Comparison of V999 and V002 L2 fluxes with CERES ERBE-like data and Met Office model simulations Richard Allan, Environmental Systems Science Centre, University of Reading Summary Versions V002 and V999 of the GERB Level 2 BARG and ARG radiative fluxes (Harries et al. 2005) for July and December 2004 are compared with CERES ERBE-like data (Wielicki et al. 1996) and from simulations performed using the Met Office global forecast model (Allan et al. 2005). Monthly-mean outgoing longwave radiation (OLR) from V999 GERB data is approximately 6 Wm-2 lower than CERES. This is of similar magnitude to the V002 data although this earlier release exhibits viewing-angle dependence to the differences which are removed in the V999 data. Monthly-mean GERB shortwave albedo is 0.02 to 0.03 brighter than CERES, approximately 0.02 larger than the mean differences seen for V002 in July 2004. Comparisons with the model simulations generally show larger model minus GERB differences for the V999 data compared to the V002 data. For July, OLR differences relate in part to regions of high optical depth mineral dust aerosol which the simulations do not include (Haywood et al. 2005). Over ocean regions, model minus GERB OLR differences are similar in both versions. However, positive model-GERB albedo differences over the oceans, thought to relate to the inaccurate simulation of low-altitude cloud radiative properties, are smaller in V999 than V002, while negative differences over north Africa, Brazil and high latitude regions are more pronounced in V999. This increase in GERB albedo from V002 to V999 is also apparent when considering clear-sky regions only. For example, the model underestimate in surface albedo over the Sahara appears more severe when analyzing V999 data compared to V002 data. Comparisons with CERES data Edition-2 ERBE-like CERES products (ES-4) combining the Terra FM1 and Aqua FM4 instruments are used to evaluate V002 and V999 GERB data. The CERES shortwave data are thought to be about 1% too dark for July and December 20041 and so CERES is likely to underestimate the shortwave albedo by around 0.005 over bright desert or cloud. Level 2 BARG data from V002 and V999 were first interpolated to a regular latitude-longitude grid (using the Met Office global model resolution; Allan et al. 2005) and averaged over each month for each 15minute time-slot. The 96 monthly mean time-slots were subsequently averaged to produce a monthly-diurnal mean (e.g. Comer et al. 2005). Figure 1 shows the V999 July 2004 mean OLR (top), albedo (middle) and calculated incoming solar radiation (ISW, bottom) for GERB V999 (left) and CERES (right). The GERB minus CERES differences are displayed in Figure 2 (V999) and Figure 3 (V002). GERB V999 OLR is generally 2-10 Wm-2 lower than CERES for July (Figure 2) and December (Figure 4) 2004. Subsatellite OLR differences are smaller in V002 data than V999 although for remaining regions the differences are similar. Overall, the V999 OLR is of order 2.5% lower than CERES ERBE-like data, consistent in sign with previous comparisons between GERB and CERES radiances (Harries et al. 2005). 1 http://eosweb.larc.nasa.gov/PRODOCS/ceres/ES4/Quality_Summaries/CER_ES4_Aqua_Edition2.html Figure 1: Top of atmosphere radiative fluxes for GERB V999 (left) and CERES (right) for July 2004: outgoing longwave radiation (top), albedo (middle) and incoming solar radiation (bottom). Figure 2: July 2004 mean GERB (V999) minus CERES differences in (top-left) outgoing longwave radiation, (top-right) reflected shortwave radiation, (bottom left) shortwave albedo, (bottom-right) incoming solar radiation. Figure 3: As Fig. 2 but for V002 GERB data. Reflected shortwave radiation (RSW) from GERB V999 is larger than CERES for July 2004, in particular over the northern subtropics where the maximum solar elevation is highest. On average, the shortwave albedo estimated from GERB V999 is about 0.02 larger than CERES. This magnitude is greater than the expected low bias in CERES data of 1% (which translates to 0.005 for bright regions of albedo=0.5). These differences are smaller in magnitude for V002 data although still positive over the northern subtropics. The GERB (V999) minus CERES albedo differences are larger still for December 2004 (Figure 4), GERB being of