5. Simulating the Atmosphere

The atmosphere is a familiar part of our daily life, and the role of the atmosphere in climate features many aspects of weather that we observe every day.¹ Figure 5.1 illustrates a schematic of the atmosphere, highlighting some of the key aspects of the atmosphere in the climate system. This figure is related to the hydrologic cycle (Fig. 2.​3) and the energy budget (Fig. 5.2, reprinted from Fig. 2.​2). Most notably, the atmosphere is where the energy input from the sun is distributed in the climate system. It features clouds and precipitation, and with evaporation from the land surface, this represents the atmospheric hydrologic cycle. The entire atmosphere is in motion with winds that are part of a large-scale atmospheric circulation. Vertical motion is driven by buoyancy: the difference in density of air. Warm air is less dense than cold air, so it rises. As air rises, it expands and cools. But if it cools enough, water vapor condenses and forms a cloud. This releases heat, and then the air may continue to rise: This process gives rise to clouds, and to deep vertical motions in the atmosphere.

There are many variations in the atmosphere that occur in both space and time. Clouds may be only a few hundred meters in size. Temperature and winds vary from place to place, and over the course of a day. We experience these variations at quite small scales compared to the global scale, or even compared to the typical scale (62 miles, 100 km) of global models. Many clouds are small scale, and precipitation events may be very localized. These small-scale variations make it difficult to simulate and predict the future state of the atmosphere.

The atmosphere is structured into a boundary layer near the earth’s surface, often unstable near the ground (which can warm up rapidly due to solar insolation), then stable at some level a few hundred to a few thousand feet (1–3 km) above the ground. Above this is the “free troposphere” (tropos = changing). At about 40,000–60,000 feet (12 km, the altitude at which a plane flies), is the top of the troposphere, the tropopause. Above this the air stops getting colder with height, and begins to warm with height due to absorption of sunlight by ozone. This region, the stratosphere (stratus = layered), is highly stable and also dry: devoid of clouds. The “weather” we experience, and most of the important climate processes, occur in the troposphere.

We have already discussed the importance of energy flows in the climate system (see Chap. 2). Many of these flows occur in the atmosphere (see Fig. 5.2). Energy comes in from the sun and is absorbed, reflected, and transmitted by the atmosphere. We think about energy flow in the atmosphere in the vertical: Sunlight hits the surface or clouds, some is reflected depending on the whiteness (albedo) of the surface, and then thermal (infrared) energy is radiated back. The greenhouse effect of CO₂ and other greenhouse gases (water vapor, methane) traps more of the thermal energy and prevents it from escaping into space, adding energy to the “Emission to the Surface” arrow seen in Fig. 5.2.

But energy also moves horizontally in the atmosphere. We see the swirling cloud patterns on weather maps and satellite images, and these large-scale weather systems move different amounts of heat and moisture around the atmosphere. Now think about what this looks like from the ground. At any one spot, nearly the same amount of energy hits the top of the atmosphere, above the clouds, where the sun always shines. The amount is broadly the same from day to day (it varies slowly with the seasons). However, the temperature and local weather (e.g., precipitation, wind) vary a lot from day to day due to atmospheric motions. This horizontal motion of energy explains why global General Circulation Models (GCMs) are good at representing climate (and weather): If we know the state of the atmosphere on a given day, we can use the basic equations of physics to estimate what the state will look like the next day, and the next. Compounding of small inconsistencies over time makes the problem of exact prediction difficult (see Sect. 5.6), but knowing the laws of physical motion and being able to conserve energy and mass helps.

Figure 5.3 illustrates why energy moves horizontally. At any one place, even over a whole band of latitude, the energy is not in balance. Energy comes into the earth system from the sun, in the form of shortwave or solar energy. This energy is usually at visible and ultraviolet wavelengths. The short wave solar energy input, minus any reflection, is called the net input. The net solar energy input peaks in the tropics. In mid-latitude regions the sun is lower in the sky for part of the year, and in polar regions there is no short wave energy (no sun) during polar night. The earth also radiates energy away in long wavelengths (longwave terrestrial energy). Longwave energy is mostly in the infrared. This is more constant and related to the temperature of the surface. As a result, there is excess energy in the tropics (i.e., at 0° latitude), and a deficit at high latitudes (e.g., at the poles), which is clearly seen in the annual average in Fig. 5.3. This gradient keeps the poles cooler than the tropics. It also means there is an energy flow toward the poles. Some of this energy is carried by water.

