ECMWF – 1© European Centre for Medium-Range Weather Forecasts Developments in the use of AMSU-A, ATMS and HIRS data at ECMWF Heather Lawrence, first-year.

ECMWF – 1 © European Centre for Medium-Range Weather Forecasts

Developments in the use of AMSU-A, ATMS and HIRS data at ECMWF

Heather Lawrence, first-year EUMETSAT fellow, ECMWF

Supervised by: Niels Bormann & Stephen English


Outline

1. Investigating the value of HIRS

2. Introducing ATMS data over land and sea-ice

3. Situation-dependent observation errors for AMSU-A channels 5 - 7

3 PARTS:


1. Investigating the value of HIRS


1. HIRS: The Instrument IR sounder with Temperature sounding CO2, CO2/N2O channels Water vapour channels

9 channels used…

Coverage: MetOp-A, NOAA-19

…over ocean & sea-ice… and land for channel 12

HIRS19 Channels


1. HIRS: Aim & Motivation

HIRS is an older instrument whose value in the ECMWF system has not been tested recently

New hyper-spectral IR sounders (AIRS, IASI) may have made HIRS redundant

AIM: Investigate the value of HIRS in the ECMWF forecasting system

WHY?


Perform 2 sets of experiments: 2 x 2 months summer and winter, T511, 38R2:

Control: 38R2 version of ECMWF model (IR, MW sounders, scatterometers, radiosondes, etc.)

HIRS denial experiments: as control but take HIRS (MetOp-A and NOAA-19) out

1. HIRS: Method


1. HIRS: Results

DEPARTURE STATISTICS: observation – 12h forecast

MHSMW humidity sounder

Improved fit of MHS, IASI, AIRS to 12h humidity & temperature forecast

IASIIR temperature sounder

AIRSIR temperature sounder

0.5 – 1% improvement 2% improvement

0.4% improvement


1. HIRS: results

FORECAST SCORES: 1 – 10 day T, Z, R, VW forecast minus analysis

Degraded forecast

Improved forecast

Lots of blue = HIRS improves (short-range) forecasts

Day 2 500hPa Day 3 500hPa

neutral to positive: e.g. 500hPa Geopotential


1. HIRS: Conclusions and future developments

HIRS improves short-range forecasts of temperature, humidity, geopotential

Future Developments: MetOp-B HIRS

Trials are underway to test the introduction of MetOp-B HIRS

So far results look promising

Improved AIRS departures


2. Introducing ATMS over land and sea-ice


2. ATMS over land and sea-ice: The ATMS instrumentMicrowave Temperature/Humidity sounder (AMSU-A & MHS combination)

10 temperature sounding channels 5 humidity sounding channels

Temperature sounding: Humidity sounding:


2. ATMS over land and sea-ice: The ATMS instrument2011: Suomi-NPP satellite launched with ATMS on board

2012: Some ATMS data assimilated operationally at ECMWF

Land, sea-ice,ocean

Channel 9 coverage (2 cycles)

Channel 6 coverage (2 cycles)

Ocean only


2. ATMS over land and sea-ice: Aim & Motivation

AIM: Add channels over land and sea-ice

• Intoducing more AMSU-A data improves forecasts• Microwave data less affected by cloud than IR: has value over land/sea-ice

Add data:Humidity sounding channelsSurface-sensitive temperature channels

MOTIVATION:


2. ATMS over land and sea-ice: Method

How can we obtain skin temperature and emissivity?

Treat ATMS like AMSU-A and MHS:

• Emissivity retrieved from window channel prior to assimilation• Skin temperature retrieved during assimilation as a ‘sink variable’

)

Desired values retrieved in analysis

We need emissivity and skin temperature inputs

Karbou et al, Di Tomaso et al (2013)


2. ATMS over land and sea-ice: Assimilation experiments

3 experiments, 1.5 + 3 months, 39R1 137 vertical levels

Control: Same as operational 39R1 at lower resolution T511 (~40km)

ATMS Land: Control + ATMS over land

ATMS Land Sea-ice: Control + ATMS over land + ATMS over sea-ice


2. ATMS over land and sea-ice: ResultsDepartures: 12h forecast – observation

0.5% improvement

1% improvement: sea-ice

AMSU-A global

standard deviation(o-b) 2x2 months

Cha

nnel

num

ber

MHS global MHS Nhem winter

standard deviation(o-b) 2 months

Improved temperature and humidity 12h forecasts fit to observations

0.05% improvement


2. ATMS over land and sea-ice: ResultsForecast scores: 1 – 10 day forecast minus own analysis

DegradedForecast

ImprovedForecast

ATMS Land

ATMS Land+ Sea-ice


2. ATMS over land and sea-ice: Results

COLD SEA ATMS data appear to have a negative impact on TEMPERATURE

Could be because adding data makes analysis more variable?

