Observations and projections of visibility and aerosol optical thickness (1956–2100) in the Netherlands: impacts of time-varying aerosol composition and hygroscopicity

In the Netherlands weather and climate observations are recorded at 42 stations. The majority of these stations date back to the 1970–1980s. However, a number of stations have a longer history. For this study we are particularly interested in long continuous records of visibility. Five stations were chosen, consisting of the five main climate stations (De Kooy, Vlissingen, De Bilt, Eelde and Maastricht). Together they form a good representation of the region (see figure S1 in the supplementary data, available at stacks.iop.org/ERL/10/015003/mmedia for additional information on the specific methods to measure visibility).

Here, four visibility classes are defined. The first class is 'fog', which is defined as a visibility of less than 1 km. This is the traditional definition and no further differentiation into moderate and dense fog is used. The second class is 'low visibility' with visibility restricted between 1 and 5 km. The third class is 'moderate visibility' with visibility between 5 and 10 km. Finally, the fourth class is 'good visibility' and applies to situations with a visibility larger than 10 km.

Visibility (Vis) is defined as

$$\begin{eqnarray}&&{\rm Vis}=-\frac{{{{\rm log} }_{e}}({\rm 0}{\rm .05})}{{{\sigma }_{{\rm ext}}}},\end{eqnarray} \tag{ 1 }$$

where the Rayleigh scatter at the same wavelength is neglected.

The aerosol optical thickness δ_aer at 550 nm is given as

$$\begin{eqnarray}&&{{\delta }_{{\rm aer}}}=H{{\sigma }_{{\rm ext}}}\end{eqnarray} \tag{ 2 }$$

The scale height for aerosols (H) is assumed to be constant over time. Its value is adjusted so as to conform to the remote sensing observations. For every year, the time series of hourly visibility values is converted to visible extinction by solving equation (1) for ${{\sigma }_{{\rm ext}}}$ and averaged over the entire year. Then equation (2) is used to retrieve a time series of yearly averaged optical thickness. The scale height for aerosols has been adjusted until the time series agreed with observations and is taken to be 900 m (see S2).

A long-term time reconstruction is made of aerosol composition valid for the center of the Netherlands (Cabauw, 52.0N, 4.9E, a site about 50 km inland located 22 km west-southwest from De Bilt). The historical part of the reconstruction (1950–2000) is based on the Community Atmospheric Model (CAM)-inventory by Lamarque et al (2010), while the projections for 2000–2100 are based on the aerosol emission scenario following the Representative Concentration Pathway 4.5 (RCP4.5, van Vuuren et al 2011). The combined data set we refer to as the CAM-inventory because we use for the scenario the aerosol composition simulated with the Lamarque et al. CAM chemistry-climate model using the RCP4.5 aerosol emissions. The CAM-inventory has been intended for use in climate model simulations within the framework of the Climate Model Intercomparison Program No. 5 (CMIP5) in support of the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment report (AR5).

The CAM-inventory has been calibrated for present-day conditions using observations of aerosol composition in the Netherlands (Matthijsen and Koelemeijer 2010, Wichink Kruit et al 2012) together with regional model simulations using the national operational Air Quality Lotos-Euros model (L-E model) (Schaap et al 2008). The aerosol composition, representative for the years 2007–2008, is taken from Matthijssen and Koelemeijer (2010). The anthropogenic component of annual country mean PM2.5 is largely composed of secondary inorganic aerosols (SIA, 46%) and total carbonaceous matter (TCM, 29%) with the remaining fraction dominated by the fine fractions of sea salt and dust. Total observed SIA is 6.7 μg m⁻³ and total observed TCM is 4.3 μg m⁻³.

