Climate Models' Predictions of Global Temperature Change

Introdutory Notes on Climate Model Forecasts

Climate models are complex computer models that are used to simulate Earth's climate and climate changes, including changes due to increased greenhouse gas concentrations. You should keep in mind that these models are not reality. There is much that we do not understand about the operation of the climate system on Earth, for example, some of the complex feedbacks mentioned on the previous page. We certainly cannot program models to be more accurate than our understanding of the underlying processes. Thus, the models are not perfect and do not make exact predictions. Simulations made by these models are one tool we have to help understand climate and climate change on Earth. Given the uncertainty in climate model predictions, we should also consider other sources of information when deciding environmental policy, such as restrictions on emissions of greenhouse gases. In other words, the results of climate model simulations should not be the sole reason for enacting environmental policy.

Several tests have been run with climate models that show they do have some predictive skill. For example, climate models have been able to closely reproduce natural, short-term fluctuations in climate due to historic events like volcanic erruptions and El Nino / La Nina. On longer timescales, some features of the Pleistocene Ice Age cycles have been reproduced in simulations. People have been able to use climate models to reasonably reproduce the measured change in global average temperature that occurred during the 1900s. In these simulations, greenhouse gas concentrations are started at year 1900 levels, then the levels are increased as they were measured to increase as the model runs for 100 years (simulating the 1900 to 2000 increase in greenhouse gases in the atmosphere). The change in global average temperature for these simulations (increase of about 1°C) over 100 years is close to the measured change in global average temperature over the period from 1900 to 2000. When these same models are run without increasing greenhouse gas concentrations for 100 years, they do not show warming. Thus, according to climate models, most of the measured increase in global average temperature since 1900 was caused by increasing greenhouse gases, and the increase would not have happened without human emissions of greenhouse gases.

However, the model's representation of Earth's climate is not reality. There no point during a simulation where the model's representation of the climate matches a known measured state of the climate. If you do try to initialize the model to some point in history, it will revert back to the model's perception of climate. Thus, while climate modelers were able to reproduce the overall change in global average temperature after adding greenhouse gases within the model world, it requires a big leap of faith to believe the same overall change would happen in the real climate system simply due to the addition of greenhouse gases. One problem with climate model simlations is that they generally fail to reproduce the temporal (time) and spatial (position) variability that is known to happen in the real world, especially over small scales. This "natural variablity" is observed in nature but not so much in climate models. In other words, the models seem to underestimate natural changes in climate that happen without changes in greenhouse gases, and may overestimate changes in climate that result from changes in greenhouse gases. It is quite possible that the observed increase in temperature over the last 100 years has been wrongly attributed to the increase in greenhouse gases by climate models. Perhaps the observed change in global average temperature was part of a natural climate change and not dominated by the increase in greenhouse gases.

In order to make this point, a few lines written by Dr. Kevin Trenberth a senior climate change researcher working for the National Center for Atmospheric Research, which was orginally posted on a blog from Nature Climate Change will be discussed. The following statements quoted from the blog, "In particular, the state of the oceans, sea ice, and soil moisture has no relationship to the observed state at any recent time in any of the IPCC models" and "None of the models used by IPCC are initialized to the observed state and none of the climate states in the models correspond even remotely to the current observed climate" both indicate that a climate model is not the real world. Therefore, starting a model in the year 1900 does not mean that the climate model looks anything like the year 1900. It just means starting the model with year 1900 greenhouse gas concentrations. It is misleading to believe that climate models have been able to reproduce the climate that happened from 1900 to 2000. The claim is that the change in global average temperature simulated by the models in response to adding greenhouse gases is the same as the change in global average temperature in the real world. Quoting from the blog, "The current projection method works to the extent it does because it utilizes differences from one time to another and the main model bias and systematic errors are thereby subtracted out. This assumes linearity." There are plenty of scientists who disagree with this last statement that the "errors" will necessarily subtract out. Finally, considering the uncertainty in current model predictions, Trenberth said, "We will adapt to climate change. The question is whether it will be planned or not?" This indicates to me that unless we are able to improve the predicitons of current models, we are just going to have to adapt to climate changes without knowing what is coming. Instructor's Note: I include the quotes from Dr. Trenberth, who is an expert in climate modeling, because I have been criticized in the past for making similar comments about climate models.

