Investor Presentaiton
Daily mean 03 SAT (DU)
R6 smooth
R5
R6 smooth
Cc1297
EUBREW NET
TOTAL OZONE MEASUREMENTS
AT ITALIAN BREWER STATIONS (ROME AND AOSTA)
Siani A.M.1, Casale G.R.1, Pedone M.¹, Scarlatti F., Diémoz H.2
Sapienza Università di Roma; 2 Arpa Valle d'Aosta
INTRODUCTION The availability of long-term records of total ozone content (TOC) and UV data represents a valuable source of information in studies on the assessment of short and
long-term changes and their impact on terrestrial ecosystem. In addition, ground-based observations provide the ground-truth for satellite-derived products, mainly in polluted and in mountain
areas where large uncertainties in space-borne estimates may be detected. To our knowledge, details about processing software packages applied to Brewer TOC measurements are usually not
specified in studies on satellite vs ground-based comparisons and on the long term TOC variability. This study analyzes the differences between TOC data processed by the Brewer Processing
Software (BPS, by Dr Fioletov and Ogyu of Environment Canada) and by O3Brewer (by Dr Stanek of Solar and Ozone Observatory of CHMI/International Ozone Service). The comparison
of BPS and O3Brewer data with Ozone Monitoring Instrument (OMI).
GROUND-BASED DATA (GB)
Aosta (Brewer #066, 45.7°N, 7.4°E, 570 m asl): 29/01/2007-31/12/2015
Rome (Brewer #067, 41.9°N, 12.5°E, 75 m asl): 01/01/1992-31/12/2015
SATELLITE DATA (SAT)
Aura OMI OMTO3(v8.5, L2 OVP): 01/10/2004-31/12/2015
(source: http://avdc.gsfc.nasa.gov)
STATISTICAL METRICS USED IN THE COMPARISON
A measure of the agreement is given by the following parameters: Spearman
correlation (RHO), Mean Percentage Error (MPE%), Mean Bias (MB) Standard
Deviation from MB, Root Mean Square Error (RMSE) and their normalized (%)
values with respect to the TOC ground-based means.
BREWER DS PROCESSING ALGORITHM
Direct Sun TOC calculation DS= (MS(9)-ETC+(XR6-R6))/(Aα mu)
MS(9)= double ratio; ETC= extra terrestrial constant; XR6 = value of R6 from the last calibration; R6=
R6 ratio for the day; Aa = weighted ozone absorption coefficient, mu
air mass.
=
-
DS measurements were processed with the rejection criteria: measurement Max_DS_std = 2.5 DU
(std=standard deviation), Max_mu = 4.0 in both software packages.
Standard lamp (SL) correction:
BPS adjusts the ETC based on the difference between SL test values at the time of the most recent
calibration and the present values. It takes the average of SL-test values from
15 consecutive days since that calibration.
03Brewer adjusts the ETC using a smoothing filter (Gaussian, 20-days width) on SL ratios creating the
SLsmooth file.
R6
LO
4000
Rome
Rome
400
4000
good
ahomalous
300
calbrations
3000
200
2000
100
1000
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
1/1/1992-31/12/2015
0
-100
8000
-200
6000
good
amalo
calbrations
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
1/1/1992 31/12/2015
600
2000
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
1/1/1992-31/12/2015
500
3000
2500
2000
1500
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
1/1/1992-31/12/2015
5000
4000
good
anomalous
3000
calibrations
2000
1000
2007
8000
400
300
200
03Brewer vs BPS. Rome
R5 R6 good
R5 R6 anomalous
calibrations
03Brewer vs BPS, Rome
80
R5&R6 good
calibrations
good RS&R603brewer
anomalous R5 R603brewer
good R5 R6 BPS
aromalis RS2R6 BPS
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
1/1/1992-31/12/2015
Aosta
400
300
Aosta
40
R5&R6 good
R5&R6 anomalous
calibrations
30
200
Diff (DU)
100
20
0
-100
10
2008
2009
2010
2011
2012
2013
2014
2015
2016
1/1/2007-31/12/2015
-200
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
1/1/2007 31/12/2015
Diff/DLIN
Diff (DU)
good
6000
anomalous
calibrations
4000
2000
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
1/1/2007-31/12/2015
03(DU)
600
500
400
-10
good R5 R603brewer
anomalous R5 R603brewer
good RS&R6 BPS
anomalous RS&R6 BPS
-20
-30
Diff (DU)
-20
-40
-60
-80
60
40
20
0
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016
1/1/1992 31/12/2015
03Brewer vs BPS, Aosta
R5&R6 good
calibrations
1900
1850
180, 2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
1/1/2007-31/12/2015
Fig. 1: Daily R6 and R5 time series and R6 smoothed (from SLsmooth file created by O3Brewer
software). SL correction is not applied by 03brewer when there are no SL measurements and when R6
and R5 values are above 400 units and 700 units, respectively in reference to the values of the last
calibration. These are indicated as anomalous R5 and R6 values (red dots). Days without both R5 and R6
values are not considered.
300
200
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
1/1/2007-31/12/2015
Fig. 2: Differences (1st and 3rd panel) in daily TOC values between 03Brewer
and BPS. The time series of TOC daily means processed with BPS and O3Brewer
(2nd and 4th panel).
-42007
2008 2009 2010 2011 2012 2013 2014 2015 2016
1/1/2007 31/12/2015
Fig. 3: Differences in TOC daily means between
03Brewer and BPS for good R6 and R5 values. TOC
data with daily std>50 DU and without SL measurements,
are excluded. Large differences (~10%) still persist
between the two processed TOCs.
