PragerU post by Happer uses flawed reasoning to claim that climate models always fail

“[Climate models] don’t work”; “the number of factors that influence climate—the sun, the earth’s orbital properties, oceans, clouds, and, yes, industrial man—is huge and enormously variable.”

All of these processes (and many others) are included in models, allowing them to simulate the climate. This assertion is simply wrong.

This is not accurate; while models can never be a perfect representation of the Earth’s system, they do an excellent job of reproducing many aspects of the Earth’s climate, from rainfall and wind patterns to storm and hurricane formation and warming of the climate. In a 2019 paper we evaluated the performance of 17 historical climate model projections published between 1970 and 2001^[1]. We found that 10 of those 17 projected a rate of future temperate change nearly identical to what actually happened in the real world in the years after they were published, while four of the models projected too much warming and three models too little warming^[1]. This is particularly impressive for the 1970s-era models, which were published at a time when evidence of observed global warming was limited (and some even thought – based on limited observations – that the world was modestly cooling).

Figure 2—Observed surface temperature change (HadCRUT5 – black line) compared to climate model projections from the years after the model was published (colored lines). Adapted from Hausfather et al. 2019^[1].

While modelling the climate system involves dealing with a lot of uncertainty, solar activity and Earth’s orbit do not contribute much uncertainty. Making climate projections often revolves around a set of assumptions about anthropogenic emissions. These are often idealized, like a 1% increase in CO₂ per year, but that does not make the results invalid. We can still learn a lot about the response of clouds and ocean circulation from studying these idealized experiments that can inform our response to the changing climate.

The number of factors influencing the climate is large, but they do not all affect the climate in equal ways. Climate models routinely include solar luminosity variations and orbital properties as external drivers. Human influences including atmospheric composition and land usage changes are similarly included as external drivers. Ocean models are part of coupled climate models such as in the publicly available CMIP global climate models. Clouds are also part of the atmospheric components in these climate models. While clouds are the largest source of uncertainty, they are simulated well enough that this statement has no merit in my opinion.

The overall claim that “[climate models] don’t work” is illogical, vague, and seems to imply, hyperbolically, that the models have no use whatsoever. Climate models are imperfect representations of nature that represent just one tool scientists use to understand how and why Earth’s climate varies. Climate models are good at simulating some physical processes and deficient in their simulation of other processes. Simply because the climate is complex does not
mean we cannot reasonably model or understand it.

“Compared to water—H₂0, carbon dioxide—CO₂—is a minor contributor to the warming of the earth.”

This is a well-worn trope that Dr. Happer unfortunately uses to mislead. Water vapor acts as a strong amplifier of warming initiated by CO₂. I think of water vapor molecules as soldiers and CO₂ as the commander. The commander decides to invade and the soldiers do most of the work. The soldiers do not randomly decide to invade. Similarly, water vapor cannot randomly decide to increase in the atmosphere. Rather, CO₂ causes warming, and the warmer atmosphere contains more moisture, further enhancing Earth’s greenhouse effect – a textbook amplifying feedback.

This statement is incorrect. This misses the point. Increases in CO₂ affect the energy balance leading to more atmospheric moisture. Climate models incorporate the radiative properties of both compounds. This statement is incorrect.

This is an incredibly misleading – and incorrect – statement. Water vapor is a powerful greenhouse gas, but it’s also one that is temperature limited. Water vapor has a very short lifetime in the atmosphere, so adding more water vapor by itself cannot effectively cause long-term climate warming. CO₂, on the other hand, lasts for centuries to millennia in the atmosphere, and accumulates. Our best estimate is that around 100% of observed warming since the late 1800s is attributable to human emissions of CO₂ and other greenhouse gases^[7].

However, water vapor does have a role in that warming, but as a feedback rather than a forcing^[8]. Higher temperatures caused by rising atmospheric CO₂ increase the amount of water vapor in the atmosphere by increasing evaporation and by higher air temperatures allowing more water vapor to be present. This additional atmospheric water vapor enhances the warming from CO₂ (and other greenhouse gases), but would not be in the atmosphere without their warming effects^[8].

