News for hackers. Stuff that matters.
Re:July 2015 was warmest month ever recorded by Afforess on Thursday August 20, 2015 @12:17PM (#10092804)
I don't see why using inferred/model data is a bad thing. It is simply not feasible to measure the temperature at every location in the world. Satellite data, combined with in-situ and model calculations can very accurately fill in for missing data. The only other option here is to sit on our hands and say "oh well, I guess we can't predict the weather because we don't have perfect data", which is completely unreasonable.
Re:July 2015 was warmest month ever recorded by pdkl95 on Thursday August 20, 2015 @12:50PM (#10093045)
There is another potential concern besides the inferred data (which, as you say, should be reasonable), though I don't know how much it would affect any of the larger conclusions (if they are affected at all).

[also, this reference is somewhat old now; I haven't checked if this problem has been addressed in more recent models]

There is evidence[1] that the climate model software doesn't handle floating point properly (which is, unfortunately, a very common problem).

    "... there exist differences in the results for different ... parallel
    libraries, and optimization levels, primarily a result of the treatment
    of rounding errors ... the ensemble spread due to the differences in
    software system is comparable to the ensemble spread due to the
    differences in initial conditions ..."
Rounding errors on different hardware, operation ordering, and optimization level screams that the software isn't properly handling[2] IEEE 754 floating point. This could be a serious problem in an iterated model that is chaotically sensitive to initial conditions (hence the use of ensemble forecasting).

/* regardless, we need a lot thorium reactors ASAP */

[1] http://journals.ametsoc.org/doi/abs/10.1175/MWR-D-12-00352.1

[2] (pdf) https://ece.uwaterloo.ca/~dwharder/NumericalAnalysis/02Numer...