One of the most fundamental mistakes which wannabee scientists make, is to attempt to apply precision to a problem which hasn’t been sanity tested for accuracy.
It should be obvious to everyone that the current data set is not sane.
What I am doing is what professionals call “sanity testing” – and the adjusted data fails miserably. The problems are too numerous to enumerate, but here are a few of them.
1. Anomalies, infilling and gridding.
The image below represents a grid cell. The U’s stand for Urban, and the R’s stand for rural. Let’s say that the U anomaly is 1.0 and the R anomaly is 0.0. The average anomaly of this grid cell is 6/8 = 0.75
Now, let’s take away one of the rural stations – as has been happening to the network. Because of infilling, the rural station now gets effectively counted as an urban station – due to it being infilled with data from its urban neighbors. The anomaly of the grid cell goes up to 7/8 = 0.875. The infilling aggravated the rural station loss, rather than improving it.
Anomalies, infilling and gridding simply smear data together. This might be a good idea with a truly random error distribution, but that isn’t what we are dealing with.
By using absolute temperatures, I am able to do sanity checking – which is impossible using Zeke’s method. Zeke skips the sanity checking, and goes directly to trying to generate false precision.
USHCN is losing station data at a phenomenal rate since 1990. If current trends continue, they will have no station data by 2020. Does Zeke plan to eventually use one station to represent the entire country, like Phil Jones does for the southern hemisphere in 1851?
Now, lets look at the raw measured data vs. the fabricated data – i.e. station data which has no raw data to back it up. The raw data shows no trend since 1990. The fabricated data shows a massive warming trend. This a huge red flag that the adjustments are garbage.
The fabricated data is diverging from the measured data at a mind-boggling rate of almost 8 degrees per century.
The next experiment is to see what happens if data is removed from all the stations with complete records. The blue line below shows temperatures for all stations with no missing data since 1990. The red line shows the same data set, with 30% of the data randomly removed.
Not surprisingly, random data loss has very little impact on the temperature. Much of science depends on that very principle. This shows us unequivocally that whatever is causing the massive data loss at NOAA is not random. It has a huge bias to it.
Zeke and crew want to smear over all this, and obtain high precision on a garbage data set with very low accuracy. I want to find out what is wrong with the data set.
Not a very subtle distinction. This is a critically important issue, and it would be very helpful if people who should be helping bring it forward would do the right thing.