Wednesday, December 5, 2007

Reanalysis Precipitation Climatology

On the MERRA WWW page, we are posting several figures showing the comparison of 5 satellite era reanalyses with GPCP and CMAP precipitation data sets. There are some similarities among the reanalyses, in their differences from the observations (Tropical precipitation, and interestingly European continental January precipitation), but also differences between the merged observation data sets (GPCP has lower tropical precipitation than CMAP, but higher January precipitation, in general). Citations are provided on the page, that provide some analysis and discussion on the sources of bias. However, there are many other aspects in comparing reanalyses to the observed data. These are only climatologies, so that interanual variability, weather scale and diurnal cycle differences are not expanded.

The WWW page is at: http://gmao.gsfc.nasa.gov/research/merra/reanalysis_precipitation_climatology.php

Please take a look, and feel free to make comments on this blog.

Thursday, November 29, 2007

Status Nov 29

Just a brief post. The spinup runs are still on hold. The physics in the system seems to be set and the output diagnostics are likewise set. The main hold up is that the adaptive bias correction of a small number of channels over land is not stabilizing even after long (coarse resolution) runs, and continues to grow ultimately leading to the rejection of observations that should otherwise be accepted. A patch seems to be working, and a clean experiment is getting underway today. The spinup experiments should be restarted soon, and that will be posted here when it happens.

Tuesday, November 20, 2007

Summary of the MERRA User's Review Group Meeting

In late 2005 a MERRA review group was formed from experts in various aspects of Earth system science and users of existing reanalyses. Their charge was to review the GMAO strategy for MERRA and the validation experimentation and results, possibly contributing some of their own analysis. The goal for GMAO was to gain a preliminary assessment of the scientific merit of the GEOS-5 data assimilation system for MERRA prior to full production. In September 2007, the validation experiments began, and on November 7, the user review group met to discuss the results of the validation experiments with the GMAO and NASA HQ representatives.


The GMAO started the day, presenting a summary of the system and critical improvements in recent months (Rienecker), the dynamical circulation, clouds and radiation (Suarez and Bacmeister), climate variability features (monsoons, hurricanes, low level jets (LLJ) and diurnal cycle - Schubert) and precipitation statistics and land hydrology (Bosilovich and Koster). Key points from the presentations are summarized below.

Michele Rienecker reviewed some major and critical changes to the system since the inception of the Review Group. These include improvements in the use of retrieved wind speed over the ocean, improvement in the radiance assimilation (through the latest CRTM radiative transfer coefficients), corrections to bias and jumps in the radiosonde observations and a fix for diurnal cycle of glacier surface temperatures.

In looking at zonal circulation, Max Suarez showed the differences between the GEOS-5 and other reanalysis systems for winds, temperature and humidity. For example, the GEOS-5 eddy heights compare with ECMWF operational analysis both in a mean sense, and in the interannual variability. With small contour intervals in the zonal cross-sections, differences in tropopause height can be identified among all the reanalyses. In addition, GEOS-5 reproduction of stratospheric ozone profiles is reasonable, and a limited comparison of the beginning of a quasi-biennial oscillation looks promising. One possible systematic problem is high upper troposphere humidity (as compared to ECMWF and NCEP operational analyses). The radiation fluxes have some bias, as well, but these are somewhat reduced compared to the existing reanalyses (Figure 1).

Siegfried Schubert reviewed some evaluations of monsoonal circulations, including the North American monsoon and Indian monsoon. GEOS-5 reproduces the low level winds (e.g. the Somali jet and the Great Plains LLJ) as well as the subseasonal breaks observed in the monsoonal precipitation. There are some apparent regional biases in the precipitation, but this is also true among all the existing reanalyses. The GEOS-5 North American monsoon circulation and precipitation compare well with the North American Regional Reanalysis (NARR) (for July 2004, Figure 2). Globally, the interannual variations of precipitation compares well with observations, and better than existing reanalyses. In addition, the monthly average water budget shows globally averaged analysis tendencies to be a small value (Figure 3). However, the diurnal amplitude of continental precipitation is large and the phase is shifted to a daytime maximum compared to observations. This is a problem for all reanalyses, and it persists in the GEOS-5 system.

Mike Bosilovich reviewed monthly mean precipitation, where GEOS-5 generally produces good fields compared with GPCP and CMAP, not only in the global mean, but also spatial correlation. In addition global P-E is generally small (near zero) indicating that the global analysis is relatively well balanced (but will be non zero). The GEOS-5 precipitation is reasonable in many regions and latitude bands. Comparisons for the Mississippi River basin precipitation against daily gauge data show the GEOS-5 was able to produce the daily precipitation events, and the no-rain days for Jan-Oct 2004 (Figure 4). However, maximum intensities in the warm season are underestimated, leading to an underestimate of the total basin precipitation. Randy Koster’s analysis of the time series of precipitation shows that the occurrence of rain during the day coincident with solar forcing causes high interception loss of water vapor, and then the runoff water is underestimated. The transition of the observing system to include SSM/I was tested in a data withholding experiment. GEOS-5 tropical precipitation increases with the inclusion of SSM/I, but the increase is less than 10% of the tropical precipitation (in contrast, JRA25 has a change in extratropical precipitation). There is also a small increase of total column water, ocean surface winds and ocean evaporation.

The overall conclusion is that the GEOS-5 system can produce many aspects of the Earth system as well or better than existing reanalyses. The quality of the data coupled with the fine temporal and spatial scale of the data should make the GEOS-5 reanalysis useful for many purposes. While there were spirited discussions among all the participants, the external user group members’ sentiment reflects this conclusion as well. As of November 2007, the reanalysis data streams are being spun up, and data should start flowing to the scientific community early in 2008. The full MERRA data product will take approximately 18 months to generate.


Figure 1 Monthly mean (Jan 2004) TOA Longwave radiation differences between CERES ERBE-like observations and several reanalyses and operational analyses.

Figure 2 Comparison of the seasonal evolution of the North American monsoon between the North American Regional Reanalysis (NARR) and GEOS5.

Figure 3 Global vertically integrated water vapor budget for July 2004 including the physical components, the analysis increment and residual.

Figure 4 Mississippi River basin area-average (over all sub-basins) daily precipitation for January – September 2004. The figures show the scatter of the daily data, the daily time series, and the accumulated precipitation. The observations are CPC daily ¼ degree gridded gauge data.

