StudentOfClimatology

What droughts are you referring to, specifically? The instrumental record is relatively short, so you run into sample-size issues when trying to analyze the proxy-instrument relationship. Interpolating it back in time can be tricky too because antecedent conditions governing tree growth can change. I'm an A.S student specializing in paleoclimate and seasonal forecasting. I've done tree ring analysis before. I'm sure you could manage it. Interpreting tree rings doesn't require an advanced degree.

The problem is there's so much contamination potential, given the slew of factors that influence tree growth, that isolating a single variable on a 1yr resolution using a tree ring proxy is extremely difficult. I would never attempt to do it myself. Tree ring spacing is influenced by precipitation, temperature variation on multiple timescales , sunshine hours, wind speed/transevaporation rates, etc, and varies with different tree species. There are a lot of factors that need to be accounted for here. What they're doing is risky and may flaw the entire study should there not be a way of isolating these phenomena and determining their role.

No tree ring proxy will capture snowpack variability on a year-to-year resolution. I'm fairly certain this paper doesn't claim otherwise.

The majority of the divergence between the satellite data and the radiosonde data can be attributed to the lack of spatial coverage in the radiosonde data. The RSS team did a thorough analysis of the this issue: To account for the lack of spatial coverage in the radiosonde data, the RSS team sub-sampled the RSS data only in the regions where the radiosonde data measures. This corrected most of the divergence between RSS and the radiosonde data, suggesting that the lack of spatial coverage in the radiosonde is responsible for the majority of the divergence between the two. http://www.remss.com/measurements/upper-air-temperature/validation

They didn't just abruptly switch from LKS to IGRA after 1997. They had to correct for the inhomogeneity between the two datasets during the transition, and the IGRA dataset was/is warmer than LKS. The last adjustment to the LKS data was in 2004, which did not sufficiently correct for the cold bias noted in the literature, which finds a cold bias in the RATPAC data through 2006. The IGRA data doesn't have this problem, and it matches the satellite data more closely than either RATPAC-A or RATPAC-B.

The majority of the divergence between the satellite data and the radiosondes can actually be attributed to the lack of spatial coverage in the radiosonde data http://www.remss.com/measurements/upper-air-temperature/validation. That said, you're referring to RATPAC-A, which matches the satellite data more closely than RATPAC-B (though the IGRA data is probably preferable). The RATPAC-B dataset runs warmer after 1997 due to the transition off LKS and onto IGRA station data from 1997-2005, as the two are relatively inhomogeneous. There have (so far) been no adjustments to account for the noted bias in the LKS data. When RATPAC switched their primary radiosonde dataset, they needed to homogenize it for continuity. The problem is the cool bias in the earlier data was left uncorrected for, so the smoothed final product depicts a somewhat unrepresentative trendline in the timeframe of interest. It's not that there are warm biases now, it's that they have yet to correct for the cool biases in the data through 2005. Do you not think that cooler biases in the data through 2005 will affect the trendline? The RSS team investigated the reasons for the differences between the satellite data and the radiosondes, and determined that over 80% of the divergence is due to the lack of spatial coverage in the radiosonde data. By sub-sampling the RSS data only in the regions where the radiosonde data measures, the trendline swerve brought back into relative agreement. http://www.remss.com/measurements/upper-air-temperature/validation In the long term, the radiosonde datasets and satellites have equal uncertainty estimates, according to AR5.

I'm sorry if I called you any names. Personal attacks are unproductive and uncalled for. That said, I'm sticking with the consensus of the peer reviewed literature regarding the error potential in the radiosonde data. In the shorter term (<15yrs), most of the divergence between the two can be attributed to inhomogeneities in the radiosonde data. Obviously this is all debatable still, and I'm sure you disagree, so I hope we can end this on a positive note.

Fine. I had a bizarre change of mind that lasted 15 minutes. The sentence in reference was wrong. I was wrong. Is that an adequate admission? Getting back on topic, my initial argument still stands. The shorter term divergence between the satellite and radiosonde data can be largely attributed to inhomogeneities in the radiosonde data. The majority of the peer reviewed literature comparing the two reaches the same conclusion.

That's fine. You don't have to believe me, but I can tell you exactly what it is I meant. There's nothing inconsistent about my re-worded quote. That's what I have and continue to imply regarding the extrapolation process, if that's what you're referring to. That post is correct. A simple extrapolation with no interpolation between grids is a gridded average. Every dataset is essentially a tuned conglomerate average. What RATPAC does is put the data into equally sized grids to account for areal bias, then extrapolates the data from these stations through the grid boxes they belong to. When there are multiple stations in a grid box, an interpolation procedure is done. When there is only only station in a grid box, no interpolation is done.

Here, I'll post my original quote, then I'll edit it to reflect what I was intending to convey. Original quote: Re-organized quote: Is that better? I'll concede that the first quote was worded poorly, but I never believed they did no gridding. That would be ludicrous.

Maybe you interpreted it that way, but that's definitely not was I was trying to say. See my post below, I was referring to the idea that macroscale homogenization was taking place. If you won't let this go, this roundabout will continue forever.

What was I wrong about? Everything I've said regarding the inhomogeneities in the radiosonde datasets is accurate, and can be verified with numerous peer reviewed analyses. If you disagree or would like to discuss the issue, I'm happy to have a respectful, open-minded conversation with you about it. However, if you'd rather mischaracterize one of my posts in order build a strawman, please don't waste my time.

For christ's sake, I never said they don't do gridding. I've explained what I was intending to say several times now. You don't have to believe me, but continuing this accusatory roundabout is pointless. Let it go. Instead of continuing this pointless back and forth, let's have a scientific discussion regarding the inhomogeneities and uncertainties in the radiosonde data. I have more than 10 peer reviewed papers that I'm ready to post and discuss, should anyone be interested.

Mears et al 2012 http://images.remss.com/papers/Mears_JGR_2012.pdf

I understand why that post could be taken out of context. The use of the word "or" was a bad idea on my part because it creates a mutually exclusive appearance between "gridding" and "spatial homogenization". However, I honestly did not intend for that to be the interpretation. I want to make that abundantly clear.