As we discussed in Chap. 2, water has a significant effect on the energy budget and is important for this heat transport. It takes energy to evaporate water into vapor. This energy is released when the water condenses into clouds. In the tropics, there is lots of water in the atmosphere over the oceans, and lots of sunlight. Much of this water condenses locally and drives cloud formation and deep towers of thunderstorms. But some of the water is transported long distances with the wind. It may condense far from its source: over a continent, for example, or closer to the poles. When it does so, it releases heat, as well as releasing water. This heat changes the atmospheric temperature and is also critical for driving storm systems (as with thunderstorms in the vertical dimension). As indicated in Fig. 5.2, this heat represents nearly one quarter of the total energy that hits the surface, for example, 90 of 340 Watts (W) of energy for every square meter. A watt is a standard unit to measure the rate of energy production (joule is the energy, and 1 W = 1 J/s), whether it is measuring light bulb output/usage or energy hitting the land surface. So 90 W/m² is the energy of a bright incandescent light bulb over 1 m² (about a square yard, or 10 square feet).

Water transformations are one of the more magical and complex parts of the atmosphere. These transformations drive weather and are critical for climate. Much of the poleward transport of heat in Fig. 5.3 occurs through the evaporation of water from the tropical oceans, and the movement of that water poleward, where it condenses at higher latitudes. This also works on more regional scales. The paths and events in which air tends to flow poleward from the tropics with lots of water vapor even have a name: atmospheric rivers. Figure 5.4 illustrates a picture of the earth with infrared wavelengths that correspond to water vapor. Dark areas have little water vapor, white areas have a lot of water vapor. The water vapor streams out of the tropics, feeding storms at mid-latitudes (e.g., over S. America). In the mid-latitudes, the winds blow from west to east, and carry water from oceans onto continents. One key aspect is the release of heat: As condensation begins, it heats air, which then typically rises because warmer air is less dense. The rising air then cools, is replaced by air from below (also rising), and condenses more water vapor. The water in clouds will eventually fall as rain, but the condensation adds extra heat that drives the storms and forms a critical part of the general circulation of the atmosphere.

In Chap. 4, we discussed the level of complexity of models. There exists a hierarchy of models used to simulate the atmosphere, illustrated in Fig. 5.5. A box model is a simple model that has a single or small number of boxes (Fig. 5.4a). A box model has inputs and outputs and one temperature. It is used to model the energy balance of the climate system. A simple energy balance box model of the earth, with one uniform temperature, essentially assumes a uniform atmosphere.

More common are energy balance models with a several-layer atmosphere in a column (Fig. 5.5b). These generally have a realistic variation of temperature with height when the flows of energy from the top of the atmosphere to the surface are taken into account. Energy balance models are useful for examining changes in the composition of the atmosphere. Typically they exist solely in a single atmospheric column, designed to represent the whole planet or a region of the planet.

Energy balance models are a type of single-column model (Fig. 5.5b). The representation does not include geographic variation. Sometimes (as with energy balance models), the intent is to understand energy flows. But single-column models can represent any number of complex processes, except that they do not have horizontal motions. In Fig. 5.5b there are movements of mass and energy through a top and bottom boundary (as from the sun at the top and the land surface below), and there is a forcing at the side boundary (often an imposed wind speed), in this case varying with height. Such a model might be used to estimate the surface temperature given solar radiation and different greenhouse gases, or to simulate cloud and precipitation processes.

