The Climate Change Deniers are Wrong: Evidence from NOAA's Barrow Atmospheric Observatory in Alaska

Indicative of the grim reality that climate change denial remains a barrier to policies that would reduce carbon emissions, a group of 39 Republican Senators has recently requested that the EPA rescind its proposed regulations to sharply reduce carbon dioxide emissions from the nation’s electric power plants

This paper challenges this apparent climate change denial by statistically analyzing the relationship between CO2 concentrations and hourly temperature using data from NOAA’s Barrow (BRW) Atmospheric Observatory in Alaska. At this location, the average annual temperature over the 2015-2020 period was about 3.37 degrees C higher than in the 1985-1990 period. At the nearby Barrow Airport, the four warmest years since 1925 occurred in 2016, 2017, 2018, and 2019. In these four years, the average annual temperature was about 3.3 degrees C higher than the maximum annual temperature from 1925 through 1939. These facts do not support the assertion by Lindzen that the recent warming is about the same as before the 1940s.

The analysis proceeds by first noting that the hourly temperature data at BRW are highly volatile at times. It is also highly autoregressive over a substantial number of hourly lags. Based on these findings, an ARCH/ARMAX ( Autoregressive Conditional Heteroskedasticity/ Autoregressive–Moving-Average with Exogenous Inputs ) modeling approach is employed. This statistical method is a form of machine learning that is very useful in analyzing data that are highly autoregressive in nature that also exhibits volatility at times. The method is virtually unknown in the atmospheric sciences, which is most unfortunate because it can be difficult to extract the “signal” from the “noisy data” when the heteroskedasticity of the conditional variance and overall autoregressive nature of the data are not explicitly recognized. The ARCH/ARMAX model includes solar irradiance and lagged CO2 concentrations as exogenous inputs. Possible non-anthropomorphic drivers of temperature are also considered. The model is estimated using hourly data from January 1, 1985 through December 31, 2015. The results are consistent with the hypothesis that increases in CO2 concentration levels have consequences for hourly temperature. The model is evaluated using hourly data from January 1, 2016, through December 31, 2021. The model’s out-of-sample hourly temperature predictions are highly accurate. However, this accuracy is significantly degraded if the estimated effects of CO2 on temperature are ignored (Figure 1). Specifically, the out-of-sample predicted temperature is substantially less than the actual temperature when the estimated CO2 effects are ignored but is essentially equal to actual temperature when the estimated CO2 effects are recognized. In short, the climate deniers are wrong when they assert that the effects of CO2 on temperature are non-existent or trivial. While some climate deniers may be tempted to reject the finding in this paper by noting “I am not a statistician”, one does not need to be a statistician to observe that excluding the estimated CO2 effects from the out-of-sample predictions renders results that are inconsistent with the claims of the climate deniers.

The implications of the statistical analysis for locations with high rates of climate denial (e.g., North Dakota) are assessed using Vector Autoregressive temperature models coupled with Granger Causality tests. These statistical tests indicate that the hourly temperature in, say, Billings, North Dakota, is affected by the lagged temperatures at BRW even though Billings, North Dakota is thousands of kilometers from BRW. In short, CO2 has significant consequences for hourly temperature throughout the USA, a result that the constituents of every Senator may be interested to learn considering the heat waves that plagued the United States in 2023.