Also, keep in mind that predictions of future climate change are much more difficult than simulating past climate changes where the answers are basically known beforehand. This allows modelers to tweak their models until they get the "known" answer. In this case the change in global average temperature that took place from 1900 to 2000 was known, and models could be tested and developed to get this answer. There is a question of whether or not the models got the "known" answer for the right reasons. Some argue that much of the warming over the last 100 years was due to natural reasons not related to greenhouse gas emissions and that the models do not properly simulate natural climate cycles. Therefore, in order for the models to reproduce the observed warming from 1900 to 2000, they were programmed to be overly sensitive to human emissions of greenhouse gases. The more sensitive the models to increases in greenhouse gases, the more warming that is predicted due to future emissions of greenhouse gases. Of course when making predictions of future climate changes, one does not know the answer beforehand. But if current models are tuned to be overly sensitive to greenhouse gas emissions, then future predictions of increases in surface temperature due to greenhouse gas emissions will be too large.

Climate models continue to evolve and improve due to increases in computing power and improved observations and understanding of the climate system. However, the predictions of current climate models should be viewed with scientific skepticism.

Predictions of Equilibrium Temperature after doubling CO₂

Keep in mind that there are many different climate models and that different models make different predictions about the future. The reason is that no model can fully represent all the complex process and feedbacks involved in the Earth system. Due to this uncertainty, we should consider all of them as possible outcomes of adding greenhouse gases. There is also a realistic possibility that all models will turn out to be wrong. Perhaps the climate changes due to adding greenhouse gases will be less severe than predicted by models or perhaps we will be surprised and climate changes will be more severe than predicted by current models.

The ability of global climate models to reproduce the observed surface temperature trends over the 20th century represents an important test of the models. While most climate models are able to reproduce the slight warming in global average surface temperature that has been measured since 1860, no model is able to correctly get the spatial patterns of temperature changes correct. In other words, the observed changes in climates at the scale of regional climate zones has not been reproduced by any climate model.

The Intergovernmental Panel on Climate Change (IPCC) has released a set of reports, which comprises the Fifth Assessment Report (AR5) on climate change. Material from the Summary for Policymakers will be referenced in this section. As reported in the 2013 IPCC Report, various modeling studies have suggested that a doubling of atmospheric carbon dioxide from it pre-industrial value of 280 ppm to 560 ppm, or its equivalent by incorporating the effects of increases in other greenhouse gases, will likely increase mean global temperatures between 1.5 and 4.5°C. It is also extremely unlikely that the increase in temperature will be less than 1°C and very unlikely greater than 6°C. According to the report, likely means probability >66%, extremely unlikely means <5%, and very unlikely means <10%. Recall that the non-feedback calculation for the increase in global average temperature after carbon dioxide has doubled from 280 ppm to 560 ppm is 1°C. This means that climate models used to produce the 2013 IPCC Report all have positive feedbacks with respect to changes in carbon dioxide, since the most likely outcome (from climate models) is for global average temperature to increase by more than 1°C. Instuctor's note. In my opinion, the IPCC places too much confidence in climate model simulations and fails to properly convey the uncertainties in the predictions made by those models.

You must understand that these studies were done by doubling carbon dioxide, then letting the model run until a new equilibrium climate state was reached. In reality, the increases in greenhouse gases happen over an extended period of time and the climate system takes some time to come into equilibrium. A big issue here is that the warming in surface temperatures tends to lag behind the increase in greenhouse gases. To a large degree, this lag is due to the large thermal inertia of the oceans -- in other words it takes a lot of energy to raise the temperature of the ocean water.

Let's try to break it down into understandable steps.

• Increasing the amount of greenhouse gases in the atmosphere, intially slows down the rate at which the Earth emits infrared radiation to space.
• Assuming no change in energy input (Energy the Earth absorbs from the Sun), the Earth is no longer in energy balance since energy in exceeds energy out.
• As long as energy in exceeds energy out, the planet warms. In particular the Earth's surface temperature will warm.
• As the Earth's surface temperature warms, the amount of radiation energy it emits increases. Since some of this extra radiation energy is lost to space, thus the amount of radiation energy that the planet emits to space starts to increase.
• The warming continues until the energy out increases to the point where it equals the energy input.
• The Earth's surface is 70% covered by Oceans. Oceans have a high thermal inertia, so the Ocean surface temperature is slow to warm up in response to the energy imbalance caused by adding greenhouse gases to the atmosphere. The two main reasons for this is that (1) water has a high specific heat (meaning you have to add a lot of energy to liquid water to change its temperature in comparison to many other materials), and (2) the surface temperature of the ocean is slow to warm up because the additional heat is able to mix somewhat through the entire depth of the oceans.
• Therefore the amount of warming of the Earth's surface (temperature increase) required to bring the system back into balance (energy input equal energy output) will take quite some time.

