regional production quarterly report on the daily analyses

Regional productionQuarterly report on the daily analyses and forecasts activities, and verification of the LOTOS-EUROS performances

December 2015 – January 2016 – February 2016

Copernicus Atmosphere Monitoring Service

Qr. report on daily analyses and forecasts activities, verification of the LOTOS-EUROS performances |

This document has been produced in the context of the Copernicus Atmosphere Monitoring

Service (CAMS). The activities leading to these results have been contracted by the

European Centre for Medium-Range Weather Forecasts, operator of CAMS on behalf of the

European Union (Delegation Agreement signed on 11/11/2014). All information in this

document is provided "as is" and no guarantee or warranty is given that the information is

fit for any particular purpose. The user thereof uses the information at its sole risk and

liability. For the avoidance of all doubts, the European Commission and the European Centre

for Medium-Range Weather Forecasts has no liability in respect of this document, which is

merely representing the authors view.



Quarterly report on the dailyanalyses and forecasts activities, and verification ofthe LOTOS-EUROS performances

December 2015 – January 2016 – February 2016

KNMI (J. Douros, H. Eskes)

TNO (A. Segers, R. Timmermans)

METEO-FRANCE (M. Pithon, M. Plu, J. Parmentier,J. Arteta, S. Guidotti, N. Assar)

Date: 04/05/2016

REF.:CAMS50_2015SC1_D50.3.2.LOTOS-EUROS-2016Q1_201605

CAMS50_2015SC1_D50.3.4.LOTOS-EUROS-2016Q1_201605CAMS50_2015SC1_D50.5.1.LOTOS-EUROS-2016Q1_201605


Contents:


1. Executive Summary.........................................................................................42. The LOTOS-EUROS model (KNMI-TNO)...............................................................5Product portfolio..................................................................................................5Availability statistics.............................................................................................5

Use for observations for data assimilation...............................................................73. Verification report..........................................................................................10Verification of NRT forecasts................................................................................10Verification of NRT analyses................................................................................15


1. Executive Summary

The Copernicus Atmosphere Monitoring Service (CAMS, www.copernicus-atmosphere.eu) is establishing the core global and regional atmospheric environmental

service delivered as a component of Europe's Copernicus programme. The regionalforecasting service provides daily 4-days forecasts of the main air quality species andanalyses of the day before, from 7 state-of-the-art atmospheric chemistry models andfrom the median ensemble calculated from the 7 model forecasts. The regional service

also provides posteriori reanalyses using the latest validated observation datasetavailable for assimilation.

This report covers the deliverables related to Near Real Time Production (NRT) for

LOTOS-EUROS: D50.3.2-2016Q1, D50.3.4-2016Q1, D50.5.1-2016Q1, for the quarterDecember 2015 – January 2016 – February 2016. Verification is done against in-situsurface observations; they are described in the report D50.1.2-2016Q1, that will bedelivered shortly. The verification of analyses is done against non assimilated

observations.

During this quarter, the performance of LOTOS-EUROS forecasts and analyses wassignificantly affected by a number of bugs associated with the reintroduction of the use

of C-IFS boundary conditions, as well as changes of format in the observational dataprovided for assimilation. The bugs were fixed and the updated model version has beenused in the LOTOS-EUROS production stream since the 1st of March 2016, i.e. just afterthe end of this quarter. It is worth mentioning however, that despite those problems,

LOTOS-EUROS was able to meet the RMSE reference targets for daily maxima andmeans in all occasions.

