopy (cross sections)

Retrieval error

SOF Composite flux alkane measurement uncertainty mass error 15% a) 12% b) 3.5% c) 6.6% d) 3.2% e) 21% f) Alkanes a) Comparing all the wind data overlapping with the 0-200 m GPS sond averages, the data spread 15 % (1σ, 68%). b) The 1σ deviation (68%) among the wind data compared to the 0-200 m sonde is 11º. For a plume transect orthogonal to the wind direction, which is always the aim, this would give a 1.9% error. On average for the sector 0-45º, the error is 12%, due to the uncertainty in wind direction of 11º. c) Includes systematic and random errors in the cross section database. d) The combined effects of instrumentation and retrieval stability on the retrieved total columns during the course of a plume transect. e) The error of the SOF alkane mass retrieval in the average Rotterdam harbor plume composition f) The composite square root sum of squares uncertainty

46

5.2 Representativeness of data and comparison to reported values The measured average alkane emissions in the studied part of the Rotterdam harbor area for the month of September 2008 was 3997 tons/month (5553 kg/h), with a measurement uncertainty of 21%. The emission value is more than 4 times (300%) higher than the reported total VOC emissions for the 6 subareas of the Rotterdam harbor which are included in the measuring campaign, corresponding to 911 tons/month (1265 kg/h). For the individual subareas the discrepancy between measured and reported values varies between 2 to 14 times, Table 5.2. The discrepancy is consistent with studies conducted in the US [De Gouw, 2009; Mellqvist, 2007; Mellqvist, 2009]. The comparison between measured and reported values is conservative since only alkanes are measured with the SOF method while total VOCs are reported. The canister sampling carried out in the project shows that the average aromatic to alkane mass fraction was about 20% (also typical for Swedish refineries) except for area 2 in which 50% of the plume was aromatic species. Total VOC emissions estimated from alkane measurements and canister sampling are thus at least 20% greater than measured alkane emissions for all areas except area 2 where the increase is 100%. However, since few canisters were taken, the representativeness of these measurements is uncertain. Table 5.2 Measured alkane and reported VOC emissions. Area 1 2 3 4 5 6 Total

Measured average alkane emission (kg/h) 792 704 964 671 768 1654 5553

Reported VOC emission (kg/h) 391 284 119 81 269 121 1265

Discrepancy Factor 2.0 2.5 8.1 8.3 2.5 13.7 4.4

The SOF emission data reflect hourly and day-to-day variations in the VOC emissions during one month, but they do not include diurnal variability. How much do the emissions then actually change from day to night and at different solar conditions and are the measurements representative for a full years operation? The main factors affecting the emissions are the a) industrial activity, b) ambient temperature, c) the wind speed and d) the solar radiation. For process area emissions only the first parameter is important, since the emissions are then driven by pressures inside the industrial processes, while all four parameters are relevant for the tank emissions. Noteworthy, is that tank emissions may correspond to 60% of the total VOC emissions from a plant [Kihlman, 2005b] while the process area emissions correspond to about 30%. The industrial activity (a) is in our mind the most difficult parameter to understand from a measurement study during a month. The emissions from a refinery can be divided into continuous emissions which occur most of the year and which vary slightly with the weather conditions and upset emissions which occur only during a short time interval but with a certain frequency. Examples of the former are emissions through leaking seals in crude oil tanks, fugitive emissions trough flanges in the process area and diffuse emission from the