An L-E model simulation for the year 2005 is used to split the observed SIA load into nitrate, ammonium and sulfate components representative for Cabauw. This yields SIA fractions representative for present-day relative amounts of nitrate (∼40%), ammonium (∼25%), and sulfate (∼35%). The TCM fractions for organic aerosol (OA) and black carbon (BC) from the L-E simulation are ∼60% and ∼40%, respectively.These values are close to the observed fractions given by Matthijssen and Koelemeijer (2010), which are representative fractions for the whole country of ∼70% and ∼30%, respectively. The split between the SIA and TCM components was taken from the observations and not from the L-E model because the L-E model is assumed better capable of simulating the individual fractions of SIA and TCM than the division of total particulate matter between SIA and TCM. The L-E model underestimates the present-day OA concentrations, mainly because of the complex formation of secondary OA. It remains difficult to adequately include this process in the simulations (e.g. Schaap et al 2008).

The CAM-inventory provides decadal mean concentrations of sulfate, OA and BC on a 2.5 × 1.9 (lon × lat) degrees model grid. Here the grid cell centered at 52.1N and 5.0E has been used. The CAM-inventory does not cover nitrate and ammonium. These components are added to the calibrated CAM sulfate component maintaining the L-E-derived SIA fractions of sulfate, nitrate and ammonium. Next to sulfate and the natural components sea salt and wind-blown dust, the CAM-inventory provides concentrations of OA and BC. The OA and BC components of CAM are further sub-divided into, respectively, hydrophobic fractions (OA1, BC1) and hydrophilic fractions (OA2, BC2). The 2000–2010 decadal mean hydrophobic and hydrophilic fractions of CAM are then used to subdivide the L-E derived OA and BC fractions. In CAM, for the 2000–2010 decade, the OA hydrophilic fraction is 58% and the BC hydrophilic fraction is 33%.

Table 1 provides the overview of the present-day calibrated CAM-aerosol components together with the observed and L-E-scaled concentrations of nitrate and ammonium as described above. The total anthropogenic aerosol hydrophilic mass (M_ahp) is the sum of SIA, OA2, and BC2. Compared to the CAM aerosol components the sulfate component in 2005 is reduced by the calibration against observations with ∼40% (from 4.0 to 2.2 μg m⁻³) while TCM is increased by ∼50% (from 2.9 to 4.3 μg m⁻³). The total anthropogenic fraction of the hydrophilic mass in CAM (sulfate + OA2 + BC2) is increased by ∼60% (from 5.5 to 8.8 μg m⁻³), mainly through the addition of nitrate and ammonium which are missing components in the CAM-inventory.

The evolution of nitrate and ammonium is not provided by the CAM-inventory. Maximum nitrate and ammonium concentrations of 20% above their observed mass in ∼2005 have been assumed for the 1990s taking into account the decrease in precursor emissions since about 1995. A linear increase from very low concentrations in the 1950s to the peak in 1995 has been assumed. Future projections of nitrate and ammonium are uncertain and dependent on future choices with respect to agricultural practices for ammonium and policy implementation of further reductions in NO_X emissions, mainly related to industry and traffic. A stabilization of the nitrate and ammonium concentrations since 2005 is assumed.

Only part of the highly soluble NaCl interacts with NH₃, NO₃ and SO₄ (Manders et al 2009), therefore an unmixed NaCl group is defined and treated separately (see S3).

The growth of aerosol with RH is dependent upon the hygroscopicity. Using the abbreviation DUS to denote mineral dust, the approach taken here is to make initially a separation between the aerosol group of (BC1, BC2, OA1, OA2, DUS) and the group (SO₄, NH₃, NO₃, Na, Cl).

For the first group OA, and BC that are insoluble the value of the hygroscopicity κ is set at zero. The groups of soluble OA, BC consists of subspecies with unknown relative composition and the actual composition will affect their hygroscopicity (Massoli et al 2010, Chang et al 2010). It appears that κ = 0.4 is a reasonable approximation for both the soluble OA and BC species so as to balance the large proportion of insoluble OA and BC, yielding an average value of κ for the entire OA–BC group ranging from 0.1 to 0.3 (Duplissy et al 2011, McMeeking et al 2011).

Even though DUS is generally thought of insoluble or marginally soluble Herich et al (2009), Hatch et al (2008) suggest that DUS is often coated with soluble compounds due to mixing with anthropogenic aerosols. DUS can therefore interact in water chemistry processes to provide a partly soluble aerosol. Here the value κ = 0.1 is used to indicate its marginal solubility.

