Technical report

Intercomparision of the High Volume Sampler PM10 MCVPM1025-CAV with the reference instrument from the Norm EN12341

Institute of Earth Sciences “Jaume Almera” CSIC

Barcelona, October 1999

Index

1. Evaluation criteria for the verification of the Norm EN12341.

2. Location of sampling stations.

3. Equipment and tests.

4. Results.

1. Evaluation criteria for the verification of the Norm EN12341 (1)

To verify the equivalence of the measurements of atmospheric particulate matter smaller than 10 microns (PM10) were done with the high volume sampler MCV-PM1025-CAV using the evaluation criteria of the Norm EN 12341. This Norm describes a procedure for the equivalence checking for the measurement of the PM10 level obtained with candidate samplers respect to a reference sampler using verifications in field. To verify the equivalence between the equipments to check and reference equipment, the Norm set out two verification procedures: a) Comparison of the candidate equipment between them, based on the double and synchronized determinations of PM10 levels with the being checked equipment. b) Comparison between the candidate equipment and the reference one, based on the determination of the function "reference-equivalence" which expresses the relation of the mass concentrations determined by the checked equipment and for the reference one. 1.1 Comparison between two candidate equipment.

To verify the comparison is statistically examined the differences Di between the values of concentration PM10 Yi1 and Yi2. In an ideal case, equipment to verify would fulfil:

Di = Yi1-Yi2 =0 The Norm for the verification procedure vary depending on the concentration of PM10 .

1.1.1. Comparison between candidate equipment for measured concentrations Yi1

<100µg/m3 In this case, the Norm exposes the following steps:

a) Calculate the mean concentration Yi of the parallel measurement i-th. b) Select all the measured concentrations Yi < 100µg/m3 c) Select the number of paired concentrations n<100 for Yi < 100µg/m3 d) Calculate the difference between Di=Yi1-Yi2 e) Calculate the absolute standard deviation Sa according: Sa =Di

2 / 2n<100µg/m3.

f) Select the corresponding Student factor tn<100µg/m3, defined as the quantile 0.975 for the confidence interval at both sides of the t-Student distribution with (n<100-2) degrees of freedom.

g) Calculate the confidence interval at both sides CI95 for the mean values Yi < 100µg/m3, under a: CI95 = Sa * t n< 100µg/m3.

The verified equipment fulfil the comparison requirement for Yi < 100µg/m3 when CI95 < 5µg/m3.

1.1.2. Comparison between candidate equipment for measured concentrations Yi1 >100µg/m3

a) Select all the measured concentrations Yi > 100µg/m3 b) Select the number of paired concentrations n>100 for Yi > 100µg/m3 c) Calculate the difference between Di=Yi1-Yi2 d) Calculate the relative standard deviation Sr according: Sr = Σ (Di/Yi)2 *

(1/2n >100µg/m3). e) Select the corresponding Student factor tn>100µg/m3, defined as the quantile

0.975 for the confidence interval at both sides of the t-Student distribution with (n>100-2) degrees of freedom.

f) Calculate the confidence interval at both sides CI95 for the mean values Yi > 100µg/m3, under: CI95 = Sa * t n> 100µg/m3, 0,975.

The verified equipment fulfil the comparison requirement for Yi > 100µg/m3 when CI95 ≤ 0,05 (≤ 5%). When the verified instruments don’t fulfil the requirements, they will be removed from the equivalence verification.

1.2. Verification of the comparison between the reference equipment and candidates.

This verification is based on obtaining and evaluating the relationship between PM10 concentrations determined by parallel and synchronous measurements of reference equipments and candidates. Ideally, the candidate equipment (x) determines the same PM10 levels obtained from the reference equipment (y), so y = x. Norm sets out the following steps to perform the verification procedure:

a) calculate the function y = f(x) between the concentrations of the equipment to verify (x) and the reference equipment (y) by linear regression analysis.

b) calculate the confidence bands on both sides (limit Cl) as: y =x±10(µg/m3) for concentrations x<100µg/m3, or y =0.9*x and y =1.1*x (µg/m3) for concentrations x>100µg/m3.

c) preparation of a graphic with nominal function lines y=x, the band of confidence on both sides (limit Cl), the measured data pairs (Xi,Yi1) and (Xi, Yi2), the reference-equivalence function y=f(x) determined in each case by a linear regression calculation with cancellation point.

Norm provides the possibility of a test of strange values, to locate and eliminate possible equipments. First, calculate the reference-equivalent function y=f(x) with all pairs of available values, and determining the typical residual deviation SyA1 as:

Sy = Σ(yi-Yf,i)2 / (n-2) Where: n, number of values yi, i-th (µg/m3) measurement concentration. YF,i, concentration determined by the function y=f(x) (µg/m3) The initial selection of potential strange values is calculated by determining residues (yi - YFi). Is considered to be a strange value every pair of values with a large residual values. Value pairs are ranked in descending order according to the magnitude of its residual values. After removing the first pair of values of the data set is calculated a new function y=f(x) with the residual dispersion SyA2. The check is then performed using a F test.

