The US Temperature Record

Social media and political propaganda sites are filled with accusations that the temperature record is unreliable. I've responded to many of these claims over the years, but I thought it might be helpful to collect links to them in one place, for easy access. Here is a list of the most common challenges to CONUS temperatures with links to my responses and short summaries of what I say in those posts. Note that each of these posts I link to cite the relevant literature and evidence, so you don't have to take my word for anything. You can verify everything I say, and if I ever get something wrong, feel free to let me know in the comments, and I'll make corrections. The subheadings below (#.#) are links to posts with more details about the following.

I. CONUS Temperature Data

1.1 Summary of CONUS Temperature Data and Trends Through 2024

In January I collected a bunch of data from various datasets on CONUS temperatures, including NOAA's two surface thermometer datasets (nClimDiv and USCRN), reanalyses (ERA5) and satellite lower troposphere data (UAH) to compare how well they agree with each other. For the new, pristine USCRN dataset, there is broad general agreement between it and the others for the dates they overlap (2005 - 2025). NOAA started the USCRN project in 2005 with a network of 114 pristine stations at ideally-sited, rural locations, and USCRN shows marginally more warming than all the other datasets. This fact alone is relevant rebutting almost all the challenges I respond to below.

1.2 The Dust Bowl Heatwaves

There are frequently claims that the US was hotter during the 1930s Dust Bowl years than it is today. And while those heatwaves were really hot, especially during the day, heatwaves today are worse.

II. Does NOAA Manipulate US Temperatures?

2.1. Accusations that NOAA Manipulates Data Depend on Data Tampering

Tony Heller and John Shewchuk have for years made accusations that NOAA manipulates the  US temperature record to cool the past and warm the present. I show in this post that both Heller and Shewchuk rely on irresponsible and dishonest handling of data to make their case. In fact, they tamper with the data to make their accusations that NOAA tampers with data. Heller has a long history of this, and I've collected resources from others making similar points here. You can see the effect of Heller's (at the time he went by the fake name Goddard) data tampering below.

The Difference Between "Goddard's Raw" and "Correct Raw" is the Data Tempering
from by Heller and Shewchuk's Method

To be clear, bias correction does cool the past for CONUS temperatures, mostly because of a well-defined and quantified bias introduced by changing the observation time of stations. This bias can actually synthesized with USCRN data, so there is independent validation of both the presence of the bias and the size of the bias in the "raw" data, so the applied correction is warranted and known to improve the accuracy of the dataset.

2.2. Accusations about Bias Correction and Homogenization are Counterfactual

NOAA (and every scientific institution on the planet) corrects biases the affect raw data. For temperature data, scientists have published numerous studies that can identify these biases, quantify them, and develop methodologies to correct them. The results of this work can be found in the scientific literature and in the documentation associated with the published datasets. As a general rule, bias correction to land temperatures increases warming trends, but adjustments to sea surface temperatures reduces warming trends by a larger amount. The net effect of all adjustments is to lower the amount of global warming. We can evaluate whether bias correction adds any spurious warming to CONUS temperatures by comparing nClimDiv (with homogenization/bias correction) to USCRN (raw temperatures). For the years they overlap, they are nearly identical, and if anything the raw USCRN data shows marginally more warming.

2.3. Accusations the Bias Correction Adds Spurious Warming Disregards the Systematic Errors in Raw Data

The largest source of bias affecting raw temperature data in the US has to do with systematic changes in observation time in the second half of the 20th century. This bias can be well-quantified and therefore corrected. It can even by synthesized with USCRN data. Correcting the bias significantly improves the CONUS temperature record.

Graph from Jim Java on X

2.4. Accusations that NOAA Reports Data from "Ghost" Station are Deceitful

The old USHCN dataset would use interpolation (based on averaging data from surrounding stations in a grid cell) to aid in area weighting when stations either close or don't report data. The interpolated values are clearly marked with an "E" and do not count as recorded data. They also demonstrably have no effect on CONUS temperatures. And USHCN was fully deprecated in 2014. The two replacements, USCRN and nClimDiv, do not use this form of interpolation. Shewchuk is dishonest to refer to "ghost" stations in the old USHCN dataset, but he's doubly dishonest for continuing to make this accusation when NOAA doesn't even use USHCN anymore.

