Impact on air quality from increasing cruise ship activity in Copenhagen port

In July 2018, 616,098 people lived in the municipality of Copenhagen with a population density of 6,937 making it one of the most densely populated regions in Scandinavia. Figure 1 shows the port and how it stretches through the city of Copenhagen including all berths in the port. There are 18 berths for cruise ships in Copenhagen and 106 for other commercial ships. More than half of the cruise ships arriving in 2018 (table 1) were mooring in the inner areas of the port (mooring sites 177, 190, 192–198) in particular at the approximately 900 m long Langelinie Quay. The study domain is 4.6 km × 6.0 km.

Zoom In Zoom Out Reset image size

During the cruise ships season of 2018, measurements of particle number, NO, NOx and NO₂ were conducted at the top of a 23 m high residential building adjacent to berth no. 192 (henceforward reference to as LQ-23). The mooring sites 192 to 198 are positioned at the Langelinie quay, where half of the cruise ships during 2018 were hotelling at berth while in Copenhagen. The measurements were conducted by FORCE Technology using a chemiliuminescence NO/NO_x/NO₂ analyzer (Teledyne T200) (Fuglsang et al 2019) in the units of μg m⁻³. The Langelinie quay is positioned North to East of the monitoring site (approximately in the 0° to 90° sector). The measurement period was from 31/5 to 19/9 2018.

The emission inventory is including both cruise ships and all other commercial ships arriving and departing in Copenhagen port during 2018. The needed input data for constructing the emission inventories are; (i) detailed port list information that contains ship IMO code, ship type, gross tonnage, time of arrival/departure, berth no., and ship manoeuvring times provided by Copenhagen Malmö Port (Nørgaard, 2020); (ii) main engine size (kW), engine stroke (2/4) and ship build year data provided by Danish Shipowners (Kristensen, 2020); (iii) auxiliary engine power (kW) functions estimated from vessel data in the SHIP-DESMO model (e.g. Kristensen, 2017); and (iv) NO_x emission factors (g/kWh) provided by MAN Energy Solutions (2012) for slow speed (2-stroke) and medium speed (4-stroke) diesel engines as a function of ship build year.

Specific main engine and auxiliary engine power loads are used for the ships during manoeuvring in the port and hotelling at berth (Jensen et al 2021).

The NO_x emissions are calculated for each vessel in the port as the product of engine size (kW), engine load factor (%), time duration at berth/during manoeuvring (hrs) and NO_x emission factor (g/kWh). The emission calculations consider increased NO_x emissions at low main engine loads during manoeuvring in the port. NO_x emission rates (g/s) for all the individual ship berthing periods are subsequently derived from the emission results, and the emissions are spatially distributed according to the time and berth locations within the port.

The cruise ship NO_x emissions for the berths at the Langelinie quay are shown in table 1, as well as total NO_x emissions for cruise ships outside Langelinie, and total emissions for other commercial ships in the port. The emissions are shown for the full year of 2018 and the measurement period (31/5–19/9 2018) coinciding with the busiest cruise ship period.

For 2018 [measurement period] the percentage emission contribution of NO_x for cruise ships at Langelinie is 21% [27%] to the total port emissions. For cruise ships outside Langelinie the 2018 [measurement period] percentage emission contribution is 31% [44%]. Thus, cruise ships contribute with 52% [71%] of NO_x emitted by ships in Copenhagen port.

Separately, for Langelinie the largest emission percentage shares of NO_x are calculated for the berth numbers 197 with 52% [49%], 192 with 24% [24%], and 193 with 14% [16%].

The OML-Multi model is a local-scale atmospheric dispersion model for multiple point and area sources (Olesen et al 2007). It is a Gaussian plume model that relies on physical parameters such as friction velocity, heat flux, and atmospheric stability, and takes plume rise, partial penetration of the mixing layer, and building and terrain effects into consideration. The spatial domain of the OML model is flexible, and within the defined domain the model calculates the dispersion for all the defined sources. For our study, a receptor resolution of 100 m × 100 m was chosen covering the port area of Copenhagen and the central part of the city.

A simplified scheme for NO_x-O₃ chemistry (Düring et al 2011) is incorporated into the OML model. Background concentrations of NO_x, NO₂ and O₃, global radiation and wind direction are obtained from measurements at the urban background monitoring station HCOE 2.5 km from Langelinie. Where data is missing from the observations, concentrations are obtained from the regional scale atmospheric chemical transport model DEHM (Frohn et al 2021). The remaining meteorological data needed in OML-Multi (e.g., friction velocity, Monin-Obukhov length, temperature, boundary layer height) are obtained from the weather research and forecasting model, WRF, which drives the DEHM model (Skamarock et al 2008).

