Ask a Climate Scientist

stephanie mcaffee

This is a platform for helping people understand and apply Alaska climate projections in their work. Here, you can ask questions about how this information is produced and used in climate change research and adaptation planning.

Stephanie McAffee (left) is a climate scientist who can help you make sense of climate data. Common questions are below. If you have another question for Stephanie, please send it to us, using the form below. 

Don't see a form below? Try this link.

Some questions we are commonly asked

What do CMIP3/CMIP5 and AR4/AR5 mean?

Short answer: These are major global climate modeling efforts. We obtain data output from these models, and then add value to by downscaling them for our region.
Long answer: The Program for Climate Model Diagnosis & Intercomparison works to develop improved methods and tools for the diagnosis and intercomparison of general circulation models (GCMs) that simulate the global climate. One of their projects, the Coupled Model Intercomparison Project (CMIP), is an ongoing effort that informs the Intergovernmental Panel on Climate Change (IPCC) Assessment Reports (AR). CMIP3 informed the IPCC AR4 and CMIP5 informed the IPCC AR5 (they skipped using the name “CMIP4” in order to align the CMIP and AR numerical values). So, CMIP3/AR4 and CMIP5/AR5 are paired up in relation to the data provided and which report it informs. And yes, there’s a CMIP6/AR6 in the works, too.

What projection is the most likely to occur?

Short answer: We really don’t know, so that’s why we always evaluate multiple emissions scenarios.
Long answer: Over the long term, projected climate change is primarily a function of the balance between greenhouse gas emissions and sequestration. Emissions levels are driven by the size of the population, how much energy each person uses, and the energy sources used (e.g., coal vs. wind). Population, per capita energy use, and energy source are influenced by economic, technological, political, and cultural considerations. Sequestration also depends on various social, physical and ecological factors. With current technologies and policies, following the lowest emissions scenario evaluated—RCP 2.6—doesn’t seem likely. Both global and per-person greenhouse gas emissions have been increasing. Surprises, both good and bad, are always possible. For example, rapid permafrost thaw could release a lot of methane, a potent greenhouse gas, or new technological innovation could rapidly improve energy efficiency or make new, low-carbon energy sources available. More about historical emissions

Which are the best historical data to use?

Short answer: This answer depends on your research question and the study area because there is no one perfect product. You can obtain downscaled climate data from the SNAP data portal.
Long answer: All gridded historical data sets include uncertainty or biases because assumptions had to be made to produce grids from local weather stations and/or satellite data, either through statistical methods or with a model.

  • If you need high spatial resolution data (i.e., <20-km), then a downscaled high-resolution product, such as SNAP-CRU or dynamically downscaled ERA-Interim, will be necessary.

  • If you need high spatial and temporal resolution data and/or variables beyond temperature and precipitation, then you require dynamically downscaled products.

  • If you are not limited by spatial resolution and/or need variables beyond temperature and precipitation, you can look at historical reanalysis data for Alaska and the Arctic regions. Many reanalyses have been tested over the Arctic, and these studies provide guidance on which might best meet your needs (e.g., Lader et al. 2016 and Lindsay et al. 2014).

New historical gridded data and reanalysis products are continuously under development and should be rigorously evaluated against available observations in your region and time period of interest to determine the data set(s) that best meet your needs.

Which are the best downscaled data to use?

Short answer: Different ways of downscaling data have different strengths and weaknesses, so what’s “best” depends on the question you are asking. You can obtain downscaled climate data from the SNAP data portal.
Long answer:

  • If you need a range of future scenarios at spatial resolutions around 2-km x 2-km and can use monthly data, the delta downscaled data might be your best option.

  • If you need daily data, you might need to consider either the the quantile-mapped data or the dynamically downscaled data. Quantile-mapped data are available for multiple models, but are only available at 2.5 degree resolution (~275-km x 120-km). At the moment dynamically downscaled data (20-km spatial resolution) are only available for two models and scenarios, so this option may not be appropriate if you need to evaluate more than 2 future climates.

Another consideration is that statistical downscaling, which includes the delta and quantile-mapping methods, assumes that relatively small-scale climate patterns remain the same over time. For example, if it is 6.5°C cooler at 4000-m elevation than at 3000-m now, statistical downscaling assumes that it will cool a similar amount as you go upslope in the future. Dynamical downscaling allows these kinds of spatial patterns to change over time.

Why don’t your data exactly match the station record from X?

Short answer: There are two main reasons for the difference in the station data record and the gridded data generated by SNAP: point vs. area observations and gaps in the data record.
Long answer: First, a weather station records the climate at a single point. In contrast, gridded data represent average conditions over a grid cell that is substantially larger in area than the area immediately surrounding a weather station; the finest-scale data we provide have grid cells that are 771-m x 771-m. A weather station may be located on a hill or in a hollow where it is often a bit cooler, warmer, wetter, or drier than the average conditions of the grid cell within which it is located.

Second, there is no way to provide perfect data in places where the weather has not been consistently monitored. Some of the gridded historical data available from SNAP were generated by downscaling climate data from the Climatic Research Unit (CRU) 0.5 x 0.5 degree monthly data to the PRISM (771-m x 771-m) climatology (30-year averages for each month of the year).

The researchers who developed these products made the best choices they could to develop quality datasets from very few weather stations, many of which do not have long temporal records, or have gaps in the data. As a result, the weather station at X location might not have been used in either dataset (because it’s a relatively new station or because the record was missing too many measurements), and the interpolated climate from the nearest stations might not be a perfect match for that location.

Can you get a Pacific Decadal Oscillation (PDO) forecast out of your climate projections?

Short answer: No.
Long answer: The PDO is, as far as we know, a kind of unforced change. The projections we have downscaled were not made in a way that allows them to correctly simulate the timing of unforced changes such as shifts in the phase of the PDO or the occurence of El Niño and La Niña events. For more discussion on the difference between forced and unforced changes, see the question below about mismatches between observations and climate model output for the early 21st century.

Why don’t the 2000-2010 projections from the climate models match what actually happened?

Short answer: Unforced changes are not always temporally accurate to a specific 10-year time frame.  
Long answer: There are two main types of climate change: forced and unforced. The long-term warming trends that accompany increases in greenhouse gases are a type of forced climate change. The warm, cold, wet, or dry year that might occur as a result of a La Niña or El Niño event, or the warm or cool period that is associated with Pacific Ocean temperatures, are examples of unforced changes.

Climate models simulate both forced and unforced climate changes, but the timing of unforced changes in climate in models do not necessarily line up with real world occurrences of these changes. So, it is possible that a climate model could be warmer or colder (or wetter or drier) than that observed for any particular decade (or even two or three decades) because the actual atmosphere or ocean conditions (unforced changes) happened to not match regional atmosphere and ocean patterns in that particular model run. However, the trends over longer periods of time are more likely to agree with observations. It’s just that we’ll have to wait until the end of this century to make that validation.