47

cleaning basins in the water treatment area. Examples of upsets are emissions occurring when loading ships, cleaning or repairing tanks or flaring. SOF data obtained over a period of a month will include both continuous and upset emissions but the latter to a variable extent. The continuous emissions can be extrapolated to annual emissions, taking into account effects of temperature and wind, while to do the same with the upset emissions one need to understand these in much more detail, in terms of how often they occur and their time duration. This requires more process oriented studies of the emissions. To understand the emission measurements better, with respect to continuous and upset emissions, it is useful to plot emission frequency distribution curves (histograms) of the data, as has been done in section 4. Figure 4.6 shows a histogram for area 2, measured on 3 separate days, in which a distribution function has been fitted to the data (Generalized extreme value distribution fitting). The distribution functions peaks at 497 kg/h and is rather skew, with measurements as high as 1300 kg/h. Our interpretation here is that the curve’s peak corresponds to the continuous emissions from the area (i.e. 500 kg/h or 4380 tons/year) while the tail of the distribution function, with the higher emission values, corresponds to upset emissions. The histogram for area 6, Figure 4.15, is in contrast to the data from area 2 much more symmetric and the maximum value of 1480 kg/h is relatively close to the average value. This indicates that few upset emission occurred here during the SOF measurements. In Table 5.2 we have used the interpreted continuous part to calculate an annual emission, see below. This is thus a conservative estimate not including the tail of upset emissions. The ambient temperature (b) and solar radiation (d) affect the surface temperature of the liquid in the tanks. This alters the surface vapor pressure which is directly proportional to the emissions from tanks. Standard emissions estimation procedures take these factors into account, typically over time scales of a month or greater. In the case of solar radiation, according to AP-42 (US EPA 2006) breathing losses from a fixed roof tank increase approximately 0.5% for a 1% increase in average solar insolation. For tanks with external or internal floating roofs, which are presumably those most in use today, and assuming vents or breakers are closed, the effect according to AP-42 (eqs.3-25 & 3-26) is less pronounced. An example for an external floating roof tank containing crude oil (RVP 5) is shown in Figure 5.1. In this case a doubling of solar radiation from the annual average for a northern European site produces a change of less than 1°C in liquid surface temperature and around 2 % increase in emissions. However, for SOF measurements in summer average solar radiation can be more than double the annual daily average and storage tank emissions under SOF daytime measurements are from 3 and 12 % higher than annually averaged emissions (this range is indicated as the blue field in Figure 5.2). The AP-42 model is based on empirical formulas developed for the estimation of monthly and yearly emissions, rather than the momentaneous emissions. For this reason we have studied the impact of solar radiation on tank emissions also in another manner by estimating how much the temperature at the liquid surface of the tank actually varies over the course of a day depending on solar radiation. This has been done by modeling the energy exchange at the surface of the tank, to the liquid surface and into the liquid. While precise estimates of liquid surface temperature depend on the amount and residence time of the liquid in the tank, wind speed, type of floating roof, thickness of the tank wall and amount of mixing in the tank, here we have generalized a one-dimensional energy balance for a non contact external floating roof tank.

48

20

18%

19

14%

18

12% 17 10% 16 8% 15 6% 14

4%

13

2% 0% 100

Average liquid surface temperature (C)

Relative Change in Emissions

16%

Emissions Liquid Surface Temp

200 300 Annual daily average

400

500 600 SOF daytime average

700

800

900 Peak 1 hour

12 1000

Solar insolation (W/m2)

Figure 5.1 Calculated change in liquid surface temperature and VOC emissions versus solar insolation for an external floating roof tank containing crude oil (RVP 5) based on AP-42. Base case is 105 W/m2/day (Göteborg, Sweden 2008 = 109 W/m2/day, Göteborgs Stad Miljöförvaltningen). Indicated in the figure is also the range for the daily annual average solar insolation at different geographical locations, the range for the average solar insolation when conducting SOF measurements over the course of a day and finally the range for the solar insolation around noon, at various locations.

First we calculated the impact of solar radiation of 800 W/m2 on the surface temperature of the liquid below a floating roof tank segment. It was assumed that the roof floats above the liquid and that a vapor pressure builds up between the liquid and the roof, that the liquid is still with the heating properties of transformer oil, that effects from the edges of the tank are negligible, that the roof is painted white and that no wind cooling and no significant free convection occurs. About 170 W/m2 of the solar light is absorbed by the floating roof and part of this penetrates down to the surface of the liquid, while the rest of the absorbed energy is reemitted as thermal radiation or through convection. The energy at the surface of the liquid is further transported downwards into the liquid by heat conduction. This results in an increase of the surface temperature of the liquid with a maximum in the early afternoon with a subsequent change in the vapor pressure. For a typical crude oil mixture [Nieto 1976], assuming an ideal solution (Raoults law), and measured gas phase data from a crude oil tank the change in vapor pressure versus temperature is about 3%/K at 20 °C (Figure 5.2). This rough calculation hence predicts that the average emission between 9 am to 5 pm, which corresponds to a typical SOF measurements time interval, will be 25% higher compared to the non solar condition at night. The solar noon emission value (when the sun is the strongest) can differ even more, around 30%. These numbers get distinctively lower if ones assumes wind cooling and free convection which has been neglected here. Hence, according to our simple tank calculation SOF measurements of a crude oil tank park conducted over the course of the day could produce a 25% higher emission value than the night time emission value. The number is higher than the emission overestimations obtained with the AP-42 model in Figure 5.1, but this is partly explained by what base case is compared against (average annual daily insolation or night time conditions). This calculation applies for an external floating roof tank, i.e. typically a crude oil tank, for which the wind 49