Often several different boxes or column models are tied together, and one box affects another. For example, a simple two-box model might try to represent the surface of the earth with a box representing the atmosphere over land and a box representing the atmosphere over ocean, as shown in Fig. 5.5c. Or there might be one box for the tropical regions and one box for the polar regions. These models may contain some sophisticated processes to represent the flow of energy, mass, and cloud formation that describes regional temperature and precipitation.

Sometimes a small number of boxes or columns are used specifically to represent conditions over large regions of the planet. There might be one column for each continent, and one column for each ocean basin, and one column over ice. An example is illustrated in Fig. 5.5d. The boxes are in balance internally and with each other, and they can exchange with each other and the boundaries. Such models are often called intermediate complexity models.² These models might have 20 or so columns over major land masses, to represent temperature and precipitation, that then drive economic models in different countries or regions. The oceans might be represented by one column for each ocean. Intermediate complexity models capture some of the basic energy budget components and limited circulation. These models are used when the needed climate variables are simple. Simple outputs might be the average surface temperature or precipitation in a region or over an entire country. These outputs can then be used to efficiently drive a country-level economic model.

Finally, we come to models that have finite elements covering the earth. These can be regional (a regional climate model³) or global. Regional climate models or limited-area models have boundaries: They do not represent the entire surface of the earth (see Fig. 5.5e). The limited area enables them to be run with finer resolution than a global model. Finer horizontal resolution is good for representing the effects of surface features like mountains (topography). If you want to understand the climate of a region near or within mountain ranges, it is critical to represent the effects of the mountains correctly. This may be easy to understand in terms of a small region like a mountain valley, or the region on the dry side of a mountain range away from the coast, such as eastern Oregon in the United States, or the high deserts on the Andes mountain range in Peru and Chile.

Limited-area models may also include more processes because they represent smaller regions with fewer boxes, a limited domain. The difficulty is that they have “edges” or boundaries: Air must blow into and out of them, along with the energy associated with the air, and water in vapor or in clouds (the “blowing” around is also called transport, or advection). Limited-area model boundaries (the region just outside the domain of the model) have to be defined from somewhere to determine what values are given to the model at the boundary. For examining present-day climate, the conditions at the boundary can be taken from observations, often using data collected for large-scale weather prediction. Indeed, many fine scale models of weather have limited domains. For climate prediction, limited-area models are difficult because the boundaries mean energy and mass can leave the model (nonconservation).

It is a challenge, however, to use these models for the future, since they need to have specified future boundary conditions. The results are often or usually strongly dependent on the boundary conditions specified. But these models, with their small grid spacing and detailed representation of processes, are good for representing details of local climate variation through weather events: extremes of precipitation and temperature, for example.

Models with a global grid are good for representing the overall patterns of motion of the atmosphere (see Fig. 5.5f). They have no horizontal boundaries. These global grids are known as General Circulation Models (GCMs). They need only a top (space) and a bottom (the surface of the earth) boundary condition. The bottom boundary is fixed, and the top boundary is generally fixed at some level between 25 and 50 miles (40–80 km) above the surface. This is well above the top of the troposphere. At these boundaries it is much easier to enforce the basic principles of conservation of energy and mass that are fundamental constraints on climate. If the top boundary is defined as a constant pressure, and air cannot flow through the bottom boundary, then the mass of air must be conserved. Energy can flow in and out at the top and bottom, but otherwise must circulate and move from one box to another. Because of these conservation properties, it has been natural to turn to GCMs for understanding and predicting climate. Conceptually they allow conservation to be achieved. In practice, this needs to be done carefully, but generally it is a tractable problem. Conservation is important because changes to climate result in small changes to the energy budget, and in a conservative model, the change in the energy budget would make the heat go somewhere: like into the thermal energy (temperature) of the surface.