Graph from "Dale Gribble" on X

The graph above shows CONUS temperatures with and without what Shewchuk wrongly calls fake data from "ghost stations." The two are indistinguishable, meaning necessarily that none of the claims about "ghost stations" adding to CONUS warming make any sense. 

Almost all of the claims about data manipulation and ghost stations made by Shewchuk, Heller, and their allies are based on the old USHCN station data. One reason why is that as stations closed in the old dataset (mostly after about 1995), the average station latitude and elevation increased (mostly seen after 2005 or so). There are enough stations in the old USHCN network that this wouldn't be a problem with proper area-weighting, but Shewchuk and Heller use simple station means of station data, which means the change in average station latitude and elevation will add a cooling bias to the their plots.

Graph from "Dale Gribble" on X

Shewchuk and Heller know this. These graphs have been shown to them on X, and yet they never seem to correct their obvious mistakes.

III. Is Warming Due to the Urban Heat Island (UHI) Effect?

Some argue that the US is warming because of the UHI effect; that is, as the US population grows, more people live in cities, and it's the growth of urban areas that are responsible for much of the warming trends in the US. It's absolutely clear that cities are significantly warmer than rural areas, putting those living in cities at greater risk to the effects of AGW than those in rural areas, but scientists can evaluate whether the UHI effect is having a significant effect on CONUS and global temperature trends.

3.1. The Effect of UHIs on both Global and CONUS Temperatures is Quantified

Wickham et al 2013 compared a "very rural" subset of station data and compared this subset to "all stations" and found that the very rural subset is warming at least as rapidly as all stations. This confirms that UHI is not having any demonstrable effect on either global or CONUS temperature trends. This is for several reasons:

1. Cities warm at the same rate as rural areas.
2. Bias correction removes biases associated with urbanization.
3. Urban areas are only ~1% of global surface area and ~3% of CONUS surface area (cities and suburbs).

While the graph below has to do with global land data rather than CONUS, Robert Rhodes at Berkeley Earth plotted global land mean surface temperatures using both all stations and only "very rural" stations from 1850-2024. The two are indistinguishable from each other, so it simply cannot be that UHI is having a measurable effect on land temperature anomalies.

3.2. A Sniff Test for the UHI effect on CONUS and Global Temperatures

If UHIs were responsible for a significant fraction of warming in the CONUS temperature record, then temperature anomalies would be much higher at cities than in rural areas, but they aren't. You can pick out cities in maps of population density, but they are invisible on maps of temperature anomalies.

3.3. A Review of Spencer's Paper on the UHI effect on CONUS Temperatures

A recent paper by Roy Spencer argues that UHI is responsible for ~22% of the warming in the raw data at US temperature stations. However, this figure is not weighted by the fraction of surface area covered by rural vs urban lands, and it's also based entirely on the raw data, before urbanization biases are removed from published CONUS datasets.  Even Spencer admits that his paper does not demonstrate that UHI has any impact on CONUS temperature trends in either USCRN or nClimDiv.

IV. Is Warming Due to Poor Station Siting?

4.1. Station Siting has a Negligible Effect on CONUS Temperatures

A common refrain from contrarians is that warming in the US is due to poor station siting - stations located on asphalt, at airports, too close to buildings, etc. A majority of stations are "poorly sited." And yet if you isolate only the well-sited and poorly sited stations and compare them before and after bias correction, the data from poorly-sited stations agrees with well-sited stations. And USCRN (where all stations are well-sited) shows marginally more warming than nClimDiv (which has some poorly-sited stations). So we can rule out the possibility that station siting is adding any spurious warming to CONUS temperature trends.

The links above give more context for these summaries as well as citations to the literature. Please don't take my word for anything. See the evidence I cite and check up on me.