Earth systems models of intermediate complexity (EMICs) are an important type of climate model. They are mainly used to study Earth's systems over long periods or when using less computer power. This is done by using lower time and space detail compared to more detailed general circulation models (GCMs). Because the connection between space detail and model speed is not straightforward, small decreases in detail can greatly increase model speed. This has allowed scientists to include systems like ice sheets and carbon cycle feedbacks that were not included before. These benefits usually come with a loss of some model accuracy. However, how much higher resolution models improve accuracy instead of just precision is still debated.
History
By the middle of the 20th century, computers had become powerful enough to create models that track how mass and energy move in both vertical and horizontal directions on a grid. By 1955, these advances led to the creation of what is now known as a basic GCM (Phillips prototype). However, even at this early stage, limited computer power made it difficult to run these models for long periods of time.
Over the next 50 years, computing power improved rapidly, but the need for more detailed models also grew quickly. To study smaller areas, models required smaller time steps due to the Courant–Friedrichs–Lewy condition. For example, if a model’s grid becomes twice as detailed in each direction, the computer needs to do 16 times more work. During this time, GCMs began solving more accurate versions of the Navier–Stokes equations and included more parts of Earth’s systems, such as ice, carbon, and clouds. These changes turned GCMs into models that connect different parts of Earth’s systems.
Because running these detailed models required powerful and expensive computers, many university research groups could not access them. This led to the development of EMICs. By simplifying key parts of the models, researchers could run climate simulations on less powerful computers or complete them much faster on similar computers. For example, the EMIC JUMP-LCM runs 63,000 times faster than the GCM MIROC4h. This reduction in computing needs allowed EMICs to simulate Earth systems over longer time periods, including slow processes like changes in ice and carbon.
Petoukhov’s 1980 statistical dynamical model is considered the first modern EMIC. Although EMICs were developed throughout the 1980s, their usefulness became more widely recognized in the late 1990s when they were included in IPCC AR2 as "Simple Climate Models." The term "EMICs" was first used publicly in 1999 at an IGBP congress in Japan by Claussen. The first model to use the term "intermediate complexity" is now one of the most well-known: CLIMBER 2. At a conference in Potsdam, Claussen identified 10 EMICs, a list that was later updated to 13 in 2005. Eight models contributed to IPCC AR4, and 15 contributed to AR5.
Classification
Climate models are classified based on their complexity, resolution, parametrisation, and integration. Integration refers to how different parts of the Earth system interact. This depends on the number of connections between components and how often they interact. Because they are faster, EMICs can create highly integrated simulations compared to more detailed ESMs. Four EMIC categories have been suggested based on how they simplify the atmosphere: statistical-dynamical models, energy moisture balance models, quasi-geostrophic models, and primitive equation models. Of the 15 models used in the IPCC's fifth assessment report, four were statistical-dynamical, seven were energy moisture balance, two were quasi-geostrophic, and two were primitive equation models. Examples of each category are provided in a case study.
CLIMBER-2 and CLIMBER-3α are versions of 2.5 and 3-dimensional statistical-dynamical models. These models use statistical knowledge of the atmosphere instead of solving equations like the Navier–Stokes or primitive equations. This method represents the atmosphere using large-scale, long-term patterns of wind and temperature. CLIMBER-3α has a much coarser horizontal resolution than typical atmospheric GCMs, with a grid size of 7.5°x22.5°. This resolution is too low to capture small-scale weather features. However, it includes detailed models for the ocean, sea ice, and biogeochemistry. Despite these details, the simplified atmosphere allows it to run about 100 times faster than similar GCMs. Both CLIMBER models can simulate current climates as accurately as modern GCMs, which is important because they use far less computing power. These models are mainly used to study past climates, especially ice sheet formation.
The UVic model simplifies how air and water move (using Fickian diffusion) and how precipitation occurs. It is based on earlier energy balance models, which reduce the atmosphere to three variables: surface air temperature, sea surface temperature, and specific humidity. By using diffusion to model heat and moisture movement, the model limits its focus to large-scale patterns over long timescales. A key feature of this approach is that the simulated climate does not show natural changes over time. Like CLIMBER-3α, it connects to a detailed 3D ocean model and includes advanced models for sea ice and land ice. Unlike CLIMBER, the UVic model has a resolution similar to modern AOGCMs (3.6°x1.8°). Its speed advantage comes only from simplifying atmospheric dynamics.
Quasi-geostrophic equations are a simplified version of the primitive equations first described by Charney. These equations apply when inertial forces are small, and the Coriolis and pressure-gradient forces dominate. This simplification reduces the primitive equations to a single equation for potential vorticity in five variables. The LOVECLIM model has a horizontal resolution of 5.6° and uses the quasi-geostrophic ECBilt atmosphere model. It includes a vegetation feedback module developed by Brovkin et al. (1997). However, the model has limitations tied to its design. It predicts an Equilibrium Climate Sensitivity of 1.9°C, which is lower than most GCM predictions. Its surface temperature pattern is overly symmetric and does not reflect the northern shift of the Intertropical Convergence Zone. The model performs less well in tropical regions. Other quasi-geostrophic models include PUMA.
The UK Met-Office's FAMOUS model blends features of detailed comprehensive models and EMICs. Designed for paleoclimate simulations of the Pleistocene, it was adjusted to match the climate of its parent model, HADCM3, by solving the primitive equations described by Charney. These equations are more complex than the quasi-geostrophic equations. Originally named ADTAN, early simulations had errors in sea ice and the AMOC, which were later corrected by adjusting sea ice parameters. The model runs at half the horizontal resolution of HADCM3, with an atmospheric grid of 7.5°x5° and an oceanic grid of 3.75°x2.5°. Atmosphere-ocean interactions occur once daily.
Comparisons and assessments
Scientists have compared EMICs since 2000, most recently through a group of scientists contributing to the IPCC's fifth report. The sensitivity of EMICs to changes in temperature and their response to warming conditions generally match the range seen in modern GCMs, which is 1.9 to 4.0 °C (compared to 2.1° to 4.7 °C in CMIP5). When tested over the past 1,000 years, the models' average results matched real-world trends, but individual models showed much more variation. Models often showed more heat absorption in the oceans than actually happened and indicated a slight slowdown. No link was found between how much the poles warmed, temperature sensitivity, and starting conditions in the models. Comparing EMICs to GCMs and ESMs does not fully show their value. Their ability to run as "fast ESMs" allows them to simulate much longer periods, up to thousands of years. In addition to running on much longer time scales than GCMs, they help in creating and combining systems that will eventually be used with GCMs.
Outlook
Future directions for EMICs may involve evaluating uncertainties and leading efforts to include new Earth systems in research. Because they work quickly, EMICs are well-suited for creating groups of models that help scientists better understand and limit Earth system parameters. EMICs have also played a key role in climate stabilization studies. In 2001, McGuffie and Henderson-Sellers suggested that EMICs might become as important as GCMs in climate modeling. Although this has not fully happened since then, EMICs remain valuable in this field. As climate science faces more attention, the ability of models to explain processes, not just predict outcomes, has become important. EMICs are helpful here because their structure makes it easier to identify and share cause-and-effect relationships, unlike complex models that produce features from many interacting parts.