The stack height influences the atmospheric dispersion, and it is estimated for the individual cruise ships based on a linear correlation between ship stack height and brutto tonnes found by Olesen and Berkowic (2005). The mean stack height of the cruise ships in Copenhagen port was found to be 51.7 m with a standard deviation of 12.7 m. In addition, building effects from the cruise ships themselves will impact the dispersion, and based on previous findings we set the height of the ship to 67% of the stack height (Olesen and Berkowic 2005). The exhaust velocity is in all cases set to 27.5 m s⁻¹, while the exhaust temperature is set to 200 °C and 250 °C, for heavy fuel oil (HFO) and marine diesel/marine gas oil (MDO/MGO), respectively (Olesen and Berkowic 2005). A roughness length of 0.5 is chosen for OML-Multi corresponding to dense urban areas.

Emissions from the individual ships are released at the mooring site during the time they are at berth. This means that the point source shipping emissions are located at the exact berth no. coordinates provided by the port authorities. We estimate 15 min for manoeuvring at arrival and departure, however, for simplicity these emissions are released at berth. Simulations were done for the entire study area at the surface (1.5 m), at 25 m and 50 m heights.

To assess the general capability of OML-Multi to simulate the dispersion of pollutants in Copenhagen port the relation between the hourly measured and modelled NO_x concentrations at LQ-23 at Langelinie quay is examined with a quantile-quantile (q - q) plot (figure 2). The modelled NO_x concentrations are extracted from a height of 25 m for the time period covering the measurement campaign. The fit indicates a strong similarity with linearity close to x = y between the observed and modelled concentrations with a few difficulties for the model to accurately mimic the highest observed NO_x concentrations. This indicates that the observed and simulated atmospheric NO_x stem from the same distribution type and that OML-Multi can simulate the given situation in Copenhagen port with realistic values from low to peak concentrations of NO_x values except for the very highest.

Zoom In Zoom Out Reset image size

To better understand the origin of the emissions giving rise to peaks in the NO_x concentrations at LQ-23 we introduce the concept of likely concentration contribution (LCC) from shipping emissions at berth upwind from the measurement site. The LCC from emissions at berth up-wind is estimated by simple assumptions to the Gaussian plume model formulation

$$\begin{eqnarray*}&&\begin{array}{c}{C}_{l}=\displaystyle \frac{Q}{2\pi {\boldsymbol{u}}{\sigma }_{y}{\sigma }_{z}}\\ \cdot \left[\exp \left(-0.5{\left(\displaystyle \frac{z-{h}_{ef}}{{\sigma }_{z}}\right)}^{2}\right)+\exp \left(-0.5{\left(\displaystyle \frac{z+{h}_{ef}}{{\sigma }_{z}}\right)}^{2}\right)\right]\\ \cdot \exp \left(-0.5{\left(\displaystyle \frac{y}{{\sigma }_{y}}\right)}^{2}\right)\end{array}\end{eqnarray*}$$

Where Q is the emissions from a specific berth no., u is wind velocity, σ_y and σ_z are horizontal and vertical dispersion parameters, z is the height of the measurement site, and h_ef is effective emission height. The equation takes reflection from the bottom and top of the mixing layer into account. For simplicity we use a mean wind velocity of 5 ms⁻¹, assume that the dispersion parameters are one-tenth of the mean velocity multiplied by the distance to the berth no., and only look at the likely concentration contribution at the plume centerline (y = 0). The direction of the LCC arising from up-wind emissions is plotted relative to the measuring site LQ-23 in figure 3 together with the measured and modelled concentration of NO_x at 25 m and wind direction used by the OML-Multi model. LCCs smaller than 50 μg m⁻³ are excluded from the figure. When LCC and wind direction are from the same direction, peak concentrations are to be expected at LQ-23.

Zoom In Zoom Out Reset image size

From figure 3 it is evident that most of the emission from cruise ships at berth causing peak concentrations at LQ-23 is originating from the northeastern sector (0°– 90°) relative to the measurement site corresponding to the Langelinie quay. Wind direction coinciding with LCC from this sector results in the largest concentration peaks increasing the NO_x concentration by up to several 100% for sources very close to the site. All peak values higher than 200 μg m⁻³ correspond to plume effects originating from the northeastern sector. In addition, emission from berth no. 177 south of the measurement site have an impact on the NO_x concentrations, but only in the first 16 days of June instances are occurring with coinciding southern winds and emissions resulting in small peaks. Moreover, the passenger ferry sailing between Copenhagen and Oslo, Norway, is positioned north of the measurement site and stays in the port during daytime hours resulting in periodic emissions, however with limited impact on the measured and simulated NO_x concentrations.

As stated, peak concentrations should occur when the wind direction and the LCC direction are similar, but this is more often the case for the simulated concentrations than the measured ones. This is because the used wind directions in OML-Multi are from HCOE, and not from LQ-23, as there are no direct measurements of wind direction here. The actual wind directions at the LQ-23 measurement site were likely different from the wind directions at HCOE 2.5 km away. Plumes close to the observation site are very sensitive to wind direction, and this wind direction discrepancy is a contributing factor to the differences in size and timing of peaks from the measured and simulated NO_x concentrations.