For the second group of relatively soluble aerosols (SO₄, NH₃, NO₃, Na, Cl) no hygroscopicity can be assigned to individual cations or anions and they are assumed to interact. The interaction is simulated by means of the water chemistry model ISORROPIA (Nenes et al 1998, see S4). For the given composition a volume-weighted hygroscopicity is found to range from 0.65 to 0.75 which is comparable to the κ-values obtained from the individual aerosol species specified in ISORROPIA (see S4).

Size distributions are reconstructed using two assumptions: (1) the distributions conform to the (lognormal, two modes) shape of the distribution and the aerosol concentrations as observed during the EUCAARI—extended period (2008–2010) at the climate station of Cabauw using an SMPS probe (Asmi et al 2011) and (2) the shapes of the distributions do not change over time, only the total aerosol concentration. Under these two assumptions (plus small correction term, see S5) the mass per species can be calculated from the third moment of the lognormal distribution, or vice versa.

Mie-calculations were performed as a function of RH and of species and stored in tables to facilitate rapid access during the scaling calculations (see S6). For the mixture of dry inorganic aerosols it is assumed that they are purely scattering with an index of refraction of Na₂SO₄: m_d = (1.50–0.0i). For BC the consensus best value of m_d = (1.97–0.79i) is used (Bond et al 2013). For the group of soluble OA it has been suggested that most OAs only have absorptive capability in the UV part of the solar spectrum (Lund Myhre and Nielsen 2004). Brown carbon is viewed as a particular subspecies of OA and is an absorber of solar radiation at visible wavelengths (Andreae and Gelencsér 2006, Alexander et al 2008, Srinivas and Sarin 2013). In the absence of any precise information it is assumed that OA is purely scattering: m_d = (1.70–0.00i)). Also, a small imaginary part was given to mineral dust (DUS) (m_d = 1.50–0.01i). See (S7) for the impact of aerosol properties on visibility.

Consider now the extinction of the aerosol species (σ_j) calculated at a base aerosol concentration (N_j), their respective hygroscopicities (k_j), and their current aerosol concentrations (N_i,j). Here the index i refers to the yearly-mean aerosol concentration for the individual species j derived from the aerosol scenarios. The index j refers to the species. Then it follows that there should be agreement between the modelled and observed values of extinction via equation (1)

$$\begin{eqnarray}&&\mathop \sum \limits_{i,j} \frac{{{N}_{i,j}}}{{{N}_{j}}}{{\sigma }_{i,j}}=-\frac{{\rm log} ({\rm 0}{\rm .05})}{V}={{\sigma }_{{\rm obs}}}\end{eqnarray} \tag{ 3 }$$

Here, σ_obs is the extinction computed from the visibility observations. Unfortunately, equation (3) has too many unknowns to arrive at a unique solution. The scaling choices will be discussed in section 3.

The results for the five stations in the Netherlands are plotted in the four panels of figure 1.

Zoom In Zoom Out Reset image size

For each of the visibility classes, the number of hours occurring in a class was counted and the mean value and standard deviation (grey region) over the five stations were plotted.

Visibility improving with time is a consistent signature in all four plots. After 1970, there is a general decrease in the number of hours with visibility less than 1 km. The fog visibility class is the only one in which a fraction of the aerosol particles will activate to cloud droplets and the data presented here indicates that this class has been reduced by 50% in the last 47 years with a further (smaller) reduction expected until 2100. Yet, the class of low visibility shows an initial increase in the number of hours from 1956 onwards reaching a plateau between 1970 and 1985 after which it decreased again. Additionally, until 2013 there has been a 25% reduction in the number of hours in the class of moderate visibility, and, since the 1980s about 1500 h per year have been added to the class of high visibility. This amounts to almost 15% of the total of 8760 h available in a year of 365 days. In the years to come the class of high visibility is projected to grow at the expense of the two adjacent classes (see S8 for further details). Depending on the visibility class in which one is interested, different conclusions can be drawn about when the upward trends in visibility have started. If one is interested in extreme conditions such as fog, the situation has continually been improving since the early 1970s. Yet, the classes of higher visibility indicate the onset of improved visibility more than a decade later. One (partial) explanation of these differences is the non-linear behaviour of visibility with RH. The reduction in visibility is gradual from RH = 30% up to RH = 85%. However, from RH > 85% visibility rapidly deteriorates. Therefore, a uniform improvement in air quality which underlies most visibility studies will impact the different classes in different ways (see also S7 for a plot of visibility versus RH).