Qr. report on daily analyses and forecasts activities, verification of the LOTOS-EUROS performances | 4


2. The LOTOS-EUROS model (KNMI-TNO)

Product portfolio

Name Forecast Analysis

Description Forecast at surface, 50m,

250m, 500m, 1000m, 2000m,

3000m, 5000m above ground

Analysis at the surface

Available for users at 4:00 UTC 10:00 UTC for the day before

Species O3, NO2, CO, SO2, PM2.5, PM10,

NO, NH3, NMVOC, PANs,

Birch pollen at surface during

season

O3, NO2, CO, SO2, PM2.5, PM10,

NO

Time span 0-96h, hourly 0-24h for the day before,

hourly

Availability statistics

The statistics below describe the ratio of days for which the LOTOS-EUROS modeloutputs were available on time to be included in the ENSEMBLE fields (analyses and

forecasts) that are computed at METEO-FRANCE. They are based on the followingschedule for the provision at METEO-FRANCE of:

- forecasts data before: 05:30 UTC for D0-D1 (up to 48h), 07:30 UTC for D2-D3 (from49h to 96h)

- analyses data: before 11:00 UTC

These schedules have been set to meet the IT requirements for ENSEMBLE products(no later than 8 UTC for 0-48h, 10 UTC for 49-96h and 12 UTC for analyses).

Indicators:

Availability_model_Forecast

Quarterly basis

D0: 99%

D1: 99%

D2: 99%

D3: 99%

Availability _model_Analysis

Quarterly basis

D: 88%



During this quarter, the following issues have been encountered by the LOTOS-EUROSproduction system:

Date Problem description

(origin, effects)

Impact

on production

From 03/12/2015 to

11/12/2015

Delay in the provision of

analysis results

Analysis results arrived too

late to be taken into account

as a member in ENSEMBLE

15/12/2015 Delay in the provision of

analysis results

Analysis results arrived too

late to be taken into account

as a member in ENSEMBLE

19/12/2015 Analysis results missing Analysis results not

participating to the ENSEMBLE

calculation

29/12/2015 New year issue (2016 input

not found) that led to a delay

in the forecast provision

Forecast results late arrival

(too late for participating to

ENSEMBLE)

The three first entries in the table above concern the same problem, which was due to

the establishment of 11:00 UTC as the delivery deadline for the daily analyses, whichwas rather early based on the runtimes of the LOTOS-EUROS analysis runs. Theproblem was diagnosed and corrective action was taken on the 21st of December 2015.

At the moment, eight (8) ensemble members are used in the Ensemble Kalman filter ofthe LOTOS-EUROS assimilation suite, instead of the “standard” twelve (12). We plan torevert to twelve ensemble members, as soon as a number of different optimizationswill be implemented in the model code, which are expected to considerably speed up

model runtimes.



Use for observations for data assimilation

Day

Use of observation for LOTOS-EUROS

December

O3 NO2 NO SO2 CO PM10 PM2.5

1 8,542

2 8,492

3 8,814

4 8,993

5 8,679

6 8,701

7 8,606

8 8,586

9 9,087

10 9,042

11 8,990

12 9,075

13 8,432

14 3,417

15 9,191

16 8,414

17 3,520

18 8,593

19 3,857

20 8,242

21 3,732

22 8,910

23 9,003

24 8,560

25 8,360

26 689

27 0

28 701

29 929

30 1,033

31 1,044

average 6,524



Day


January


1 1,527

2 1,843

3 286

4 3,831

5 6,078

6 8,527

7 1,347

8 7,845

9 1,328

10 4

11 2,623

12 6,341

13 9,108

14 8,907

15 3,559

16 2,546

17 2,458

18 9,416

19 9,150

20 9,358

21 9,327

22 9,308

23 9,171

24 8,955

25 9,300

26 4,249

27 9,005

28 9,059

29 9,208

30 8,705

31 8,903

average 6,170



Day


February


1 9,149

2 2,533

3 9,214

4 8,989

5 9,027

6 9,018

7 8,892

8 8,907

9 8,765

10 8,416

11 8,730

12 7,936

13 7,733

14 7,977

15 8,348

16 8,418

17 7,867

18 4,528

19 0

20 8,062

21 8,096

22 8,141

23 7,040

24 8,156

25 8,132

26 5,980

27 212

28 210

29 0

average 6,844

Comments: days with zero or limited number of assimilated observations are due tolimited availability of observations.



