search

ReFRACtor / ABSCO

Software that can generate a set of ABSoprtion COefficient tables that span the thermal IR to UV spectral range
Other
9 stars 4 forks source link

Efficiency Improvements #6

Closed pernak18 closed 6 years ago

pernak18 commented 6 years ago

These guys are not necessary, but if we have time we should address them so that we can reduce run times:

  1. PT grid: we got LBLRTM to run when including the entire profile in the associated TAPE5, rather than just doing 1 layer at a time. The latter was giving me problems in the LBLRTM runs, but the former added some complexity as well, because we only want ABSCOs for a subset of of the 15 temperatures (dependent on the pressure level) we are using in this study. As a workaround, we are just running LBLRTM for all 15 temperatures -- 180-320 K -- for entire profiles at a time, then parsing down to what we need when calculating ABSCOs. The disadvantage of this approach is that it will include some unnecessary OD calculations that might add a significant amount of time to the LBL runs.
pernak18 commented 6 years ago

we are no longer producing one TAPE5 per molecule per channel per temperature. these TAPE5s were causing unexplained segmentation faults (LBL couldn't handle certain temperatures at a given pressure, perhaps?). now we generate one TAPE5 per allowed temperature at each pressure (per channel per molecule).

so no only allowed pressure/temperature combinations are run, lifting any concerns about efficiency.

see https://github.com/pernak18/ABSCO/commit/2697366806b7f1dbb568062ddcbce1a4e85299bf