MPE (%) MBE DU (%)
RMSE DU (%)
Rome
550
Daily SAT vs GB (R5&R6 good)
500
450
400
350
300
Daily mean 03 SAT (DU)
550
500
450
400
350
300
250
Rome
+
Daily SAT vs GB (R5&R6 good)
Daily mean 03 SAT (DU)
n
RHO
03Brewer vs BPS
Rome
Rome
550
All data
Daily SAT vs GB (R5&R6 good)
Data with good R5 &R6
6453 0.919
5994 0.994
1.15
-0.18
3.60 (1.17)
-0.56 (-0.17)
21.34 (6.53)
3.64 (1.11)
500
450
Aosta
All data
400
Data with good R5 &R6
2315 0.982
2162 0.999
0.20
-0.04
0.47 (0.14)
-0.16 (-0.05)
11.09 (3.39)
2.02 (0.62)
350
300
250
Table 1: Summary of the statistics 03Brewer vs BPS (n= number of pairwise measurements).
Data with std >50 DU per day and without SL measurements, are excluded. In brackets the
normalized values are in percentages (%).
250
Stand.dev.
OMI_vs_GB (good R5 & R6)
200
200
200
200
250
300
350
400
450
500
550
200
250
300
350
400
450
500
550
n
200
250
300
350
400
450
500
550
TOC (BPS) (DU)
TOC (103Brewer-BPS|<=2.5DU)
TOC (03Brewer) (DU)
Aosta
550
Daily SAT vs GB (R5&R6 good)
500
Aosta
550
Daily SAT vs GB (R5&R6 good)
500
Aosta
550
Daily SAT vs GB (R5&R6 good)
500
03Brewer
BPS
2452 0.964
MPE
RHO (%)
Rome
daily std ≤50 DU
-0.76
MBE
DU (%)
from MBE
RMSE
DU (%)
DU (%)
2449
0.978
-1.14
-2.56 (-0.80)
-3.73 (-1.15)
8.49 (2.64)
7.26 (2.25)
8.87(2.75)
8.16 (2.52)
daily std ≤2.5 DU
03Brewer
BPS
693 0.976
692
0.980
-1.16 -3.62 (-1.17)
-1.52 -4.71 (-1.52)
6.65 (2.15)
5.95 (1.92)
7.57 (2.45)
7.58 (2.45)
daily std 50 DU
450
450
450
Daily mean 03 SAT (DU)
400
350
Daily mean 03 SAT (DU)
400
350
Daily mean 03 SAT (DU)
03Brewe-BPS ≤ 2.5 DU
1830 0.975
400
03Brewe-BPS ≤2.5 DU
540
0.976
350
-0.93 -3.11 (-0.96)
daily std 2.5DU
-1.30 -4.07 (-1.31)
Aosta
7.64 (2.37)
8.25 (2.55)
6.38 (2.06)
7.56 (2.44)
daily std ≤50DU
300
300
300
03Brewer
BPS
1816 0.983
1815
-2.37
-7.76 (-2.36)
0.984
-2.41
-7.89 (-2.40)
7.23 (2.20)
6.98 (2.13)
10.60 (3.23)
10.53 (3.21)
250
250
250
daily std ≤2.5DU
200
200
250
300
350
400
450
500
550
200
200
03Brewer
BPS
200
250
300
350
400
450
500
550
200
250
300
350
400
450
500
550
TOC(103Brewer-BPSI<=2.5 (DU))
TOC (03Brewer) (DU)
TOC (BPS) (DU)
529 0.984
528 0.985
-2.40
-2.39
-7.28 (-2.37)
-7.25 (-2.36)
6.08 (1.98)
5.92 (1.93)
9.48 (3.09)
9.36 (3.05)
daily std ≤50 DU
03Brewer-BPS ≤ 2.5 DU
1718
0.984
Fig. 3: Scatter plots OMI vs ground-based TOC (03Brewer, on left panel; BPS on the central panel, on the right when |03brewer-BPS| ≤2.5) for R6 & R5 good and std <50 DU.
|03Brewer-BPS
≤2.5 DU
527
0.984
-2.44 -7.92 (-2.43)
daily std ≤2.5 DU
-2.39 -7.27 (-2.37)
6.82 (2.09)
10.45 (3.20)
6.09 (1.98)
9.48 (3.09)
Table 2: Summary of the statistics OMI vs ground based TOC (03Brewer and BPS). TOC were selected with std
≤50 DU and std ≤2.5DU per day (good R5 and R6 values). The comparison was also made with |03brewer-BPS|
<2.5 DU. In brackets the normalized values in % .
CONCLUSIONS This study analyzed the difference between total ozone column processed by two different software packages in Rome and Aosta. When anomalous SL values occurr, the smoothing
filter correction in 03Brewer is not applied, producing TOC anomalies. The difference between BPS and O3Brewer can be less than 1% when R5 and R6 differ by about 20% with respect to the values of
the last calibration. However, large differences (up to ~10%) can be still observed, probably due to the larger SL correction applied by BPS with respect to the smoothed correction of O3Brewer. In both
sites, the comparison between OMI and ground-based data (taking into account data with good R5 and R6 values) shows a a systematic underestimation of satellite TOC, although some differences between
the two processing software packges were found (sligthly higher if TOC data are processed with BPS). This result suggests that a standardized processing protocol (COST Action ES1207-EUBREWNET)
is necessary to avoid the inconsistency in SL correction application between different Brewer data processing software packages. Alternately, users could use more than one package as a cross validation of
own data.
CC
BYView entire presentation