(Needs editorial input) Weather  Hindcasting - Using models to look backwards ACRE   and   its   partners   undertake   the   recovery   and   digitisation   of   historical   surface   terrestrial   and marine   weather   data   from   around   the   globe.   The   data   they   rescue   is   stored   in   international   databanks which   are   used   by   a   host   of   climate   reanalysis   systems   which   estimate   past   weather   activity.   Generally speaking,   climate   reanalysis   systems   take   the   concepts   of   weather   forecasting   models   and   reverses them   to   do   look   backwards   instead   of   forwards.      This   weather   ‘backcasting’   recreates   what   past weather   probably   looked   like.   ( http://mc-stan.org/manual.html,   p.57 ).      A   more   nuanced   description   of     the difference between climate models and weather forecasting models can be found here . Products of Reanalysis Systems ACRE    data    are    described    as    “observational”    as    it    originates    from    human    or    machine-recorded observations   of   weather   instruments   such   as   thermometers,   barometers,   anemometers,   etc.      Using limited   sets   of   these   observational   data   as   a   starting   point,   climate   reanalysis   systems   simulate   the   full array   of   the   Earth’s   weather   patterns.   They   do   this   over   several   decades   or   longer,   and   can   cover   the entire   globe   from   the   Earth’s   surface   to   well   above   the   stratosphere.      By   calculating   the   overall historical   state   of   the   climate,   they   generate   estimates   of   many   more   weather   variables   beyond   the original    limited    set    of    observational    data    assimilated    by    them.    These    interpolated    variables,    or reanalysis   “products”,   can   cover   50   or   more   weather   conditions   including   atmospheric   temperature, pressure,   wind   and   humidity   at   different   altitudes,   surface   rainfall   and   soil   moisture   content,   and   even ocean-surface   temperature   and   salinity. They   also   create   an   array   of   derived   variables,   such   as   fluxes. Thus,   limited   historical   observations   are   used   to   create   a   much   richer   perspective   of   our   climate heritage. With   this   perspective,   climate   scientists   use   reanalysis   products   to   assess   climate   variability   over   a long   period   of   time,   so   gaining   an   insight   into   how   it   may   be   changing.   Specific   outputs   from   the systems    are    also    useful    in    other    fields    of    study    including    ecology    (climate    impact    on    species), commercial   risk   analysis   (insuring   for   storm/flood   damage)   and   the   social   sciences   (human   reactions to   climate   extremes).      See   here   for   a   more   comprehensive   review   of   how   reanalyses   products   are used. (URL) Differences Between Systems Reanalysis   systems   can   be   distinguished   by   the   geographical   span,   temporal   coverage   and   individual characteristics   of   their   output   products.         Most   systems   have   a   global   reach,   they   cover   at   least   the   last 30-50   years   of   weather   and   they   generally   produce   50+   estimates   covering   ground,   air   and   water weather   products.   An   important   feature   is   the   level   of   accuracy   in   each   of   the   products,   considering that   they   have   been   interpolated   from   a   narrow   base   of   observational   data.      Some   models   claim   to have the same accuracy as current three day weather forecasts. Those   systems   focussed   on   reanalyses   after   1940s-50s   generally   use   large   amounts   of   automatically recorded   data   by   radiosondes,   ocean   buoys,   automatic   ground   weather   stations,   etc.      These   late   20th Century    devices    have    automated    data-logging    capabilities    that    generated    a    very    rich    set    of observational   weather   data.      However,   before   the   1940s-50s,   weather   observations   are   generally   only available   from   physical   recording   media   like   paper.      Some   of   the   observations   were   made   by   machines onto   paper   while   others   were   recorded   by   humans,   resulting   in   far   less   data   being   available   for reanalysis.      What   data   does   exist   (and   has   been   discovered)   has   to   be   digitised   before   it   can   be assimilated   into   the   computer   models.      This   is   done   either   by   humans   or   by   character   recognition software,   both   laborious   processes   resulting   in   a   further   reduction   in   available   data.      Addressing   this deficiency is one of ACRE’s core activities.