Climate models: applications to understand past climates and climate change

Release Date:

This article is part of the Fall 2015 issue of the Earth Science Matters Newsletter

Effective climate change research requires a thorough understanding of the characteristics and interactions of the Earth’s physical, chemical and biological components as they affect climate on varying temporal and geographic scales. However, many important climate-related variables and processes in the Earth system are sparsely observed (e.g., evapotranspiration and soil moisture) or are unobservable (e.g., aspects of the global atmospheric circulation). Therefore, researchers rely on climate models to simulate climate information. Numerical climate models simulate processes and feedbacks within the Earth systems. While they have limitations, they are a valuable tool for discovering and understanding atmospheric processes and mechanisms, for investigating interactions of the Earth system, and for simulating climates of the past and future.

modelled topography of North America

Figure 1: Topography as represented by a GCM and RCM. The larger grid boxes in the map illustrate the topography of North America 11,000 years ago in a GCM that has a resolution of about 300 km × 300 km in latitude and longitude. The inset box, which is the domain of the RCM, illustrates topography on 45 km × 45 km grid boxes. The area within the RCM domain includes the Rocky Mountains in the southwest, the greatest heights of the Laurentide Ice Sheet (LIS) in the northeast and the large proglacial Lake Agassiz (blue arrow) along the southwestern margin of the ice sheet as reconstructed from the geologic record. The Great Lakes are located south of the eastern margin of the ice sheet. From Hostetler et al. (2000).

(Public domain.)

Climate models range in complexity from conceptual to highly detailed numerical models that are programmed in computer codes and run on supercomputers. Numerical climate models are derived from weather‑forecasting models which are built upon the physics of thermodynamics and fluid dynamics. The first numerical weather model was developed in 1904 by Norwegian physicist and meteorologist Vilhelm Bjerknes. Most models include fully coupled land surface components to simulate surface processes and their feedbacks to the atmosphere.  Many atmospheric‑land surface models also are coupled with atmospheric chemistry models and ocean models that simulate oceanic circulation, sea ice and sea-surface interactions with the atmosphere. With the emerging addition of ice sheet models, future climate models will achieve the full ability to simulate the atmosphere and surface of the Earth.

Two model types that are used widely in climate research are general circulation models (GCMs) and limited area or regional climate models (RCMs).  GCMs simulate the global circulation of the atmosphere and oceans and typically have a spatial resolution on the order of 100s of kilometers. They simulate global climate in response to prescribed, global-scale features, or boundary conditions, that include atmospheric greenhouse gas (GHG) concentrations (e.g., carbon dioxide and methane), incoming solar radiation (insolation) based on Earth-Sun geometry, land mass distribution, sea level and continental ice sheet extent. RCMs include physics that are similar to GCMs but are designed to be run over smaller areas (e.g. North America) at much higher resolution (10s of km or less).  RCMs are driven with output from GCMs to obtain higher spatial resolution of climate that captures, for example, the effects of mountain ranges, coastlines, vegetation, and lakes (Figure 1).

Climate models are used to reconstruct past climates when the models are run with prescribed global boundary conditions covering the period of interest. These reconstructions are known generally as paleoclimate simulations. For example, many paleoclimate simulations have been run of the Last Glacial Maximum (LGM) which occurred about 21,000 years ago. The boundary conditions for LGM simulations include carbon dioxide and other GHG concentrations that were less than half of present levels, insolation that was similar to present, expanded continental ice sheets that covered much of North America and Europe, and sea level that was roughly 120 m lower than present (Figure 2).  In response, the models simulate an annual average LGM climate that was globally about 5 °C colder and regionally drier than present. In contrast, simulations of the deglacial period, from 12,000 to 6,000 years ago, include increasing GHG levels, smaller continental ice sheets and Earth-Sun geometry that amplified seasonality of the Northern Hemisphere (increasing summer insolation and decreasing winter insolation), and sea level approaching present conditions.  The simulated global climate 6,000 years ago was globally about 1 °C warmer and regionally wetter or drier than present.

modelled Earth surface 21,000 years ago

Figure 2: The Earth surface at 21,000 years ago as represented in a GCM. The colors indicate the depth of the oceans, the elevation of the land and the elevation of the continental ice sheets over North America, Europe, Greenland and Antarctica.  (From Alder and Hostetler, 2015)

(Public domain.)

