CHANGE LOG: Real-time, experiment HRRR Data-Assimilation System (HRRRDAS) and Ensemble forecast (HRRRE) on Jet Recommended periods for analysis, when system configurations were stable: ------------------------------------------------------------------------- (1) 2017-04-14 to 2017-06-04 (2) 2017-07-22 to 2017-08-17 (3) 2017-08-19 to 2017-10-30 (4) 2017-11-04 to 2018-02-04 (5) 2018-03-29 to 2018-09-05: A long dataset was produced with a relatively fixed HRRRDAS and HRRRE configuration. However, the analyses and forecasts during this period had a significant daytime cold bias at low levels, resulting in a low bias in daytime deep, moist convection. The cold bias was attributed to excessive clouds. This problem was largely eliminated starting on September 6, when GSI radar- and satellite-based cloud and hydrometeor clearing was added to the hourly analysis. (6) 2018-09-06 to 2018-12-25 (7) 2019-01-28 to 2019-02-08 Real-time runs from mid February to late May 2019 were not stable and at times contained significant errors. For this period, retrospective runs are available for the VORTEX-SE IOPs and for the Spring Forecast Experiment (April 28-May 31). (8) 2019-06-12 to 2019-08-13 (9) 2019-09-25 to 2020-03-18 (10) 2020-03-20 to 2020-04-16 (11) 2020-05-04 to 2020-08-04 (12) 2020-08-06 to 2020-10-31 ======================================================================== 2020-08-05 HRRRDAS: LSM perturbations turned off in initial conditions, to match HRRRDAS configuration on WCOSS. ======================================================================== 2020-06-03 HRRRE: “Member 0” added to forecast, which is an unperturbed member initialized from the ensemble mean. ======================================================================== 2020-05-04 HRRRE: Stochastically perturbed parameterization tendencies (SPPT) turned back on, with reduced magnitude. ======================================================================== 2020-04-24 HRRRE: Stochastically perturbed parameterization tendencies (SPPT) turned off owing to unphysical wind anomalies. ======================================================================== 2020-04-17 HRRRE: Stochastically perturbed parameterization tendencies (SPPT) turned on. The forecasts also continue to use SPP. ======================================================================== 2020-04-08 HRRRE: HAILCAST output turned on. ======================================================================== 2020-03-19 HRRRDAS and HRRRE: Change to latest HRRRv4 version of WRF 3.9, which includes some minor bug fixes. ======================================================================== 2019-09-24 (2100 UTC) HRRRDAS: "Rolling" ensemble design turned on. 18 of the ensemble members (odd-numbered members) are initialized at 0900 UTC, and the other 18 (even-numbered members) are initialized at 2100 UTC. Each member is cycled for 24 hours before being replaced. (Previously, all 36 members were initialized at 0900 UTC.) ======================================================================== 2019-09-24 (0900 UTC) HRRRDAS: Source of IC perturbations changed from GDASv14 to GDASv15. ======================================================================== 2019-09-06 (1700 UTC) HRRRDAS: Turned off hourly, non-variational cloud building in ensemble members. Although it was producing a desirable increase in low-level water vapor, it was also causing spurious convection at short lead times. ======================================================================== 2019-08-23 (0900 UTC) HRRRDAS: Atmospheric initial conditions (ensemble mean) now come from RAPX at 0900 UTC. Previously, HRRRDAS was initialized from GFS at 0300 UTC. ======================================================================== 2019-08-20 (2100 UTC) HRRRDAS: Turned on hourly, non-variational cloud building in ensemble members. The method is like that in HRRRX. ======================================================================== 2019-08-14 (0400 UTC) HRRRDAS: Discontinued EnKF assimilation of METAR-based qv pseudo obs, which had been used for cloud building. ======================================================================== 2019-06-12 HRRRDAS and HRRRE: Following a long development period, stable runs begin of a new version of HRRRDAS and HRRRE. Significant changes include (1) full CONUS 3-km domain; (2) WRF 3.9 (HRRRv4 version) including hybrid vertical coordinate, IEVA, and no limit on microphysics temperature tendency; (3) data assimilation with GSI and EnKF instead of GSI and DART; (4) graupel included in list of analysis variables when radar-reflectivity data are assimilated; (5) cloud building with EnKF assimilation of METAR-based qv pseudo obs; (6) posterior inflation on 3-km domain set to 1.08 after the assimilation of conventional observations and 1.08 again after the assimilation of reflectivity observations; (7) SPP applied to PBL, LSM, and microphysics schemes during HRRRE forecasts. ======================================================================== 2019-02-08 to 2019-06-05 HRRRDAS and HRRRE: Development period with numerous changes and debugging. Real-time output from this period has at times significant errors. More recently, retrospective runs have been made for VORTEX-SE IOPs and for the Spring Forecast Experiment (April 28-May 31). ======================================================================== 2019-01-28 HRRRE: Ensemble forecasts resumed. ======================================================================== 2018-12-26 to 2019-01-27 HRRRDAS and HRRRE: Partial government shutdown. HRRRDAS cycling continued on most days. No HRRRE