Day 1 Temperature 1000hPa


2. ATMS over land and sea-ice: Conclusions

ATMS temperature and humidity sounding data was introduced over land and sea-ice

Departure statistics were improved for AMSU-A and MHS

Forecast scores were neutral to positive for ATMS over land data

Geopotential Forecast scores were neutral for ATMS over sea-ice

Short-range Temperature forecasts appeared degraded over Southern Ocean when sea-ice data introduced


3. AMSU-A Observation Errors


3. AMSU-A observation errors: The Instrument

10 Temperature sounding channels 7 satellites: good global coverage

Microwave Temperature Sounder


Tropospheric channels 5 – 7:

Important for NWP But cloud contamination/surface sensitive

3. AMSU-A observation errors: The Instrument


3. AMSU-A observation errors: Aim & MotivationChannels 5 – 7 observation errors should contain:

AIM: Develop situation-dependent observation errors

=

Observation error = surface term + cloud term + noise

Situation-dependent constant

stdev(o-b) MetOp-A AMSU-A channel 5: ALL DATA

NOT CONSTANT


3. AMSU-A observation errors: Surface term

𝝈𝒔𝒖𝒓𝒇𝒂𝒄𝒆𝟐=(𝑻 ¿¿ 𝒔𝚪𝟐)𝟐𝝈𝒆𝒎𝒊𝒔𝒔𝒊𝒗𝒊𝒕𝒚

𝟐+(𝜺𝚪)𝟐𝝈 𝒔𝒌𝒊𝒏𝒕𝒆𝒎𝒑𝒆𝒓𝒂𝒕𝒖𝒓𝒆𝟐 ¿

Do not include skin temperature term: skin temperature retrieved as sink variable

in analysis

Include emissivity term

Surface type

Sea 0.015

Sea-ice 0.050

Snow-free land 0.022

Snow-covered land 0.050

=

(S. English 2008)


3. AMSU-A observation errors: Liquid water path term

𝝈𝒄𝒍𝒐𝒖𝒅= 𝒇𝒖𝒏𝒄𝒕𝒊𝒐𝒏(𝒍𝒊𝒒𝒖𝒊𝒅𝒘𝒂𝒕𝒆𝒓 𝒑𝒂𝒕𝒉)

𝝈𝒄𝒍𝒐𝒖𝒅=𝟐 .𝟎𝟎𝒍𝒘𝒑𝟐+𝟎 .𝟕𝟗 𝒍𝒘𝒑

𝝈𝒄𝒍𝒐𝒖𝒅=𝟎 .𝟓𝟒𝒍𝒘𝒑𝟐+𝟎 .𝟑𝟎𝒍𝒘𝒑

𝝈𝒄𝒍𝒐𝒖𝒅=𝟎 .𝟐𝟎𝒍𝒘𝒑

Channel 5:

Channel 6:

Channel 7:

LWP (kg/m2)

Std

ev(o

-b)

Data screened for cloud but may still have some contamination…


3. AMSU-A observation errors: Noise term

LWP (kg/m2)

Std

ev(o

-b)

Channel 5: 0.25 KChannel 6 – 7: 0.20 K

=


3. AMSU-A observation errors: New Observation Errors

Metop-B AMSU-A channel 5 observation errors: used data

Nadir angles have higher valuesHigh lwp = higher value

=


3. AMSU-A observation errors: Assimilation Trials

Situation- dependent observation errors:

Weight data differently Allows the introduction of more data in ‘difficult’ areas: cloudy, high orography

Assimilation trials (2 months):

Control: version 40R1 with some 40R2 contributions at T511 (40km) resolution, 137 vertical levels

New observation errors: Control + new observation errors Extended coverage over cloud: Control + new observation errors + relaxed cloud

screening Extended coverage over high orography: control + new observation errors + relaxed

orography screening


3. AMSU-A observation errors: Extended coverage

Add cloud-screened data

Metop-B AMSU-A channel 5

Add data over high orography


3. AMSU-A observation errors: ResultsControl vs Observation errors experiment

Neutral Impact on forecast accuracy

degradation

Temperature 850hPa Geopotential 500hPa

improvement


Ctrl – obs errorCtrl – ext. cloud

3. AMSU-A observation errors: ResultsControl vs Extended coverage in cloudy regions

ATMS over sea Observation - 12h forecast

0.4% improvement

Improved fit to ATMS, neutral forecast scores: results encouraging

degradation

improvement

Ctrl – obs errorCtrl – ext. cloud


3. AMSU-A observation errors: ResultsControl vs Extended coverage in high topography

3 day geopotential fc - an

3 day temperature fc - an

Blue= Improved forecastRed/green= degraded forecast

Positive impact in northern hemisphere

Mixed positive/negativeOver Antarctica

Mixed positive/negative results


3. AMSU-A observation errors: Conclusions

Situation-dependent observation errors were derived for AMSU-A channels 5 -7

This gave neutral results with screening as-is

Introducing data previously screened for clouds improved fit to ATMS instrument

Introducing data over high orography had mixed positive/negative results

Work is ongoing


Summary of Findings

The HIRS instrument has a small positive impact on short-term T, Z, R forecasts

Introduction of ATMS data over land improves temperature/humidity forecast accuracy

Introduction of ATMS data over sea-ice has mixed results – further investigation needed

Situation-dependent observation errors for AMSU-A channels 5 – 7 have the potential to improve forecasts by introducing more data (work ongoing)

ECMWF – 1© European Centre for Medium-Range Weather Forecasts Developments in the use of AMSU-A, ATMS and HIRS data at ECMWF Heather Lawrence, first-year.

Documents

european centre

ecmwf land

value of hirs slide

hirs metop

hirs data

ecmwf system

motivation hirs

seaice slide