So what does this general circulation look like? By circulation, we mean air motions that eventually must return to their starting point (circulate) because there are no horizontal boundaries, and there are vertical boundaries to the atmosphere. Because the air is not created or destroyed, it has to go somewhere, and other air takes its place; hence, it circulates. The earth has general circulation patterns (hence the term General Circulation Models). The general circulation gives rise to the basic distribution of rainfall and temperature: wet, dry, hot and cold regions, and, as a result, vegetation patterns (see Fig. 5.6). The easiest way to see the general circulation is to simply look at the earth from space. The tropical land regions within about 15° latitude of the equator are mostly dark and green: West Africa and the Congo, Indonesia and the Amazon basin are tropical rainforests, or what we call jungles. In the tropics, sunlight drives evaporation, and winds come together to cause air to rise and the water to rain out. Then the air sinks on either side of the equator in the subtropics. Rising air gets colder and promotes condensation and cloud formation; sinking air is dry, and does not form clouds or rain. The dry regions are dark in the water vapor image of Fig. 5.4, which was taken at the same time as the visible image in Fig. 5.6. In Fig. 5.6 you can see the brown, dry desert regions at the same latitudes on either side of the equator. The desert regions of North America are the same latitude (~30°N) as the Sahara in Africa, and in the Southern Hemisphere, the deserts of Australia are at the same latitude south of the equator (~30°S). This is also the latitude of the Atacama Desert in South America (see Fig. 5.6). At higher latitudes, the land gets green again in the middle latitudes (40°–60° north and south of the equator): Europe, North America, and Asia. Then, of course, in polar regions a new color emerges: white for snow, ice caps like Antarctica, and sea ice–covered ocean.

These vegetation bands generally result from upward and downward motion, conditioned by the basic laws of physics that govern the climate system (see Chap. 4). The exact latitude is a result of the energy input, and the rotation speed of the earth. Air moves upward in the deep tropics, causing clouds to form. It circulates poleward at upper levels into the subtropics. There air cools and descends in the desert regions of the subtropics. This is known as the Hadley circulation, after George Hadley, who came up with the basic framework of why the “trade winds” blow westward near the equator.⁴ Another broad cell extends over mid-latitudes, with rising motion there and sinking motion over the poles. The same effect creates the banded cloud patterns seen on other planets such as Jupiter and Saturn: bands of different colors, representing clouds at different altitudes due to the motion of the atmosphere in cells up and down, as on earth. GCMs exist for other planets as well, and these same equations can produce a tolerable representation of the banded cloud and wind structure on Jupiter, another useful test of the basic physics in a climate model.

Although the latitude bands are a guide, the circulation is not strictly by latitude: Not all regions at the same latitude have the same climate. Denver, Madrid, and Beijing all lie at the same latitude (40°N), but they have very different climates. Dry and moist regions are localized regionally, due to the contrasts between the land and ocean. Denver and Madrid are at relatively high altitudes, and Denver is in the middle of a continent, whereas Beijing is on the eastern edge of a continent. These geographic differences give rise to different climates.

In the tropics, upward motion of air is favored where it is warmer over land, and where the ocean basins are warmer. Air moves up in some regions, and down in others. Warm water near Indonesia and tropical land masses over Africa and South America have preferentially upward motion, with downward motion favored in the eastern Pacific. The tropical circulation pattern of upward and downward motion in particular regions is called the Walker Circulation after Sir Gilbert Walker, the longtime head of the Indian Meteorological Department in the early 20th century, who first charted many of the tropical circulations. Air also moves poleward at upper levels due to this heat input in the tropics, and eventually it descends in the subtropics.

Most modern GCMs do a good job of broadly reproducing these different dry and wet regions, or climate regimes. The driving forces are from the largest scales: land and ocean contrasts, the rotation of the earth. But getting the details right is critical for understanding how things evolve and will change at any given place. If the locations of the regions of precipitation (such as the edge of the tropical wet region) is off by a “small” amount (a few hundred miles or kilometers), this may mean a vastly different climate at a particular place. Billions of people live in the subtropics, which lie both north and south of the equator (between ~10° and ~30° latitude). The subtropics include large swaths of India, Southeast Asia, China, Africa, and North America. We discuss the potential shifts in climate regimes further in Chap. 10, when we consider uncertainty. But society’s vulnerability to climate—or to say it positively, the extreme adaptation of human populations to climate—makes understanding and being able to simulate climate critical.