order 10% higher than CERES. Note that differences in ISW between the model calculated values for the GERB data coverage and CERES calculated values are sensitive to missing data since no interpolation is applied to the GERB data in this analysis. Figure 4: As Fig. 2 but for December 2004. Comparison with model simulations Figure 5 shows the 1200 UTC mean model minus GERB differences in OLR (left) and albedo (right) for July 2004 for V002 (top) and V999 (bottom). The OLR difference fields for V002 and V999 appear similar, highlighting the model errors due to cloud, primarily negative differences over the inter-tropical convergence zone which over land relates to the early onset of convection by midday in the model. Over the west Sahara, positive differences relate to model inadequacies in the representation of aerosol (Haywood et al. 2005) and land surface temperature. These differences are larger in V999 than V002. Also note that the small negative OLR differences over the east Sahara in V002 data are not present in V999 data. Positive model minus GERB V002 albedo differences over the ocean are reduced somewhat in the V999 data while negative differences over Europe, the Sahara and Brazil in V002 data are larger in V999 data. Similar results are applicable to December 2004 (not shown) although the OLR differences over the west Sahara are much smaller, which is likely to be in part due to the lower aerosol optical depth and cooler surface temperatures across the Sahara during December. Figure 5: July 2004 1200 UTC mean model minus GERB outgoing longwave radiation (left) and broadband albedo (right) for GERB versions V002 (top) and V999 (bottom). Cloud-cleared radiative flux comparisons with model simulations Assessments of GERB clear-sky fluxes are now conducted by sampling only grid-points where both the RMIB and model cloud fraction are zero. This allows comparison of consistently sampled model and GERB clear-sky OLR (OLRc) and clear-sky albedo for daylight hours. The June, July and December composite of model minus GERB OLRc differences are displayed for V002 (left) and V999 (right) in Fig. 6. The positive differences over the west Sahara are smaller in V002 than V999 (see also Comer et al. 2005) while the negative model minus GERB OLRc differences over the Kalahari desert are reduced in V999. The lower panels of Figure 6 show the scatter between model minus GERB OLRc difference and surface temperature. This highlights the large positive OLRc differences over the hot land regions. The light blue dots denote ocean points; the agreement with the model is generally good, as noted in Allan et al. (2005), and appears slightly improved in V999 data. Part of this improvement may relate to the improved geolocation (e.g. Comer et al. 2005); for example note the reduction in OLRc differences off the coast of Namibia. The negative model minus GERB clear-sky albedo differences over the Sahara are increased in V999 relative to V002. This represents a known underestimate in model surface albedo which was partially corrected in January 2005. It will be interesting to repeat the comparisons with recent model data with the correction implemented. Over the Atlantic, clear-sky albedo differences are reduced in V999 compared with V002. However, the reverse is true over the Mediterranean and also off the western coast of north Africa where the negative model minus GERB differences may be explained by the advection of reflective mineral dust aerosol over the dark ocean as illustrated in Harries et al. (2005) which is not represented by the model. Figure 6: model minus GERB clear-sky OLR differences for June, July and December daylight hours using V002 (left) and V999 (right) data. The bottom panels show the differences scattered with model surface temperature; ocean points are blue. Figure 7: As Fig. 6 but for clear-sky albedo Time-series and current data Time-series of the ocean-mean OLR and albedo and the model minus GERB (V002) differences are presented (Figure 8). The variability in model and GERB OLR is remarkably consistent with mean differences never greater than 5 Wm-2 (about 2% of mean OLR). The albedo time-series shows large differences which vary in time. Prior to January 2005, the model overestimated albedo by 0.02 (more than 10%). This was