forecasts were made. ======================================================================== 2018-12-12 HRRRDAS: Hourly GSI cloud and hydrometeor clearing in 15-km domain, based on radar and satellite obs, turned on. ======================================================================== 2018-12-10 HRRRDAS: Assimilation of Texas Tech StesoNet data in AL and TN turned on. Surface temperature, humidity, and pressure data are assimilated hourly in the same manner as METAR observations. ======================================================================== 2018-09-06 HRRRDAS: Hourly GSI cloud and hydrometeor clearing in 3-km domain, based on radar and satellite obs, turned back on. This analysis step existed in HRRRDAS in 2017 and early 2018 but was eliminated on 9 March 2018 when the analysis system was simplified. Unfortunately, analyses and forecasts from March 9 to September 5 had excessive clouds, resulting in a daytime cold bias at low levels, and a low bias in daytime convective precipitation. ======================================================================== 2018-08-20 HRRRDAS: Bug fix. Aerosols (QNWFA and QNIFA) are now being initialized as intended from climatological data. ======================================================================== 2018-04-26 HRRRE: 3-km forecast domain at 0000 UTC expanded from 60% CONUS to full CONUS. For this forecast, ensemble members are initialized from a combination of 15-km and 3-km HRRRDAS analyses. ======================================================================== 2018-03-28 HRRRDAS and HRRRE: EnKF update of QRAIN, QSNOW, and QICE turned back on after some recent debugging. Although QGRAUP was analyzed previously by the EnKF in HRRRDAS, it is not included in the current list of analysis variables; this change was made so that the EnKF will run quickly enough in real time. Also, diff_6th_factor in the WRF namelist decreased from 0.25 to 0.1. ======================================================================== 2018-03-09 HRRRDAS and HRRRE: Two-domain configuration introduced. A large 15-km domain allows more observations to be assimilated and also moves the random boundary-condition perturbations far away from the 3-km domain. The inner 3-km domain remains 60% CONUS. Also, GFS 3-h forecasts are now used to initialize the ensemble mean in HRRRDAS. Previously, RAP analyses were used. ======================================================================== 2018-02-28 HRRRDAS: Magnitude of LSM SPP perturbations decreased in HRRRDAS initial conditions. ======================================================================== 2018-02-05, 2018-02-13, 2018-02-15, and 2018-03-07 HRRRDAS: Minor adjustments to GSI observation preprocessing parameters. Also, assimilation of precipitable water (ob type 153) turned on. ======================================================================== 2017-10-31 and 2017-11-03 HRRRDAS and HRRRE: Decreased magnitude of boundary-condition perturbations, to see if precipitation anomalies can be reduced near the lateral boundaries ======================================================================== 2017-08-30 HRRRDAS: Prior ensemble mean used as input to GSI observation preprocessing. Previously, ensemble member 1 was used. ======================================================================== 2017-08-18 HRRRDAS: Cycling changed to the following schedule: initialization at 0300 and cycling to 1200 UTC, and then initialization at 1500 UTC and cycling to 0000 UTC. ======================================================================== 2017-07-22 HRRRDAS and HRRRE: Full-CONUS HRRRDAS and HRRRE turned off. HRRRDAS cycling on the 60% CONUS domain changed to the following schedule: initialization at 0000 and cycling to 0900 UTC, and then initialization at 1200 UTC and cycling to 2100 UTC. ======================================================================== 2017-07-18 HRRRDAS and HRRRE: Completion of increased perturbations. During the past month, changes have been introduced gradually that (1) increase the magnitude of random boundary-condition perturbations and (2) increase the posterior inflation from 1.05 to 1.20 during cycling. ======================================================================== 2017-06-09 HRRRDAS and HRRRE: Domain size increased to full CONUS for the cycling from 2100 to 0000 UTC and the forecast from 0000 UTC. Also, PBL SPP has been turned off for the 0000 UTC forecast. ======================================================================== 2017-05-03 HRRRDAS: Initialization changed to twice per day. Cycling occurs from 0900 to 2000 UTC and then from 2100 UTC to 0000 UTC. A hypothesis being tested is that the HRRRE forecasts at 0000 UTC can be improved by starting the cycling with the more recent large-scale analysis in the RAP at 2100 UTC. ======================================================================== 2017-04-14 HRRRDAS and HRRRE: Regular runs begin, using WRFv3.8 over a 60% CONUS domain. HRRRDAS is initialized daily at 0900 UTC from a recent RAPX analysis or forecast, to which GDAS perturbations are added. HRRRDAS is cycled hourly until 0000 UTC. Soil perturbations are introduced into the HRRRDAS members during the first few timesteps of WRF integration at 0900 UTC. Then, these perturbations evolve through model processes only during the HRRRDAS cycling and HRRRE forecasts. The HRRRE forecasts from 0000 UTC include stochastic physics: SPP applied to the PBL scheme. The HRRRE forecasts initialized at other times do not include stochastic physics. ========================================================================