So how is a GCM constructed? The basic concept involves coupling the different physical processes (clouds, energy flows, exchanges with the surface) with the basic laws of motion as discussed in Chap. 4. Some of these processes are transformations (like clouds), and some are external processes that push (force) the model. The laws of motion provide the resulting distribution of winds (motion, kinetic energy) and temperatures (thermal energy). The motion moves water vapor, clouds, and chemicals in the air. Physical and chemical processes describe transformations that determine the sources and loss processes of water vapor, cloud water/ice, and chemicals. Loss processes are sometimes called sinks for the water that flows down a drain. As described in Chap. 4, the first step is to calculate all the processes and determine the sources and sinks of critical parts of the atmosphere (water, clouds, chemicals). Each process in each box is parameterized, shown in Fig. 5.7a. The result is sources and sinks, along with energy transfer by radiation and by surface fluxes (shown in Fig. 5.7b). These sources and sinks are then used to “solve” the equations of motion, energy, and mass conservation on a rotating sphere (Fig. 5.7c). This step provides winds and temperatures that can be used to recalculate the processes; hence, the model marches forward in time.

As discussed in Chap. 4, an atmosphere model is a component of the climate system. It contains a representation of the equations of motion and conservation (Fig. 5.7c), as well as a representation of physical processes in the atmosphere (Fig. 5.6a, b). These physical processes must be represented in individual boxes (Figs. 5.7 and 4.​3) at each location on the planet based on a grid of points (Fig. 5.7).

Now think for a second about the global grid in Fig. 5.7. The latitude-longitude grid has about 30 different latitudes (or a resolution of about 6°). At this scale, there are four grid boxes that encompass all of Japan, with each including part of the ocean around it. A typical modern model has a finer grid than this: maybe 1 or 2 degrees of latitude. But that is still 68 miles (110 km) on a side. Now think about a sky: The entire sky you can see from near the ground (from a hilltop or a tall building) in all directions is not much more than that distance. So if that represents one column of air, with one value for each layer in it, what happens when the sky is partly cloudy? What happens when there is a thunderstorm that occupies only part of the grid box? What happens when it rains in only part of a grid box?

More generally, what do we do with variations across the grid box, when one number will not do? Many processes in climate models have ways of dealing with this below-grid-scale (“sub-grid”) variability. It is one of the most difficult problems of all scales of modeling: What do you do with variations that occur below or even near the grid scale?

The sub-grid scale problem is mostly variations with scales near the grid scale. If the grid is 62 miles or 100 km (1° latitude), then sub-grid variations are greater than 0.6 miles (1 km). Very fine scale variations (much less than 1 km in this example) can often be represented with statistics, because the scales of interest are well separated from the grid. A 100-km grid has 10,000 one kilometer square elements (10,000 km²). It has 1 million elements that are 100 m by 100 m (about two American or European football fields next to each other). Often we can represent a distribution of elements (a probability distribution function of small-scale features in space). But cloud systems like thunderstorms are often 5–20 km on a side and there might be just a few small-scale features in a grid box. So it becomes difficult to generate good statistics with only 25 elements in a 100-km grid box. Even at the 1 km scale, clouds will vary quite a bit. Even over the football field scale (300 ft, 100 m), small clouds have important variations. But they may be captured statistically since they are now quite smaller than the grid scale, and there are 1 million of them in 10,000 km².

If the statistics are not well known, because there are a few elements due to large size or rarity, then the statistics are not well sampled. This often happens when the scales of variability are not well separated from the grid scale (they are a significant fraction of the grid size). This is particularly a problem for clouds in the atmosphere, and clouds are particularly important for climate.

We return to this problem of subgrid variability later, but for now it is apparent that this variability often makes determining the tendency and forcing terms in models very complex, and it is one of the central challenges to climate modeling. It is an acute challenge for all components, but especially for the atmosphere. We often can only approximate the terms (statistically, for example) and cannot represent them explicitly: Hence, we build parameterizations or submodels of the different processes (also based on physical laws) and link them together.

