Evaluation of Proposed Winter PM Concentration Reduction Strategies Using the MM 5 and CAMx 4 Modelling System-Christchurch , New Zealand , 2005 – 2013

Mesoscale Model (MM5) and Eulerian Comprehensive Air quality Model (CAMx4) were used to evaluate dispersion of particulate matter (PM) generated by “Total” emissions for Christchurch (New Zealand) for winter 2005. “Total” emissions consist of the “Domestic”, “Transport” and “Industry” emissions. A composite chemical scenario generated from transport-related (day-time) and domesticrelated (night-time) chemical scenarios was shown to be an optimal chemical split of input gridded emissions for predicting PM concentrations with minimal error when compared with ambient data. Reduction of gridded emissions of fine (PM2.5) and total (PM10) aerosol from domestic and transport sources can be achieved by linear reduction of the PM emissions in the emissions groups, as well as by non-linear reduction in the groups by varying the percentage of each chemical component of the scenario used to chemically split the PM input gridded emissions. Results of comparison of the linear and non-linear reduction for winter 2005 heavy smog episodes support the reliability of the 1999 inventory. The predicted linear and non-linear reduced PM values belong to the same population with correlation coefficients of 0.88 to 0.98. Based on these results, a sequence of experiments has been conducted to evaluate the potential decrease of PM winter concentrations over the 2005–2013 time period, using proposed reduction of PM winter emissions (in both the “Domestic” and “Transport” groups) using the linear reduction scheme. Two different abatement strategies outlined by the New Zealand Ministry for the Environment (MfE) to reduce aerosol concentrations and to achieve compliance with the PM reduction plan (target year 2013) were studied numerically using proposed aerosol emissions reductions.


INTRODUCTION
The city of Christchurch is situated on the coastal edge of the Canterbury Plains in the South Island of New Zealand.It has significant wintertime air pollution that is dominated by smoke generated by domestic fires (Scott, 2005).The emissions consist mostly of fine particulate matter (PM 2.5 ) that represents about 75-85% of the total PM 10 (Aberkane et al., 2004).The "Total" group emissions consist of the "Domestic", "Transport" and "Industry" aerosol emissions with a dominant contribution from the domestic sector in the development of winter-time aerosol peak concentrations.A composite chemical scenario generated from transport-related (day-time) and domesticrelated (night-time) scenarios (Titov et al., 2007c) is considered to be an optimal one for accurately predicting concentrations of ambient PM.
The key aims of this research are: To check the reliability of the 1999 emissions inventory using 50% linear and nonlinear reduction of PM emissions in the principal emission groups, and to verify that predicted ambient aerosol concentrations can be considered to come from the same general population (based on winter 2005 smog episodes); To undertake a sequence of experiments to predict future winter PM 2.5 and PM 10 concentrations (over the 2005-2013 time period) using proposed reduction of winter PM emissions (Ministry for the Environment, 2003;Scott and Gunatilaka, 2004) in the "Domestic" and "Transport" groups applying the linear reduction method to input emissions calculated initially from the 1999 inventory.

THE CHRISTCHURCH AREA WINTERTIME METEOROLOGY
Christchurch has significant wintertime aerosol air pollution that is dominated by smoke generated by domestic fires burning fossil fuel during cold nights (Spronken-Smith et al., 2005) and by secondary aerosol predominantly from vehicular exhaust systems (Senaratne et al., 2005).PM emissions consist mostly of fine particulate matter (PM 2.5 ) that contributes about 75-85% of the total PM 10 during wintertime pollution episodes (Aberkane et al., 2004;Scott 2005).From winter 2005 observations PM 2.5 provided 78.9% of PM 10 for a 45-day research period (1 May-15 June 2005) and 82.1% for May-July 2005 (PM 2.5 was not measured in August).A principal aim of this study is to investigate a proposed decrease of PM 2.5 (PM 10 ) ambient concentrations over the 2005-2013 time period using the MM5-CAMx4 numerical modelling system and an optimal chemical scenario used to split PM input gridded emissions for input to CAMx4 (Titov et al., 2007c).The Christchurch gridded emissions data for PM 2.5 and PM 10 have a resolution of 1 km and were prepared using a Geographic Information System for the "Domestic", "Transport" and "Industry" emission groups.To account for the diurnal variation in PM 2.5 -PM 10 emissions, the Christchurch "mean winter day" emissions from the 1999 inventory were split into four time intervals (6 am-10 am, 10 am-4 pm, 4 pm-10 pm, 10 pm-6 am) and the emission values in any grid point were constant within each time interval (Scott and Gunatilaka, 2004).Table 1 shows a temporal scenario obtained from the 2002 inventory and provides the distribution of percentage hourly emissions over a winter day    2), with total PM 10 emissions for winter 2005 expected to be about 54% of the total emissions for winter 1999.The ratio 0.50 was therefore accepted for the "Total" group emissions (Table 2) and was used for linear reduction in the "Domestic" and "Transport" emission groups.It is important to stress that it is impossible to properly separate "Domestic" and "Transport" emissions (within "Total" (50%) = PSO (as PSO

