A partner project of the Hawai‘i Climate Data Portal

Home About Top Climate Datasets Future Climate Projections

Future Climate Projections

Future climate projections are essential for understanding what Earth’s future climate will look like. Researchers have used future climate projections to help answer questions like How much rainfall will we receive in the future? Will average temperatures change? Could flooding become more frequent? Future climate projections are simulations of Earth’s climate in the future, and they are vital for researching climate change and its potential impacts. Future climate projections are also an important part of understanding what climate change adaptation and mitigation strategies can be most effective.

The ASCDP provides several downscaled datasets for future climate projections in American Samoa

What is downscaling?

Global circulation models (GCMs) are used to project future climate across the world, although, they are not accurate for small, local areas. GCMs typically cover 100-300 kilometer areas per pixel, which is too large provide detailed projections of future climate across the unique features of local areas like the islands of American Samoa. 

High resolution climate projections are used instead to project future climate across local areas and include the diverse landscapes and climate patterns. Unfortunately, there are very few high resolution climate projections exist for American Samoa. 

Downscaling is used to create more high resolution climate projections. There are multiple types of downscaling including statistical downscaling and dynamical downscaling. Downscaling takes GCMs and makes their projections more detailed, so they can be used for smaller areas.

Overview of the process of downscaling. A GCM projection has a downscaling method applied to it, which allows the projection to have more information about what climate is like locally. The process of downscaling results in a higher-resolution projection of climate that is better able to project climate for small, local areas.

Existing Datasets

CHELSA CMIP5 and CMIP6

CHELSA CMIP5 and CMIP6: Climatologies at high resolution for the earth’s land surface areas (CHELSA) is a free, statistically downscaled high-resolution climate dataset which is available for multiple climate variables and time periods. Global scale climate projections from Coupled Model Intercomparison Project (CMIP) Phase 5 and 6 have been statistically downscaled to a 30 arc sec, 1 km spatial resolution. CMIP is a World Climate Research Programme project that designs and compares global climate model projections. Climate models are then assessed for quality and accuracy. CMIP has released groups of climate models which have been endorsed because of their accuracy. CMIP5 and CMIP6 are the most recent groups of models or ‘phases’ of CMIP. 

CMIP5 is the older phase of CMIP and uses different future climate conditions or scenarios, than CMIP6. Both CMIP5 and CMIP6 models are presented here, so we have a range of models with different strengths and weaknesses. By including CMIP5 and CMIP6, we can compare how different their projections of future rainfall are as well. 

If you’d like a short explanation of CMIP watch this video: A Short Introduction to Climate Models – CMIP – YouTube

Wang and Zhang, et al. (2016)

Wang and Zhang, et al. (2016): Wang and Zhang et al., 2016 dynamically downscaled CMIP5 model projections for two future climate scenarios (RCP 4.5 and RCP 8.5); producing island-scale maps of future precipitation and temperature for the island of Tutuila and Guam. The output climate projection data has a spatial resolution of 0.8 km.

Download CHELSA Data

Download Data

CHELSA CMIP6

CHELSA CMIP6 precipitation projections were processed and clipped to the extent of Tutuila. Projections are based on a 1981-2010 reference period, under the scenarios Shared Socioeconomic Pathway (SSP)1-2.6, SSP3-7.0, and SSP5-8.5, for the periods 2011-2040, 2041-2070, and 2071-2100. Projections of annual, wet, season, and dry season precipitation for American Samoa are listed below

2011-2040

2041-2070

2071-2100

CHELSA CMIP5

CHELSA CMIP5 precipitation projections were processed and clipped to the extent of Tutuila. CHELSA CMIP5 precipitation projections are based on a 1979-2009 reference period, under the scenarios Representative Concentration Pathway (RCP)4.5 and RCP8.5 for 2061-2080. Projections of annual wet season and dry season precipitation for American Samoa are listed below: 

CHELSA high-resolution, downscaled climate projections are available for multiple climate variables besides precipitation. CHELSA version 1.2 provides climate projections for average daily temperature, average minimum daily air temperature and average maximum daily air temperature, as well as bioclimatic variables.

2080-2099

Download Wang and Zhang Data

Download Data

Wang and Zhang

Wang and Zhang CMIP5 precipitation projections were processed and clipped to the extent of Tutuila. The output layers are linked below. Wang and Zhang's projections of precipitation are for the 2080-2099 period and are based on a 1990-2009 reference period.

