Climate Models

Models attempt to represent systems and processes. Models are critically important for scientific inquiry and for providing guidance to help inform decisions. While not all models are 'correct', they can often provide invaluable information. For example, weather forecast models to help shape decisions over the next several days. Climate models are used by scientists to help gain a more complete understanding of the climate system and its response to a variety of factors ranging from El Nino events, volcanic eruptions, and increases in greenhouse gas concentrations. Likewise, climate models are the primary tool used to anticipate longer-term changes in climate that can help inform adaptation strategies.

What are climate models?

Global climate models use mathematical equations that describe physics, biochemistry, and processes across the atmosphere, ocean, land surface, and ice. These models are like those used in developing weather forecasts. However, unlike weather forecasts, climate models are not designed to predict the weather on any single day or year. Rather these models are used to inform how climate will change under different future scenarios. These efforts include a coordinated set of experiments allow for comparisons across models.

Since models are simplifications of real-world systems and processes, we should not treat any individual model as being 'correct'. Several dozen climate models have been developed by government and university research groups across the globe. We have evaluated the reliability of a broad set of models to capture important climate variables across the northwest, and limited our focus to models for which all necessary data was archived. Our products allow users to explore scenarios from each model; however, we also summarize averages across models as this has been shown to provide a more representative view of projected climate change.

Usable Climate Projections

Climate models tend to simulate important variables that are important to shaping regional and local climate. Unfortunately these models to be coarse in their resolution and do not resolve factors important in determining climate at individual locations . However, there are ways that scientists have successfully sought to translate coarse resolution information from climate models to local scales thereby making this information more usable for decision-making. One approach that climate scientists use to develop more usable climate projections is a statistical process, called downscaling. The data presented here use the historical gridMET data as a baseline for downscaling to create a seamless set of data.


​Perhaps, but there is limited evidence that suggests "poorer" performing models simulate changes in climate different from "better" performing models. Rather, it is typically preferable to use a multi-model average when projecting future changes in climate (e.g., change in temperature). Climate models are evaluated for accuracy based on their ability to simulate historical climate. A study by Rupp et al. (2013) examined the ability of climate models to simulate characteristics of climate across the broader northwestern US.

A total of 20 climate models were selected from those that archived daily data for all variables considered and both low emission (RCP45) and high emission (RCP85) experiments as of January 2013. An overall ranking of the credibility of models are provided (link to table) to help inform sub-selection of models. However, we note that model credibility over the historic time period does not necessarily provide information regarding the credibility of future projections.


We advocate for the use of at least 10 climate models or an average across models inform future conditions. If resources are limited to choosing only a few models, it is useful to examine projected changes in climate for models you choose relative to the complete set. This can be facilitated by using our scatterplot tool. While we provide a range of projections from across 20 climate models, a couple caveats worth mentioning:

  1. The 20 models presented here are not intended to capture the full range of possibilities;
  2. Variations in climate projections stems from changes in greenhouse gas concentrations, the response of the climate system to those drivers, and natural variability.


The most up-to-date view of future climate come a coordinated set of experiments among climate models from the fifth phase of the Coupled Model Inter-comparison Project (CMIP5). Each modeling group performed the same experiment to facilitate an inter-model comparison. These experiments include both historical conditions and future conditions. The latter consider a range of potential changes in socioeconomic scenarios such as changes in population, energy, land use and globalization called Representative Concentration Pathways (RCPs). For the purposes of anticipating future changes in climate we considered both a low emissions future (RCP4.5) and a high emissions future (RCP8.5).


The Multivariate Adaptive Constructed Analogs (MACA) statistical downscaling approach was used as it provides several potential advantages in capturing characteristics of surface meteorology, particularly in regions of complex terrain. Learn more about the method by visiting the MACA homepage.


There are several ways to access the catalog of data. Depending on your exact interests we suggest the following:

  1. Access time series for selected locations or aggregated region (e.g., county-average) using Climate Engine:
  2. Download the full data over CONUS in netCDF format.
  3. ​Download a subset of the data using your favorite software program (i.e., MATLAB, Python, R, IDL) and an OPeNDAP connection to the data through NKN's THREDDS server. See our example on how this works.