Influence of emissions groups on a proposed future PM scenario
The effects of a proposed linear decrease of aerosol emissions in the "Domestic" and "Transport' groups and some increase in the "Industry" group (see emissions had a tendency to be faster than PM 10 after year 2007 (Table 4).This is associated with a slower reduction (or no reduction at all) of the coarse PM 10-2.5 emission fraction (crustal elements and soil-dust particles), which has mostly a natural source.2 and Table 5), it is important to analyse PM 10 and PM 2.5 concentration ratios in the "Total" group for the  6).This analysis will provide a better understanding of the basic rates of reduction of modelled PM concentrations obtained from the MM5-CAMx4 output (Titov et al., 2007a).The numerically obtained concentrations were converted to ratios (%), reducing effects of PM time trends, and were compared with the proposed decrease of emissions (Table 5).
Table 6 presents modelled versus observed average aerosol concentration, mean absolute error (at the Coles Place observation site), correlation coefficient and the ratio of modelled to observed PM 10 and PM 2.5 in the "Total" group for 1999 and 2007 winter emissions; Part 3: modelled PM 10 and PM 2.5 in the "Total" group for 1999 and 2010 winter emissions; Part 1: modelled PM 10 and PM 2.5 in the "Total" group for 1999 and 2005 winter emissions; Part 4: modelled PM 10 and PM 2.5 in the "Total" group for 1999 and 2013 winter emissions; From Part 1 of Table 6 it is apparent that there was no serious decrease of total (80%) and fine (77.5%) modelled average aerosol concentrations in spite of the proposed (and modelled) 50% reduction in "Domestic" emissions and, as a result of residential sector dominance, a nearly 44% reduction in the "Total" group (Table 5) for the year 2005.Part 2 of the Table 6 shows that there was a significant decrease for total (48%) and fine (51%) modelled average aerosol concentrations in response to the proposed (and modelled) 70% reduction in "Domestic" emissions and, as a result of residential sector dominance, nearly a 60% reduction in the "Total" group (Table 5) for the year 2007.Results from Table 6 suggest that there should be a 3-fold decrease of residential aerosol emissions and a nearly 50% reduction of vehicular emissions to cause a 50% reduction of average concentrations of PM 10 and PM 2.5 .This was true for modelled average aerosol concentrations only (long-term PM exposure), while for short-term exposure (night-morning peak PM hourly concentrations) a 50% decrease in the group "Transport" was sometimes enough to reduce the PM maximum 2-fold.
Part 3 of Table 6 demonstrated a more noteworthy decrease for total (64%) and fine (65.5%) modelled average (over 7 episodes) aerosol concentrations in response to a proposed (and modelled) 82% reduction in "Domestic" emissions and, as a result of residential sector dominance, a more than 70% reduction in the group "Total" (Table 5) for the year 2010.Part 4 of Table 6 showed a third considerable decrease for total (78%) and fine (79.5%) modelled PM average concentrations (for the year 2013) as a result of the proposed (and modelled) 88% reduction in "Domestic" emissions and, as a result of residential sector dominance, a 75% reduction in the "Total" group (Table 5).This was the first time that the proposed reduction of PM emissions (Table 2) was less than the level of the modelled PM average concentration decrease.From the 4 parts of Table 6 it is clear that a 3-fold decrease of the initial (year 1999) residential aerosol emissions and nearly 50% reduction of vehicular emissions was required to halve the average PM concentrations (Part 2, Table 6).Received for review, February 24, 2008Accepted, April 30, 2008 During winter time under anticyclonic stagnant synoptic conditions, strong nearsurface temperature inversions increase the air pollution potential, resulting in high PM concentrations especially close to midnight when there is minimal near-surface ventilation (McKendry et al., 2004).The winter season is characterized by frequent occurrence of severe nocturnal aerosol smog episodes and the WHO health guideline (50 µg/m 3 daily) is exceeded 30-40 nights each winter (Aberkane et al., 2004).Topographically induced local air circulation over the Christchurch region plays an essential role in the accumulation and dispersion of PM (Kossmann and Sturman, 2004).The development of local winds includes a thermally generated day-night (sealand) breeze and nocturnal cold air drainage wind from the Canterbury Plains and Port Hills (McKendry et al., 1986).Nocturnal drainage winds enhance the strength of the near-surface night-time temperature inversion (Jonhstone 2000; McKendry et al., 2004).The dataset, including several in-town observational sites, obtained during the Christchurch Air Pollution Study 2000 (CAPS2000 -Kossmann and Sturman, 2004) was used to validate the MM5-CAMx4 numerical modelling system (Titov et al., 2007b)..7.3 is a fifth generation nonhydrostatic limited-area mesoscale model using a terrain-following sigma-coordinate system and up to 5 nested grids to simulate mesoscale atmospheric circulations.It was developed by the National Center for Atmospheric Research (NCAR) and University Corporation for Atmospheric Research (UCAR, Dudhia et al., 2000).The Comprehensive Air quality Model (CAMx4.4) is a Eulerian photochemical dispersion model that is designed to unify all of the technical features required of 'state-of-the-science' air quality models into a single efficient system.CAMx4.4 simulates the emission, chemical reactions and removal of pollutants (ENVIRON, 2005).The combined MM5-CAMx4 numerical system provides advantages in predicting pollution dispersion over complex terrain, since the mesoscale model calculates surface layer meteorological fields, which are used as input-nudging ones to drive CAMx4.4 to estimate the dispersion of aerosol air pollution.Maps of the location of the city of Christchurch and the MM5 four grids with spatial resolutions of 27, 9, 3 and 1 km are presented in Fig. 1.The coarsest resolution grid covers nearly all of New Zealand, while the 4 th grid covers just the Christchurch area.Topography and land-use distribution were obtained from the USGS global database with a spatial resolution of 30 seconds.NCEP/NCAR re-analysis global data were used for initialization and nudging of MM5 during each run.
Fig. 1.(a) Location of the Christchurch region where the black dot represents the location of the Coles Place monitoring site; (b) the MM5 grid setup used in this study.
A flow chart of the logical structure of the model experiment is presented in Fig. 2. A 50% reduction in emissions of fine (PM 2.5 ) and total (PM 10 ) aerosol in the "Total" group (proposed by ECan for winter 2005) from domestic and vehicular sources can be achieved in two ways: 1.A 50% linear reduction in emissions of fine (PM 2.5 ) and total (PM 10 ) aerosol from domestic and transport sources can be achieved by dividing (by a factor of two) the domestic and transport emissions from the Christchurch 1999 inventory (Scott and Gunatilaka, 2004).A linear reduction in the groups "Domestic' and "Transport" results in a linear reduction in the "Total" group.The procedure involves arithmetic manipulation of 2-dimensional arrays of gridded PM hourly emissions for all 24 hours of a winter day; 2. Non-linear reduction of the proposed "Domestic" emissions only (no change to transport emissions).For the night-time part (Titov et al., 2006) of the optimal chemical scenario (7 pm to 6 am), the chemical split of particulate species was: EC (50%) = EC*0.75given that domestic and transport emissions of EC are approximately the same magnitude (Scott, 2005); POC (50%) = POC/2; SOA (50%) = SOA/2; PSO is associated with industry only and industrial emissions were nearly the same for the years 1999 and 2005 -Table 2); PNO 3 where EC is elemental carbon, POC is primary organic carbon, SOA is secondary organic aerosol, where P = particulate sulphate, nitrate or ammonia).Primary and crustal elements (mostly natural origin, Scott, 2005) remain unchanged as there is no serious land erosion, car fleet increase or any mining works in the Christchurch area.The particulate species listed above are the only particulate species (except particulate water) defined in the CAMx4 chemical mechanism (ENVIRON, 2005).Nonlinear reduction of the proposed "Transport" emissions (included in "Total" emissions) for the day-time (7 am to 6 pm) part of the optimal chemical scenario (Titov et al., 2006) has the following chemical split: EC (50%) = EC/2; OC (50%) = OC/2; SOA (50%) = SOA/2; PSO 4 (50%) = PSO (industry only)with primary and crustal elements unchanged.Reduction of the input emissions by 50% was based on a proposed Environment Canterbury (the local environment agency) strategy of decreased emissions over the years 2005-2013 (Table