In general, there is a poor correlation (r = 0.23) between the observed and modelled concentrations, although they have similar averages (17.6 μg m⁻³ and 20.5 μg m⁻³), median values (10.5 μg m⁻³ and 11.0 μg m⁻³), and standard deviations (37.3 μg m⁻³ and 42.9 μg m⁻³). Figure 3 shows that peaks only arise when there are considerable emissions from cruise ships in the port, but also indicates problems with capturing the timing of the peaks. If wind data had existed from the LQ-23 as input for the OML-Multi model better compliance between observed and modelled NO_x would likely have been obtained.

According to air quality regulations the hourly NO₂ limit value of 200 μg m⁻³ can be exceeded 18 times during a year, and thus the 19th-highest hourly values of NO₂ at ground level are shown in figure 4(a). The hourly values are well below the limits. In two separate areas of the port, the highest concentrations are found—one is the outer industrial port, and the other is the inner part, where the cruise liners are moored while in the port. The latter is also the area where the city is expanding with residential and office buildings in the port. The maximum 19th-highest hourly concentration of NO₂ at ground level in the port is 133.8 μg m⁻³. The mean annual NO₂ concentrations are in the range of 14 μg m⁻³ to 18.1 μg m⁻³ and have a similar spatial pattern as the 19th-highest hourly concentrations. Hence, the emissions from cruise ships and other commercial ships in the port do not cause an exceedance of the annual EU limit value of 40 μg m⁻³ for NO₂ in 2018 (figure 4(b)).

Zoom In Zoom Out Reset image size

Since most new buildings in the harbour will be multiple storages, the concentrations are examined at a higher level (50 m) close to the average stack height (51.7 m) of the cruise ships visiting Copenhagen port (Jensen et al 2021). At this height, exceedances are obtained for the hourly limit value of NO₂, but only very locally at the quays in the outer industrial port area, at Langelinie and a few curise ship berths outside Langlinie (figure 5). The 19th-highest hourly value of 1048 μg m⁻³ is obtained at 50 m, while the highest annual average of NO₂ is 24 μg m⁻³.

Zoom In Zoom Out Reset image size

In comparison, the urban background concentration in Copenhagen measured at HCOE is 13 μg m⁻³ and 63 μg m⁻³ for the annual average and the 19th-highest hourly concentrations of NO₂ in 2018 (Ellermann et al 2020). Thus, our calculations indicate a substantial augmentation of the urban background level in some areas of Copenhagen Port for both the annual aveage of NO₂ (up to 31% at ground level and 86% at 50 m) and the 19th-highest hourly concentration ( up to 100% at ground level and a factor 10 at 50 m). On the other hand, the measured concentration of NO₂ at the busiest street in Copenhagen in 2018 had an annual average of 39 μg m⁻³, and a 19th-highest hourly value of 117 μg m⁻³ (Ellermann et al 2020). Thus, the maximal annual mean concentration in the port is not uncommonly high when compared to other places in the City. However, the 19th-highest hourly concentration in the port area is unprecedented in Copenhagen.

Although bottom-up emission inventories are better than top-down (Eyring et al 2009) which typically are based on fuel-based approaches, they are not without uncertainties. Cooper and Gustafsson (2004) estimate an uncertainty range of 20% to 50% on emission factors, while Moreno-Gutiérrez et al 2015 find that in particular engine load factors, emission factors and specific fuel consumption factors contribute to a 30% difference in total emissions among different inventory methodologies. Studying the health impact of shipping emissions in the Sydney port area, Broome et al (2016) consider an uncertainty of 30% to be realistic for shipping emission inventories in port areas. Our port emission inventories for Copenhagen are state-of-the-art. They are based on the best available data to determine auxiliary engine power functions and NO_x emission factors, in addition to being spatiotemporal explicit with exact arrival and departing time together with mooring site position. They offer a great improvement in comparison to a previous study of Copenhagen port, where all emissions were gathered in seven points and cruise ship emissions averaged over the year (Saxe and Larsen 2004). However, an uncertainty in the range of 30% is not unlikely. To reduce the uncertainty further, improvements to auxiliary engine power functions and NO_x emission factors would be the way forward.

Greater uncertainties might stem from the use of wind speed and direction from the HCOE monitoring site. This is illustrated at LQ-23, where both the measured and modelled concentrations of NO_x are very sensitive to the plumes emitted very close to the site, which in turn are very sensitive to the wind direction. The difference between the wind direction used in OML-Multi and the actual wind direction (which was not measured at the site) results in peak concentation values of NO_x that are not always well correlated.