Paleoclimate simulations are linked with geologic data in several ways.  First, the boundary conditions for the simulations are derived directly from geologic records: GHG levels are known from analyses and dating of ice cores; the extent of the continental ice sheets is known from the geologic evidence left behind after the ice sheet’s retreat; and sea level is known through analyses of past shorelines and geophysical models. Insolation, based on Earth-sun geometry, is known precisely from computations developed in the 1920s by Milutin Milankovitch, a Serbian astronomer and mathematician. 

The second way that paleoclimate simulations are linked with geologic data is through comparisons of the simulations with records obtained from lake and marine sediments, ice cores, speleothems, sand dunes, glacial moraines, tree rings, packrat middens, and other natural sources that preserve proxy information about past climate conditions.  Such data-model comparisons are a cornerstone of paleoclimate research. They provide a well-established method for quantifying the accuracy of simulations of known past climates and thus improve our confidence in their ability to simulate future climate conditions. Additionally, data-model and model-to-model comparisons are fundamental for studying the mechanisms of climate change and for testing climate hypotheses that lead to an understanding of, for example, the sources of wet and dry cycles over North America during the Holocene (11,700 years ago to present).

Regional climate models are well suited to applications that benefit from higher resolution of climate processes than is typically possible with GCMs. They are particularly useful for exploring feedbacks between the Earth’s surface and the atmosphere.  The example illustrated in Figure 3 is from the study introduced in Figure 1 that focused on using an RCM to isolate the effect of large proglacial Lake Agassiz on the regional climate over the Laurentide Ice Sheet 11,000 years ago. The extent and depth of Lake Agassiz prescribed in the model was based on reconstructions from geologic records of shorelines, land surface elevation changes associated with the melting of the ice sheet, and the margin of the LIS. The patterns of the simulated winter and summer temperature and annual precipitation climatologies clearly reflect the high resolution of the RCM and the ability of the model to capture the influences and feedbacks associated with the complex paleogeography in the region.

modelled regional temperature and precipitation 11,000 years ago

Figure 3: Regional temperature and precipitation climatologies 11,000 years ago as simulated by the RCM (model domain indicated by the inset box in Figure 1).  From left to right: a) average January air temperature, b) average July air temperature and c) average annual precipitation.  From Hostetler et al. (2000).

(Public domain.)

GCMs and RCMs similar to those illustrated here are also applied extensively by international modeling centers to simulate future climates under prescribed changes in GHGs, land use, ice sheets and vegetation.  Many of the GCMs that were used to simulate future climate for the current Fifth Climate Model Intercomparison Project (CMIP5) have also been applied to simulate the climates of the LGM and mid‑Holocene (6,000 years ago) under the Paleo Model Intercomparison Project (PMIP3).

The broad range of research conducted in the  U.S. Geological Survey’s Climate Research and Development program (R&D) focuses on understanding local to global interactions of the hydrosphere, cryosphere, geosphere, biosphere and atmosphere over time periods ranging from millions of years ago into the future. The paleoclimate research of the R&D program contributes to our understanding of the rates and magnitudes of past climate change which provides a context within which to view present and potential future climates.  Combining climate models with data-based climate reconstructions provides a way of discovering and understanding the mechanisms of climate change and explicitly tests the ability of the models to capture known changes in climate. In turn, this approach leads to model improvements and heightened confidence in the ability of the models to simulate future climates. 

References Cited

Hostetler, S.W., P.J. Bartlein , P.U. Clark, E.E. Small, and A.E. Solomon (2000), Simulated interactions between Lake Agassiz and the Laurentide Ice Sheet 11,000 years ago. Nature 405:334-337.

Alder, J. and S. Hostetler (2015). Global climate simulations at 3,000-year intervals for the last 21,000 years with the GENMOM coupled atmosphere-ocean model.  Climate of the Past 11, 449-471 doi: 10.5194/cp-11-449-2015.

<< Back to Fall 2015 Newsletter

Related Content