2061-2080

Dataset Processing Steps

CHELSA CMIP5, CHELSA CMIP6 as well as Zhang and Wang’s projections of future rainfall were clipped to the extent of American Samoa in ongoing research titled “Analyzing the Usability of a Global Statistically Downscaled Dataset for Projecting Future Precipitation in Guam and Tutuila”. Each dataset’s projections were used to calculate absolute change and future percent change in rainfall across the island of Tutuila. This was done to understand the differences between dataset projections of future rainfall in Tutuila.

The outputs from this research are map layers showing future percent change in rainfall across Tutuila. These files can be downloaded from the Download section above, the American Samoa government ArcGIS page, or directly from GitHub. All the code used to complete these steps can also be found on GitHub. 

Read more about dataset processing steps

Processing Steps

  • Downloading Data
    • CHELSA CMIP5 and CMIP6 future rainfall projections were downloaded from the CHELSA website.
    • For CHELSA CMIP5, monthly rainfall data for 1979 - 2009, were downloaded, along with monthly projections of future rainfall from each model, time period, and scenario. It is important to note that all of the CHELSA CMIP5 data only projected rainfall across land areas and not over the ocean as well.
    • For CHELSA CMIP6, monthly rainfall climatologies for the historical period 1981-2010 were downloaded, as well as monthly rainfall projections for each of the three future periods and scenarios.
  • Clip all datasets to the extent of Tutuila
    • All of the downloaded layers were clipped to the extent of the island of Tutuila using a code written in the software RStudio.
  • Calculate Historic Rainfall Averages
    • A baseline or ‘historical’ average of rainfall across Tutuila was calculated first so that the amount of change in rainfall in the future could be determined.
    • A historical average of rainfall was calculated for each dataset, by adding monthly rainfall data.
    • Historic averages were calculated for annual rainfall, wet season rainfall and dry season rainfall. 
    • Each dataset has a different span of years (reference period) that is represented by the historical data. For example: The reference period for CHELSA CMIP5 is 1979-2009, which means that historical rainfall data is representing the average rainfall experienced from 1979 to 2009
CHELSA CMIP5CHELSA CMIP6Zhang and Wang
1979-20091981-20101990-2009

CHELSA CMIP5 and CMIP6 - Calculating Multi-Model Ensembles

Multi-model ensembles (MMEs) were developed for CHELSA CMIP5 and CMIP6 projections of rainfall. An MME is a group of climate model projections that are averaged together to make a single model projection. Developing an MME starts with a group of independently made models that project future climate. The climate projections of each model are then averaged to create an MME. The main goal of using MMEs is to improve the reliability of future climate projections.

  • All models have some level of uncertainty. By combining models, uncertainties and model weaknesses can be balanced by another model’s strengths.
  • Combining model outputs, like in an MME, can improve the accuracy of climate model projections.

CHELSA CMIP5

  • For the scenario RCP4.5, 36 CMIP5 model projections of rainfall were downscaled by CHELSA. The monthly data were added to create an annual average for each model output. The annual averages for each model were then averaged to create an MME for CHELSA CMIP5 (projecting rainfall for 2080-2099 under RCP4.5)
  • For RCP8.5, the same process was used but 37 CMIP5 model outputs of rainfall were downscaled by CHELSA.
    • MMEs were also developed for wet season as well as dry season rainfall so that average projections of rainfall during those seasons could be understood. 
CHELSA CMIP5 RCP4.5CHELSA CMIP5 RCP8.5
MME of annual rainfall under RCP4.5MME of annual rainfall under RCP8.5
MME of wet season rainfall RCP4.5MME of wet season rainfall RCP8.5
MME of dry season rainfall RCP4.5MME of dry season rainfall RCP8.5

CHELSA CMIP6

  • Only five model projections of rainfall were downscaled by CHELSA for CMIP6. For each of the three scenarios and three future periods, the same methodology as CHELSA CMIP5 was followed to create MMEs.
  • Nine MMEs were developed for each season, meaning twenty-seven MMEs were developed for CHELSA CMIP6 in total: 
SSP1-2.6SSP3-7.0SSP5-8.5
2011-2040Annual, wet season, dry seasonAnnual, wet season, dry seasonAnnual, wet season, dry season
2041-2070Annual, wet season, dry seasonAnnual, wet season, dry seasonAnnual, wet season, dry season
2071-2100Annual, wet season, dry seasonAnnual, wet season, dry seasonAnnual, wet season, dry season
Total MMEs = 27= 9= 9 = 9
  • Calculate the Absolute Change in Rainfall
    • Absolute Change: Measurement of the absolute change in rainfall between the reference (historical) period and future rainfall measured in millimeters (mm).
  • Calculate Future Percent Change in Rainfall
    • Future Percent Change: Change in rainfall between historic rainfall and future rainfall represented as a percent of the historical rainfall.
CHELSA Methodology Technical Details

To develop high-resolution precipitation data to 1 km spatial resolution, a statistical downscaling method was applied to CMIP5 and CMIP6 model outputs.