Friday, November 9, 2007

Status Nov 9

The validation of the system has been somewhat time consuming between this post and the previous, and much has happened. At least 20 GMAO staff (or more) spent several weeks interogating the validation experiments each focusing on various Earth system components. On Oct 11, the GMAO held an internal review of the validation experiments. On Nov 7, the summary of these results were presented to our User Review group in a meeting at GSFC. My interpretation of the Review is that the system has more than enough scientific merit to proceed to production phase, weighing the advances and advantages against the limitations and some weaknesses. When any formal writing from the Review are made available for public posting, I'll put it on the blog. This is a significant milestone for the MERRA project and the GMAO.

There were many very positive results that came out of the MERRA validation experiments. Too many to easily synthesize into blog posts. A validation document is under development, but should take some time. Some results will be posted here as time goes on. In the near term, however, validation pointed out a serious flaw in the system. When the CERES science team evaluated the data, they found that Antarctica and glaciers did not have a diurnal cycle of surface temperature. The reason ended up being a thick glacier layer. Some new code, including a thinner layer and revisions to the energy budget code have produced very reasonable results. So, this fix will be added to the MERRA system. (see the Figure)

Figure: Time series of 2m air temperature at two Antarctica stations. The green line indicates GEOS5 Patch 15, Blue is patch 20 (including the fix) and the read is ECMWF operational analysis. Model data are the nearest gridpoint to the stations. Station data is marked with a black box.

So, the spin up of MERRA production runs are on hold until the system is updated. Some testing of convection parameterization coefficients has been going on through this process. A decision is pending on which, if any, will go into MERRA. The issue to be resolved are, updating the system with new glacier surface temperatures, finalizing the MERRA output routines and final evaluation of the convection parameterization. Spinup runs will restart once these issues are resolved.

Thursday, September 27, 2007

Status Sept 27

Just a quick post on current jobs and activities.

The validation runs for Jan-Jul and Jul-Oct 2004, Jan and Jul 2006, Jul and Aug 1987 are complete. Jan and Jul 2001 are going, and should wrap up next week. For the next two weeks a summary of the results will be pulled together.

For production, the system will run in three data streams. Each stream will be spun up for two years at a coarse (2 degrees resolution) then the native MERRA resolution for 1 year. The streams will each start at January of 1979, 1989 and 1998.


Presently, the coarse (2 degree) runs are complete and the native (1/2 degree) spin up runs have started. They are all either completed January or into February. Each stream is producing around 10 days/day when the computers are up. With reasonable up time, they should reach the nominal beginning of production by early November.

Friday, September 7, 2007

Incremental Analysis Update

GEOS5 uses an Incremental Analysis Update (IAU) to constrain the atmospheric numerical model by observations. The following figure shows the schematic of the procedure. Starting at 09Z, a 6 hour forecast is run, and forecast data from 09Z, 12Z and 15Z are used to create the analysis (blue diamond). From the analysis and the forecast, a tendency is calculated. This tendency is applied to another model forecast cycle in the prognostic equations (green arrow and light blue box). This is called the corrector segment, so that the tendencies are nudging the model forecast in the direction of the observations at every time step.

MERRA will have two primary products. First, the analyses will consist of the model state variables written instantaneously after the analysis every 6 hours (00Z, 06Z, 12Z and 18Z). There will be model level (72 eta levels) and pressure interpolated (42 pressure levels) for each analysis time. Second, the diagnostic fields are written from the model corrector segment. These include 1 hourly average 2 dimensional (1/2 deg latitude by 2/3 degree longitude) surface, single level (e.g. H500), radiation, land specific and vertically integrated fields. In addition, 3 hourly average 3 dimensional coarse resolution 1.25 deg x 1.25 degree) atmospheric diagnostics are produced from the corrector segment. These include all the tendencies for the state variables, as well as fluxes and budget terms.



One advantage of IAU is that it allows the corrector segment data to be written. This data is exposed to the observational forcing spread out in time, rather than a large change in the initial conditions. The spin up spin down problems in the forecast, associated with initializing a forecast system with an analysis data, are much smaller. Essentially, this allows the production of 1 hourly precipitation and other physics fields. The figure below shows a global average precipitation time series (data is written at every model time step, ~30 min) using the synoptic analysis as initial conditions, IAU and a pure model forecast. Reinitializing the forecasts with the analysis causes jumps in the time series. The free running model tries to have a global precipitation rate of ~3 mm/day. The analysis tried to reduce that, but after the initial time the forecast starts to drift back to it's preferred climate state. The IAU provides forcing at every time step, constraining the system with the observations.

Monday, September 3, 2007

Ocean Surface Winds and Fluxes

The ocean atmosphere interactions are one of the crucial elements in climate variability. The MERRA system does not have a coupled ocean model and data assimilation, but future reanalyses will likely go in this direction. In MERRA, seas surface temperatures are prescribed and ocean surface wind observations (from buoys and satellites) are analyzed. Downward components of the radiation would be related to the parameterized clouds and radiation calculations, as well as the input observations (radiances, temperature and moisture). The following figures prepared for validation compare some winds and fluxes with GEOS5 experiments and other reanalyses.

This following figure shows the daily time-series for Jan 2004 and 2006 U10M and V10M winds at the TAO mooring location of 165E on the equator. GEOS-5 shows good agreement with TAO and matches the minimum and maximum values everywhere. In Jan 2004 and 2006, GEOS-5 is more highly correlated with QSCAT than NCEP CDAS or JRA-25. The NCEP-CDAS analysis shows several periods of larger bias against the observations.

The next figure shows maps of monthly latent heat flux for GEOS5, JRA-25 and NCEP CDAS. GEOS-5 has much less evaporation out of the ocean than JRA-25 and NCEP CDAS, especially in the western boundary currents: Gulf Stream and Kuroshio. Mean and RMS differences between GEOS-5 and JRA-25 and GEOS-5 and NCEP CDAS are much larger than that of NCEP CDAS and JRA-25 in the above region. Similar patterns are seen in January 2004.

In three validation periods investigated so far (Jan/Jul2004 and Jan 2006), GEOS-5 net radiation is more highly correlated with TAO in the Eastern Pacific that other reanalyses.
NCEP-CDAS generally is biased low in most time-periods and TAO locations. GEOS5 also correlates well with the TAO incoming shortwave radiation observations.


