Les jeux de données exploités dans le cadre de cet exercice de comparaison ont été édités en fonction des périodes précisées dans l’email de Pierre mais aussi et surtout, en fonction de la disponibilité des données de l’analyseur Magee Aethalometer AE33 (CELINE) qui a été installé à la station de Herstal durant la période du :
Il est important de signaler que durant la période 2, l’analyseur BC AE22 (sn 1082) titulaire a été retiré (raisons : maintenance et calibration ISSeP) de la station pour être provisoirement remplacé par un analyseur de remplacement (sn 1154). Cet échange a été réalisé dans la matinée du 22 décembre 2016. Le mean ratio de l’analyseur titulaire était de 0.66 alors que celui de l’analyseur de remplacement était de 0.84. La variation du mean ratio ayant un impact sur la réponse de l’analyseur, il est préférable de diviser le set de données “période 2”" en deux sous set :
2.1 du 15 octobre 2016 au 21 décembre 2016;
2.2 du 22 décembre 2016 au 03 janvier 2017.
Concernant l’analyseur AE33, les données brutes (fournies par Ph. Maetz) sont des valeurs minutes. Elles ont été ensuite agrégées dans un premier temps sous forme de moyennes semi-horaires puis sous forme de moyennes horaires.
Données BC issues de l’analyseur AE33 (données corrigées)
ae33.1 <- read.csv2("D:\\ISSEP_LSP\\BC\\AE33_DATA_R221\\ae33_r221_vsh_2013_14.CSV", dec = ",")
ae33.1$date <- as.POSIXct(strptime(levels(ae33.1$date)[ae33.1$date],
format = "%d/%m/%Y %H:%M", tz = "GMT"), tz = "GMT")
ae33.1.h <- hourlyMean(x = ae33.1, col.date = "date", rounding = F, verbose = F)Données BC issues de la base iséo XR (données non corrigées)
df1 <- subset(hm, as.Date(date) >= as.Date("2013-09-30") & as.Date(date) <= as.Date("2014-03-20"),
select = c("date", "Herstal_BC", "Herstal_NOX" ))Jointure des deux fichiers de données
tmp1 <- merge(df1, ae33.1.h, by = "date", all = T)Données BC issues de l’analyseur AE33 (données corrigées)
ae33.2 <- read.csv2("D:\\ISSEP_LSP\\BC\\AE33_DATA_R221\\ae33_r221_vsh_2016_17.CSV", dec = ",")
ae33.2$date <- as.POSIXct(strptime(levels(ae33.2$date)[ae33.2$date],
format = "%d/%m/%Y %H:%M", tz = "GMT"), tz = "GMT")
ae33.2.h <- hourlyMean(x = ae33.2, col.date = "date", rounding = F, verbose = F)Données BC issues de la base iséo XR (données non corrigées)
df2 <- subset(hm, as.Date(date) >= as.Date("2016-10-15") & as.Date(date) <= as.Date("2017-01-03"),
select = c("date", "Herstal_BC", "Herstal_NOX" ))Jointure des deux fichiers de données
tmp2 <- merge(df2, ae33.2.h, by = "date", all = T)Subdivision du set de données associé à la période 2
tmp2.titulaire <- tmp2[as.Date(tmp2$date) >= as.Date("2016-10-15") &
as.Date(tmp2$date) <= as.Date("2016-12-21"), ]
tmp2.rempl <- tmp2[as.Date(tmp2$date) >= as.Date("2016-12-22") &
as.Date(tmp2$date) <= as.Date("2017-01-03"), ]Ci-dessous sont présentés le summary de chaque modèle linéaire
mod.her <- lm(formula = Herstal_NOX ~ Herstal_BC , data = tmp1)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ Herstal_BC, data = tmp1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -137.550 -6.558 -1.287 4.697 191.374
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.8830 0.4062 2.174 0.0298 *
## Herstal_BC 24.6309 0.1397 176.302 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 19.09 on 4020 degrees of freedom
## (106 observations deleted due to missingness)
## Multiple R-squared: 0.8855, Adjusted R-squared: 0.8854
## F-statistic: 3.108e+04 on 1 and 4020 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ Herstal_BC - 1 , data = tmp1)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ Herstal_BC - 1, data = tmp1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -140.165 -5.922 -0.643 5.183 190.182
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## Herstal_BC 24.8348 0.1036 239.7 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 19.1 on 4021 degrees of freedom
## (106 observations deleted due to missingness)
## Multiple R-squared: 0.9346, Adjusted R-squared: 0.9346
## F-statistic: 5.747e+04 on 1 and 4021 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ bc.ae33 , data = tmp1)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ bc.ae33, data = tmp1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -141.380 -7.605 -1.435 5.218 187.033
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.5959 0.4438 3.596 0.000327 ***