cooling of the floating roof is small, since it is exposed to the wind mostly when the tank is full. For a fixed roof tank with internal floating roof the sun will have less impact, since there are two ceilings protecting the liquid surface from radiation and since the cooling by the wind at a typical SOF measurement at 5m/s is quite strong. Measurements in Sweden [Kihlman, 2005b] show that crude oil tanks and product tanks correspond to about 1/3 each of the total emissions from the refinery. The impact of the ambient temperature on the surface temperature of the liquid in the tanks depends on how long time the liquid stays in the tank, i.e. the flow through rate. For tanks with a high flow through rate, such as crude oil tanks, the bulk temperature stays almost constant and the surface temperature will in this case be somewhere in between the ambient temperature and the bulk temperature. For tanks with low flow-through rate the bulk temperature will on the other hand be affected. The SOF measurements in the campaign were conducted at about 18oC, which is 6oC warmer than the annual average temperature. As an upper estimate, the measured values during the campaign therefore had 18% higher emissions, compared to the annual average. (6oC·3%/K). This value would be true if the liquid surface in the tanks has ambient temperature, but is rather a conservative error estimate since the bulk temperature in most tanks is rather constant and since most tanks have floating roofs that isolates the liquid surface. 50%

Change in vapor pressure

40% 30% 20% 10% 0% -10% -20% -30% -40% -50% 10

12

14

16

18

20

22

24

26

28

30

Temperature C Figure 5.2 Calculated change in vapor pressure (Raoults law) versus temperature for a typical crude oil mixture (5% ethane, 48% propane, 33% butane, 8% pentane, and 6% hexane).

The wind (c) has a potential impact on tank emissions since it affects the removal speed of the gas from the tank. For instance O’Connor et al. [1999] reports a square dependence on the wind speed on tank emissions. Such a strong correlation is not clear from studies that we have conducted [Kihlman, 2005a]. In Figure 5.2 the relative emissions of alkanes from area 3, 4 and 5, corresponding mostly to tank emissions, are shown versus wind speed, together with a linear fit. The emissions changes by 5.5% for a 1 m/s change in the wind speed but the

50

correlation is poor (R2 = 0.13). The same analysis for data only from area 3 yields a slope of 8% and higher correlation (R2 = 0.28). The SOF measurements during the campaign were carried out at an average wind speed of 6.9±2.5 m/s (area 1: 5.9 m/s, area 2: 5.4 m/s, area 3: 4.8 m/s, area 4: 5.6 m/s, area 5: 8.4 m/s, area 6: 10 m/s) to be compared to the average wind speed in Rotterdam of 5 m/s (data for Rotterdam airport from http://www.windfinder.com). The data in Figure 5.3 yields the following potential overestimations of the emissions, compared to the annual average: area 1: +5%, area 2: +2.2%, area 3:- 1%, area 4: 3.3%, area 5: 19% and area 6: 28%). This is an upper estimate of the wind effect, since process area emissions typically constitute one third of a refinery’s emission, and are not as affected by wind forces as external roof tanks.

Emission rate (relative to average)

250%

200%

150%

100%

50%

0% 0

2

4

6

8

10

12

wind speed m/s Figure 5.3 Relative emission rate versus windspeed for area 3, 4 and 5 and a linear fit to all data points. Data corresponds to 6 individual days, 27 transects.

51

In this section we have gone through factors which influence how representative the measured SOF data for the month of September are for estimating the yearly emissions. In Table 5.3 annual emission values of VOCs have been estimated for the various areas by correcting the measured alkane values according to the discussion above. In general we have used upper estimates of the influence of the various factors. Firstly the continuous alkane emissions have been obtained as the peak of the measured emission distribution functions. These values have been multiplied by 8760 hours and then they have been corrected for the aromatic fraction (20%), wind effect as shown in the table, the maximum effect of ambient temperature (18%) and effect of daily average solar insolation (25%), as discussed above. Note that the derived annual VOC emission values are underestimates of the true annual emissions, since upset emissions are missing to a large degree and since the effects of ambient temperature and solar insolation are upper estimates. Table 5.3 Estimated annual alkane emissions, excluding upset emissions. Area

Continuous Wind Annual VOC Reported VOC alkane factor emission* Emission Tons/year Tons/year (kg/h) 5560 1 819 1.05 3421 3466 2 497 1.022 2486 7056 3 1000 1.01 1045 3339 4 484 1.033 709 4696 5 784 1.19 2355 8241 6 1480 1.28 1059 32356 Total 5094 11075 * Annual emission=alkane·af·8.76/(wf·sf·Tf) where alkane: continuous alkane emission, af: aromatic mass factor (1.2), wf: wind factor, sf: solar factor, Tf: ambient temperature factor

Discrepancy Factor

1.6 1.4 6.8 4.7 2.0 7.8 2.9

It may be somewhat puzzling that the there is a factor 1.4 to 7.8 discrepancy between the SOF measurements and the values reported. The reported values are mostly calculated and we believe that these methods assume far too ideal conditions. For instance, SOF and mobile extractive FTIR measurements that have been conducted in Sweden show that leaks on oil tanks are one or two order of magnitude larger than what is calculated and the measurements also show that a few malfunctioning tanks may stand for a large portion of the total emissions. Hence, if a few tanks are leaking two orders of magnitude more than calculated than there will easily be a significant discrepancy between the calculated emissions for a tank park and the measured ones.

