You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Corrections for tipping bucket and nudging in very long simulations (#2063)
TYPE: bug fix
KEYWORDS: precipitation, tipping bucket, nudging, spectral nudging, analysis nudging, grid nudging, regional climate, dynamical downscaling, downscaling
SOURCE: Tanya Spero (U.S. EPA)
DESCRIPTION OF CHANGES:
Problem:
Several processes in WRF are currently triggered at periodic intervals
(such as reading/writing files and managing certain bookkeeping
processes). Some of those processes had been coded to identify those
time triggers by referencing a variable that contains the time elapsed
since the model simulation was initialized. That variable, XTIME, is a
Fortran single-precision real variable that counts the number of elapsed
minutes since initialization. However, single-precision real numbers
become imprecise (i.e., cannot accurately resolve "whole" numbers) after
they exceed 2^24, which is 16,777,216. In long simulations, that occurs
just before 32 years of simulation period.
Solution:
Two new variables are introduced based on existing variable CURR_SECS2: CURR_SECS2_R8, and double precision version of CURR_SECS2, to address bucket tipping and CURR_MINS2, based on CURR_SECS2_R8, to address issues in nudging code.
LIST OF MODIFIED FILES:
M phys/module_diag_misc.F
M dyn_em/module_first_rk_step_part1.F
M dyn_em/solve_em.F,
M phys/module_fdda_psufddagd.F
M phys/module_fdda_spnudging.F,
M phys/module_fddagd_driver.F
M wrftladj/wrftladj/solve_em_ad.F
TESTS CONDUCTED:
1. Modified code has been tested extensively using a long simulation (using a restart at more than 31 years into the simulation) with WRFv4.5.1 and with shorter (3-day) simulations using WRFv4.6.
2. The Jenkins tests are all passing.
RELEASE NOTE: Corrected algorithms in the tipping bucket for precipitation and in the nudging routines to adjust for imprecision in single-precision real numbers exceeding the resolvable values in long (>23-year) continuous simulations.
0 commit comments