Water is more abundant and a more powerful greenhouse gas. However, the residence time of water vapor in the atmosphere is very short compared to CO₂, and the spatial distribution of water is much less homogenous than atmospheric CO₂. If humans were to emit a lot of water instead of CO₂ it would be rained out rather quickly. With large and constant water vapor emissions we probably would see a change in the localized climate. None of this contradicts the fact that CO₂ can also have an impact on the climate. Our CO₂ emissions will elevate the atmospheric concentration for hundreds of years and contribute a small, but persistent, downward radiative flux over the whole planet. That small contribution is enough to be concerned about even if you can find other things that make it seem „too small“.

This is misleading and excessively vague. Water vapor is the primary and most abundant greenhouse gas in the atmosphere, and water in its liquid form in the oceans has an enormous capacity to retain and redistribute heat. Without water vapor, the planet would be much colder. But the increase of CO₂ and other greenhouse gases due to human activity is the primary reason why the planet has warmed since the middle of the 20th century.

“We can’t predict what effect the atmosphere is going to have on future temperatures because we can’t predict cloud formations.”

Clouds are one of the big areas of uncertainty in projecting future climate change, as they form on scales too small to directly simulate in climate models. Uncertainties in how clouds will change in a warming world is one of the main reasons why we are uncertain if doubling CO₂ concentrations will warm the world by 2C or as much as 4.5C at equilibrium. However, suggesting that these cloud uncertainties mean that we can’t predict future changes is quite misleading. As discussed earlier, it’s clear that our climate models have performed quite skillfully in predicting the changes we’ve actually seen in the real world after they were published. A somewhat cloudy crystal ball is, after all, much better than no crystal ball at all.

The biggest source of error in climate projections is how certain cloud types will react. This is a very active area of research right now, and probably will be for the next decade. However, I do not know of any credible climate scientist that feels that clouds could reverse the warming trend we expect from CO₂. It’s possible that clouds will make the climate less sensitive than we thought, but that should not alleviate our concerns about the impact of global warming. For example, even a strong negative cloud feedback will not remedy the problem of ocean acidification from CO₂, which is likely to be a very serious problem in the future.

Dr. Happer has correctly identified a key uncertainty in models’ predictions of future temperature – how clouds will respond to warming. This is a big reason why the scientific community tries to constrain models’ predictions using observations of how clouds respond to warming. We do not blindly accept model results as truth – we constantly evaluate them against observations, allowing us to hone our estimates of future warming. That said, we do not know *nothing* about future temperature — instead we have a range of plausible outcomes, all of which suggest a warmer future.

Recent state of the art cloud system resolving models do very well in simulating intense storms and the cloud formations associated with them.

This is false. We know that Earth’s temperature will continue to rise as greenhouse gas emissions increase unabated. The most advanced, state-of-the-art predictions of the sensitivity of Earth’s climate to increasing carbon dioxide are based on multiple lines of evidence: observations, the paleoclimate record, theory, and models of varying complexity^[9]. Based on this evidence, climate scientists estimate a likely range of the planetary warming resulting from a doubling of atmospheric carbon dioxide: 2.6-3.9 C (or 4.7-7 F). Climate models produce a wider range of the severity of planetary warming in response to increasing carbon dioxide^[10]. This is primarily because their simulation of cloud processes is highly variable, with some models performing better or worse than others. However, a variety of independent evidence taken together reveals how clouds throughout the planet will likely behave as the climate warms, allowing scientists to predict future temperature changes with more precision than climate models simulate^[9].

“A major aspect of climate involves the complicated interaction between two very turbulent fluids: the atmosphere, which holds large amounts of water (think rain and snow), and the oceans, which cover fully 70% of the earth’s surface…We can’t predict either side of the atmosphere/ocean equation.”

We can (and we do) predict both of these systems reasonably well. Coupled models often produce strong regional errors, but this does not invalidate the basic conclusion about the concerning amount of warming we expect from elevated CO₂.

