Interesting point you bring up. I wasn't aware of this specific adjustment that you describe.
I found a paper from a Canadian researcher on this topic:
http://www.ocean-sci.net/9/683/2013/os-9-683-2013.pdf
If you want a detailed explanation of your question regarding the adjustment for earlier temperature data taken using buckets, you will find it there.
Three main methods of sampling water for measuring sea surface temperature were described:
Buckets
Intake water readings
Surface readings of the submerged hull
Use of bucket: A bucket is lowered into ocean, allowed to fill then pulled by rope up to the deck. Temp of water is measured shortly thereafter. This is the oldest method and most subject to errors. Most of these errors cannot be factored out because the error can be either up or down, which causes larger error bars for earlier data. For example thermometer calibration was seen as an issue -- calibration errors can affect the reading upward or downward and so cannot be corrected out of the data set.
The author of the paper found that sampling with a bucked cause an error bias due to cooling once the bucket is brought onto the deck. Wind causes evaporation cooling, especially canvas buckets, which were more commonly used after 1900. Buckets can be wood, canvas or tin. Each affect the reading in a different degree.The bias is between 0.1 and 0.5 degrees C.
From the paper:
Bucket temperatures have generally been found to average a few tenths of ◦C cooler than simultaneous intake temperatures in field studies, although with considerable scatter amongst the individual bucket-intake differences (e.g. James and Fox, 1972)
The author also discusses error when measuring the temperature of intake water: From the paper:
Systematic warm error in intake temperatures is also a plausible explanation for negative average bucket-intake differences. For instance, Tabata (1978b, d) found EIT to average 0.3 ± 1.2 ◦C warmer than accurate in situ temperatures on a research vessel, while Brooks (192 found EIT to be overly-warm by 0.7 ◦C on average on an ocean liner. Given the large magnitude of these errors, it is possible that the principal cause of the 0.3 ◦C average intake-bucket difference found by James and Fox (1972) is EIT error rather than bucket cooling.
OK, now I'm confused.
So, you are right. It is a bit dicey what they are doing with surface water temperature readings on ocean going vessels. The bias error is about a half a degree. This is a valid cause for doubt. If by adjusting the data we change our conclusions or greatly change the forecast then I'd be skeptical of these forecasts too.
So, let's look at the data. The figure below show Sea Surface Temp with bias adjustments and error estimates: Prior to 1940 practically all readings were taken using a bucket of one type or another. The conclusion here is that Sea Surface Temperatures are definitely warmer after 1970.
What happens if we don't apply the bias? Sea surface temperatures are summarized below with and without the bias adjustment. The conclusion from this graph is that without the bias adjustment (blue line), we see a greater difference before 1970 and the oceans appear much warmer after 1970.
View attachment 3616803
This was lifted from: ftp://ftp.wmo.int/Documents/PublicWeb/amp/mmop/documents/JCOMM-TR/J-TR-13-Marine-Climatology/REV1/joc1171.pdf
The various agencies involved in the decision regarding what to do with historical data decided to adjust the earlier data upward. Without correcting the data, we would conclude that the oceans have warmed much more than if we used corrected data. The adjustment upward appears to be something like 0.5 C. The paper referenced in the link above goes into deadly detail about how they tested this correction. It is important to the overall global surface temperature estimates and so it's reasonable to ask if the bias adjustment is valid. What I see is that the researchers did their due diligence. They compared results from studies on measurement methods, satellite data and computer simulations. They concluded that the correct adjustment to be applied is to adjust earlier data upward.
I can see how anybody looking at this hot mess would not be so sure.
I admit that the scope of this is beyond me. So, I'm going to leave right now and think on this.