52

6. Acknowledgements This work was funded by the DCMR (Environmental Protection Agency, Rijnmond). The authors would like to thank Rob Van Doorn and Willem-Jan Okkersee DCMR for assisting with the SOF measurements and analysis.

7. References Buhr, M., Alvarez, S., Kauffmann, L., Shauck, M., Zanin, G., Alkene/NOy Emission Ratios observed from the Baylor Aztec during the 2006 TexAQS II study and Comparison with results obtained during 2001-2002, Rapid science synthesis workshop, Austin, October 12, 2006 Börjesson, G., Samuelsson, J., Chanton, J., Adolfsson, R., Galle, B., Svensson, B.H., A national landfill methane budget for Sweden based on field measurements, and an evaluation of IPCC models, Tellus, 61B, 424-435, 2009. De Gouw, J.A., et. al., Airborne Measurements of Ethene from Industrial Sources Using Laser PhotoAcoustic Spectroscopy, Environmental Science and Technology, 43, 2437-2442, 2009. DCMR Emissie-overzicht-2008, (DCMR emission overview 2008), Available at http://www.dcmr.nl/binaries/publicatie/2008/lucht/emissie-overzicht-2008.pdf, 2008. Galle, B., Mellqvist, J., et al., Ground Based FTIR Measurements of Stratospheric Trace Species from Harestua, Norway during Sesame and Comparison with a 3-D Model, JAC, 32, no. 1, 147-164, 1999. Galle, B., Samuelsson, J., Svensson, B.H., Börjesson, G., Measurements of methane emissions from landfills using a time correlation tracer method based on FTIR absorption spectroscopy, Environ. Sci. Technol. 35, 21-25, 2001. Griffith, D.W.T., Synthetic calibration and quantitative analysis of gas-phase FT-IR spectra. Applied Spectroscopy, 50(1), p. 59-70, 1996. Hurley, P.J., Physick, W.L., Luhar, A.K., TAPM: A practical approach to prognostic meteorological and air pollution modeling, Environmental Modelling and Software, 20(6), p. 737, 2005. Kihlman, M. (2005a), Application of solar FTIR spectroscopy for quantifying gas emissions, Technical report No. 4L, ISSN 1652-9103, Department of Radio and Space Science, Chalmers University of Technology, Gothenburg, Sweden, 2005a. Kihlman, M., J. Mellqvist, and J. Samuelsson (2005b), Monitoring of VOC emissions from three refineries in Sweden and the Oil harbor of Göteborg using the Solar Occultation Flux method, Technical report, ISSN 1653 333X, Department of Radio and Space, Chalmers University of Technology, Gothenburg, Sweden, 2005. Mellqvist, J., Application of infrared and UV-visible remote sensing techniques for studying the stratosphere and for estimating anthropogenic emissions, doktorsavhandling, Chalmers tekniska högskola, Göteborg, Sweden, 1999.

53

Mellqvist, J och Galle, B., Utveckling av ett IR absorptionssystem användande solljus för mätning av diffusa kolväteemissioner (Development of an IR absorption system for utilization of sunlight for measurement of fugitive hydrocarbon emissions, in Swedish), Report to Preems miljöstiftelse Preem Environmental Foundation), july 1999. Mellqvist, J., Arlander, D. W., Galle, B., Bergqvist, B., Measurements of Industrial Fugitive Emissions by the FTIR-Tracer Method (FTM), IVL report, B 1214, 1995. Mellqvist, J., Samuelsson, J., Rivera, C., Lefer, B., Patel, M., Measurements of industrial emissions of VOCs, NH3, NO2 and SO2 in Texas using the Solar Occultation Flux method and mobile DOAS, Project H053.2005, Texas Environmental Research Consortium., Texas. (http://www.harc.edu/projects/airquality/Projects/Projects/H053.2005), 2007. Mellqvist, J., et al., Measurements of industrial emissions of alkenes in Texas using the Solar Occultation Flux method, accepted to JGR, 2009. O’Connor, S.J, Walmsley, H.L. et al., Measurement of the VOC emission from from the Shell raffinaderi, Göteborg, Shell Global Solution, report OP.99.47110, 1999 (available at provincial government of Västra Götaland, www.o.lst.se). Potter, A., “Analysis Method for Ozone Precursor Volatile Organic Compounds”, Report to the Swedish Provincial government of Västra Götaland, ordered by Swedish EPA, IVL Rapport U1121, Available at http://www.lansstyrelsen.se/NR/rdonlyres/C9EB9EF6-D062-45F893FA-C1A11A73D8EB/100156/U2245Stenungsundrapport_slutversion0804181.pdf, 2005. Rinsland, C. P., Zander, R., Demoulin, P., Ground-based infrared measurements of HNO3 total column abundances: long-term trend and variability, J. Geophys. Res., 96, 9379–9389, 1991. Rothman, HITRAN 2000, http://www.hitran.com, 2005 Samuelsson, J., et. al., VOC measurements of VOCs at Nynas Refinery in Nynäshamn 2005 (Utsläppsmätningar av flyktiga organiska kolväten vid Nynas Raffinaderi i Nynäshamn 2005, in Swedish), Bitumen refinery official report to provincial government 2005, Available at: http://www.fluxsense.se, 2005. Sharpe, S., et al., Gas-Phase Databases for Quantitative Infrared Spectroscopy, Applied Optics, 58(12), 2004. Tucker, S., Marine boundary Layer dynamics and heights during TexAQS 2006: HRDL measurements from the RV Brown, Principal Findings Data Analysis Workshop TexAQS II, Austin, May 29-June 01, 2007. (available at www.tceq.state.tx.us) USEPA Emission Factor Documentation for AP-42 Section 7.1 Organic Liquid Storage Tanks Final Report. Available at http://www.epa.gov/ttn/chief/ap42/ch07/bgdocs/b07s01.pdf, 2006. Walmsley, H. L., O’Connor, S. J., The accuracy and sensitivity of infrared differential absorption lidar measurements of hydrocarbon emissions from process units, Pure Appl. Opt., 7, 907–925, 1998.

