1. Turn off assimilation of surface temperature data over land in the GSI analysis
2. Turn off MSU satellite data
3. Tighten the gross error checks for all conventional data except winds.
4. Increase iteration number from 50 -> 75
5. +/- 15 minute window for all surface obs (was +/- 1.5 hours)
6. Begin assimilation of NEXRAD Level 2.5 radial wind data in the GSI analysis
7. The amplitude part of the background error was retuned to improve treatment of winds (reduced penalty function)

1. Changed the computation of the difference between the 24h sum (12Z-12Z) of edas precip input and the daily gauge analysis in the pcpbudget code to inflate the daily gauge analysis data by 10% to account for under-catch of precip before the difference is computed. Previous version of the code for the WRF-NMM NDASX had no inflation of the analysed daily gauge data, while the ops NDAS has a 10% inflation.
2. Revert back to pre-3/17 vertical background errors for humidity in the GSI analysis.
3. Begin use of GPS-IPW data in GSI analysis

E = 1-.exp(-0.75*D),

as defined over a contiguous cloud layer was changed to

E=1.-exp(-1.66*1.5*D)

The 1.66 factor agrees with the equation in the WRF RRTM, while the extra factor of 1.5 accounts for the larger absorption coefficients used in NCAR's RRTM to convert from cloud mixing to optical cloud depth.

1. Begin use of latest (March 2006) GSI release
2. Use NMM-based background errors computed using an ensemble/Monte Carlo approach

1. Changed the partial cycling code in NAMY (on 00z 3/4) so that it wrote out the z0 computed by SI/Real in the GDAS-based wrfinput file, instead of using z0 from the best available cycled NDAS WRF-NMM restart file. Since the NMM logical variable RESTRT=FALSE if partial cycling is invoked, the model was adding 1/2 of the vegetation z0 component to z0 every time the NDAS cycle was coldstarted. This led to z0 values 3-4x too big over a large portion of the NAM domain.
2. Since NCO does not have the partial cycling capability working yet, the only way to get this fix in the NAMX parallel on white is to restart the NCO parallel from a wrf restart file from NAMY with the correct roughness lengths.

1. Parallel NDAS on white was restarted from the EMC NDASX on blue at 18z 2/27
2. NAMX run on white will be done 4x/day

1. Turn on tropospheric AMSUA radiance channels
2. Updated NMM-based background error covariances

1. doubled snow thermal conductivity
2. placed an upper limit of 0.98 on the snow cover fraction

1. Reduction of ZT over elevated terrain is only active in stable surface layers.
2. A baseline lower CZET-like value of 3.0 was imposed, regardless of surface elevation
3. PBL height computation changed to make it consistent with MYJPBL code and post

1. The array SH2O was inproperly dimensioned in its declaration in subroutine TURBL. SH2O is a 3-D state array but in TURBL it is dimensioned as only 2-D. It is passed from solve_nmm through TURBL and the surface driver into the land surface package. SH2O is properly dimensioned in both the surface driver and the LSM package.
2. Variable DZ was incorrectly calculated in subroutine CUCNVC. Two 1-D local working arrays, namely T(K) and Q(K), were being used to generate DZ(I,K,J) but T and Q are never being updated for each different I,J. DZ is subsequently passed into the BMJ code where it is used to calculate the layer pressure thicknesses.

1. Fixed an error in subroutine SFCDIF in which the first guess z0T is used to compute z+z0T instead of the converged value of z0T obtained iteratively.

1. Longwave temperature tendencies from the lowest two model layers are averaged together in order to address concerns that the layers may be too thin for the longwave scheme to provide accurate heating and cooling rates.
2. Because the GFDL longwave (LW) radiation calculates heating rates and net radiative fluxes (rather than separately calculating the upward and downward components), the downward (incoming) LW is calculated as the sum of the net and the upward LW at the surface. The upward LW at the surface was based on the ground (skin) temperature, but now this has been changed to be the average of the skin and lowest model layer temperatures. The objective is to slow the rate that downward LW decreases when stable PBL conditions form in clear skies overnight.
3. Cloud optical depths were reduced by the square of the cloud fraction in order to increase solar transmission through clouds, but now they are reduced by a factor proportional to the cloud fraction (as is the case for the NAM).

1. Turn off multi-variate correlation between temperature and humidity above sigma level =0.15
2. Install small bug fixes from official GSI directory
3. Restart cycle from ops GFS atmospheric states and ops NAM land states

1. Turned off AMSUA channels 1-5 and 15 for NOAA 15, 16 and 18. These channels, which influence the temperature near the surface, were causing analysis increments to be in the opposite direction from the raob increments in Northern Canada (sfc emissivity issues over snow).
2. Increased the thinning of satellite data (currently about one data point in each grid box). The large amount of satellite data led to the raobs being effectively thrown out in certain regions.

1. Model top pressure (NAMXF = 50 mb, NAMXF2=2 mb)
2. The GSI in NAMXF uses NMM-based background errors, while NAMXF2 uses GFS-based background errors for now.