## bc.ae33 28.6382 0.1796 159.422 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 20.84 on 4009 degrees of freedom
## (117 observations deleted due to missingness)
## Multiple R-squared: 0.8638, Adjusted R-squared: 0.8637
## F-statistic: 2.542e+04 on 1 and 4009 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ bc.ae33 - 1 , data = tmp1)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ bc.ae33 - 1, data = tmp1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -146.227 -6.572 -0.206 6.219 184.794
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## bc.ae33 29.0715 0.1334 217.9 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 20.87 on 4010 degrees of freedom
## (117 observations deleted due to missingness)
## Multiple R-squared: 0.9221, Adjusted R-squared: 0.9221
## F-statistic: 4.748e+04 on 1 and 4010 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ Herstal_BC , data = tmp2)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ Herstal_BC, data = tmp2)
##
## Residuals:
## Min 1Q Median 3Q Max
## -146.739 -10.673 -1.264 8.094 272.954
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.7285 0.9739 0.748 0.455
## Herstal_BC 45.4475 0.4434 102.507 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 28.33 on 1669 degrees of freedom
## (273 observations deleted due to missingness)
## Multiple R-squared: 0.8629, Adjusted R-squared: 0.8629
## F-statistic: 1.051e+04 on 1 and 1669 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ Herstal_BC - 1 , data = tmp2)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ Herstal_BC - 1, data = tmp2)
##
## Residuals:
## Min 1Q Median 3Q Max
## -148.697 -10.162 -0.717 8.586 272.138
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## Herstal_BC 45.6806 0.3154 144.8 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 28.32 on 1670 degrees of freedom
## (273 observations deleted due to missingness)
## Multiple R-squared: 0.9262, Adjusted R-squared: 0.9262
## F-statistic: 2.097e+04 on 1 and 1670 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ bc.ae33 , data = tmp2)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ bc.ae33, data = tmp2)
##
## Residuals:
## Min 1Q Median 3Q Max
## -239.207 -11.097 -1.886 8.479 283.317
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.9994 0.9096 3.298 0.000994 ***
## bc.ae33 28.9787 0.2682 108.057 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 28.3 on 1866 degrees of freedom
## (76 observations deleted due to missingness)
## Multiple R-squared: 0.8622, Adjusted R-squared: 0.8621
## F-statistic: 1.168e+04 on 1 and 1866 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ bc.ae33 - 1 , data = tmp2)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ bc.ae33 - 1, data = tmp2)
##
## Residuals:
## Min 1Q Median 3Q Max
## -249.216 -9.265 0.330 10.204 280.203
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## bc.ae33 29.5924 0.1936 152.9 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 28.38 on 1867 degrees of freedom
## (76 observations deleted due to missingness)
## Multiple R-squared: 0.926, Adjusted R-squared: 0.926
## F-statistic: 2.336e+04 on 1 and 1867 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ Herstal_BC , data = tmp2.titulaire)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ Herstal_BC, data = tmp2.titulaire)
##
## Residuals:
## Min 1Q Median 3Q Max
## -149.793 -10.811 -0.536 8.328 258.159
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -1.3607 1.1092 -1.227 0.22
## Herstal_BC 47.9942 0.5302 90.528 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 28.49 on 1366 degrees of freedom
## (264 observations deleted due to missingness)
## Multiple R-squared: 0.8571, Adjusted R-squared: 0.857
## F-statistic: 8195 on 1 and 1366 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ Herstal_BC - 1 , data = tmp2.titulaire)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ Herstal_BC - 1, data = tmp2.titulaire)
##
## Residuals:
## Min 1Q Median 3Q Max
## -147.457 -11.599 -1.427 7.529 259.902
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## Herstal_BC 47.5262 0.3682 129.1 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 28.49 on 1367 degrees of freedom
