j. k. gietl * and o. klemm institute of landscape ecology, university of münster, germany

1
J. K. Gietl * and O. Klemm Institute of Landscape Ecology, University of Münster, Germany •Corresponding author: [email protected], http://kli.uni- muenster.de T06A236P, European Aerosol Conference - August 2008 - Thessaloniki, Greece Weekday Weekend PC HDV Ri PC HDV Ri 6 nm - 100 nm Number Conc. 0.7 4 0.96 - 0.81 0.87 0.57 -0.12 100 nm - 225 nm Number Conc. 0.7 9 0.92 - 0.76 0.54 0.32 0.32 PM 10 Mass Conc. 0.7 9 0.96 - 0.87 - 0.41 - 0.016 - 0.063 Parameter Rank MLP 10-13-1 Temperature 1 2.42 Time of Day 2 2.38 Relative Humidity 3 2.08 Wind Direction 4 1.27 Pressure 5 1.26 Wind Speed 6 1.16 HDV 7 1.15 Precipitation of Last 24 Hours 8 1.11 PC 9 1.09 Day of Week 10 1.05 Acknowledgement This study was financially and logistically supported by the municipality of Münster. Tab.1: Spearman rank correlation coefficients for mean diurnal particle concentrations (1-h values), Richardson number (Ri), and vehicle numbers. To gain knowledge of the variables determining PM 10 a pre-diction model was calculated by artificial neural networks. The predicted concentration by the final multilayer perceptron model correlated well with the observed data (Spearman rank correlation = 0.72, p< 0.05, RMSE for the validation data = 9.0). Input variables which did not have an effect on the prediction quality (sensitivity analysis result ≤ 0) have been excluded: Richardson number and precipitation. The rank of importance for the used input variables is found in Tab 2. Temperature, time of day and relative humidity have the greatest influence on the variability of PM 10, whereas traffic numbers play subordinate parts. Spearman rank correlations between meteorological data and particle concentrations showed different behaviour for mass and number concentrations. Particle number concentrations of diameter < 100 nm correlated best with relative humidity (0.3), greater particles (diameter > 100 nm) with wind speed (-0.4). The PM 10 mass con- centrations showed highest correlations with precipitation (-0.4) and wind speed (-0.4). Fig. 1: Mean variability of PM 10 concentrantion, vehicle numbers and Richardson niumber at weekdays Tab. 2: Rank of importance of the input variables for predicting PM 10 During the measurement period the rise in atmospheric turbulence coincided on working days with the increase of traffic numbers and vice versa. This simultaneous change makes it difficult to state the effect of the atmospheric stability on the PM 10 concentration. On weekends the correlation is weaker due to decreased traffic intensity. 1 Motivation The aim of this study was to qualify the influence of traffic and meteorological parameters on the PM 10 mass concentration in Münster, NW Germany. The long term objectives were to prove whether the exceedances of the European PM 10 limit values can be predicted and whether PM10 can be reduced by street traffic measures. 2 Method Between March 2006 and September 2007, the PM 10 mass concentration was measured with a TEOM / FDMS and the fine particle number concentration (6 to 225 nm) with a SMPS at a 4 to 6-lane road in Münster. Traffic intensity was counted, separated in passenger cars (PC) and heavy-duty vehicles / busses (HDV). The meteorological data were measured at two sites nearby the traffic site, each about one km away. For statistical analysis Spearman rank correlation and artificial neural networks were applied. 3 Results The traffic numbers, especially of HDVs, correlated well with the particle mass and number concentrations on weekdays, whereas on weekends the lower traffic numbers seem to be insufficient to clearly affect the PM 10 concentration (Fig. 1, Tab.1). Instead PM 10 seemed suppressed by other sources and meteorology. 4 Conclusion Both traffic and meteorological parameters influence the particle concentrations but in different manner. The meteo-rorological parameters contribute highly to the variability of PM 10 , whereas the effect of traffic with its diurnal and weekly cycle was of more regular nature and provided a more or less constant PM 10 pattern, of which the absolute magnitude is governed by the meteorology. -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0 5 10 15 20 25 30 35 0:00 4:00 8:00 12:00 16:00 20:00 R ichardson num ber N um berofvehicles PM 10 [µg m -³] H eavy-duty vehicles in 10 Passengercars in 100 PM 10 R ichardson num ber

Upload: zenda

Post on 03-Feb-2016

41 views

Category:

Documents


3 download

DESCRIPTION

Influences of traffic and meteorology on the PM 10 concentration in Münster , Germany. J. K. Gietl * and O. Klemm Institute of Landscape Ecology, University of Münster, Germany Corresponding author: [email protected], http://kli.uni-muenster.de. 1 Motivation - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: J. K. Gietl *  and O. Klemm Institute of Landscape Ecology, University of Münster, Germany

J. K. Gietl* and O. Klemm

Institute of Landscape Ecology, University of Münster, Germany•Corresponding author: [email protected], http://kli.uni-muenster.de

T06A236P, European Aerosol Conference - August 2008 - Thessaloniki, Greece

Weekday Weekend

PC HDV Ri PC HDV Ri

6 nm - 100 nm

Number Conc.0.74 0.96 -0.81 0.87 0.57 -0.12

100 nm - 225 nm

Number Conc.0.79 0.92 -0.76 0.54 0.32 0.32

PM10 Mass Conc. 0.79 0.96 -0.87 -0.41 -0.016 -0.063

Parameter Rank MLP 10-13-1

Temperature 1 2.42

Time of Day 2 2.38

Relative Humidity 3 2.08

Wind Direction 4 1.27

Pressure 5 1.26

Wind Speed 6 1.16

HDV 7 1.15

Precipitation of Last 24 Hours 8 1.11

PC 9 1.09

Day of Week 10 1.05

Acknowledgement This study was financially and logistically supported by the municipality of Münster.

Tab.1: Spearman rank correlation coefficients for mean diurnal particle concentrations (1-h values), Richardson number (Ri), and vehicle numbers.

To gain knowledge of the variables determining PM10 a pre-diction model was calculated by artificial neural networks.The predicted concentration by the final multilayer perceptron model correlated well with the observed data (Spearman rank correlation = 0.72, p< 0.05, RMSE for the validation data = 9.0).Input variables which did not have an effect on the prediction quality (sensitivity analysis result ≤ 0) have been excluded: Richardson number and precipitation. The rank of importance for the used input variables is found in Tab 2. Temperature, time of day and relative humidity have the greatest influence on the variability of PM10, whereas traffic numbers play subordinate parts.

Spearman rank correlations between meteorological data and particle concentrations showed different behaviour for mass and number concentrations. Particle number concentrations of diameter < 100 nm correlated best with relative humidity (0.3), greater particles (diameter > 100 nm) with wind speed (-0.4). The PM10 mass con-centrations showed highest correlations with precipitation (-0.4) and wind speed (-0.4).

Fig. 1: Mean variability of PM10 concentrantion, vehicle numbers and Richardson niumber at weekdays

Tab. 2: Rank of importance of the input variables for predicting PM10

During the measurement period the rise in atmospheric turbulence coincided on working days with the increase of traffic numbers and vice versa. This simultaneous change makes it difficult to state the effect of the atmospheric stability on the PM10 concentration. On weekends the correlation is weaker due to decreased traffic intensity.

1 MotivationThe aim of this study was to qualify the influence of traffic and meteorological parameters on the PM10 mass concentration in Münster, NW Germany. The long term objectives were to prove whether the exceedances of the European PM10 limit values can be predicted and whether PM10 can be reduced by street traffic measures.

2 MethodBetween March 2006 and September 2007, the PM10 mass concentration was measured with a TEOM / FDMS and the fine particle number concentration (6 to 225 nm) with a SMPS at a 4 to 6-lane road in Münster. Traffic intensity was counted, separated in passenger cars (PC) and heavy-duty vehicles / busses (HDV). The meteorological data were measured at two sites nearby the traffic site, each about one km away. For statistical analysis Spearman rank correlation and artificial neural networks were applied.

3 Results The traffic numbers, especially of HDVs, correlated well with the particle mass and number concentrations on weekdays, whereas on weekends the lower traffic numbers seem to be insufficient to clearly affect the PM10 concentration (Fig. 1, Tab.1). Instead PM10 seemed suppressed by other sources and meteorology.

4 Conclusion Both traffic and meteorological parameters influence the particle concentrations but in different manner. The meteo-rorological parameters contribute highly to the variability of PM10, whereas the effect of traffic with its diurnal and weekly cycle was of more regular nature and provided a more or less constant PM10 pattern, of which the absolute magnitude is governed by the meteorology.

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

0

5

10

15

20

25

30

35

0:00 4:00 8:00 12:00 16:00 20:00

Ric

ha

rds

on

nu

mb

er

Nu

mb

er o

f ve

hic

les

P

M1

0 [

µg

m-³

]

Heavy-duty vehicles in 10

Passenger cars in 100

PM10

Richardson number