As a result of the delay induced by the oceans, some climate scientists do not expect the Earth to warm by the full 1.5-4.5°C (2.7-8.1°F) by 2060, even though the level of CO₂ is expected to have doubled by that time. Assuming this delay is real, we can make several conclusions here.

1. This delay provides an opportunity to reduce the magnitude of anthropogenic climate change before it is fully realized, if appropriate action could somehow be taken to reduce greenhouse gas concentrations.
2. Even if we were to stabalize greenhouse gas concentrations today, this delay means that the surface temperature will continue to warm until a new equilibrium is reached (this may take hundreds of years).
3. If we wait to act, still greater climate change will be expected, which will be more difficult or impossible to avoid.
4. The ocean surface temperature would also be slow to cool. Suppose in the future climate change gets out of control and we decide that we need to lower the average surface temperature of the Earth by somehow removing greenhouse gases from the atmosphere. The cooling would be very slow and lag behind the decrease in greenhouse gases. Thus anthropogenic global warming, if it takes place, cannot be quickly reversed.

Model Predictions of Temperature Changes for the 21st Century

In the 2013 IPCC Summary for Policymakers it is stated that the global average temperature change for the period 2016 - 2035 relative to the 1986 - 2005 period will likely be in the range 0.3°C to 0.7°C warmer. The models predict an average rate of warming of 0.17°C per decade over this period. The 2007 IPCC report stated that even if concentratrations of all greenhouse gases had been kept constant at year 2000 levels, a further warming of about 0.1°C would still be expected [due to the ocean delay]. Therefore, the near term prediction is for warming even if we reduce (or eliminate) emissions of greenhouse gases. The 2013 IPCC document addresses the issue of climate change commitment and irreversibility. If climate model predictions are correct, then each year that CO₂ increases means a commitment to some higher level of temperature in the future, which cannot be easily reduced or reversed without some way to remove large quantities of CO₂ from the atmosphere. Quoting from section E.8 of the 2013 IPCC Summary for Policymakers:

Cumulative emissions of CO₂ largely determine global mean surface warming by the late 21st century and beyond. Most aspects of climate change will persist for many centuries even if emissions of CO₂ are stopped. This represents a substantial multi-century climate change commitment created by past, present, and future emissions of CO₂.

The first IPCC report in 1990 projected a warming of global average temperature of between 0.15 and 0.30°C per decade from 1990 to 2005. This can now be compared to the observed value of 0.20°C per decade as determined from satellite measurements (see figures below). The success of this early prediction through the end of the 1990s led many to have confidence in the ability of climate models to predict changes in global average temperture due to increasing greenhouse gases. However, the relatively rapid warming trend that took place in the late 1990s seemed to slow down in the early 2000s, and this was not predicted by most of the IPCC climate models. The figures below are designed to compare the climate model predictions of global average temperature changes with the observed temperature changes from satellite observations and surface thermometers. The figure on the left is from a climate scientist who believes that climate models overestimate warming due to increased greenhouse gases, while the figure on the right is straight from the latest IPCC report, which tends to give significant weight to climate model predictions. Notice that the climate model temperature changes and observed changes are fairly consistent through the early 2000s but have since diverged. Most climate models have predicted far too much warming of global average surface temperature since the early 2000s. The observed slowdown in global warming through at least 2013 is sometimes referred to as "the pause in global warming." If you do a web search of that phrase, you will find a spirited debate about its existence. There are several plausible reasons for this discrepancy between climate model predictions and observed temperature changes after 2000. One is that the lack of warming since 2000 is due to natural variability. We know climates are always changing even without considering the influences of humans. There may just be a natural cooling cycle that is masking the model-predicted warming from additional greenhouse gases. A second possibility is that energy or heat is currently being stored in the deep oceans instead of warming the surface temperature. Since the deep ocean has such a high heat capacity, this would not immediately cause much of a change in ocean temperature, but could slow the rise in surface air temperature. It is possible that the models are incorrectly distributing where the extra energy (due to the radiation imbalance resulting from increasing greenhouse gases) is going in the climate system. A third possibility is that the models are wrong. Specifically, the real climate is not as sensitive to human emissions of greenhouse gases as the models predict. In other words, the real climate response to additional carbon dioxide may not be dominated by positive feedbacks.