The three reanalyses are fairly different in their net heat fluxes. GEOS-5 has less heat loss than both NCEP and JRA in the Kuroshio and Gulfstream areas, and more heat gain off Western Australia. In the 45S-45N band, GEOS5 and JRA have substantial differences.

Tuesday, August 28, 2007

Update and TOA Radiation

Vacations in August have limited the posting, but the validation experiments are continuing. The experiments are moving well, and the system performance will be discussed in another post. Here are some of the validation comparisons for TOA radiation of some reanalyses to CERES ERBE-like data from Terra and Aqua. In the figures, the GEOS5 validation experiment is labeled d5_b10p15 (d is the 1/2 degree resolution, beta10 patch15 is the version of the data assimilation system).

The first figure is the zonal mean of the TOA LW and SW fluxes in Jan and Jul 2004. CERES observations are in red, GEOS5 is blue, JRA-25 is black and NCEP RII is green. GEOS5 shows somewhat smaller bias in the tropics LW, and also mid latitude SW. The July upward SW in the tropics seems biased high compared to the obs and other reanalyses, but otherwise, it seems in the range of the reanalyses. (Click on thumbnails to see the full figure)



The next two figures show the monthly mean maps of the comparisons. to CERES Terra. The differences between the CERES-Terra and CERES-Aqua are provided as a reference. In TOA LW, JRA-25 seems systematically biased high, while NCEP RII has strong positive and negative variations. GEOS5 leans to a high bias, but not as high as JRA. JRA and NCEP also show large positive bias in the tropics, which impling a dry upper troposphere or low cloud top. Newman et al (2000) have evaluated the interal consistency of several reanalyses, between OLR, precipitation and upper level divergence. In addition, they note that the correspondence among the reanalyses is quite low. The reanalyses OLR are all different from each other.



The shortwave biases of the reanalyses generally are similar. The exception seems to be the polar warm season. For example, TOA SW in Antarctica January is high for NCEP RII and GEOS5, but low for JRA.




Newman M., P. D. Sardeshmukh, J. W. Bergman, 2000: An Assessment of the NCEP, NASA, and ECMWF Reanalyses over the Tropical West Pacific Warm Pool. Bull. Amer. Meteror. Soc. 81, 41–48.

Smith, G. L.; Wielicki, B. A.; Barkstrom, B. R.; Lee, R. B.; Priestley, K. J.; Charlock, T. P.; Minnis, P.; Kratz, D. P.; Loeb, N.; 2004: Clouds and Earth Radiant Energy System (CERES): An overview, Advances In Space Research, 33, 1125-1131.

Wielicki, B.A., B.R. Barkstrom, E.F. Harrison, R.B. Lee, G. Louis Smith, and J.E. Cooper, 1996: Clouds and the Earth's Radiant Energy System (CERES): An Earth Observing System Experiment. Bull. Amer. Meteor. Soc., 77, 853–868.

Friday, August 10, 2007

Historical Satellite Assimilation

When evaluating one of our first analyses in the 1980s, we found some results very different from what was apparent in the recent (2001-2006) experiments. The experiment was run to test the impact of SSMI on the reanalysis time series (See this report). THe increments were drying the lower tropospehre continuously, and precipitation was small and getting smaller. Several aspect of the system were studied to pin down the problem. One hypothesis was that older coefficients (for the radiative transfer model) for the historical satellite data were causing some of the problems. NOAA NESDIS was in the process of creating new coefficients, and were able to make the new values available for testing, and now for the whole reanalysis period.

The figure below shows the time series of analysis increments of temperature and moisture for the old coefficients compared to the new coefficients. The original coefficients lead to large drying increments in the lower troposphere, as well as dramatic jumps at the when the satellites change. NOAA10 data starts at the end of Oct 86, and a couple channels drop out for several days in January 1987. The new coefficients work much better with the radiance assimilation, and the increments are much more uniform.


Thursday, August 9, 2007

New JRA-25 Citation

Subject: JRA-25 paper
Date: Thu, 09 Aug 2007 13:02:25 +0900
From: JRA-25 user administrator

Dear JRA-25 reanalysis data users,

Sorry the address written in the previous mail was wrong.
Please refer the next address.

JRA-25 standard reference paper titled 'The JRA-25 Reanalysis' is aviailable from
http://www.jstage.jst.go.jp/article/jmsj/85/3/369/_pdf/-char/ja/

Please forward this mail to researchers who are interested in reanalysis.

Sincerely yours,

Secretariat of JRA-25
CPD/Japan Meteorological Agency

Saturday, August 4, 2007

2004 Precipitation

Precipitation from a couple cases has been posted previously. Below are the monthly mean differences from GPCP for the GEOS 5 2004 January and July validation experiment, compared with JRA25 and NCEP. Qualitatively, the patterns of difference maps seem quite similar among the reanalyses. Though, the GEOS5 US precipitation seems low where the others are high. The GEOS5 tropics are also lower than the other reanalyses, a positive result. It is interesting to note that all the reanalyses underestimate GPCP in Europe in January. GPCP applies a snow undercatch correction, especially in Europe, which increases the precipitation. The GEOS5 southern hemisphere and North Atlantic biases appear quite low compared to the other reanalyses.

To expand on the comparisons, we are also producing Taylor Diagrams (Taylor 2001) for precipitation. The preliminary figures follow the mean difference maps. These show the spatial correlation of the reanalysis compared to the normalized standard deviation. The reference data set is GPCP, so the reanalyses are correlated to GPCP, and the standard deviation is normalized to GPCP. The closer the point is to 1:1, the closer a match to the reference data set. Here, CMAP is also included to show a portion of the uncertainty in the observations. By this metric, GEOS5 is improving the spatial pattern on the monthly precipitation. Spatial resolution of the analysis likely plays a role.

A more detailed examination of these statistics is underway, including more localized regions (such as the NA and SA continent). These relate to the monthly precipitation. The higher frequency precipitation is also being examined, compared to TRMM 3b42 3 hourly 1/4 degree. There is some uncertainty with those observations (satellite swaths are occasionally apparent in the spatial structure), but the comparisons so far have been promising where weather systems are reasonably reproduced.