Fig. 2 .
Fig. 2. Flow chart of the experimental numerical system.Input fields are indicated by filled arrows and output fields by empty arrows.
groups (EC, OC, PNO 3 ).The dominance of "Domestic" emissions during night-time and "Transport" emissions during daytime(Titov et   al., 2007c)  allows application of a non-linear 50% reduction to "Total" emissions during the half of the day when domestic activity dominates particulate production (7 pm to 6 am), and the other half when vehicular activity dominates (7 am to 6 pm).The "Industry" group emissions were unchanged, as industry emissions hardly changed between 1999 and 2005 (Table2).Inter-comparison of the linear and non-linear approaches will allow evaluation of the quality of the 1999 emissions inventory for all 3 emission groups if it is shown that modelled total PM 2.5 and PM 10 concentrations for linear and non-linear reduction methods belong to one general population.The null hypothesis (that the two samples of PM concentrations belong to the same population) is applied to the 50% non-linear and linear emissions reductions in the "Transport", "Domestic" and "Domestic + Transport" groups for 7 smog episodes in winter 2005.Table3presents the ratio between reduced and initial (1999) concentrations, Pearson correlation and Spearman rank test (level of reliability) for linear and non-linear reduced fine and total PM concentrations in the groups "Domestic", "Transport" and "Domestic + Transport".Pearson correlation and Spearman rank test statistics were applied to modelled PM (Table3) to indicate whether the two different approaches to emissions reduction produce aerosol concentrations from the same population.From Table3it is evident that the linear and non-linear reduced ambient concentrations associated with the "Domestic" and "Transport" emissions reduction belong to the same population with a significance level 0.01 < P < 0.05.The results supported the quality of the 1999 emission inventory, and provided confidence in the use of a linear reduction of emissions to investigate the possible impact of proposed reductions of PM 2.5 and PM 10 emissions over the years 2005-2013 (Table2).The meteorology from seven smog episodes in winter 2005 was used to provide the basis for this investigation.