Changes in global average temperature relative to the 1979 to 1983 average as predicted by 90 climate models (thin colored lines), one satellite-derived estimate, labeled as UAH (bold blue line), and one based surface temperature measurements (bold green line). The bold black line is the average temperature change for all 90 models. Surface and satellite temperature change estimates are through 2012 (Source)

This is figure 11-25A from the IPCC's Fifth Assessment Report, which shows changes in global average temperature relative to the 1986 to 2005 average. The bold black line shows the average changes for four different measured datasets of global average temperature. The individual colored, thin lines represent predictions from different climate models used by the IPCC after 2005. The light grey lines show predictions made prior to 2005. The last year for the observational data is 2012.

The figures shown above do not include the warm years of 2014 through 2016. Including these years does bring the observed temperature changes more in line with the mean predictions of the models (see this update to IPCC figure 11-25A from the Climate Lab Book). However, in 2015/2016 there was a strong El Nino event. Without going into details now, it is common for global average temperature to spike upward during strong El Nino years and then come way back down in following years. We will have to wait and see if that is what happens this time. It now appears that the peak warm temperature anomaly for this El Nino may have peaked early in 2016. Here is a link to a figure that shows monthly changes in the global average temperature of the lower atmosphere based on satellite observations from 1979 to February 2017. Notice the decrease in global average temperature following the El Nino warming events of 1998/1999 and 2010/2011. There has also been a decrease in global average temperature following the warm peak of the recent El Nino, but it does not appear to be falling as fast as it did following the previous El Nino events. We will have to wait and see what happens over the next year or so. More discussion of the record warm years of 2015/2016 is included in later lectures on observed changes in global average temperature.

Currently, the Intergovernmental Panel on Climate Change (IPCC) projects a warming of 0.3-4.8°C (0.5-8.6°F) in global average temperature for the last two decades of this century (2081 - 2100) relative to the average temperature for 1985 - 2005. Note that this is toned down somewhat compared with the previous (2007) IPCC report which predicted temperature increases in the range of 1.1-6.4°C. This updated estimate is based on the latest runs of what are considered to be the best global climate models AND the several estimates of future emissions of greenhouse gases. Keep in mind that we are never certain of future emissions of greenhouse gases, so various emission senarios are run by modeling groups. Part of the reason for the large spread in predictions of future warming is due to uncertainty in future emissions of greenhouse gases. The figure below shows the future changes in carbon dioxide concentration for the 4 different emission senarios considered by the IPCC. The model predictions shown in the figures above, include all four of the emission senarios shown below.

Future concentrations of greenhouse gases (in units of carbon dioxide equivalent greenhouse forcing) based on the 4 different emission senarios tested by the IPCC. Prediction of future climate changes depend heavily on the emission senario selected, especially after 2030 when the emission senarios diverge greatly.

The latest IPCC Report also claims that there is now higher confidence in projected patterns of warming and other regional-scale features, including changes in wind patterns, precipitation, and some aspects of extremes. In previous semesters, I would point out that while all climate models predict warming of the global average temperature (giving us relatively high confidence in that prediction), individual models were all over the place when you looked at regional (small spatial scale) changes in temperature and precipitation patterns (giving us low confidence in the ability of climate models to project regional climate changes). Since I do not have time to evaluate the claim that regional projections from climate models are more confident, I have just presented the material from the IPCC report. In summary, IPCC Fifth Assessment projects that warming in the 21st century will continue to show geographical patterns of warming similar to those observed over the past several decades, i.e., warming is expected to be greater over land than ocean, and the high northern hemisphere latitudes (Arctic) will warm more rapidly than the global average. In spite of what the IPCC report claims, climate models have major problems in reproducing the multi-decadal climate variability that is known (through observations) to take place at regional (ecosystem-level) scales. Since global scale changes can be considered as a summation and interaction of regional climate changes, many question whether current climate models are even capable of accurately predicting future climate changes until they can better simulate natural variability and regional-scale climates. There is a local example of the inability of climate models to accurately simulate regional climate. The North American Monsoon is an important feature in the climates of Northwestern Mexico and parts of the southwestern US, including Arizona. The monsoon happens every year, although the timing and intensity of the monsoon varies from year to year. Current climate models do a poor job of representing this important and persistent climatological feature. Some of reasons for this are briefly discussed in the article Monsoon Modeling. There are those who question the ability of climate models to accurately predict future regional-scale changes in the summer climate of the North American Monsoon region given their inability to properly simulate the monsoon itself. In spite of the known difficulties, news headlines like this are common: Phoenix's Summer High Temperatures Projected to Be 10 Degrees Higher by 2100