The equipment to verify fulfil the requirements as equivalent provided if: 1. The coefficient of variance R2 0.95 be above the range of relevant concentration and 2. The reference-equivalence line calculated within the band of confidence on both sides (limit Cl) be above the relevant concentration range.

F Test: The residual dispersions SyA1 and SyA2 of the two lines are examined to check if there are significant differences. Check value

PW = [(n -2)s – (n -2)s ] / sA12

yA1 A22yA2

2yA2

Purchased with the table value F (f1=1, f2=nA2-2; P=95%. When PW>F are facing a strange value, and value pair is removed from the dataset. This test is performed for all pairs of values. With the teams to check duplicate determinations are made and then calculated the reproducibility RD under:

RD= y / (Sa*tn-1;0.95)

Where: Y, class average (µg/m3) Sa, standard deviation of duplicate determinations (µg/m3: Sa = Di

2 / 2n Di difference between levels Yi,1 and Yi 2 of each pair of values. N, number of pairs of measured values. tn-1; 0.95, Student's factor for P=95%.

2. Location of sampling stations The assays were carried out at four sampling stations: Monagrega (Teruel), Onda (Castellón), Barcelona and Madrid (chart 1). They were chosen to cover a wide range of emissions. The station of Monagrega is located in a semi-arid rural zone in the desert of Calanda (SE of the province of Teruel). In this location, there are a lot of suspended particles of natural origin. For this reason, the PM10 registered levels are relatively low. The sample was done during the months of May and June, with a low relative humidity (35 –70 %). On the other hand, the station of Madrid (Escuelas Aguirre) and Barcelona’s metropolitan area (L’Hospitalet) are located in big cities with high levels of anthropogenic PM10 emissions. However, while Barcelona is a seaside city, highly influenced by sea breeze and with high levels of relative humidity, Madrid is an example of a continental weather city. As expected, in both locations we have registered quite high levels of suspended particles. The sample was done during the months of January (Barcelona) and September (Madrid). Finally, the station of Onda (Castellón) has a semi-urban location but influenced by contributions of particulated material from the ceramic industrial estates of Alcora, Vilareal y Onda. The sample was done in the second fortnight of July with very low levels of relative humidity. Levels are intermediate between Monagrega and the urban stations.

Chart 1. localization of sampling stations Name of the station

Province Zone Type Longitude Latitude Altitude

Monagrega TE Rural Agricultural 00º18’20W 40º56’50”N 600 m Onda CS Semi-rural Residential-

Industrial 00º15’W 39º58’N 193 m

Escuelas Aguirre

M Urban Residential-Commercial

03º40’52” W 40º25’32”N 671 m

L’Hospitalet B Urban Residential-Commercial

02º06’57” E 41º22’12”N 30 m

3. Equipment and tests The determination of the levels of PM10 was made for 24-hour mean (start at 8:00 a.m.) with the gravimetric method. For comparative measurements under the Norm EN12341 (1) there were used the following equipment: 3.1. Reference sampler HVS-PM10 Grasbey-Andersen (Figure 1), with high volume

sample pump CAV-A/HF (68m3/h). 3.2. High volume sample MCV CAV-A with head for PM10 particles model

PM1025-CAV (30m3/h, figure 2 and 3). In the PM10 measurements, quartz filters (QF20 Schleincher & Schuell) were used as filtration material. The determination of the mass of airborne dust accumulated on the filter was performed in the laboratory by the weight difference. Before it, the initial weight of filter is determined with analytical scale (Ohaus GA110), the filter was balanced during at least 48 hours at 21 ºC and 55% hr. After the sample, the filter is balanced again at least 48 hours, and with the same scale, it is determined the weight. From the mass and filtered air volume, it is determined the concentration of suspended dust in µg/m3. Apart of the obvious precautions during the starting and ending of the sampling, after making the sampling, the filters has been folded inwards and protected with non-stick paper, and externally with tinfoil, cardboard folder and labelling.

Figure 1. Reference sampler HVS-PM10 Grasbey-Andersen, with high volume sample

pump CAV-A/HF.

Figure 2. High volume sample MCV CAV-A with head for PM10 particles model

PM1025-CAV

Figure 3. PM1025-CAV diagram

Protector Nozzle plate Impactor plate Filter Filterholder

4. Results The results of the intercomparision is presented in the table 1 and 2 and figures 4 and 5. In the figures is represented the resulting correlations of the parallel measurements between: 1) the candidate equipment and 2) between one candidate equipment and the reference one. Also, in the figures is indicated the ideal function y=x, the linear regression y=f(x) with the annulation point according the Norm EN12341 (1), and coefficient of variation R2 from the regression line. The experimental results that we have obtained for both intercomparision with the sampler for PM10 MCV PM1025CAV and the reference equipment, allow us to affirm that both candidate equipment fulfil the requirements of the Norm EN 12341 (1).

Table 1. Obtained values for candidate equipment MCV 1 and MCV2.

Figure 4. Intercomparision between both candidate equipment MCV1 and MCV2. tstudent=0,1655 n=38

Table 2. Registered values with the high volume sampler PM10 MCV and the PM10 Graseby-Andersen

Figure 5. Intercomparision between the equipment high volume PM10-MCV and PM10 Graseby-Andersen for n=38.