Monthly mean precipitation differences from GPCP for NCEP R1, R2 , GEOS5 and JRA25. January 2004 (above) July 2004 (below).
Monthly Taylor diagrams for precipitation for the globe, NH, SH and Tropics using GPCP as a reference data set for NCEP R2 , GEOS5, JRA25 and CMAP. January 2004 (above) July 2004 (below). The normalized standard deviations increase with radial distance from the origin. All standard deviations are normalized to GPCP so that a value of 1.0 matches GPCP. Spatial correlation are plotted as the radial lines, so that the 1,1 point is identical to GPCP.

Tuesday, July 31, 2007

Status July 31

Several validation runs have started, and results are only just becoming available.

For 2004, two experiments have begun, one starting in January 2004, and the other starting in July 2004. These will run forward through 6 months to provide a full year of analysis. So far, the first months (January and July) have recently completed, and results for those months should be available soon.

January 2006 was run, completed, and evaluations begun. However, between the time it was started, and its evaluation, it was determined that the initialization of the satellite bias corrections was a problem. A new experiment has begun and should be ready for evaluation soon. This will also serve as a test on our bias initialization procedures, which need to be done each time a new satellite becomes available.

The problems with initialization of the satellite bias corrections became apparent in evaluating the SSMI experiment, discussed in the July 10 status update. The NOAA 10 data began using a bias correction that was determined from a previous experiment, with different physics, and hence, different biases. Large parts of the SSMI experiment are being re-run, including some half degree experiments and data withholding experiments. In addition, new coefficients for the historical NOAA radiances are available. Presently, these are being tested. It is worthwhile to mention that the science in the system has been frozen, and these tests are more related to the input data for the 1980s.

File Spec: Thanks to those who have taken time to look at the document! Reviews and intercomparisons of the doc with the output data have identified some inconsistencies. The developers are well into resolving the discrepancies. A new file spec document should be posted soon.

The initial conditions and input data for spin up runs for the MERRA production streams are being developed. When operations personnel get some time, these will be started so as to not lose time while we validate. This expects a favorable result from the validation of the system. The spin-up and steams for MERRA production will be posted separately.

AMS abstracts are due August 10, and the WCRP reanalysis conference abstracts are due today. There will be several GMAO presentations at each of these on MERRA validation.

Thursday, July 19, 2007

Pressure levels intersecting the surface

Many modern atmospheric numerical models use terrain following vertical coordinates, meaning that the pressure of the lowest model level tracks the topography and does not intersect the surface. ERA40 and NCEP reanalyses have produced pressure level data extrapolated downward beneath the Earth’s surface. The result is that for 850, 925 and 1000 mb levels etc, continuous grids are available. Previous versions of GEOS models and assimilation systems have not extrapolated data beneath the surface, favoring to provide undefined values when the surface pressure is lower than a given pressure level.

For instantaneous analyses, comparing GEOS5 pressure levels to other reanalyses would be straight forward, once the undefined value is considered. However, monthly averages pose a problem. There are some regions and pressure levels where the number of valid values may be available for a fraction of the times. If all valid values of GEOS5 are averaged and reported, the average would not be representative or comparable to NCEP or ERA40 reanalyses which made averages of all times.

Figure 1 850 mb temperature RMS error between GOES5 and NCEP analyses for different criteria of the sampling of missing data in the GEOS5 time series. At the left of the graphs, lower criteria allow undersampling of the monthly time series to be compared with NCEP complete monthly mean. Far right, rejects points that have missing data in the time series, so there are fewer data points, but the comparisons to NCEP are more completely sampled. (Click figure to enlarge)

This can lead to an increase in the squared error and systematic bias between GEOS5 and other reanalyses because of the temporal sampling at the edges of topography. This is also noticeable in global and regional map comparisons. We computed global monthly averages testing a range of criteria for rejecting a monthly average. The criteria are applied at each grid point and are based on the percentage of valid data over the month. In Figure 1, on the far left, if data are valid only 1% of the time during a month, a valid monthly mean value is saved. Moving right, at 20%, a grid point with valid data 20% of the month produce a monthly mean (fewer than 20% are reported as undefined). At the farthest right, the strictest criteria requires that for each gridbox to produce a monthly average much have gridpoints that have valid data 100% of the time. The two figures are global land only and North America (20-70, -170--60). At higher pressure, there are more points affected by sub-sampling, and the errors are most noticeable in these large area averages. For higher altitudes, the large scale error drops slowly for criteria greater than 20% (more points valid 100% of the time).
Figure 2 Comparison between GEOS5 and NCEP for different criteria, and a map of the sampling percentage. At 20% criteria (data is valid only 20% of the month) large differences are apparent. These are reduced at 80%. At 100% the data should be showing only differences between full monthly averages, no effect of sampling. There are some artifacts because these figures have interpolated NCEP to the GEOS5 ½ degree resolution. Differences near topography can be significant and misleading (to one not knowing about the character of the data). (Click figure to enlarge)

To address this issue in the monthly mean MERRA products, only means which include counts that exceed a threshold of 20% valid data are included in the mean. Otherwise, the monthly mean value is reported as undefined. This low value is defined to provide as much information as possible. The monthly mean 3D pressure files will also include a variable that counts the valid data at each pressure level. The data user can then screen data to suit their needs. This can also be used to screen other data sets for comparison purposes, and also zonal averaging.

One difficulty that may arise is the lack of a 1000-500 mb thickness diagnostic. This was produced in some previous versions of GEOS5. However, in revising the pressure level interpolation code for MERRA, the calculation of 1000 mb height has been left out, and so, 1000-500 mb height is not available. Also, consider that the 1000 mb analyses will have undefined data over large areas of the globe (land and ocean). Lowest model level data are also available that may be suitable for some purposes, instead of the 1000mb level.

Tuesday, July 10, 2007

Status July 10

The January 2006 validation experiment is underway. This period has been rerun enough in recent weeks that much of the preparatory work had already been done. The primary validation experiment for 2004, is still being prepared. Scripts, code and data need to be in place and working together, soon was the latest update.

There has been much work on the mid- to late- 80s experiment (referred to as the SSMI experiment). Originally, this was established to investigate the impact of the availability of SSMI observations in July 1987 on the time series. The experiment was initialized in Mid-Dec 1983, and run almost through 1990. The spatial resolution is coarse (2 x 2.5), and the version of the system is slightly behind the expected version for MERRA validation (a subsequent test shows that the physics/statistics difference do not change the main results of the SSMI experiment).