Figure 5.7c shows the step where the changes due to all the different processes are applied to basic physical equations. These equations are used to determine the winds and temperatures at every model grid box that will result from all the different processes, like radiation and clouds. This is often called the dynamical core of the model (Fig. 5.7c). The laws of physics and fluid motion describe what happens to a compressible fluid (air) when it is pushed, or heated. The temperature of the air helps determine where there is higher pressure (more density, often colder) and where there is less density (warmer). The equations describe how air tends to move from higher density to lower density. The wind we feel is just air motion.

Changes in water (condensation, evaporation) are important in altering the heat content of the air (temperature). Other forces that pull on the air include drag, or friction from the surface (more over mountains and rough terrain than over smooth terrain, and varying with vegetation). This list of forcing terms is put into a set of equations to determine the response of the air to these changes (e.g., in winds or in temperature). Conservation of mass and energy are also applied. The winds are used to transport substances in the air: cloud drops, water vapor, and other substances such as ozone or small particles like dust (small particles in the air are known as aerosol particles).

Each of these substances then has a balance equation. Think of a box with a pile of balls in it, with each ball representing a molecule (or many molecules) of water, dust, or a chemical. Some are pushed in and out of the box in different ways (moving to a box adjacent in the horizontal, or moving up or down in the vertical), and some balls may move out of the bottom of the atmosphere (or come from the surface into the atmosphere). Some times the balls change category: from water vapor to cloud water for example. But all the balls need to be accounted for using the calculated wind and the laws governing fluid flow. The laws take into account the rotation of the earth (which affects the weather patterns on large scales, by altering the momentum of the air).

Let us put this all together and walk through a single time increment, or time step, in an atmosphere model. The basic structure of an atmosphere model is to break up the atmosphere into pieces representing a part of the earth’s surface. For each grid box on the earth’s surface, there is a column of air (see Fig. 5.7). The processes or forcing terms due to clouds, precipitation, chemistry, and turbulence local to a grid box are estimated using parameterizations (Fig. 5.7a). The physical laws for radiative transfer (the flow of energy) and conservation of mass are applied locally to the column (Fig. 5.7b). Then the changes in the local quantities are applied as forcing terms to equations that describe the motion of air on a rotating sphere, or the dynamics of the model (Fig. 5.7c). The equations provide estimates of the change in temperature and wind, in addition to the changes of each substance such as water or cloud drops. The updates are applied and the process begins again.

So what are these forcing terms that push the model dynamics? They are the input (and extraction) of heat, and the changes (and transformations) in the different chemical species, such as water vapor, represented in Fig. 5.6a, b. In an atmosphere model, we need to understand the sources and sinks of water vapor and the flows of heat in the system that drive the laws of motion. These forcing terms are made up of different physical processes: each one itself a complex set of different effects. The ones we discuss in the sections that follow are clouds, radiative transfer, and chemistry.

We discussed in Chap. 1 some of the differences between climate (the probability distribution) and weather (the location on the distribution). So what is the difference between a weather model and a climate model? The main difference is in the way the models are set up and run (integrated forward in time). The climate model starts up with some state of the atmosphere. For “climate” purposes, the initial state is not important: If you are looking for the distribution of possible precipitation events in 50 years, then it is independent of the initial state. But if you are looking for the distribution of precipitation events in space at a particular time (like tomorrow), then the starting point is important.

Both weather and climate models use similar equations, and similar processes. Because the simulation time is shorter, weather models can be more complex. Weather models often have finer horizontal resolution to better resolve topography and often contain more details of processes. They are often just regional models. If you are predicting the weather only a few days in advance, knowledge of the winds at the edges of the model at the present time is often sufficient. They also need not have absolute conservation of energy and mass, since small leakages over a few days do not affect the weather (if they are not related to weather events, like clouds). In this respect, many weather models have more detailed processes at higher resolution, but they are not good climate models because they do not conserve energy or mass.