## (264 observations deleted due to missingness)
## Multiple R-squared: 0.9242, Adjusted R-squared: 0.9241
## F-statistic: 1.666e+04 on 1 and 1367 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ bc.ae33 , data = tmp2.titulaire)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ bc.ae33, data = tmp2.titulaire)
##
## Residuals:
## Min 1Q Median 3Q Max
## -143.099 -10.965 -1.288 8.406 270.453
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.8762 1.0156 0.863 0.388
## bc.ae33 30.4831 0.3140 97.068 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 28.37 on 1577 degrees of freedom
## (53 observations deleted due to missingness)
## Multiple R-squared: 0.8566, Adjusted R-squared: 0.8565
## F-statistic: 9422 on 1 and 1577 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ bc.ae33 - 1 , data = tmp2.titulaire)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ bc.ae33 - 1, data = tmp2.titulaire)
##
## Residuals:
## Min 1Q Median 3Q Max
## -144.831 -10.475 -0.516 8.936 269.410
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## bc.ae33 30.6758 0.2208 139 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 28.37 on 1578 degrees of freedom
## (53 observations deleted due to missingness)
## Multiple R-squared: 0.9245, Adjusted R-squared: 0.9244
## F-statistic: 1.931e+04 on 1 and 1578 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ Herstal_BC , data = tmp2.rempl)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ Herstal_BC, data = tmp2.rempl)
##
## Residuals:
## Min 1Q Median 3Q Max
## -82.067 -7.686 -2.262 6.897 131.022
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.7738 1.6947 1.637 0.103
## Herstal_BC 39.6611 0.6475 61.248 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 22.28 on 301 degrees of freedom
## (9 observations deleted due to missingness)
## Multiple R-squared: 0.9257, Adjusted R-squared: 0.9255
## F-statistic: 3751 on 1 and 301 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ Herstal_BC - 1 , data = tmp2.rempl)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ Herstal_BC - 1, data = tmp2.rempl)
##
## Residuals:
## Min 1Q Median 3Q Max
## -87.300 -6.588 -0.127 8.294 128.066
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## Herstal_BC 40.3556 0.4905 82.27 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 22.35 on 302 degrees of freedom
## (9 observations deleted due to missingness)
## Multiple R-squared: 0.9573, Adjusted R-squared: 0.9571
## F-statistic: 6768 on 1 and 302 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ bc.ae33 , data = tmp2.rempl)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ bc.ae33, data = tmp2.rempl)
##
## Residuals:
## Min 1Q Median 3Q Max
## -159.241 -10.929 -2.836 8.044 90.064
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6.0655 1.7086 3.55 0.00045 ***
## bc.ae33 25.0612 0.4119 60.84 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 22.36 on 287 degrees of freedom
## (23 observations deleted due to missingness)
## Multiple R-squared: 0.928, Adjusted R-squared: 0.9278
## F-statistic: 3702 on 1 and 287 DF, p-value: < 2.2e-16mod.her <- lm(formula = Herstal_NOX ~ bc.ae33 - 1 , data = tmp2.rempl)
summary(mod.her)##
## Call:
## lm(formula = Herstal_NOX ~ bc.ae33 - 1, data = tmp2.rempl)
##
## Residuals:
## Min 1Q Median 3Q Max
## -172.955 -6.431 1.861 11.810 90.474
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## bc.ae33 25.9944 0.3234 80.37 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 22.81 on 288 degrees of freedom
## (23 observations deleted due to missingness)
## Multiple R-squared: 0.9573, Adjusted R-squared: 0.9572
## F-statistic: 6459 on 1 and 288 DF, p-value: < 2.2e-16En comparant les pentes obtenues avec les deux méthodes de mesures, on peut constater que ce coefficient est stable d’une période à l’autre avec l’analyseur AE33 :
\[NOx = 29 \times BC\]
\[NOx = 30 \times BC\]
Pour l’analyseur AE22, la pente a très nettement augmenté :
\[NOx = 26 \times BC\]
\[NOx = 46 \times BC\]
L’impact du Mean ratio ne se traduit pas vraiment au niveau de l’équation de régression linéaire pour l’analyseur AE22. D’ailleurs, on peut observer un comportement assez semblable pour l’analyseur AE33..
Pour l’analyseur AE33 :
\[NOx = 31 \times BC\]
\[NOx = 26 \times BC\]
Pour l’analyseur AE22 :
\[NOx = 48 \times BC\]
\[NOx = 40 \times BC\]