The main points to be discussed on the SSMI experiment are:

1. Global time mean precipitation bias
2. Water cycle time series
3. Impact of SSMI

1.) As stated on the previous post, global time mean precipitation, where the SSMI experiment (version b10p9 read - beta 10 patch 9) is 2.2 mm/day compared to 2.6 mm/day for GPCP and ~3mm/day for JRA25 and ERA40. An experiment at the full resolution and fallback MERRA version of GEOS5 (b10p14) shows that the precipitation to be ~0.2mm/day higher than the coarse resolution experiment. Most of this increase is a result of the spatial resolution. Much of the difference in the precipitation is over the tropical oceans where reanalyses are typically much to high. As the system stands now, global mean precipitation is lower than GPCP in the mid to late 80s.

2.) The time series of the SSMI experiment also showed some features that are currently being investigated more closely. Figure 1 shows the time series of precipitation anomalies (mean annual cycle from 1984-1987 removed) for the GEOS5 SSMI experiment, JRA25, ERA40 and GPCP.
Figure 1 (click on the figure to expand it)

Both GEOS5 and ERA40 show a decreasing tendency in the precipitation from 1984 throught the end of 1985. In Jan 1986, ERA40 tendencies reverse and start increasing. In Nov 1986, GEOS5 drops sharply, but stabilizes. NOAA10 data begins in Nov 1986, SSMI begins in Jul1987, NOAA11 begins in Jan 1989. The two issues being investigated are the sudden downward jump of precipitation with NOAA10, and the slight downward tendency early in the experiment. Also the introduction of SSMI is noticeable in the ocean only average (and increase for GEOS5, and a decrease for JRA25).

Figure 2

Figure 2 shows the time series of GEOS5 monthly analysis increments of water vapor (the incremental analysis updates that drive the diagnostic ouput, such as precipitation) at 4 levels in the lower troposphere for latitudes 60S-60N. At the lowest model level (not shown) the shipborne observations of moisture lead to positive increments almost every where and when. Above the surface layer, the lower tropospheric analysis is largely negative increments, acting to dry the atmosphere. The negative increments appear correlated to the precipitation anomalies, though there is also a period between Jul1985 and Jan1988 where TPW increases (see Figure 3). A large jump in the increments and precipitation (Figure 1) is associated with the introduction of NOAA10 (and shutdown of NOAA6 MSU). These are the focus of some ongoing evaluations.

First, an experiment with the latest MERRA system has been run over the start of NOAA10, and it shows less sensitivity in the increments than Figure 2. However, the precipitation in these experiments is similar. To test strictly the sensitivity of the system to NOAA10, a new experiment is being run forward but constraining the NOAA10 observations to passive mode (an analysis is made, but the increments will not contribute back to the system). Secondly, NOAA NESDIS is generating new coefficients for the historical periods polar orbiting satellites (to which we are grateful). It is not clear at this point what the impact of that will be, but will be thoroughly tested prior to MERRA production. Lastly, bias corrections are being made in the system (e.g. for view angle). The procedures for initializing and carrying these bias corrections are being reviewed. This is one possible source of error, but it is not yet clear a problem exists.

In summary for point 2, it seems we have some sensitivity to the observing system (regarding precipitation) on the same order of magnitude as previous reanalyses in the 80s. We are using this opportunity to flesh out an problems in the system that may be exacerbating the discontinuity of the analysis during observing system changes. These tests are intended as checks on the system before production, though, we expect that there will be noticiable changes in the MERRA time series as a function of the observing system.

Figure 3 Monthly anomalies from the Jan84-Dec87 mean annual cycle.

3.) The impact of SSMI was not immediately apparent in GEOS5 global precipitation (Figure 1), and ocean only average precipitation increases. The SSMI impact on evaporation and surface wind speed over the ocean is also apparent (Figure 3), likely related to the SSMI wind speed. The GEOS5 anomalies and JRA25 seem to be tracking really closely. In the mean, JRA is ~0.2m/s higher wind speed than GEOS5 (of 4.4m/s). So, it seems that in some ways GEOS5 is sensitive to SSMI, like JRA. However, in ocean precipitation, GEOS5 is more sensitive to the NOAA transitions than JRA (Figure 1).

Just a caveat regarding these results. In comparisons of the 2 degree resolution with 1/2 degree, the monthly precipitation increases at finer resolution, especially in mid-latitudes. We have not yet run multiple years of the 1/2 degree system, and don't have a grasp on the interannual variations. It should be interesting to see how similar the coarse and fine resolution analyses are over long periods.


Figure 4 Zonal time series of precipitation anomalies (after ENSO removal) for JRA and the GEOS5 SSMI experiment. The impact of SSMI on JRA is apparent in the southern hemisphere (30-60S). GEOS5 low frequency decreasing precipitation is focused in the tropics.

Thursday, June 28, 2007

Status June 28

Since the last post, work has progressed on setting up the validation experiments. Initial conditions are being created and assorted other tasks to make sure the system is configured. The jobs should be close to starting. However, there are some ongoing experiments that may affect the setup. First, the data group has applied some corrections to the sonde data for known biases (e.g. radiation). I don't have the citations handy but I'll try to get them posted before long. With the corrections applied to the data, the quality control code needs to be modified to reflect that. The experiment is just a sanity check to make sure that the data and code are implemented properly.

While investigating the "sawtooth" problem discussed in earlier posts, the analysis group looked more closely at some of the statistics for the temperature assimilation. A feature in the lowest atmospheric layers of the statistics may partly contribute to the sawtooth, so they are testing a correction to the statistics. Preliminary results seem positive, and a full resolution was being started. This correction would be much more broad than the target correction already applied to fix the sawtooth. So, we will need to look closely at the experiment before it is implemented for validation.

In parallel, the code is being engineered to work on a newer supercomputer at GSFC called Discover. This will be where MERRA production is carried out. The system is running well there. However, occasional random crashes are being investigated. A likely cause is a memory leak. These things happen, and the folks working on the problem should have it fixed without impact on production.