A key difference between weather and climate models lies in how they are started, called the initialization. For short-term weather, it matters very much that the state of the atmosphere is as realistic as possible, and for this, weather forecasters worry a great deal about how observations (from surface stations, weather balloons, and satellites) are fed into the model. The evolution of weather systems depends on accurately understanding the state of the atmosphere. Errors in the description of the temperatures and winds contribute to errors in weather forecasts. So great care is taken to minimize the errors in the observed state. Errors are minimized by correcting observations and by taking into account as many observations as possible. In specific cases, new observations are taken in critical areas. A good example of additional observations are the “hurricane hunter” aircraft that fly around tropical cyclones (hurricanes in the Atlantic) and take additional observations. These observations are fed immediately into weather models to try to better predict the near-term evolution of a particular storm by reducing the uncertainty in its current state.

There is a spectrum of forecasts, ranging from short-term weather (1–5 days), to medium-range weather (7–12 days), to “seasonal” (next 3–6 months) or “interannual” (1–3 years) to “decadal” (5–10 years) forecasting. Beyond the short-term weather scale, most of the work is done by carefully initializing global climate models with observations. Basically, climate and weather models share similar sets of equations, and as computation power increases, they look more and more alike. Climate models are run at higher resolution more typical of weather models, and some weather models are run longer (7–12 days or even months) to do seasonal forecasting. The big difference is in the initial conditions: They are very important for weather models, but not so important for climate models. Climate models, if they are set up with detailed observations similar to weather models, do a comparable job of predicting the weather. And each set of models learns from the others. Climate models are adding more detail to their process representations (like cloud parameterizations), often using techniques from weather forecast models, and weather models are using techniques for conservation of mass and energy and transport schemes to better predict events longer than 5 days away. These parameterizations (such as for radiative transfer) often come from climate models.

The discussion of an atmosphere model here has been very reductionist. You would assume that simply getting each process correct, and adding them all together with sufficient resolution, would produce a reasonably correct result and that it is simply an exercise in “turning the crank” forward following the basic laws of physics. But this, of course, is not the case, and large uncertainties are present in simulating the atmosphere. These uncertainties come from several different sources: uncertain and unknown processes, representing scales, and the complication of feedbacks and interactions between processes in the system. Many of these challenges are also important for the ocean (Chap. 6) and the land surface (Chap. 7).

At the end of this and subsequent chapters we will present a section describing a particular application and use of climate models relevant to the topics of the chapter.

In the atmosphere, most of the impacts of climate change are felt through extreme events. Today’s extreme events provide a framework for developing integrated planning for climate change. In particular, tropical cyclones are one of the most difficult types of extreme events to predict and understand in the atmosphere. Short-term weather forecasting is difficult enough. It is especially difficult to estimate the future climate characteristics of storms. Because the events are extreme and rare, we really do not have enough statistics to generate a reliable present-day distribution of storms. So storms often surprise us, such as in the example below.

In late October 2012, Hurricane Sandy made landfall centered on the New York and New Jersey coastline. The tropical storm merged with a mid-latitude cyclone, forming a large storm system with impacts across large portions of the northeastern United States and Canada. Characteristically, this merger of weather systems leads to wide-scale damage away from the coastline.

Hurricane Sandy received widespread attention for a number of reasons. The storm had a large geographic extent and affected a heavily populated region. The intensity of Sandy was not extreme (Category 3 of 5, winds of 115 mph or 185 km/h). The storm surge in New York City and northern New Jersey was record breaking, however. The resulting monetary damages in the United States were second only to Hurricane Katrina, which hit New Orleans in 2005. Sandy took an unusual path, with a sharp westward turn. This was well predicted a few days in advance by different forecast models. The forecast model of the European Center for Medium-range Forecasts predicted the storm track change two days earlier than the Global Forecast System of the U.S. National Weather Service. This discrepancy in forecast skill was widely reported, with public and policy consequences. Furthermore, there was much public discussion that the unusual path of the storm was consistent with hypotheses that the movement of weather systems is responding to global climate change. These factors helped to invigorate public discussion about climate change, including the discussion of science-based uncertainty, and a discussion of using models for prediction—in this case, weather forecast models.