Figure 2 shows the retrieval of De Bilt station. Here a three-point smoother is applied to facilitate comparison with the remote sensing observations. The remote sensing observations consist of ground-based retrievals for the period 1959–1971 at De Bilt (Frantzen 1977), ground-based retrievals from radiation equipment at Cabauw, and retrievals from the standard MODIS AQUA en TERRA data sets (Savchenko et al 2003). Despite some year-to-year variability there is good agreement between the data and the optical thickness retrieval. The fact that only a single scaling height could be used to achieve agreement between the retrieval and the different observations indicates that this is a robust result.

Zoom In Zoom Out Reset image size

The time series from the aerosol emission scenarios from European emission inventories from 1956 to 2100 of the eleven species at Cabauw are shown in figure 3.

Zoom In Zoom Out Reset image size

There is a strong reduction in BC and OA aerosols in the period 1955–1990. In the same period SO₄ peaks around 1980 and then decreases, while the masses of NH₃ and NO₃ both increase. When observing the relative proportions of the various species it is apparent that the water chemistry in the Netherlands is undergoing a transition from a sulfur-based chemistry as was the case in 1960–1980 to a nitrogen-based chemistry from present day onwards until 2100. Overall there is a decrease in aerosol mass throughout the full 145-year period. Half of this decrease takes place in the first 55 years, the other half in the last 90 years.

Based on the aerosol properties treated in Section 2.4 the grouping of aerosols can be further simplified into three. The first group comprises the natural aerosols consisting of NaCl-unmixed and DUS. They remain constant in time and their combined impact on visibility and extinction is constant throughout the entire period. This group is colour-coded as green in the figures to follow. The second group consists of the soluble inorganic mixture which includes the Na released from NaCl upon interaction with NO₃, NH₃ and SO₄. As a fraction of the total mass this group is increasing rapidly, and will by the end of the period be the dominant aerosol group in the Netherlands. This group is colour-coded as blue. The last group consists of the marginally soluble and insoluble OA and BC combined. This is the group that has undergone the largest reduction in aerosol mass and it is colour-coded as brown.

There are two options to scale the modelled extinction to the observed extinction. Option 1 is defined by attributing all differences between the measured and modelled visible extinction to the group of highly soluble aerosols. Option 2 is defined by attributing all differences exclusively to the group of aerosols of low solubility. Most confidence is endowed to the specification of the group of inorganic soluble aerosols (blue coloured), as their time dependence is confirmed by local observations in the Netherlands. This means that option 2 can be considered the most plausible reconstruction option for the Netherlands and will be pursued here.

Figure 4 shows the reconstruction/scenario for the aerosol mass, number concentration and aerosol optical thickness. A three-point smoother was applied to the data in order to facilitate the separation in colour schemes. The red line indicates the unscaled original mass reconstruction/scenario (see S9 for the original scenario-plot in three colours). Assuming option 2, where any difference is owing to the low solubility aerosols, it appears that the total mass of BC and OA for the reconstruction is considerably higher than in the scenario, on the averaged roughly doubled. A factor of around 2 is used as a continuous scaling fraction for BC–OA after 2013 to allow for continuity between the scaling before 2013 and the scenario thereafter. Most reduction in mass, concentration and optical thickness occurs after 1985. The relative contributions of the various aerosol components to the optical thickness are different from the relative contributions of the mass and concentration. Even in the early period of 1956–1990 when BC–OA assumes at 70–80% of the total mass a dominant position in the aerosol grouping, the contribution of the inorganic mixture to the aerosol optical thickness is at 40–60% quite considerable. This means that the changing aerosol composition has a non-negligible impact on the evolution of aerosol optical thickness even though there is a reduction in BC–OA by 50% in the period 1956–2013. The overall hygroscopicity is increased from 0.3 to 0.4 over this period which again is mostly due to the increased importance of the soluble inorganic mixture in the aerosol composition. Interestingly, the values around 2000–2010 are very close to values of 0.4–0.5 as reported by Pringle et al (2010) in their estimates of κ for the region of the Netherlands derived from atmospheric chemistry model simulations.