The SSMI 2 degree experiment is nearly complete (Jan 1984 - Dec 1989). The main purpose was to checkout the potential impact of the introduction of SSMI (radiances and surface ocean winds) to the data stream. Several people are still looking at it. The preliminary results show that the SSMI does not seem to have considerable impact on atmospheric moisture and precipitation. However, it is plainly apparent in the global time series of surface wind, moisture and evaporation (all increasing). There is a slight upward trend in global precipitation after the introduction of SSMI, but it does not exhibit a a shock or step (to the eye, statistical analysis would probably reflect the introduction).

One feature that this experiment has demonstrated is that the operational NOAA satellites, and their variations do seem to be influencing the precipitation time series. Firstly, the global mean precipitation (averaged for Jan 1984 - Dec1987) is lower than any experiment we have run to this point, 2.2 mm/day compared to 2.6 mm/day for GPCP and ~3mm/day for JRA25 and ERA40. Resolution might be part of the deficit. In some recent tests, we have not seen a large change in precipitation going from 1/2 to 2 degrees. However, some previous experiments did show ~0.2mm/day global deficit in the 2 degree system compared to 1/2. So, we are putting a new experiment into the validation experiments, July 1986. This will run at 1/2 degree with the latest system, and be included in the validation experiments.

Another issue from the SSMI experiment is that the precipitation bias is not uniform across the period of analysis. In 1984, precipitation is quite close to GPCP at2.4 mm/day, but it has a downward drift over the next few years, from early 1985 (NOAA 9 starts in Dec 1984) until Jan 1987 when NOAA 10 begins. The analysis increments are negative (removing water vapor) and getting more in time. There are jumps in the time series of moisture increments when the NOAA satellites change. We still have a bit of work to do on this, as well as check the 1/2 degree experiment. A more detailed evaluation is being prepared, and we have much more to investigate.

The primary validation experiments will be Jan and Jul 2004 (expected to run for six months each to make a full annual cycle). The Jul 1986 is also being put on a fast track. They should start soon, and more info will be posted then. The secondary runs, Jan/Jul 2001 and Jan 2006 will begin when CPUs become available.

Saturday, June 16, 2007

Sawtooth Update June 15

The experiment testing the impact of not using AMSUA for land snow conditions progressed several more days overnight. The results show that a weaker version of the Sawtooth shows up as we get deeper into winter, even without AMSUA. AIRS radiances are still being assimilated. Further analysis shows that in Siberia (where the Sawtooth is worst) the synoptic conditions change, decreasing clouds, and the Sawtooth simultaneously appears. The figure below shows the original Sawtooth (b5_b10p9), the latest experiment restricting AMSUA (exp06g) and NCEP Operational analysis for Jan 1-11, 2006. It is important to note that the weakened Sawtooth in exp06g somewhat compares to a semi-diurnal cycle in the NCEP operational analysis (between Jan 9-11, should have been still using the SSI analysis).

The working hypothesis for why the weakened sawtooth appears is that AIRS is still prevalent in this region. Some are looking into the radiative transfer model. Also, we are not using the latest NCEP CRTM. There was some discussion on updating the CRTM to the latest, however, it quickly determined that this would require a complete update of the GSI in the GEOS5 system. Of course, that would incur significant delays.

For now, we are going to continue to monitor the experiment. But the major portion of the problem seems to be under control. Since theGEOS5 is in range of NCEP operational analyses, plans for validation will go forward. In validation we will examine further whether this feature adversely affects the science quality of the data set. The good news would be that the problem has finite bounds in, only over land snow, and only occur then when AIRS is available (starting 2002).

Just a quick note and reminder of the positive aspects of this system we have previously reported (we tend to hammer at the problems). The precipitation from the configuration for MERRA seems to be well in line with GPCP, and a good contribution compared to other reanalyses. TPW looks reasonable compared to NVAP data (though those obs may have bias). There will be data supporting the stratosphere community that should be very unique. At one-hourly intervals, the surface diagnostics should be very useful for diurnal cycle studies. These and a few other aspects should help make MERRA a reasonable contribution.

Thursday, June 14, 2007

Status June 14

Just a brief update, as our regular meeting was relatively brief and focused.

Progress has been made on the sawtooth problem. The results presented today show that rejecting all AMSUA radiance over land snow conditions eliminates the sawtooth. There are some experiments ongoing to test which channels (surface peak and high peak) might be included without the instigating a sawtooth. Also, there is an experiment with all AMSUA, but with much stricter quality control. We should have the results from those experiments in a couple working days. However, the fallback position would be to not use AMSUA over snow.

During these tests, some other minor bugs and updates to the system have been found and fixed. An experiment pulling the system all together is being configured and run over the weekend. That should be evaluated next week along with the AMSUA quality checks. If all goes as expected there, validation experiments will begin soon thereafter.

The 2 degree SSMI experiment is ongoing, and producing a month every 14 hours or so. Presently it is at 16 February 1988. While we're monitoring some of the data as it progresses, some analysis of SSMI impact will be done when we have 1 year after SSMI starts (probably try to get 2 years after, eventually).

Wednesday, June 13, 2007

Land Fractions

One feature in the GEOS5 GCM that is different previous reanalyses is fractional surface tiles, as opposed to a land/sea mask. Each grid box contains a fraction of land, water, lake or land ice. The fractions are based on the 1km Global Land Cover Characteristics (GLCC) database. Along coast lines, the land processes are blended with sea surface processes. Inland, lakes and rivers are also included. This will require some attention of users calculating precise budgets. For example, the grid box average evaporation that would be appropriate for an atmospheric budget, may not be appropriate for in-situ data comparisons and land surface budget studies, if the lake fraction is substantial. Figures showing some examples of land and lake fractions are shown below.

To support the analysis of land water and energy budgets, a land only output data collection will be produced (tavg1_2d_lnd_Nx in the MERRA File Specification Document). These grids would be for the land only fraction of the surface. Users will be able to differentiate land evaporation from total evaporation in grid boxes near the coast and near inland water bodies. Soil water (GWET variables) will show up in grid boxes that appear to be oceanic. The model does provide an integer land/ocean mask (variable name LWI). However, this is simply a 50% cutoff between land/land ice and ocean fractions. For some specific purposes, this may not be appropriate, and users should use the land fraction to develop their own mask.

Additionally, the land fraction is subset into tiles based on the Catchment hydrology. The land collection includes data from the catchments. For more information, the interested reader should review Koster et al. (2001, J. Geophys. Res. Vol. 105 , No. D20 , p. 24,809, 2000JD900327).