The trajectory of Sandy at landfall was verifiably unusual, and it contributed to the large amount of damage associated with the storm. There was major environmental and infrastructure damage. There was significant environmental damage due to failures of sewage systems, making explicit the relationship of infrastructure, environment, and weather. There was extraordinary damage to transportation systems, residential structures, and commercial property. Because New York City is important to the global economy, economic impacts were amplified. The broad extent of the hurricane impacts revealed different levels of preparedness and response as well as relationships between policy and practice.

Much of the damage of the storm was related to the flooding, and specifically the storm surge of wind-driven seawater into New York City and towns along the New Jersey coastline. The region has seen sea-level rise on the order of 1.5 feet (0.4 m) since 1900. Some of this sea-level rise is attributable to climate change. Other factors include sinking of the land and variability of oceanic and atmospheric processes (see Sect. 6.​7). In the next decades, the rate of sea-level rise is expected to increase; climate change will come to dominate the other causes of sea-level rise. The certainty of sea-level rise and the role of flooding due to hurricanes can help to manage planning for future tropical cyclones. Sandy is also a good example of how the different parts of the climate system (oceans and atmosphere) interact to create impacts.

There is no doubt that tropical storms will continue to have severe impacts on both the built and natural environment. There is no doubt that increasing sea levels will amplify the risks related to tropical storms. The risk due to sea-level rise is not limited only to tropical storms. Wintertime mid-latitude storms (called Nor’easters in the eastern United States) have some similar impacts to hurricanes. Hurricane Sandy demonstrated that the path of a geographically large, but not necessarily intense storm, can have large and costly impacts. Planning for increasing risk and frequency of storm-related flooding due to sea-level rise is warranted. Detailed scientific arguments about the frequency and intensity of tropical storms have only incremental impacts on most planning.

Sandy also exposes the importance of monitoring the emerging observational climate research to understand how it aligns with model projections. For example, two recent Pacific cyclones, in November 2013 (Typhoon Haiyan) and March 2015 (Cyclone Pam), had record winds at landfall, providing anecdotes of possible strength increases. The flooding of the island nation of Vanuatu by Cyclone Pam is viewed by many as a preview of a future with higher sea levels. The large diameter of Sandy and a number of other recent storms is consistent with the prediction that the geographic extent of storms is increasing.⁷ In addition, evidence indicates that, in the Northern Hemisphere, tropical storms are having influence farther north than in the past.⁸ Similarly, there is growing discussion that tropical storms are occurring earlier and later in the year,⁹ consistent with a warming planet. There is additional risk if tropical storms are more likely to be present in areas where they were previously absent. These risks need to be taken into account by decision makers in their planning, and scientists need to focus on scientific investigations that can quantify and reduce the uncertainty in future projections of the characteristics of tropical cyclones.

Climate models of the atmosphere come in many shapes and sizes, ranging from simple idealized models with no dimensions, to complex models with high spatial resolution over particular regions or the whole planet. Both regional and global climate models are valuable. Regional models are often coupled to other scales to perform detailed experiments. The processes represented in an atmosphere model include the equations of motion and the various forcing terms that come from energy flows, motion, and transformations. Moisture is critical in the transformation and storage of energy.

Models have to approximate processes occurring on many scales with parameterizations to derive the source and sink terms for the equations of motion and heat that are iterated forward in time. Key processes include clouds, radiative flows of heat, and chemistry, as well as small-scale forcing of motions. Representing the variations in processes on small scales correctly can be difficult. Most of the uncertainty in the models results from missing or incorrect processes and particularly from the interactions across scales that cannot be represented. Feedbacks between processes can be complex and are not easy to observe. Cloud feedbacks are highly uncertain in atmosphere modeling. We return to many of these ideas in Chaps. 6 and 7, on the other model components of the climate system.