Possible use of Copernicus MACC-II modeling products in EEAs

assessment work

Leonor Tarrasón, Jan Horálek, Laure Malherbe, Philipp Schneider,

Anthony Ung, Bruce Denby, Linton Corbet and Peter de Smet

18tn EIONET workshop on Air Quality Assessment and ManagementDublin, Ireland, 24th and 25th October 2013

Outline

1. Mapping assessments at EEA2. Use of MACC-II ensemble products 3. Results for PM10 and O34. The role of spatial resolution5. Conclusions and recommendations

Outline

1. Mapping assessments at EEA2. Use of MACC-II ensemble products 3. Results for PM10 and O34. The role of spatial resolution5. Conclusions and recommendations

Approaches to AQ assessments

Three different approaches to air quality status assessments:

1. Mapping based on air quality observations2. Mapping based on air quality modelling 3. Mapping based on a combination of

models and measurements

Increasing accuracy in the description of the extent of exceedances in a certain area

Why mapping? Support to national assessments

PM2.5 Source Contribution Grid Cell (10,12)

0

10

20

30

40

50

60

01-N

ov-0

303

-Nov

-03

05-N

ov-0

307

-Nov

-03

09-N

ov-0

311

-Nov

-03

13-N

ov-0

315

-Nov

-03

17-N

ov-0

319

-Nov

-03

21-N

ov-0

323

-Nov

-03

25-N

ov-0

327

-Nov

-03

29-N

ov-0

301

-Dec

-03

03-D

ec-0

305

-Dec

-03

07-D

ec-0

309

-Dec

-03

11-D

ec-0

313

-Dec

-03

15-D

ec-0

317

-Dec

-03

19-D

ec-0

321

-Dec

-03

23-D

ec-0

325

-Dec

-03

27-D

ec-0

329

-Dec

-03

31-D

ec-0

3

Con

cent

ratio

n ( g

/m3 )

PM2.5 - Cell: {10,12}. All sources and BC included

PM2.5: Background EMEP

PM2.5 - Cell: {10,12}. Source: Domestic wood combustion

PM2.5 - Cell: {10,12}. Source: Road traffic

PM2.5 - Cell: {10,12}. All other sources

PM2.5 source contribution in Oslo

Mapping for regulatory purposes

Different mapping approaches in different countries

The recent ETC/ACM report contains an overview of situation on the use of models for regulatory purposes in 2010 and European composite maps

http://acm.eionet.europa.eu/reports/docs/ETCACM_TP_2013_3_CompAQModMaps_v2.pdf

Current Mapping at ETC/ACM• Unified methodology to

estimate background concentrations in Europe in 10x10 km2

• Useful for population exposure analysis

• Includes an analysis of uncertainties and probability of exceedance of limit values

EEAs current mapping methodology• Airbase background rural and

urban stations• EMEP model calculation in

50x50km2

• Meteorological and site parameters

• Residual kriging for rural and urban maps

• Merging Rural and Urban kriging results

PM10, O3 annual averages PM10 36st highest daily average O3 26th highest daily max 8-h avg SOMO35, AOT40

Combined rural and urban concentration map.

PM10 – 36th maximum daily average values, year 2009 Resolution: 10x10 km.

Probability of exceedance of limit value based on standard interpolation error.

Outline

1. Mapping assessments at EEA2. Use of MACC-II ensemble products 3. Results for PM10 and O34. The role of spatial resolution5. Conclusions and recommendations

Possible improvements of background mapping: example for NO2

1. Airbase station data 2. Ordinary kriging of Airbase station data3. Residual kriging of Airbase station data using OMI satellite obs.