Wednesday, June 6, 2007

Status June 5

We've had some good news and bad news. The good news is that efforts by the analysis group to revise the error statistics have had the expected positive impact on a systematic zonal wind bias in the tropics at upper levels. Further, the adjustments in the convective parameterizations by the model group have held up through this testing, and precipitation looks be reasonable compared to other reanalyses and observation data sets at the monthly time scales.

The bad news is that the "saw tooth" problem persists, but progress is being made in understanding it. In one of the data sweeper runs, we find that the analysis is reasonable, through Dec 2000, then the sawtooth begins in January 2001. This is at the same time NOAA 16 data begins to be used in the input data stream. Presently, the satellite bias corrections used in the system are being reexamined. The current thinking is that, if an inappropriate bias is included, the analysis rejects the radiance observations (near the surface), then the bias is never properly adjusted. To test this, an experiment started with zero bias is being started, this should spin up a new bias correction. Most of the system testing has simply carried bias corrections along from older systems.

So, while the system is getting closer and we are certainly learning more on the system with every experiment, the important validation experiments have not yet begun. Some minor corrections have also been identified. For example, it was found that snow was reaching the surface in regions too warm and equatorward. The snow production algorithm was not including surface temperatures, so now a check has been added.

The 2 degree experiment to test SSMI in the late eighties has progressed into 1987, and is closing on the onset of SSMI in July. However, the experiment is being removed from the priority queues. The GMAO is gearing up for operational support for TC4, which needs high priority, and also the experimentation on the sawtooth problem takes priority.

To emphasize the critical nature of the sawtooth problem, the figure below shows an example of how it looks at the worst. The time series is 850 mb temperature at a point every 6 hour analysis time. The points closest to the NCEP operational analysis are the when sondes are present (00Z and 12Z) and the very warm temperatures are when only satellite data are available. At this point, the GEOS5 surface temperature is much colder than NCEP and the surface channel radiances are being rejected by the quality controls.

June 11 Follow up: Recent experiments have shown that the AMSU channels over ice may be the root of the sawtooth problem. Further work is aimed at identifying specific channels. Result should be posted in the next status report, probably Friday June 15.

Monday, May 28, 2007

Latest on the 3rd WCRP Reanalysis Conf.


A recent Message from JRA-25 User Admin:

Dear JRA-25 reanalysis data users,

I am pleased to inform you that registration of the
Third WCRP International Conference on Reanalysis will
begin on 1 June 2007.

Please visit the conference page
(http://jra.kishou.go.jp/3rac_en.html)
and move to Instruction page.
Please note the instruction page will be available
from 1 June.
The deadline for submission of abstracts is 31 July 2007.

WCRP, GCOS and GEO may support travel costs for
participants from developing countries.

Please inform researchers who are interested in the conference of the conference.




Sincerely yours,

Secretariat of 3rd Reanalysis Conference
CPD/Japan Meteorological Agency

Saturday, May 26, 2007

Status May 25

I'm on my way out to the LandFlux Workshop, so this will be a quick post.

The "sawtooth" problem first discussed in the May 18 status update is still being worked on. The bias correction that seemed like it might be a solution did not fix the problem. A short 1/2 degree experiment with limiting very strong stability in NH winter did not have an immediate response, though, GCM seasonal simulations show that this should make a big step to correct the cold bias thought to be the root of the saw tooth. This problem is also found in 2 degree experiments, so that new seasonal 2 degree experimetns are being started to assess our understanding of the problem and show that correcting the surface temperature bias corrects the sawtooth. 2 degree runs turn around much more easily than 1/2 so results should be available next week.

Unfortunately, being on travel, I'll miss the meetings that review the results, but I'll spin back up as soon as possible.

The SSMI experiment is moving forward. It is in late 1985, looking forward to July 1987 when SSMI becomes available. There should be alot to work with when I get back from Landflux (June 4).

In a general sense, we (a concensus of the GEOS5 status meetings) are very pleased with where the systems is, and were it not for the sawtooth would be moving on with the validation experiments. There is one other science issue that has come up. The Relocator (corrects the location and dynamics of tropical cyclones) was run last summer for MAP06. Somewhere along the way, the numerous updates and changes have "broken" the Relocator code. It shouldn't be difficult to fix, but will need to be tested before we go to validation.

As far as the schedule goes, a critical milepost will be the start of the validation experiments. Once validation experiments start, then we should have good ideas on the beginning of production.

Friday, May 18, 2007

Upcoming Runs and Plans

An update to the file spec was posted on the MERRA WWW page this week. No major changes, but it is still a draft.

We fully expect to have the system ready for production by mid-July. As such we are gearing up for the validation runs. The reanalysis system is also being configured for the new Discover super computing system. This provides a significant performance increase. A new partition for MERRA is being prepared.

The runs will be:

Jan/Jul 2001, Pre-AIRS period where we have legacy experiments, including the GEOS4 validation

Jan-Apr and Jul-Oct 2004: We want a full annual cycle, but that will take a substantial amount of time to run. As a shortcut, we'll run two experiments that will provide the central month of the four seasons, so most focus will be on Jan, Jul then Apr, Oct. This should show any red flags that might require us to hold production. However, we plan to keep the Jan 2004 run going beyond April, as computing cycles permit, with the idea that we will get a full annual cycle (the start of production will not be held up for that to finish, but we will evaluate it nonetheless).

Jan 2006: We have been using this case for several experiments so far, and have some familiarity with it.

We have been formulating a list of comparisons and issues to consider in validation. I'll post that separately in a couple days.

Status May 18

The SSMI 2 degree test runs is back up and running. It was started on Jan 1 1984 and completed 1984 yesterday. It will be a couple weeks before it catches up to the first SSMI observations. In the mean time we are scanning it's precipitation, temperature and heights. Most look reasonable. However, we know from previous experiments that the GEOS5 precipitation is much improved at 1/2 degrees resolution.

The system put toward validation is being finalized. We are checking and rechecking the implementation of a correction to the observational error statistics. Also, the system to be used in validation should have the entire file spec being produced. One component still being tested is the grid resolution reduction. The most voluminous data are the 3-D tendencies. I think just about every physical tendency has been included in the file spec. The 72 layer 1/2 degree data would be far too enormous for anyone to use, ( and uncomfortable to store). So, we are reducing the tendencies to 1.25x1.25 degrees and 42 pressure levels. The resolution reduction is being build inline to conserve the time doing I/O.

