The RSS dataset is what it is. A model-heavy reconstruction from satellite data with sparse coverage in time and space (since satellites are constantly moving with respect to the ground, and not over the same place at the same time each day), with fuzzy vertical resolution (an inevitability given the geometry of their observations). They’re reasonably “true” in that the average temperatures they report for different altitude bands seem to match up quite well with reconstructions based on independent data (radiosondes). But they’re simply not well equipped to deal with the tropopause - a layer on the order of 1/10th as thick as their minimum resolution, with a swap from positive to negative temperature trends across it, and varying in altitude by almost 10km from equator to poles. To handle that, you need a measurement method with better vertical resolution. Surely that’s not that hard to understand?
The satellite observations are still useful, to those who understand their limitations. But they need to be considered in context with all the other data - radiosondes in particular, plus the other satellite reconstruction bands from altitudes less affected by the complications of the tropopause. Attempting to force the conversation to focus solely on the TLS dataset says quite clearly that you’re much more interested in “winning” than any real understanding.
