Imperial College London Dust Longwave Forcing from GERB

Imperial College London Dust Longwave Forcing from GERB and SEVIRI Vincent Gimbert, H. E. Brindley, J. E. Harries Thanks to Rainer Hollmann, Nicolas Clerbaux, Alessandro Ipe and Richard Allan, Andy Smith GIST 26, 03 May 2007, RAL, Abingdon

Motivation § Use GERB and SEVIRI to measure TOA LW radiative forcing of Dust aerosols over the Sahara § Model Clear-sky TOA LW Radiances Using ECMWF Analyses to separate out effect of Dust from that of variable meteorology § Test the ECMWF model against observations and provide uncertainties in estimate of Dust forcing

Outline of Presentation ü Method, measurements, model ü March 2004: LW forcing of dust storm from GERB ü Model comparison with Clearsky observations from GERB and SEVIRI IR Channels ü Dust spectral forcing ü Results from March 2006 ü Conclusion and future work

GERB LW radiance 200403 12 UTC ü Strong OLR perturbation associated with dust event ü Daytime SAFNWC Dust detection

RT modelling / GERB-SEVIRI data Spectral Radiances modelled using MODTRAN 4 from 3. 5 μm to ∞ • • Temperature, humidity, ozone from ECMWF analyses Unity surface emissivity 1 * 1 Degree resolution - 60/91 vertical levels Spectral integration over appropriate instrument/channel response Comparison with GERB and SEVIRI measurement • • GERB L 2 ARG SEVIRI L 1. 5 Radiances (7 IR channels, 062, 073, 087, 097, 108, 120, 134) 6 -hourly 00: 00, 06: 00, 12: 00, 18: 00 (No GERB data at 00: 00) March 2004 and 2006

12: 00 GERB - Model difference § GERB/model anomaly coincides with dust front on 3 rd March 2004 § Anomalies over clear sky as well

GERB and Model 12 UTC through March Cloud/dust free pixels averages over dust front region (~10*10 deg square) Over dust front (20040303) : LWRF ~ 20(+/- 3) W/m 2

Time dependence of clearsky errors • Model warm bias at 06 and 18 • Larger errors at 12 (larger Std Dev than 06 and 18) • Use SEVIRI to test model. . .

SEVIRI/model Water Vapour channels WV 6. 2 μm, High/mid atm Time WV 7. 3μm, Mid/Low 00 UTC 06 UTC IR_062 Mean (1σ), K -1. 67 (0. 8) -1. 59 (0. 6) IR_073 Mean (1σ), K 0. 95 (0. 8) 0. 95 (0. 7) 12 UTC 18 UTC Channel -1. 51 (0. 6) -1. 53 (0. 6) 2. 05 (1. 0) 1. 34 (0. 8)

Window channels • Persistent model warm bias at night (σ~1 -2 K) • Daytime cold bias (~3 K), σ~3. 3 K • Discrepancies over LAND • Same results from 10. 8μm channel

The case of 8. 7μm Channel 12. 0μm 8. 7μm Night Day

• Strong SEVIRI/model differences in 8. 7μm channel • Same pattern at all time steps • Surface emissivity?

• Assume emissivity (12. 0μm)~1 and T(8. 7 μm) ~T(12. 0 μm) • Separate surface emissivity signal from 8. 7 μm channel Method useful for removing consistent bias No a priori knowledge of surface spectral emissivity needed

Latitudinal dependence of Clearsky errors • No dependance in WV channels 12. 0μm: Surface Temperature 8. 7μm: Emissivity + Tsurf

Dust LW Spectral forcing No correction: unrealistic forcing, No uncertainties 6. 2 um: non- physical 7. 3 um: Negative at 12 UTC… 8. 7 um: Too high (emissivity bias)

March 2006 Dust event 20060307 12 UTC EUMETSAT Dust RGB colour scheme

March 2006 Dust event 20060308 12 UTC EUMETSAT Dust RGB colour scheme

Dust LW Spectral forcing • Maximum forcing on 20060308 • Similar spectral dependence as March 2004 • Consistent with models

Conclusion and future work ü ü Mean 200403 Dust storm OLR forcing: ~20 W/m 2 Diurnal cycle of Clear-sky GERB-model errors Good agreement in SEVIRI WV channels (small biases) Land Emissivity and Temperature Regional Biases in Window Channels ü Very similar results for March 2004 and 2006: ü Dust Spectral Forcing Future work: ü Validation at Niamey ARM site: ü Dust detection reliability ü Diurnal cycle of Tsurf ü Consistence of GERB-SEVIRI vs Model biases

Diurnal cycle of forcing

Daytime Conditional Bias Model does not reproduce the range of GERB OLR at 12 Similar finding as Trigo and Viterbo (2003) with ECMWF model compared to MS 7 Window channel We look at possible sources of errors at the surface