OMI NO2 satellite data from NASA Godard GESC DISC 2012 ETC/ACM Task 1.0.2.8

Possible improvements through use of Copernicus MACC-II products

MACC –II ensemble Operational products with 25x25 km2 and 10x10 km2

http://macc-raq.gmes-atmosphere.eu/som_ensemble.php

MACC-II ensemble performance

Background model requirementsModel

requirementsEMEP MACC-II

ensembleSustainability Long term UNECE Long Term Copernicus

Documentation Scientific reviews Scientific reviews

Availability Every year Y-2 Every Year Y-2 ( Y-1 interim)

Model validation Yearly evaluation reports daily evaluation of forecasts

Yearly evaluation reports, daily evaluation of forecasts

Robustness 10 years experience with Unified model

Ensemble approach with EMEP model included

Accuracy CTM state of art Data assimilation approach implies significantly increased accuracy

Adequate spatial resolution

50x50 km2 25x25 km2 10x10 km2 from 2010

Model requirements

EMEP MACC-II ensemble

Sustainability = =Documentation = =Availability +Model validation = =Robustness +Accuracy +Adequate spatial resolution +

for urban background status assessments

Indicator maps evaluated

Rural and urban background maps For years 2009 and 2010

Outline

1. Mapping assessments at EEA2. Use of MACC-II ensemble products 3. Results for PM10 and O34. The role of spatial resolution5. Conclusions and recommendations

Evaluation of EEA mapping results with use of different background models

Identification of meaningful comparisons from available data, avoiding data assimilation issues

20091. Residual kriging using EMEP model background (50x50 km2)2. Residual kriging using EC4MACS hindcast (50x50 km2) 3. Residual kriging using MACC-II ensemble hindcast (25x25km2)4. Residual kriging using EC4MACS hindcast (7x7 km2)

20105. Residual kriging using EMEP model background (50x50 km2)6. Residual kriging using MACC-II ensemble hindcast (10x10km2)

… and allowing for an evaluation of the effect of scale

ETC/ACM residual krigingEvaluation of CTM models used

EMEP MACCenshindcast

Similar mapping results both for O3 and PM10 indicators, with different background model used

Kriging driven by observations

Larger differences between rural and urban kriged maps independently of the models used

EMEP MACC_ens_hindcast

RURA

LUR

BAN

PM10 - rural areas -2009

Significant improvement in performace inherent to residual kriging as the method is developped to optimize RMSE

PM10- urban background - 2009

Significant improvement in performace inherent to residual kriging as the method is developped to optimize RMSE

model vs DA vs kriged

Largest differences in

areas

with few observations,

kriging driven by observations

Is residual kriging better than DA ? Not really, there are caveats in the comparison

Caveats in the present comparison !!!

MACC-II DA does not use the same stations

Not all stations are in MACC-II data DA but all are included in ETC/ACM kriging

Outline

1. Mapping assessments at EEA2. Use of MACC-II ensemble products 3. Results for PM10 and O34. The role of spatial resolution5. Conclusions and recommendations

SOMO35 –rural 2009

50x50 km2 25x25 km2 7x7 km2

SOMO35 –urban background 2009

50x50 km2 25x25 km2 7x7 km2

PM10 – rural areas -2010

MACC-II ensemble in 2010 with 10x10 km2

Improvement with MACC-II ensembleBoth in rural and urban areas

Effect of increased resolution

PM10 general performance

Finer resolution improves the results MACC generally better results for PM10 in 2010 because of incresed resolution

RMSE bias R2 regr.equation RMSE bias R2 regr.equationEMEP 4.63 0.30 0.531 y = 0.567x + 8.67 5.81 -0.04 0.728 y = 0.727x + 7.89MACC-ENS hindcast 4.55 0.15 0.553 y = 0.620x + 7.50 5.74 -0.09 0.730 y = 0.724x + 7.75EC4MACS hind. 20x20 4.43 0.13 0.568 y = 0.577x + 8.30 5.93 -0.08 0.703 y = 0.733x + 7.56EC4MACS hind. 7x7 4.21 0.14 0.612 y = 0.645x + 7.00 6.04 -0.06 0.692 y = 0.732x + 7.49

RMSE bias R2 regr.equation RMSE bias R2 regr.equationEMEP 7.99 0.51 0.550 y = 0.577x + 14.51 11.38 -0.04 0.715 y = 0.712x + 14.23MACC-ENS hindcast 7.55 0.36 0.599 y = 0.638x + 12.36 11.25 -0.08 0.722 y = 0.716x + 13.67EC4MACS hind. 20x20 7.61 0.54 0.591 y = 0.603x + 13.69 11.29 0.18 0.710 y = 0.731x + 13.26EC4MACS hind. 7x7 7.00 0.55 0.656 y = 0.693x + 10.72 11.69 0.25 0.691 y = 0.714x + 13.98

rural urban background

PM10 annual average

PM10 36th highest value

rural urban background

RMSE bias R2 regr.equation RMSE bias R2 regr.equationEMEP 4.63 0.30 0.615 y = 0.642x + 7.28 5.98 -0.19 0.776 y = 0.773x + 6.38MACC-ENS hindcast 4.49 0.29 0.620 y = 0.667x + 6.88 5.78 -0.16 0.789 y = 0.795x + 5.79

