• No se han encontrado resultados

Capítulo 2. Propuesta metodológica para la dirección del pensamiento táctico en

2.4. Diagnóstico en el pensamiento táctico para determinar las regularidades46

The  WRF  simulations  of  2001  summer  climate  over  the  CONUS  indicates  that  the   dry   bias   in   SE   US   summer   precipitation   is   most   likely   caused   by   the   inaccurately   simulated   NASH   western   ridge   and   associated   circulation   due   to   the   erroneous   distribution   of   zonal   winds   in   the   tropical   oceans,   i.e.,   errors   in   circulation   dynamics.  

                                                                                                               

2  The  null  hypothesis  for  the  Hotelling’s  t-­‐‑square  test  is  that  the  WRF  simulated  NASH  western  

ridge  does  not  differ  significantly  from  that  in  reanalysis  datasets.  According  to  the  test,  the  null   hypothesis   can   be   rejected   with   a   99.99%   confidence   level,   suggesting   that   the   erroneous   northwestward  extension  of  the  ridge  is  significant.    

Thus,  an  improved  simulation  of  large-­‐‑scale  circulation  (especially  the  NASH  western   ridge)  could  potentially  reduce  the  RCM  bias  in  SE  US  summer  precipitation.  

To  verify  the  importance  of  circulation  dynamics  in  generating  the  SE  US  summer   precipitation   bias   and   to   assess   the   potential   improvement   in   precipitation   simulation   from   an   improved   large-­‐‑scale   circulation,   two   sets   of   WRF   experiments   utilizing   the   FDDA   are   performed.   The   FDDA,   i.e.   interior   grid   nudging   technique,   continuously   nudges  the  WRF  simulated  thermodynamic  and  dynamic  variables  towards  the  driving   reanalysis   datasets   during   the   simulation   (Stauffer   and   Seaman   1990).   The   FDDA   has   been   widely   applied   in   regional   climate   downscaling   and   has   significantly   improved   climate  downscaling  skills  over  the  US  (Bowden  et  al.  2013;  Lo  et  al.  2008;  Otte  et  al.  2012).   In   this   analysis,   however,   the   application   of   the   FDDA   is   not   for   the   purpose   of   improving  precipitation  simulation  skills  but  rather  for  identifying  the  potential  sources   of  RCM  skills  in  SE  US  summer  precipitation.    

The  two  sets  of  FDDA  experiments  are  designed  as  follows:  thermodynamic  FDDA   and   dynamic   FDDA.   In   the   thermodynamic   FDDA   experiment,   the   temperature   and   specific   humidity   are   nudged   towards   NCEP-­‐‑R2   at   each   6-­‐‑hr   interval   during   the   simulation,   while   the   wind   fields   are   generated   by   WRF.   In   the   dynamic   FDDA   experiment,   the   WRF   simulated   three-­‐‑dimensional   wind   fields   are   nudged   while   the   temperature  and  specific  humidity  are  not.  The  previous  experiment  without  an  FDDA   is  defined  as  the  control  experiment.  We  run  both  thermodynamic  and  dynamic  FDDA  

with  the  four  different  cumulus  schemes  as  in  the  control  experiment.  The  improvement   of   the   simulated   precipitation   in   thermodynamic   (dynamic)   FDDA   is   attributed   to   the   correction  of  atmospheric  thermodynamic  (dynamic)  structures.  Thus,  by  comparing  the   simulated   precipitation   in   thermodynamic   and   dynamic   FDDA   with   that   from   the   control   experiment,   the   relative   importance   of   thermodynamic   and   dynamic   contribution  to  SE  US  dry  bias  can  be  compared  (Li  et  al.  2013b;  Seager  et  al.  2010).    

Figure   4.12   shows   the   CONUS   JJA   precipitation   in   the   FDDA   experiment.   By   correcting   the   WRF   simulated   circulation   fields,   the   dynamic   FDDA   experiment   substantially   reduces   the   bias   in   SE   US   summer   precipitation.   In   the   dynamic   FDDA   experiment,  summer  precipitation  increases  to  about  5  mm/day  over  the  SE  US  domain.   The  domain-­‐‑averaged  bias  is  reduced  to  -­‐‑0.3  mm  day-­‐‑1,  indicating  that  about  80%  of  the  

original   dry   bias   in   control   experiment   has   been   corrected   (Figure   4.12a   and   c).   Furthermore,  the  spatial  distribution  of  precipitation,  especially  the  southeast-­‐‑northwest   oriented   gradient,   is   also   reasonably   simulated   in   the   dynamic   FDDA   (Figure   4.12a).   Thus,   the   dynamic   FDDA   experiment   further   verifies   that   the   errors   in   wind   fields   generated   during   the   WRF   simulations   are   responsible   for   the   SE   US   dry   bias   in   the   control  experiment.    