54

Appendix I: SOF Transects The Table shows emission in kg/h, time when the SOF transect was conducted, wind speed and wind direction as well as the average emission for the days measured. Note! Raw data is presented here, i.e. it does not include adjustments for the ground mast meters to the 0-200 m average as obtained from the balloon sondes (according to Section 3). Area 1  Day 080829  Emission Time Wind  851.0 kg/h 165705‐170800 (Wind 4.3m/s 326deg)    677.0 kg/h 171721‐172740 (Wind 3.9m/s 319deg)    828.2 kg/h 173404‐174515 (Wind 4.0m/s 337deg)    Average: 785.4±94.6 kg/h (12.04%)     Day 080831  Emission Time Wind  613.6 kg/h 145411‐145651 (Wind 4.4m/s 152deg)   Average: 613.6    Day 080914  Emission Time Wind  515.0 kg/h 92506‐93415 (Wind 6.0m/s 54deg)    940.0 kg/h 133120‐133929 (Wind 9.4m/s 76deg)    618.5 kg/h 90437‐92153 (Wind 5.3m/s 56deg)  744.6 kg/h 101654‐102843 (Wind 5.9m/s 59deg)    760.3 kg/h 103045‐104143 (Wind 7.2m/s 67deg)    651.3 kg/h 105337‐110346 (Wind 8.2m/s 78deg)    697.2 kg/h 110346‐111138 (Wind 8.3m/s 83deg)    754.9 kg/h 120703‐121900 (Wind 8.0m/s 82deg)    759.7 kg/h 124423‐125725 (Wind 7.7m/s 86deg)    753.7 kg/h 125743‐130916 (Wind 8.1m/s 78deg)    782.0 kg/h 132100‐133124 (Wind 9.1m/s 78deg)    Average: 725.2±107.5 kg/h (14.83%)  

  Day 080916  Emission Time Wind  1081.0 kg/h 120110‐121012 (Wind 4.1m/s 50deg)    1023.1 kg/h 121210‐122226 (Wind 4.0m/s 45deg)    987.0 kg/h 123449‐124930 (Wind 4.0m/s 47deg)    925.7 kg/h 151906‐152957 (Wind 4.2m/s 43deg)    Average: 1004.2±65.1 kg/h (6.48%)     Day 080917  Emission Time Wind  769.9 kg/h 115629‐120829 (Wind 4.2m/s 59deg)    858.0 kg/h 120902‐122051 (Wind 4.6m/s 75deg)    Average: 813.9±62.3 kg/h (7.66%)     Day 080918  Emission Time Wind  1053.7 kg/h 135150‐135440 (Wind 5.5m/s 67deg)    749.9 kg/h 135150‐135440 (Wind 5.5m/s 67deg)    384.5 kg/h 135150‐135440 (Wind 5.5m/s 67deg)    786.4 kg/h 135150‐135440 (Wind 5.5m/s 67deg)    698.0 kg/h 135150‐135440 (Wind 5.5m/s 67deg)    951.4 kg/h 152144‐153102 (Wind 6.7m/s 75deg)   

55

782.0 kg/h 135101‐140224 (Wind 5.6m/s 67deg)    1088.8 kg/h 151154‐152150 (Wind 6.2m/s 69deg)    Average: 811.8±225.4 kg/h (27.77%)  