RMSE bias R2 regr.equation RMSE bias R2 regr.equationEMEP 8.53 0.19 0.646 y = 0.643x + 12.83 11.38 -0.33 0.782 y = 0.773x + 11.14MACC-ENS hindcast 8.64 0.53 0.639 y = 0.663x + 12.44 11.20 -0.21 0.787 y = 0.799x + 10.00

rural urban background

PM10 annual average

PM10 36th highest value

rural urban background

Ozone general performance

MACC general better performance, specially in 2010 with finer resolution

RMSE bias R2 regr.equation RMSE bias R2 regr.equationEMEP 8.23 -0.01 0.681 y = 0.709x + 33.43 9.33 0.09 0.642 y = 0.668x + 36.96MACC-ENS hindcast 7.84 0.02 0.710 y = 0.714x + 32.97 9.08 0.12 0.660 y = 0.678x + 35.82EC4MACS hind. 20x20 8.57 0.13 0.656 y = 0.696x + 35.13 9.48 0.06 0.631 y = 0.672x + 36.40EC4MACS hind. 7x7 7.87 0.01 0.708 y = 0.728x + 31.34 9.23 0.06 0.649 y = 0.671x + 36.51

RMSE bias R2 regr.equation RMSE bias R2 regr.equationEMEP 1627 9 0.629 y = 0.642x + 1999 1476 -1 0.613 y = 0.619x + 1701MACC-ENS hindcast 1567 -11 0.655 y = 0.661x + 1876 1456 -3 0.624 y = 0.636x + 1623EC4MACS hind. 20x20 1665 16 0.613 y = 0.648x + 1973 1496 -7 0.603 y = 0.620x + 1690EC4MACS hind. 7x7 1626 2 0.629 y = 0.646x + 1969 1460 -5 0.622 y = 0.628x + 1657

rural urban background

Ozone, 26th highest daily 8-hourly maximum

Ozone, SOMO35

rural urban background

RMSE bias R2 regr.equation RMSE bias R2 regr.equationEMEP 8.94 0.13 0.682 y = 0.722x + 32.28 9.18 0.04 0.709 y = 0.729x + 30.45MACC-ENS hind. 8.91 0.13 0.684 y = 0.727x + 31.79 9.15 0.04 0.711 y = 0.734x + 29.97

RMSE bias R2 regr.equation RMSE bias R2 regr.equationEMEP 1582 11 0.629 y = 0.652x + 1899 1270 4 0.651 y = 0.667x + 1441MACC-ENS hind. 1566 15 0.637 y = 0.663x + 1843 1272 5 0.650 y = 0.666x + 1445

rural urban background

Ozone, 26th highest daily 8-hourly maximum

Ozone, SOMO35

rural urban background

Conclusions Residual kriging results largely driven by observations Therefore there are larger differences in the urban vs rural maps than

between models used and differences between mapping results are larger where we have

fewer stations!

Similar mapping results independently of the models used although in the analysed cases the MACC-II ensemble shows general better performance

MACC-II ensemble appears to be better when used in 10 x 10 km2, specially for urban background mapping

Recommendations

1. ETC/ACM mapping activities will benefit from the regular use of MACC-II ensemble products both for rural and urban background assessments

2. MACC-II ensemble has long-term sustainability, can be available regularly (yearly assessments) and can provide accurate model results with increased spatial resolution. However, the capabilities of MACC-II ensemble data assimilation results have not been assessed in this context.

3. It is recommended to carry out a dedicated study of the capabilities of the DA data assimilated products from MACC-II ensemble in comparison with EEAs urban and rural background assessment mapping routines.

Thank you for your attention!