Recently, we found an oscillation between 6 hour analyses, with and without the radiosondes. The problem is local to high winter latitudes over land. It appears that the model has a cold surface bias, and the surface radiances are being rejected. Mid troposphere radiances still think the atmosphere is cold, and try to warm it up, but in the next analysis, the radiosondes draw the temperatures back. So, it seems to be a disagreement between a model bias (at high winter land latitudes) and differences in the radiosondes and radiances. Here, this has become known as the "sawtooth" problem, as the 500 mb time series shows a 6 hour zigzag pattern.

One possible source of the model bias is that there is no limit on the stability of the surface layer. So, Richardson numbers are very high in these conditions. In free running model, it doesn't appear to be a problem, but it decouples the land and atmosphere when observations are included. So, a limit on the Richardson number is being tested. There is also a strict criteria that limits the surface radiance increments. So that when the forecast and observation differences are large, the increment is rejected. This is strict for very good reasons (cloud clearing problems, for example). We are cautiously looking into relaxing that restriction. However, this could be risky and would only go forward if remarkable improvements can be obtained.

We also had some good news in this respect. The analysis group turned on the bias correction routines for surface temperature in GSI. This will apply a bias correction to the model background, before the surface temperatures are analyzed. So, the GSI sees a better surface temperature. These bias corrected temperaturesdo not feed back to the model forecasts (a much more difficult problem). After 15 days, the sawtooth has been corrected in many regions. East Siberia is spinning up to this a bit slower. In any event, it seems to be a reasonable fix as well.

We had not seen a problem in the system like this until recently, when running Jan 2006 as a test case. The previous winter case that was examined in detail was Jan 2001. In Jan 2006, we also include AIRS radiances. While these are thinned, AIRS still contributes 500K observations of the 1.2M observations assimilated. Away from the AIRS period, we still see the problem when we are looking for it, but it is reduced. We can also see this periodically in other operational analyses, but not nearly the degree seen in our Jan 2006 experiment.

Wednesday, May 9, 2007

Status May 10

We are still rebuilding the system with the latest patches and output diagnostic code. Part of this is also building on the Discover systems. It is taking time. The next round of systems tests should be the last before validation experiments. The 80's SSMI Impact tests will be run with the latest build, as well. For a timeline, Th May 3, decisions on the configuration of the system were made. The system build was completed Tue May 8. Computer systems were down May 9 for scheduled maintenance. They system is built and the jobs may start by the end of the week (May 11).

We met again today (May 9) with the GES DISC personnel working on the MERRA data distribution system. It is called MDISC, where M is for modeling (not MERRA specific). There will be several portals to the data for access and visualization. FTP will likely be the most general route people take, but there is also a subsetting routine with the FTP with a graphical interface that looks very good so far. I'll do some testing on that in the next couple days.

The working schedule for getting MERRA production going is not earlier than July 1. There are a lot of small items on the checklist for production, as well as a couple major gateways. The major gateways are: first we need a systems test that holds up in some short (one month) experiments, second take that system to run validation experiments, then analyze the results of validation and present those to both HQ and our User review group. The results of the review should indicate that we can go forward with productions. All told, it feels pretty ambitious at this point. Regardless, we are checking off the items needed to begin production.

Monday, May 7, 2007

Reanalysis paper by L. Bengtsson

An interesting paper led by Lennart Bengtsson is in this month's BAMS on the limitations, need and future role of reanalyses in climate studies. Found at AMS Online.

SSMI test delayed

The test of the MERRA system during the mid-1980s, enveloping the initiation of SSMI observations, had been running with a critical bug in the system. It was caught after the third of 5 planned years of running. The bug was a known problem in an old tag, but that tag was not adequately updated in Operations, despite being fixed in the research and development systems for months.

The purpose of this experiment is to test the impact of a significant change in the observing system, namely SSMI, on the GEOS5 time series. The system is running at a very coarse 2x2.5 degrees, 3 hourly output, which allows for much quicker throughput than we will see with the full half degree system. While the configuration is close to, but not exactly, that of the MERRA validation experiments, it should be close enough to contribute to the system validation. It will assess how sensitive the system is to the SSMI data when it first becomes available.

Thursday, May 3, 2007

MERRA Status

At the development meeting today, we reviewed proposed fixes for the GSI statistics. This effort addresses a bias in the upper tropospheric zonal wind. These have been tested at coarse resolution. One fix has proved promising and the system is being rebuilt. This will be tested at the full MERRA spatial resolution. This also impacts certain aspects of the convective parameterization. So, a set of tests are being prepared and executed over the upcoming weekend. Early (but possibly incomplete) results of the tests are expected by Thursday May 10.

These tests will be used to decide the final configuration of the GEOS5 system to execute the MERRA reanalysis. Full system validation will begin with this configuration.

Also, Queen Elizabeth II will visit Goddard, May 8.

Wednesday, May 2, 2007

File Specification Document

The MERRA File Specification Document describes every aspect of the output data files that users will need to use the data. The file formats, grid structures, frequency and variables are all included. There are also some brief discussions on the physics and analysis.

The file can be found at: http://gmao.gsfc.nasa.gov/research/merra/MERRA_FileSpec_DRAFT_4-10-2007.doc

Full documentation of the GCM science and analysis methods are in development.

Thursday, April 19, 2007

MERRA Current Status

4/19/2007

The science aspect of the MERRA is frozen. There are some final settings in the system that are awaiting a final decision. Validation of the whole system should be under way in 1-2 weeks.

This is an example of the precipitation being produced in the system.

The figure shows the difference of monthly reanalysis precipitation from GPCP. The numbers over each panel show the mean global difference (near zero is good) and the global standard deviation of the difference (smaller is better).

Wednesday, April 18, 2007

MERRA Introduction

NASA's Modern Era Retrospective-analysis for Research and Applications (MERRA) is intended to be a full reanalysis of the satellite era. In general, the data provided will support science research and applications.

The official MERRA WWW site is: http://gmao.gsfc.nasa.gov/merra/

This Blog is intended to be an interface between the developers and user community, and a place for status updates and Q&A regarding the MERRA production and data.