     

Area 2  Day 080830  Emission Time Wind  823.0 kg/h 173730‐174524 (Wind 7.6m/s 90deg)  Average: 823.0±246.9 kg/h (30.00%)  

  Day 080909  Emission Time Wind  1354.8 kg/h 101929‐102654 (Wind 4.7m/s 150deg)  897.4 kg/h 105825‐110746 (Wind 4.2m/s 150deg)    1048.4 kg/h 110746‐111600 (Wind 4.2m/s 152deg)    456.4 kg/h 120550‐121422 (Wind 4.3m/s 148deg)    361.7 kg/h 121815‐122357 (Wind 4.4m/s 148deg)    520.1 kg/h 122637‐123353 (Wind 4.8m/s 148deg)    449.8 kg/h 123722‐124201 (Wind 4.6m/s 149deg)    380.8 kg/h 125106‐125630 (Wind 4.9m/s 142deg)    611.7 kg/h 130210‐130743 (Wind 5.4m/s 142deg)    583.2 kg/h 163314‐163931 (Wind 6.4m/s 150deg)    565.1 kg/h 164104‐164835 (Wind 6.3m/s 152deg)    Average: 657.2±312.9 kg/h (47.61%)  

  Day 080911  Emission Time Wind  579.5 kg/h 100608‐102642 (Wind 4.8m/s 151deg)    657.4 kg/h 145149‐145959 (Wind 6.3m/s 143deg)    892.0 kg/h 150246‐151511 (Wind 6.1m/s 144deg)    660.1 kg/h 151509‐152322 (Wind 6.3m/s 142deg)    419.2 kg/h 152626‐153425 (Wind 6.2m/s 149deg)    577.8 kg/h 153606‐154222 (Wind 6.2m/s 148deg)    Average: 631.0±154.9 kg/h (24.55%)  

     

Area 3  Day 080825  Emission Time Wind  884.2 kg/h 151954‐153108 (Wind 8.6m/s 241deg)  Average: 884.2    Day 080911  Emission Time Wind  717.0 kg/h 163926‐164956 (Wind 4.5m/s 144deg)    824.7 kg/h 171249‐171740 (Wind 4.5m/s 154deg)    Average: 770.8±76.2 kg/h (9.88%)     Day 080914  Emission Time Wind  707.3 kg/h 182332‐183106 (Wind 5.4m/s 72deg)  Average: 707.3   

56

Day 080915  Emission Time Wind  842.1 kg/h 134247‐135711 (Wind 6.3m/s 73deg)    510.3 kg/h 141322‐142754 (Wind 6.0m/s 104deg)    1254.7 kg/h 150204‐151315 (Wind 5.8m/s 94deg)    960.1 kg/h 152049‐154102 (Wind 4.9m/s 83deg)    Average: 891.8±307.9 kg/h (34.52%)  

  Day 080918  Emission Time Wind  639.4 kg/h 174834‐180105 (Wind 4.2m/s 96deg)    526.7 kg/h 180120‐181508 (Wind 3.8m/s 82deg)    Average: 583.0±79.7 kg/h (13.67%)     Day 080920  Emission Time Wind  751.6 kg/h 110043‐111032 (Wind 3.4m/s 72deg)    609.1 kg/h 111055‐111856 (Wind 3.2m/s 90deg)    724.2 kg/h 114541‐115148 (Wind 2.8m/s 62deg)    692.0 kg/h 120330‐121724 (Wind 3.1m/s 72deg)    904.4 kg/h 124504‐125400 (Wind 4.5m/s 74deg)    1173.2 kg/h 125529‐130459 (Wind 4.7m/s 80deg)    1290.8 kg/h 133237‐134129 (Wind 4.5m/s 78deg)    1151.9 kg/h 134303‐135306 (Wind 5.4m/s 81deg)    951.3 kg/h 142157‐143135 (Wind 5.2m/s 74deg)    1005.4 kg/h 142429‐143227 (Wind 5.1m/s 73deg)    Average: 925.4±230.7 kg/h (24.93%)  

     

Area 4  Day 080825  Emission Time Wind  252.5 kg/h 102640‐103251 (Wind 4.4m/s 228deg)    364.9 kg/h 111432‐111933 (Wind 5.4m/s 236deg)    343.2 kg/h 115239‐120111 (Wind 5.7m/s 236deg)    383.1 kg/h 124350‐124653 (Wind 5.9m/s 243deg)    330.3 kg/h 133933‐134258 (Wind 7.1m/s 244deg)  778.4 kg/h 141559‐142003 (Wind 7.5m/s 235deg)    959.7 kg/h 143800‐144343 (Wind 7.8m/s 238deg)    538.1 kg/h 145213‐145656 (Wind 7.6m/s 236deg)    577.9 kg/h 151558‐152020 (Wind 8.2m/s 243deg)    Average: 503.1±235.3 kg/h (46.76%)     Day 080830  Emission Time Wind  515.5 kg/h 132058‐132523 (Wind 4.2m/s 97deg)  Average: 515.5±154.6 kg/h    Day 080901  Emission Time Wind  555.0 kg/h 153514‐153951 (Wind 7.6m/s 244deg) 