reduced following changes in the model parametrizations, in particular relating to boundary layer cloud. Recent GERB data (from November 2005) shows a higher albedo than the model. Again it will be interesting to re-assess the radiative properties of model cloud over the ocean following the reprocessing of the GERB data. The reduction in GERB albedo for November 2005 also applies for 0600 and 1800 UTC and for land (clear and cloudy) regions although is not evident for clear-sky time-series over the oceans where albedo is low (see http://www.nerc-essc.ac.uk/~rpa/GERB/gerb.html ). Conclusions Outgoing longwave radiation (OLR) from GERB (V002 and V999) is approximately 2.5% lower than CERES ERBE-like data. Viewing angle-dependence of the differences is reduced substantially in V999. Negative differences between model and GERB OLR over the tropical belt relate to errors in model cloud cover and are not strongly affected by the version of the GERB data. Ocean-mean OLR agrees to within 2% of the Met Office model although larger regional differences are evident. Model minus GERB OLR differences of up to 50 Wm-2 over the west Sahara during July 2004 are mainly determined by clear-sky conditions and are larger in the V999 data. Small negative differences in model minus GERB V002 OLR over the east Sahara are removed in V999 data. Shortwave radiative fluxes from V999 GERB data are of order 10% higher than CERES, showing worse agreement that the V002 data. However, the scene-dependence of the differences appears smaller for V999. Negative model minus GERB albedo differences over the Sahara are larger in the V999 data than the V002 data. However, there is a known underestimate in model surface albedo over the Sahara which has since been improved. A large model minus GERB V002 albedo difference of up to 0.2 for 1200 UTC data over the oceans is reduced but still present in V999 data. This relates to an overestimate in reflection from model low-altitude clouds, which has also since been reduced following improvements to model parametrizations. It will be interesting to reassess the model-GERB differences after January 2005, when the model changes were implemented, using V999 data. It is important to note that model clouds have in the past been tuned to produce better agreement with ERBE data so comparison with independent observations from GERB are particularly important in re-assessing the radiative properties of clouds, aerosol and the surface in the model simulations. Model simulated clear-sky fluxes over the ocean continue to show excellent agreement with GERB data with a modest improvement over the open-ocean in V999 data. A negative model minus GERB clear-sky albedo difference of approximately 0.03 is present in the V999 data over the Mediterranean and the Atlantic coast of north Africa when considering clear-sky composites for daylight hours. This is not present in the V002 data although may be explained by reflection by mineral aerosol dust which is not represented by the model at present. Over many coastal regions, there is an improvement in flux comparisons with the model, in particular for clear-skies, due to improved geolocation. Figure 8: Time-series of ocean mean (a) OLR, (b) albedo and Met Office unified model (UM) minus GERB differences (c-d) for 1200 UTC data. References Allan, R. P., A. Slingo, S.F. Milton I. Culverwell (2005), Exploitation of Geostationary Earth Radition Budget data using simulations from a numerical weather prediction model: Methodology and data validation, J. Geophys. Res., 110, D14111, 10.1029/2004JD005698. Comer, R.E., A. Slingo and R.P. Allan (2005), Diurnal cycle in GERB data – reprocessed data comparison: report to the GIST, ESSC, University of Reading, UK. Harries, J. E. and coauthors (2005), The Geostationary Earth Radiation Budget (GERB) Experiment, Bull. Amer. Meteorol. Soc., 86, 945-960. Haywood, J. M., R. P. Allan, I. Culverwell, A. Slingo, S. Milton, J. M. Edwards and N. Clerbaux (2005), Can desert dust explain the outgoing longwave radiation anomaly over the Sahara during July 2003? J. Geophys. Res., 110, D05105, doi:10.1029/2004JD005232. Wielicki, B.A. and coauthors (1996), Clouds and the Earth’s radiant energy system (CERES): an earth observing system experiment, Bull. Amer. Meteorol. Soc., 77, 853-868.