Instructor's note. Simulation of regional climate changes continues to be an area of active debate. I attended a seminar recently in which the speaker argued that regional scale climate changes have a significant random component to them, i.e., over short time scales (decades), random and unpredictable changes in circulation patterns have a large influence on regional climate changes. If this is true, there may be a fundamental limit to how well regional scale climate changes can be predicted over a specific decadal time period, and perhaps some are expecting more from climate models with repsect to regional climate change than they can possibly deliver. However, one may be able to separate the random components of regional climate change from the global climate change forced by adding greenhouse gases. This again raises the possibility that climate models may be able to somewhat accurately predict the increase in global average temperature due to increasing greenhouse gases, but have little skill in predicting changes over smaller regional areas. In fact, the speaker was running "ensembles" of climate models, similar to ensemble weather forecasting that we covered earlier in the semester, which indicated a whole range of equally possible outcomes at regional spatial scales as greenhouse gases increase. The point is that we may have to change our expectations about the capabilities of climate models. The models may be able to accurately predict the change in global average temperature, but at best only be able to give a range of possible outcomes at regional scales. This regional uncertainty, of course, presents a problem for those who want to plan ahead for known climate changes. The information most applicable for future planning would be regional scale as opposed to global scale climate changes. Please consider these ideas when reading the next section and the next page.

Climate Model Predictions for Extreme Weather Events

Recall when we started this section on global warming and climate change, it was pointed out that the frequency and intensity of extreme weather events is probably more influential on the types of plants and animals that can survive in a given ecosystem than the average conditions. Therefore, any changes in the distribution of extreme events is an extremely important thing to monitor and predict.

Predicting the distribution of extreme events is a very difficult problem for climate models to answer. For one, extremes are by definition rare, which makes statistical conclusions far more difficult to draw. Another reason is that the wildest weather is often confined to areas that are smaller than global climate models can predict (typical horizontal resolution of climate models is about 150 km). In a sense this is just a re-statement of the problem mentioned above: climate model projections are more uncertain over small regional scales.

Because predicting and monitoring changes in the distribution of extreme weather events is so important to our understanding of the effects of global warming and climate change, research groups working with climate models are beginning to look at this issue. The following 10 indicies for extreme weather have been identified as target issues for climate prediction models (where available, I have added information available from the latest IPCC report):

• Total number of frost days
• Intra-annual extreme temperature range -- difference between the highest and lowest temperatures over a calendar year
• Growing season length -- begins after daily average temperature is above 41°F for 5 consecutive days; ends at start of 5 consecutive days when daily averge is below 41°F
• Heat wave duration index -- longest stretch of days with high temperatures at least 9°F warmer than the 1961-1990 average high temperature IPCC claims it is very likely (>90% Confidence) that there will be an increase in the number and severity of heat waves worldwide.
• Warmth of daily lows -- Percentage of time daily low temperatures fall above the 90th percentile of daily minimum temperature for 1961-1990 average
• Wet days -- number of days with at least 0.40 inches of precipitation
• Dry days -- maximum number of consecutive days with no precipitation
• Wet stretches -- maximum 5-day precipitation total
• Average daily precipitation intensity -- Annual precipitation divided by the number of days in which precipitation fell
• Influence of very wet days on annual total -- Percentage of annual total precipitation occurring on days that exceed the 95th percentile from 1961-1990 average. IPCC predicts a very likely increase in extreme precipitation events over most mid-latitude land masses as well as wet tropical areas.

Currently, we have less confidence in the ability of climate models to accurately predict the above indices as compared with predicting changes in global average temperature. Hopefully, with lots of hard work, scientists can improve climate models to better answer the important question: how might the distribution and intensity of extreme weather events change in response climate change resulting from human activities?