57

Average: 555.0    Day 080914  Emission Time Wind  540.6 kg/h 162949‐163944 (Wind 5.0m/s 81deg)    721.7 kg/h 174225‐175110 (Wind 5.3m/s 69deg)    730.1 kg/h 181604‐182432 (Wind 5.4m/s 67deg)    Average: 664.2±107.1 kg/h (16.12%)     Day 080920  Emission Time Wind  331.6 kg/h 105448‐110019 (Wind 3.4m/s 65deg)    312.3 kg/h 112144‐112604 (Wind 2.5m/s 70deg)    308.7 kg/h 135427‐135953 (Wind 3.7m/s 74deg)    480.7 kg/h 141836‐142348 (Wind 4.8m/s 75deg)    Average: 358.3±82.2 kg/h (22.93%)  

   

Area 5  Day 080823  Emission Time Wind  1203.4 kg/h 125418‐130227 (Wind 9.5m/s 343deg)  1220.5 kg/h 131345‐132127 (Wind 9.7m/s 343deg)   1794.5 kg/h 132159‐132746 (Wind 9.4m/s 342deg)  1275.7 kg/h 140341‐141156 (Wind 8.0m/s 342deg)  892.2 kg/h 141246‐142332(Wind 7.8m/s 341deg)  1083.2 kg/h 142425‐143606 (Wind 7.8m/s 339deg)  1147.1 kg/h 144129‐145047 (Wind 7.7m/s 335deg)  Average: 1230.9±278.1 kg/h (22.59%)     Day 080825  Emission Time Wind  367.9 kg/h 102102‐102752 (Wind 7.1m/s 256deg)    572.1 kg/h 112010‐112529 (Wind 8.5m/s 241deg)    827.6 kg/h 114450‐115304 (Wind 8.7m/s 237deg)    936.5 kg/h 123611‐124346 (Wind 10.3m/s 238deg)    863.6 kg/h 131304‐131826 (Wind 10.1m/s 237deg)    857.3 kg/h 133451‐133936 (Wind 11.3m/s 236deg)    894.5 kg/h 141844‐142333 (Wind 11.1m/s 240deg)    923.0 kg/h 143424‐143903 (Wind 11.0m/s 237deg)    1071.8 kg/h 145656‐150211 (Wind 11.1m/s 237deg)    875.5 kg/h 151240‐151655 (Wind 11.2m/s 240deg)    Average: 819.0±201.5 kg/h (24.60%)     Day 080831  Emission Time Wind 

58

749.5 kg/h 173448‐174215 (Wind 5.6m/s 199deg)  Average: 749.5 kg/h    Day 080901  Emission Time Wind  716.3 kg/h 161000‐161245 (Wind 11.2m/s 250deg)   611.3 kg/h 162409‐162812 (Wind 10.6m/s 250deg)    1069.0 kg/h 163049‐163404 (Wind 10.5m/s 248deg)    Average: 799±240 kg/h (30.01%)     Day 080906  Emission Time Wind  573.4 kg/h 155104‐155515 (Wind 10.0m/s 232deg)  Average: 573.4 kg/h    Day 080920  Emission Time Wind  137.3 kg/h 104632‐105442 (Wind 2.4m/s 108deg)  345.9 kg/h 112545‐113140 (Wind 2.7m/s 88deg)  277.4 kg/h 113257‐113820 (Wind 2.7m/s 84deg)  343.1 kg/h 140027‐140602 (Wind 5.6m/s 64deg)  1080.9 kg/h 140553‐141238 (Wind 5.6m/s 68deg)  Average: 436.9±369.8 kg/h (84.64%)  

     

Area 6  Day 080823  Emission Time Wind  1737.5 kg/h 112420‐113457 (Wind 9.8m/s 329deg)    1415.5 kg/h 113530‐114736 (Wind 10.1m/s 335deg)    2645.5 kg/h 115526‐120159 (Wind 10.2m/s 340deg)    2100.9 kg/h 122130‐123600 (Wind 9.4m/s 341deg)  2091.6 kg/h 152157‐153333 (Wind 7.5m/s 339deg)  1144.2 kg/h 153414‐154723 (Wind 7.2m/s 338deg)  Average: 1855.9±539.2 kg/h (29.05%)     Day 080825  Emission Time Wind  1219.0 kg/h 113810‐114444 (Wind 8.9m/s 236deg)  1790.5 kg/h 122813‐123611 (Wind 10.0m/s 238deg)  1910.9 kg/h 131822‐132507 (Wind 10.7m/s 237deg)  1571.0 kg/h 142331‐142812 (Wind 10.9m/s 240deg)  1867.0 kg/h 142858‐143427 (Wind 11.0m/s 240deg)  1938.8 kg/h 150208‐150656 (Wind 11.3m/s 237deg)  1651.2 kg/h 150829‐151237 (Wind 10.4m/s 239deg) 

