1. - 200 hPa geopotential heights in the GDAS analysis are lower than in CDAS between 20 o N to the South Pole. - 200 hPa geopotential heights are consistently lower for GDAS compared to CDAS since at least November 2004 over the global Tropics. -200 hPa Errors in the GFS forecasts in excess of 0.5 std. dev. are observed after Day-6 & in the Topics & near the poles. -The CDAS monthly precipitation differences from CAMS_OPI are similar to May and June 2005: CDAS < observations over the near-equatorial Tropics in the NH, the northern Indian Ocean and “coastal” western Pacific, and = or > observations over much of the Southern Hemisphere near-equatorial Tropics -GFS precip. >> than obs. over the Atlantic & Pacific ITCZs and the Arabian Sea, and < over parts of SE Asia and Indonesia at all forecast projections (1-15 days) -Vertical velocity and upper-level divergence fields are consistent with the precipitation differences over the Pacific near-equatorial regions in the GDAS/GFS. -CDAS 2m temperature anomalies are generally within 1K of observed surface temperature and 1-3K cooler where they differ. GDAS sfc. temps much closer to obs. than CDAS. July 2005 Summary of CDAS/GDAS/GFS

2. 200 hPa geopotential heights in the GDAS analysis are lower than in CDAS between 20 o N to the South Pole. The magnitude of the differences over the Tropics are equivalent to about 1 standard deviation in the CDAS monthly means over the 1971-2000 base period. This is the the same basic pattern since March 2005.

3. Time series of 200 hPa heights for GDAS and CDAS over the Tropics indicate that heights are consistently lower for GDAS compared to CDAS since at least November 2004, and that this difference occurs over land as well as ocean, although the difference is largest over the oceans.

4. The zonal 200 hPa hgt errors in the GFS grow with the length of the forecast projection. During July 2005, negative hgt errors in excess of 10m are observed for the day-4 forecasts and beyond at most latitudes. Errors in excess of 0.5 std. dev. (right panel) are observed after Day-6 that grow with forecast projection in the Topics & near the poles. The error south of 60 o S are much less than in June 2005. http://www.cpc.ncep.noaa.gov/products/fcst_eval/html/maps_mrf.html

5. Similar to the past 2 months, the CDAS monthly precipitation accumulations are less than the satellite- gauge (“CAMS_OPI”) observations over most of the near- equatorial Tropics in the Northern Hemisphere and the southern portion of the SPCZ. In contrast, CDAS indicates much heavier accumulations than the observations over the hurricane development region in the tropical Atlantic west of the Caribbean in July 2005. May 2005 June 2005

6. The precipitation anomalies between CDAS and the satellite estimates agree well over the Indian subcontinent (positive anomalies) and the Pacific ITCZ (negative anomalies). As obs- erved in the accumulation map (previous slide) CDAS has a 4-5 Mm/day positive anomaly for July 2005 that is not in the CAMS-OPI estimates.

7. GFS precip. much larger than obs. over equatorial South America and the Arabian Sea at all fcst. projections. GFS rainfall is much less than the satellite estimates over the much of SE Asia and the equatorial Indian Ocean. Very similar to June 2005 Precip. is much too strong in the Atlantic and Pacific ITCZs at all forecast proj- ections – same as all previous months.

8. GFS precipitation forecasts generally too wet (by ~ 100 mm) east of the Mississippi for the Day1 forecasts, but the sign of the diff- erences is not uniform for the other fcst. projections. Contrary to previous months (but consistent with June), these biases do not shrink or grow steadily with forecast projection.

9. The CDAS OLR is generally cooler over the continents compared to observed OLR, although this difference may be due to the fact that observed OLR is from measurements at only a few times of day (0200/1400 LST at the equator). GDAS OLR is substantially closer to the observed OLR compared to CDAS, and the differences with observed OLR are almost completely positive.

10. The evolution of near-equatorial OLR anomalies during Nov 2004 through July 2005 in CDAS shows negative anomalies for the entire period near the date line. That contrasts with the observed OLR which indicates negative anomalies there during Dec 2004 – Feb 2005 and mid-March-May 2005. The relatively dry period During Jan-Mar 2005 on either side of the date line agrees well With the observed OLR. Note that GDAS anomalies are not plotted because a reliable climatology is not available from GDAS due to the many model changes that have occurred during the GDAS record.

11. The GDAS upward motion at 500 hPa is considerably stronger than CDAS over the ITCZs in the Atlantic and Pacific which is consistent with the higher rainfall in GDAS over those areas compared to CDAS (see earlier precip. figs – buttons below). This is also consistent with differences in the upper-level Divergence (button below). CDAS precip GDAS precip Divergence

12. GDAS vertical velocity (500 hPa) is consistently higher than CDAS over tropical land regions, in good agreement over the tropical oceans, although consistently lower there since April 2005. Both are in good agreement over the NH oceanic storm tracks. We presume that the vertical motion over the Pacific ITCZ increases with forecast projection to be consistent with the heavy precipitation in the 5-15 day GFS forecasts.

15. CDAS 2m temperatures are generally 1-3K cooler over the land surfaces than is observed; regions where CDAS is up to 1K warmer than the obs. are generally in areas with high terrain. In contrast, GDAS 2m temps. are 1-3K warmer than CDAS over the majority of the continental regions.

16. In general, the CDAS temperature anomalies are within 1K of the observed anomalies; where they differ, the CDAS anomalies are generally cooler than the CAMS (station) data.

17. Global Mean precip from GFS is about 25% higher than GPCP.