3. Verification report

This verification report covers the period December 2015 – January 2016 – February2016. The LOTOS-EUROS skill scores are successively presented for four pollutants:ozone, NO2, PM10 and PM2.5. The skill is shown for the entire forecast horizon from 0 to96h (hourly values), allowing to evaluate the entire diurnal cycle and the evolution of

performance from day 0 to day 3. The forecasts and the analyses cover a largeEuropean domain (25°W-45°E, 30°N-70°N). The statistical scores that are reported arethe root-mean-square error, the modified mean bias and the correlation. The surfaceobservations that are acquired by METEO-FRANCE and used for verification are

described in D50.1.2-2016Q1, that will be delivered shortly.

Verification of NRT forecasts

The following figures present, for each pollutant (ozone, NO2, PM10, PM2.5):

- in the upper-left panel, the root-mean square error of daily maximum (for ozone and

NO2) or of daily mean (PM10) for the first-day forecasts with regards to surfaceobservations, for every quarter since DJF2014/2015, a target reference value isindicated as an orange line,

- in the upper-right panel, the root-mean square error of pollutant concentration

forecasts with regards to surface observations as a function of forecast term,

- in the lower-left panel, the modified mean bias of pollutant concentration forecastswith regards to surface observations as a function of forecast term,

- in the lower-right panel, the correlation of pollutant concentration forecasts with

regards to surface observations as a function of forecast term.

The graphics show the performance of LOTOS-EUROS (black curves) and of theENSEMBLE (blue curves).



LOTOS-EUROS forecasts: ozone skill scores against data from representativesites, period December 2015 - January 2016 - February 2016

The performance of LOTOS-EUROS as regards ozone during this quarter was slightlyworse than that of the same quarter of the previous year, but manages nevertheless to

meet the target reference value for the RMSE of the daily maxima. The modelunderestimates more strongly during the early afternoon hours, which however is thecase for the ENSEMBLE mean as well. Correlation is considerably lower that of theENSEMBLE mean and decreases with forecast time.

To a certain extent, the performance of LOTOS-EUROS during this quarter was affectedby a bug in the C-IFS BCs code implementation which was responsible for latitudes ofthe C-IFS domain to be read in the wrong order. This has led to inconsistencies in the

influxes from the lateral boundaries. The bug was corrected in a code update whichtook place on the 1st of March 2016, therefore after the end of the quarter underdiscussion.



LOTOS-EUROS forecasts: NO2 skill scores against data from representativesites, period December 2015 - January 2016 - February 2016

The behaviour of LOTOS-EUROS as regards NO2 during this quarter was very similarwith the same quarter of the previous hear and manages to meet the target reference

value for the RMSE of the daily maxima. Bias (MMB) is consistently negative and veryclose to the values of the ENSEMBLE mean, except during the early morning hourswhen the underestimation is relatively stronger. Correlation is slightly better and withless fluctuation compared to last years values, although about 0.1 points lower than

those of the ENSEMBLE. The primarily local character of NO2 concentrations couldprovide a reasonable clue as to why the impact of the bug in the C-IFS BCsimplementation mentioned before, does not appear significant in this case.



LOTOS-EUROS forecasts: PM10 skill scores against data from representativesites, period December 2015 - January 2016 - February 2016

The LOTOS-EUROS forecast performance for PM10 deteriorated considerably during thisquarter. This is exhibited by a moderately positive bias (instead of the negative bias of

the ENSEMBLE and LOTOS-EUROS during the same quarter last year) and much higherRMSE, again compared to the ENSEMBLE and last year's results. The reason behindthis deterioration is the reintroduction of the use of C-IFS boundary conditions on the3rd November 2015. Further to the aforementioned bug that was responsible for

reversed lateral boundary conditions, the use of C-IFS allowed for unrealistically largeamounts of sea salt to enter the western boundary of the computational domain, whichdepending on the prevailing meteorological conditions lead to overestimatedconcentrations over mainland Europe for both PM10 and PM2.5. As a temporary measure

to ameliorate this effect, the use of C-IFS BCs for sea salt was discontinued on the 1st

of March 2016 (i.e. after the end of the quarter under discussion), at the same timewhen the aforementioned bug was fixed.



