Evaluations. aerosol modules in global models. Stefan Kinne. MPI-Meteorology, Hamburg Germany. Michael Schulz, Christiane Textor, Sarah Guibert

Transcription

1 Evaluations aerosol modules in global models Stefan Kinne MPI-Meteorology, Hamburg Germany Michael Schulz, Christiane Textor, Sarah Guibert CE, Gif-sur Yvette, France

2 Anthropogenic climatic impacts our understanding is based on models aerosol introduces one of the largest uncertainties low understanding reflects deficiencies in modeling: let us have a closer look at aerosol modules in global models AEROSOL illustration of forced changes to the radiative energy budget at the top of the atmosphere

3 Aerosol Climate - Modeling winter the Earth s climate is a global issue global aerosol is complex (variable by region, season, year) concentration (aot ) absorption size spring summer fall MODIS/ SR 21 composite for seasonal aerosol optical depth

4 Aerosol Modeling a 4 STEP process process (lifetime) Step 2 MASS STEP 1 ESSION convert Step 3 AOT radiative transfer Step 4 FORCING point of validation most aerosol measurements

5 Modeling: OLD vs. NEW OLD aerosol = sulfate 2 3 SU 1 4 low absorption focus on industry globally incomplete NEW aerosol = many types OC 2 3 SS 2 3 SU 1 4 BC DU better characterization more processes more errors?! despite better representation in new aerosol modules the associated climate uncertainties remain large!

6 aerosol optical depth (STEP 3).2 12 models simulated global yearly averages for the visible aerosol optical depth (aot) total aot (55nm) modeled global yearly averages are similar so let us look at details behind differences

7 .8.6 opt. depth (STEP 3).4.2 by type dust aot (55nm) notice the different make-up different properties mean sulfate aot (55nm) differences in size (e.g. water uptake) differences in absorption differences in aerosol forcing!.4 sea-salt aot (55nm) total aot (55nm) org.carbon aot (55nm) black carbon aot (55nm)

8 simulated aerosol - by type emission mass opt. depth DU - emission (Tg/yr) SU emission ('S' Tg/yr) SS - emission (Tg/yr) OC - emission (Tg/y) BC - emission (Tg/y) dust mass (g/m2) sulfate mass (g/m2) sea-salt mass (g/m2) org.carbon mass (g/m2) black carbon mass (g/m2) dust aot (55nm) sulfate aot (55nm) sea-salt aot (55nm) org.carbon aot (55nm) black carbon aot (55nm).

9 lifetime (days) mass ext. eff. (m2/g) aerosol processing by component Transformation in 12 diff. component aerosol modules in global modeling lifetime STEP 1 STEP 2 emission mass mass ext. eff. STEP 2 STEP 3 mass opt.depth STEP 1 ESSION DU - lifetime (days) SU - lifetime (days) SS lifetime (days) OC - lifetime (days) BC - lifetime (days) STEP 2 MASS DU - mass ext. eff. (m2/g) SU - mass ext. eff. (m2/g) SS - mass ext. eff. (m2/g) OC - mass ext eff. BC - mass ext. eff. (m2/g) STEP 3 AOT

10 first impressions despite similar yearly global aot totals there are significant differences in aerosol composition also from significant differences in component processing large differences on regional and seasonal scales! modeling skill could be based on offsetting errors aot evaluations tell only part of the story but data-sets are available for aot consistency / sensitivity tests are needed to clarify issues in aerosol processing Experiment B: prescribe emission sources for models

11 aot datasets for evaluations global available data (suggest that most models underestimate aot ) Satellite Data AVHRR TOMS POLDER MODIS SR composite WHAT TO USE? AERONET IS SAMPLING REONAL REPRESENTATIVE? results of 12 global models with comp. aerosol modules component combined aot (.55um) global yearly averages GBAL MODE MODIS SR TOMS POLDER AVHRR,g AVHRR,n Aeronet

12 aot seasonal AERONET data

13 global fields of monthly aot deviations of an average models with respect to AERONET data model too small zero model too large

14 aot ( average model AERONET) blue simulations too small yellow / red simulations too large discrepancies biomass peak Euro summer Asian dust

15 global fields of seasonal aot deviations of individual models with respect to AERONET data model too small zero model too large

16 aot (individual models AERONET) many non-green colors = larger discrepancies ( but regional non-representation has NOT been filtered)

17 aot Satellite Data choice: MODIS complemented by SR (MM)

18 aot (average model vs multi-year satellite data (AVHRR, TOMS) vs year 21 satellite data (MODIS, SR) MODEL- DATA simulations suggest smaller AOT than satellite retrievals in remote regions! (transport? sources?) note: extreme satellite data at high latitude winters are in error sub-pixel snow ground cover MODIS, SR year 21 AVHRR, TOMS multi-yr

19 aot (average model vs multi-year satellite data (AVHRR, TOMS) vs year 21 satellite data (MODIS, SR) best AOT retrieval over ocean: MODIS best cloud detection (using 25m pixels) thus less pot. contamination lowest ocean aot (but still up to twice as large than simulations) note: an unusual trend [MODIS >TOMS, AVHRR] off-asia because 21 had unusual strong dust transports from Asia MODIS, SR year 21 AVHRR, TOMS multi-yr

20 aot (average model vs multi-year satellite data (AVHRR, TOMS) vs year 21 satellite data (MODIS, SR) best AOT retrieval over land:? TOMS: biased high MODIS incompl. cover SR: temp. sparse MODIS/SR combo? note: SR (with a more complete spat. coverage than MODIS) suggests smaller (!) optical depths over urban regions smaller sizes? MODIS, SR year 21 AVHRR, TOMS multi-yr

21 aot (average model vs multi-year satellite data (AVHRR, TOMS) vs year 21 satellite data (MODIS, SR) SEASONLITY both simulations and retrievals underestimate seasonality compared to ground data statistics e.g. biomass maxima in tropics too weak (from Aug-Nov in S.Ame/S.Afr) note: summer / fall simulations exceed satellite data near urban sources (outdated inventories?) MODIS, SR year 21 AVHRR, TOMS multi-yr

22 global fields of seasonal aot deviations of global satellite data-sets with respect to AERONET data model too small zero model too large

23 aot (Satellite data AERONET) significant differences to AERONET (even for MODIS/SR choice)

24 global fields of seasonal aot deviations of individual models with respect to MODIS/SR 21 data model too small zero model too large

25 aot Jan (indiv.models - MODIS/SR 21)

26 aot Apr (indiv.models - MODIS/SR 21)

27 Jul aot (indiv.models - MODIS/SR 21)

28 aot Oct (indiv.models - MODIS/SR 21)

29 final remarks optical depth data are an insufficient benchmark when evaluating aerosol modules in global models a better understanding on the data quality of global or regional aerosol measurements is needed comparisons of simulations to component combined column aerosol data-sets do not tell the entire story strong differences in lifetime and for (mass to optical depth) conversions for all aerosol components (off-setting errors?) coordinated sensitivity studies are needed to understand differences in aerosol processing simulations with prescribed emission sources etc.