1. Direct LW radiative forcing of Saharan dust aerosols Vincent Gimbert, H.E. Brindley, J.E. Harries Imperial College London GIST 25, 24 Oct 2006, UK Met Office

2. Outline of Presentation Mineral aerosols – Radiative effects Case study March 2004: dust storm OLR perturbation Cloud and Dust detection GERB / RT Model comparisons over dust storm Analysis of clearsky OLR Conclusion and future work

3. Primary aerosols emitted from desert surfaces Lifted in atmosphere by strong surface winds Residence time 1 day ~ 1 week Present in the Lower troposphere but can travel 1000s of km Mineral dust aerosols Composite image from 3 SEVIRI Visible channels Composite image from 3 SEVIRI Infra-Red channels Credit EUMETSAT

4. Dust LW radiative effects Dust are large aerosols (~ 1 micron) so also interact with IR thermal radiation. Absorption of LW radiation and re-emission at level temperature Scattering of LW radiation Both Absorption and Scattering decrease the TOA OLR ΔOLR (dust type, height, AOD, surface/atmosphere)

5. Strong decrease of OLR Strong daytime surface temperature anomaly associated with dust event GERB, 3 March 2004 12:00

6. Cloud / Dust detection issues Dust misidentified as clouds by the RMIB GERB cloud product (visible detection) Similar problem with MPEF cloud mask (Visible/IR detection)

7. SAFNWC dust detection Daytime dust detection algorithm developed at Meteo France for SAFNWC Empirical algorithm based on SEVIRI 10.8 μm and 12μm channels thresholds. Generated operationally at SEVIRI spatial and temporal resolution

8. Strategy for estimating Direct Radiative Forcing Use of ECMWF analyses and assume the data assimilation provides adequate information on state of the atmosphere Assume that the dust feedback effects via modification of surface temperature and atmospheric profiles are accounted for in analyses (e.g. strong daytime surface temperature drop picked up by analysis on 3rd March 2004) Compare modelled clearsky OLR to GERB measurement over dust and Estimate direct LW forcing as the difference. Use clear sky scenes to test the model.

9. Radiative transfer modelling / GERB measurement Modelling of radiances using MODTRAN 4 v3r1 from 3.5 μm to ∞ Minor gases, heavy molecules from std profiles Surf Temp, Temp, Humidity, Ozone profile from ECMWF analysis 1 Surface type spectral emissivity (Dunesand) 1 * 1 Degree resolution – 60 vertical levels Comparison with GERB measurement GERB L2 ARG product (Not Edition) Use radiances because of suspected problems with ADM GERB viewing zenith angle form BARG products Comparison at 06:00, 12:00, 18:00 (No GERB data at 00:00) for the Month of March 2004

10. 12:00 GERB - Model difference  GERB/model anomaly coincides with dust front on 3rd March 2004  Anomalies over clear sky as well (cloud identification problem? representation of orography in model?)

11. GERB and Model 12UTC through March Cloud/dust free pixels averages over dust front region Over dust front (20040303) Mean diff = -7W/m2/sr Over clear sky: Mean diff = -0.4W/m2/sr σ = 0.9 W/m2/sr

12. Strong reduction of OLR diurnal amplitude on 3rd March 2004 OLR decrease smaller at 18 and not significant at 06 Model Warm bias at 06 and 18 (No dust detection) Need to understand clearsky GERB-Model differences to understand model deficiencies

13. GERB/Model Clearsky comparisons Study the Sahara region over clearsky pixels at 06, 12 and 18 for March 2004 Apply 1 pixel security margin to cloud detection (GERB PSF not accounted for in Cloud Mask)

14. Model warm bias at 06 and 18 Larger errors at 12 (larger Std Dev than 06 and 18) Time dependence of clearsky errors

15. 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 MS7 Window channel We look at possible sources of errors at the surface

16. Land surface emissivity Land surface emissivity retrived from ASTER/MODIS (Ogawa et al., 2004) Varies from 0.86 to 0.99 in the 8-13.5μm region Sensitivity study shows that it could change the OLR by up to 6W/m2/sr for hot and dry scenes. Early results show that model warm biais is reduced at 18UTC Model is offest at 12 as well In both cases, slight reduction of GERB-model variance Models need better surface emissivity over Sahara (effect on surface energy budget)

17. Land SurfaceTemperature Early study on 2 days of March 2004 at 12 and 18 with SEVIRI 10.8μm channel (‘Hottest’ SEVIRI channel) Tb(10.8μm) > Tskin (ECMWF) over some regions indicate Pb in model Tsurf How does that relate to GERB-model differences? 12:00

18. Temperature differences correlated to GERB/Model errors at 12:00

19. 18:00 Temperature differences much smaller at 18:00 Some strong OLR errors (e.g. orography) can be explained by a cold biais in model surface temperature Early comparisons suggest model errors in the diurnal cycle of surface temp

20. Conclusion and future work Dust storm was associated with OLR anomaly (wrt clearsky) of 7W/m2/sr at 12:00 with max forcing reaching 12W/m2/sr Strong reduction of OLR diurnal amplitude associated with storm Clearsky GERB/Model comparisons: Land surface emissivity has significant effect on OLR Evidence for surface temperature problem in ECMWF model, responsible for strong local model inconsistencies with GERB, Future work: Analyse Tb(10.8) for the whole of March 2004 and look for systematic regional biases over North Africa Model comparisons with SEVIRI WV channels Use ECMWF profile uncertainty data to estimate clearsky OLR error budget

21. Many thanks to: Helen Brindley for providing some of the SEVIRI data Alessandro Ipe for the RMIB cloud reprocessing Nicolas Clerbaux for running the SAFNWC software

23. 3rd March 2004 12:00, Srong decrease in Surface Temperature Surface Temperature anomaly from the 1st to the 18th of March 2004 at12:00 wrt March 2004 12:00 average. ECMWF operational model

24. Precipitable Water (mm) from the 1st to the 18th of March 2004 at12:00 ECMWF operational model