To   confirm   that   the   improved   simulation   of   precipitation   in   the   dynamic   FDDA   experiment   results   more   from   circulation   dynamics,   the   effects   of   the   thermodynamic   FDDA  are  also  compared.  Generally,  the  thermodynamic  FDDA  does  not  improve  the  

simulation   of   SE   US   summer   precipitation   as   significant   as   the   dynamic   FDDA   when   compared   to   the   control   experiment   and   observations.   Specifically,   in   the   thermodynamic   FDDA   experiment,   the   SE   US   dry   bias   is   not   meaningfully   reduced   (Figure   4.12b   and   d).   The   areal-­‐‑averaged   rainfall   bias   reaches   -­‐‑2.0   mm   day-­‐‑1   in   the  

thermodynamic  FDDA  experiment,  compared  to  the  bias  of  -­‐‑1.3  mm  day-­‐‑1  in  the  control  

experiment.  In  addition,  the  rainfall  amount  decreases  over  the  coastal  regions,  and  the   spatial  gradient  of  rainfall  further  weakens  (Figure  4.12b).  Thus,  unlike  the  atmospheric   dynamical   fields,   correcting   the   atmospheric   thermodynamic   fields   is   insufficient   to   generate   a   satisfactory   skill   in   SE   US   summer   precipitation   simulations.   More   importantly,   the   comparison   between   the   thermodynamic   and   dynamic   FDDA   experiments   indicates   that   the   improvements   of   the   simulations   due   to   the   dynamic   FDDA  are  most  likely  from  direct  dynamic  contributions  instead  of  indirect  thermally   driven   circulation   dynamics.   In   other   words,   if   the   thermally   driven   circulation   (i.e.   circulation   component   determined   by   atmospheric   thermal   structure)   contributes   significantly   to   rainfall   simulation,   the   thermodynamic   FDDA   should   generate   similar   corrective   effects   that   the   dynamic   FDDA.   However,   since   the   thermodynamic   FDDA   fails  to  improve  the  precipitation  simulations,  it  is  the  direct  dynamic  contributions  in   dynamic   FDDA   that   provides   the   ultimate   sources   of   simulation   skills   for   SE   US   summer  precipitation.  

 

Figure  4.  12:  2001  JJA  summer  precipitation  (shaded,  unit:  mm  day-­‐‑1)  as  simulated  in  a)  

Thermodynamic  FDDA,  and  b)  Dynamic  FDDA  experiment;  and  the  precipitation  bias  in  c)   Thermodynamic  and  d)  dynamic  FDDA.  The  results  are  shown  as  the  average  of  the  four  

cumulus  schemes.  

 

The  experiments  utilizing  FDDA  collectively  suggest  that  the  atmospheric  dynamics   plays   a   direct   and   predominant   role   in   regulating   SE   US   summer   precipitation   at   seasonal  scales.  The  results  from  FDDA  experiment  are  consistent  with  the  results  of  SE   US  summer  precipitation  based  on  the  regional  moisture  budget  (Chapter  2).  Over  the   SE  US,  large-­‐‑scale  circulation  contributes  to  more  than  90%  of  the  variance  in  moisture   transport   for   SE   US   summer   precipitation,   whereas   thermodynamic   (temperature   and   specific  humidity)  contribution  accounts  for  the  less  than  10%  (Chapter  2).  The  observed   characteristics  of  SE  US  hydrological  cycle  indicate  that  errors  in  large-­‐‑scale  circulation   could  easily  translate  into  summer  precipitation  bias  due  to  its  active  role  in  atmospheric  

a) Thermodynamic FDDA b) Dynamic FDDA

moisture   balance.   Thus,   the   distortion   of   the   large-­‐‑scale   circulation   along   the   NASH   western  ridge  during  the  WRF  simulation  could  result  in  the  dry  bias  over  the  SE  US   (Figures.  4.9  and  4.10).    

Overall,  the  FDDA  experiments  as  well  as  the  diagnostic  analysis  suggest  that  the   WRF  simulated  dry  bias  in  SE  US  summer  precipitation  probably  originates  from  errors   in  modeled  large-­‐‑scale  circulation.  Thus,  a  better  representation  of  large-­‐‑scale  dynamics,   especially  that  associated  with  the  NASH  western  ridge  circulation,  likely  improve  the   WRF  performance  in  simulation  summertime  climate  over  the  SE  US.