This statement is misleading. Climate science is a statistical one and we can simulate the statistical behavior of the relevant aspects of the climate quite well. In fact, simulations of future climate made in the 1990s have been shown to predict the present day warming very well. Furthermore, climate models are extensively independently evaluated by analysts not involved in model development. Clearly some aspects of climate model simulations can be improved, but the global energy budget is well simulated by nearly all models.

This is false. Many aspects of both the atmosphere and ocean are predictable, though the degree of predictability varies depending on the process in question. A recent study finds that “climate models published over the past five decades were generally quite accurate in predicting
global warming in the years after publication”^[1].

If models can’t predict the path of a hurricane, they can’t predict climate over the last 30 years, “one climate prediction after another – based on computer models – has been wrong.”

This is another well-worn trope. Dr. Happer has unfortunately confused weather and climate in this statement. The accuracy of weather forecasts – and in particular extreme events like hurricanes – can be degraded by poor data about the current state of the atmosphere that gets fed into the models. These issues are very different from those that affect climate model predictions – namely, how clouds and humans will respond in the future. Secondly, the nature of weather forecasts („what is the probability of rain in my city tomorrow afternoon?“) are very different from climate projections („how much warmer on average will the 2050s be in California?“). It is also worth mentioning that hurricane track forecasting is becoming extremely skillful, in contrast to Dr. Happer’s assertion. Cherry picking a bad hurricane track forecast to throw shade on climate model projections is like claiming Tom Brady isn’t going to the Hall of Fame because he burned his breakfast yesterday.

I think that the National Hurricane Center would beg to differ. Hurricane track forecasting has improved dramatically over the past few decades and has led to substantially reduced fatalities^[11,12]. The statement is a (false) assertion, without any evidence.

Modern weather models – which share many parts of their code with our long-term climate models – are actually quite good at predicting hurricane tracks^[13]. Moreover, the fact that our prior model projections have proven quite accurate at estimating how much warming actually occurred gives us confidence that our models – while imperfect – are accurate enough to get a good sense of how much warming to expect in the future if we keep emitting CO₂ and other greenhouse gases.

Figure 3—Historical temperatures (colored lines) and the last generation of climate models (black line shows the model average, with the grey shaded area representing the range across all the models). From: https://www.carbonbrief.org/state-of-the-climate-2020-ties-as-warmest-year-on-record

Let’s think about this idea that „computer models are always wrong“. Computer models do amazingly well at a lot of things, but it is also very easy to find little things that they do not do well at. It is not fair to judge all weather and climate models based on only the things they do wrong. One model might not rain enough in one area of the globe, but even if it gets it mostly correct everywhere else people will criticize the model for being „wrong“, which is completely unfair. Hurricane forecasts are another good example where they can be right most of the time, only to then be harshly judged when they get the hurricane track wrong. We care about these small errors because they can potentially mean an unexpected loss of life or property, but to use blanket statements about all computer models being wrong is simply dishonest or naïve.

This is an illogical statement and misleading. The author is conflating weather forecasting of a hurricane with long-term climate prediction, and more fundamentally, he is conflating weather with climate. As the American Meteorological Society states, climate is defined as “The slowly varying aspects of the atmosphere–hydrosphere–land surface system. As distinguished from climate, weather consists of the short-term (minutes to days) variations in the atmosphere.”

Simply because some weather forecasts are inaccurate does not imply that changes in climate are unpredictable. Climate model projections are not intended to forecast changes in the atmosphere as precisely as the particular path of a single hurricane. Some of the main elements of the climate that models are used to predict are changes in large-scale patterns of temperature, humidity, precipitation, and wind.

It’s false to say that “Over the last 30 years, one climate prediction after another —
based on computer models — has been wrong”. Models accurately predict some climate changes and imprecisely predict others, with some overpredictions and some underpredictions. For example, a recent study finds that “climate models published over the past five decades were generally quite accurate in predicting global warming in the years after publication”^[1]. To take another example, climate models tend to underestimate the rapid Arctic sea ice decline observed in recent decades^[14].