LOTOS-EUROS forecasts: PM2.5 skill scores against data from representativesites, period December 2015 - January 2016 - February 2016

The LOTOS-EUROS forecast performance for PM2.5 deteriorated during this quarter in amanner that resembles the performance of PM10. The reasons for this are explained in

the comments on PM10 performance.



Verification of NRT analyses

The following figures present, for each pollutant (ozone, NO2, PM10):

- in the upper-left panel, the root-mean square error of daily maximum (for ozone andNO2) or of daily mean (PM10) for the analyses (solid line) and for the first-day forecasts(dashed line) with regards to surface observations, for every quarter since

DJF2014/2015, a target reference value is indicated as an orange line,

- in the upper-right panel, the root-mean square error of pollutant concentration of theanalyses (solid line) and of the first-day forecasts (dashed line), with regards tosurface observations as a function of forecast term,

- in the lower-left panel, the modified mean bias of pollutant concentration forecasts ofthe analyses (solid line) and of the first-day forecasts (dashed line), with regards tosurface observations as a function of forecast term,

- in the lower-right panel, the correlation of pollutant concentration of the analyses

(solid line) and of the first-day forecasts (dashed line), with regards to surfaceobservations as a function of forecast term.

The graphics show the performances of LOTOS-EUROS (black curves) and of theENSEMBLE (blue curves). The superposition of the analysis scores (solid lines) and of

the forecast scores (dashed lines) computed over the same observation dataset ishelpful to assess the added value of data assimilation.



LOTOS-EUROS analyses: ozone skill scores against data from representativesites, period December 2015 - January 2016 - February 2016

The LOTOS-EUROS ozone analyses scores show a rather negative bias and RMSE thatis clearly larger than the that of the ENSEMBLE forecasts. However, RMSE for the daily

maxima is marginally below the reference target. Correlation is lower than that of theENSEMBLE. The reason for the relatively poor performance of the LOTOS-EUROSanalyses lies to a bug that was discovered in January (2016) and was responsible formisidentifying species other than ozone in the observational data (in most cases PM2.5),

as ozone. The effect of the bug was most pronounced during summertime, whenobserved ozone concentrations are generally high and at the same time PM2.5 wasgenerally lower than in the wintertime. The bug was corrected in the operationalproduction stream on the 2nd of February.



LOTOS-EUROS analyses: NO2 skill scores against data from representativesites, period December 2015 - January 2016 - February 2016

LOTOS-EUROS analyses did not use ground based NO2 observations for the periodsunder consideration.

The LOTOS-EUROS FC RMSE of daily maximum value meets the target value for thisquarter, while the diurnal profiles generally convey the same picture as detailed in theprevious set of graphs for NO2.



LOTOS-EUROS analyses: PM10 skill scores against data from representativesites, period December 2015 - January 2016 - February 2016

LOTOS-EUROS analyses did not use ground based PM10 observations for the periodsunder consideration.

The LOTOS-EUROS FC RMSE of daily maximum value meets the target value but theeffects of CIFS Bcs are evident, and have obviously led to a deterioration ofperformance since the previous quarter.



Analysis of LOTOS-EUROS performances for the quarter

The performance of the LOTOS-EUROS model this quarter was hurdled by twoimportant bugs in the model code used to pre-process input data. In conjunction withan intrinsic C-IFS overprediction of sea salt concentrations over the Northern Atlantic,these have led to unrealistically large forecast concentrations or PM10 and PM2.5 and

poor performance of the daily analyses for ozone. Even with those problems however,LOTOS-EUROS was able to meet the RMSE reference targets for daily maxima andmeans in all occasions.

The bugs were corrected on the 1st of March 2016 and this expected to provideimproved performance for the coming quarters. A number of other improvements inthe model are also planned, including the implementation of data assimilation ofadditional ground-based observations and OMI NO2 for the daily analyses which are

also expected to contribute to an enhancement of the LOTOS-EUROS performance.


regional production quarterly report on the daily analyses

Documents