Zoom In Zoom Out Reset image size

There are considerable differences in aerosol mass, concentration and relative contribution to optical thickness by the different aerosol groups if option 1 (scaling with the group of highly soluble aerosols) were chosen. In that option, there are simply not as many aerosols needed to make up the difference in extinction because hydrophilic aerosols are more efficient in taking up water (thus reducing visibility). For option 1, this amounts to a 25% reduction in aerosol mass/concentration in the early part of the period (1960–1985) when compared to option 2. Furthermore, the contribution of these highly soluble aerosols to the overall optical thickness would have been double that of option 2. The implication is that the choice of scaling option is a sensitive factor in determining the relative contributions of aerosol mass and optical thickness.

Since aerosol hygroscopicity and concentration are the dominant factors contributing to aerosol optical thickness and visibility they are plotted in a single plot of figure 5, both on a log scale.

Zoom In Zoom Out Reset image size

The essential point of this plot is to demonstrate that the change in aerosol composition implies that visibility has improved less than would be expected if it were based on a reduction in aerosol concentration while leaving the composition the same. This plot shows a number of isolines. The slanted straight red lines are isolines of equal optical thickness, the slanted blue lines are those of equal visibility. For this plot we chose the RH of 87% to compute the isolines. The reason for choosing a fixed value of RH is because optical thickness and visibility are both functions of RH, while in figure 5 the only two variables are aerosol concentration and hygroscopicity. Optical thickness (visibility) is increasing (decreasing) towards the upper right, the darker region of the plot. The isolines of the two variables are parallel to each other, as is to be expected based on the fact that one is the inverse of the other. Note that the individual points on this plot do not adhere exactly to the optical thickness values of figure 4 because in this figure, optical thickness and visibility are computed at a fixed RH, while the points in figure 4 are based on data obtained at all values of the RH as present in a year's worth of data. The reconstruction starts at 1956 in the lower right, and ends in 2013 in the middle. To facilitate comparison between pathways a 10-point smoother was applied on the data.

The general slant in the curve from lower right to upper left is an illustration of the importance of the hygroscopicity change in shaping the optical thickness time varying signature: during part of this period the mass of soluble inorganic matter increased while at the same time the relatively insoluble BC/OA aerosols decreased. While for the first group of aerosols visible extinction is increased due to their enhanced capability to take up water, for the second group visible extinction is decreased as this group is reduced in size. The optical thickness resulting from the simultaneous increase and decrease of the different aerosols is thus only slowly decreasing for the period 1956–1985 at the end of which it crosses the 10 km visibility isoline in figure 5. Thereafter, it more rapidly decreases until 2013 in accordance with the rapid reduction in BC/OA during that period. The rapid reduction in BC/OA combined with an increase in the soluble inorganic mixture results in a general increase in hygroscopicity so the general movement of the pathway trace is towards the upper left (i.e. towards the left to reflect a lower aerosol concentration, and towards the top to reflect increasing hygroscopicity). The pathway as indicated in figure 5 highlights its dependence on the relative strength in trends of the different aerosol types involved in shaping the optical thickness. This point is often overlooked in studies on trends in 20th century optical properties.

The pathway 2013–2100 is plotted, as derived from the aerosol mass scenario in figure 4. The plot indicates that future changes in visibility and optical thickness will not be as large as has been observed over the last 50 years (see also figure 1). The pathway for the period 2013–2100 shown by the trace in figure 5 has an identical scaling fraction for the BC–OA group as for the year 2013 to allow for a continuous transition towards the 2013–2100 scenario at the year 2013. However, for comparison purposes we included the 2100 end point of the unscaled pathway as the black triangle in the upper left. This illustrates the amount of uncertainty in simulating the pathway into the future.