59

2264.4 kg/h 154159‐155057 (Wind 11.6m/s 239deg)    Average: 1776.6±306.9 kg/h (17.28%)  

  Day 080901  Emission Time Wind  1259.1 kg/h 131433‐132049 (Wind 10.5m/s 252deg)    1334.5 kg/h 132458‐133222 (Wind 10.5m/s 250deg)    1591.1 kg/h 133545‐134406 (Wind 10.5m/s 248deg)    1449.7 kg/h 134420‐135121 (Wind 10.2m/s 248deg)    1134.5 kg/h 135442‐140517 (Wind 10.4m/s 250deg)    1418.5 kg/h 140642‐141635 (Wind 10.9m/s 249deg)    1658.8 kg/h 161245‐161744 (Wind 11.2m/s 250deg)  1306.6 kg/h 161823‐162539 (Wind 10.3m/s 250deg)   Average: 1394.1±173.0 kg/h (12.41%)     Day 080910  Emission Time Wind  1592.7 kg/h 131635‐132300 (Wind 7.9m/s 233deg)  Average: 1592.7 

60

Appendix II: Reported emission calculations Table AII.1 shows reported [DCMR Emissie-overzicht-2008] annual emissions of VOCs for each area in tons. Table AII:2 shows the corresponding values converted to hourly emissions in kg. Table AII.1. Reported [DCMR Emissie-overzicht-2008] annual emissions of VOCs for the subareas included in the SOF study in tons.

Area 1  tons  SNR  Nerefco Pernis  SNV Pernis  SNC  Sum     Area 2  Vopak Botlek  Esso  Exxon Rop  Exxon RRP  Sum     Area 3 

    

   2757     207     82    

Exxon RAP  MET  Vopak  Europoort  Sum 

tons  220 444 39

375    

Akzo‐neste 

6

3421           192    

709

173     126    

Sum    Area 5  ADM  SNR  Europoort  TEAM  Sum      Area 6  Eastman‐ voridian  BP‐Nerefco 

746    

Sum 

2266  14  14  2486    ton 

Area 4  TEAM  KPE  Caldic 

                 

1045    

  

      2073 282 220 2355          53 1006 1059   

Table AII.2. Hourly emissions (kg/h) calculated from the data in Table AII.1.

Corresponds:  Area 1  Area 2  Area 3 

kg/h 

   391     284     119    

   Area 4  Area 5  Area 6 

kg/h  81 269 121

61

Appendix III: Canister sampling In the following the setup and results from respectively canister sampling at the different areas are presented in detail. The data is presented by area, and each section starts out with an overview of the sample location. Most samples correspond to about 60 minutes integration time. For these the blue line correspond the path where the canister was continuously sampled, repeatedly back and forth. The black arrows show the variability of the wind direction during sampling. Plume location was determined with SOF to guide the integration distance of the sample. For some areas additional long time samples were collected, corresponding to 5 hours integration time. For these the plume location during each of the five hour segments have been indicated by a semi-transparent rectangle, originating at the sample location and centered along the wind direction average for respectively hour, pointing up in wind. Following the sampling setup overview, the pie charts show the result of the GC lab analysis, with proportions between the VOC groups alkanes, alkenes and aromatics, as well as the speciation into compounds within each VOC group. In the original project outline, 4 main areas were identified for measurements. In the end, these were split into 6 areas, since the SOF data set was solid enough to allow for it. Hence, canister data can only be presented for areas 1-4 and 6. On average the plume sample canisters contained a VOC concentration of 139 μgm-3 (range 20-256 μgm-3), in excess above background concentration.

Area 1 FS5 080918 10:11-11:23 and FS11 080918 12:06-13:29

62

63

FS14 080917 300 min. Start 21:38.

64

Area 2 FS4 080919 19:58-20:59

65

Area 1+2. FS8, 080918 14:02-15:16

66

Area 3 FS12 080918 17:48-19:19

67

Area 4 FS2 080919 10:12-11:14

FS13 080917 300 min. Start 22:07.

68

69

Area 6 FS1 080919 18:29-19.11

FS6 080919 17:21-18:23

70

71

FS9 080919 300 min. Start 19:59.

72

Background samples FS7 080919 09:39-10:39 and FS10 080919 11:38-13:40.

73

74