Fig. 3
Fig. 3 represents the spatial distribution of modelled PM 10 concentrations (the optimal chemical scenario was applied to chemically split the gridded input emissions) for peak night-time PM concentration (11 pm on 4 10

The
hours, which is the typical synoptic time scale of a winter pollution episode for the Christchurch area.Fig.4presents the spatial distribution of predicted PM 10 concentrations based on a proposed reduction of 70% (see Table5) of the "Total" emissions for winter 2010 (Fig.4a) and winter 1999 (Fig.4b) based on "Total" emissions from the 1999 inventory.The proposed (by Environment Canterbury) 70% reduction of PM emissions is considered to be achieved in the "Domestic" and "Transport" groups only.This proposed PM reduction includes primary and secondary aerosol components.The example in Fig.4is typical of the midnight heavy smog episodes that are based on winter 2005 meteorology for the city of Christchurch.The maximum modelled hourly PM 10 concentration for the proposed winter 2010 70% reduction of total emissions was about 70 µg/m 3 (Fig.4a) compared with more than 440 µg/m 3 (Fig.4b) obtained from the initial (winter 1999) total emissions.The modelled hourly PM 10 concentration (peak time) is therefore predicted to decrease by more than 6 times, reducing the short-term exposure of local inhabitants.
Fig. 5. From the document extract it is apparent that there is no serious reason to apply the CLiP strategy and the rest of the document is dedicated to development of the SLiP concept only.However, from the previously described MM5-CAMx4.4modelled output it was absolutely obvious that the SLiP abatement strategy was an unrealistic one for aerosol pollution reduction.The authors of the key strategy document just relied on linear statistical methods, including the box model through the rest of the document (Fisher et al., 2005).However, 4-D dynamical variation of ambient air pollution reflects a non-linear Ndimensional system (where N tends to infinity) and would never comply with straight-line reduction.From Table 6, it is clear that numerically modelled PM proposed decreases (over the 2005-2013 time period) could follow the CLiP graphic only, with the PM pollution reduction gradient increasing after year 2006.

Table 1 .
Distribution of percentage hourly emissions through the day for "Domestic", "Transport" and "Industry" groups (after Environment Canterbury, Scott, 2005).

Table 3 .
Reduced PM concentrations divided by 1999 inventory values, Pearson correlation and Spearman rank test: linear and non-linear reduced modelled PM in groups "Domestic", "Transport" and "Domestic + Transport" (7 winter 2005 smog episodes).The two numbers show the range of values for the seven events.

Table 4 .
Proposed changes in the "Total" PM 2.5 and PM 10 emissions group over the 2005-2013 time period expressed as proportions of the 1999 values(Scott, 2005).

Table 5 .
Ratio of winter day PM 10 emissions in the "Domestic", "Transport" and "Industry" groups (grams/area-day, Christchurch area = 17,680 ha) for 2005-2013 to the 1999 winter day emissions.