CHELSA CMIP5 (V1.2): 

Model projections were downscaled using an algorithm that incorporates orographic predictors including wind fields, boundary layer height, and valley exposition, as well as bias correction. To develop the precipitation projections of CHELSA CMIP5 included on the American Samoa Climate Data Portal, the ‘CHELSA_CMIP5’ and ‘Monthly precipitation climatologies’ products were used (as listed in V1.2 technical specifications). In their methodology, monthly precipitation climatologies were bias-corrected using GPCC Climatology Version 2015 gridded datasets (Meyer-Christoffer et al. in press). ‘CHELSA_CMIP5’, does not account for changing wind patterns, or temperature lapse rates and assumes them to be constant over time. Please see CHELSA V1.2 Technical Specifications (sections 2.4, 2.6, 3.1, and 3.4 specifically) for complete details on methods, bias correction, and algorithms used.

CHELSA CMIP6 (V2.1): 

Model projections were downscaled using an algorithm that incorporates orographic predictors including wind fields, boundary layer height, and valley exposition, as well as bias correction. CHELSA only downscaled the outputs of a select number of CMIP6 model outputs due to the large number of existing models and scenarios. Models were selected based on the Intersectoral Impact Model Intercomparison Project (ISIMIP). Model outputs were bias-corrected using a trend-preserving bias correction (Lange, 2019) before they were downscaled to 1 km spatial resolution. Some CMIP6 models show artifacts of spatial interpolation. These were caused by statistical downscaling of the models by ISIMIP3b bias adjustment, not CHELSA. The artifacts could not be removed. CHELSA V2.1 used an improved bias correction method for precipitation that wraps around the dateline and corrects for systematic gauge undercatch using (Beck et al. 2020).Please see CHELSA V2.1 Technical Specifications (sections 1, 2 and 6 specifically) for further details on methods, bias correction, and algorithms used.

Technical Validation

The results of CHELSA downscaling algorithm and the various bias correction steps applied were validated using statistical cross-validation. Statistical cross validation was used to compare CHELSA results with independent meteorological station data as well as comparable products that are available at similar temporal and spatial resolutions. Climatologies were validated in two steps. First, the CHELSA time series product that was used to create climatologies was validated, then the final climatological products were validated.

Results of validation showed CHELSA is an improvement over existing high-resolution climatologies, though errors do exist and are quantified in several ways.

  • Steps to validate precipitation climatologies included: 
    • Cross-validation of bias correction method using monthly station data
    • Validation of the orographic precipitation patterns used in the CHELSA algorithm
    • Comparison of fit between small scale stations and final climatologies through linear regression models
    • Validation of results using independent precipitation station data (unique from observational data typically used to develop other gridded datasets)
      • Independent precipitation stations used:
        • FAO Data: 2,316 stations
        • Mexico Data: 2,950 stations
        • Australia Ehyd Data: 877 stations
        • South AFrica SAEON Data: 14 stations
        • Scandinavia Nordklim Data: 11 stations
        • China CMA Data: 241 stations
    • Large-scale spatial comparison of precipitation patterns
    • Small-scale spatial comparison of precipitation patterns 

Links to CHELSA Resources

The complete methodology for CHELSA CMIP5 and CMIP6 products can be found in Climatologies at high resolution for the earth’s land surface areas. 

To download CHELSA CMIP5 and CMIP6 high resolution downscaled climate data directly:

Processing Zhang and Wang Data

  • Zhang and Wang pre-processed high-resolution dynamically downscaled monthly rainfall climatologies were downloaded. Monthly rainfall climatologies for the historic period of 1990 - 2009 (referred to as the present-day period in the final report) were downloaded along with projections for the period 2080-2099 for the scenarios RCP4.5 and RCP8.5. 
  • Calculate historical average:
  • Calculate future rainfall averages
  • Calculate Absolute Change
  • Calculate Future Percent Change

Appendix

Read more

Global Scale versus Local Scale Spatial resolutions

Spatial resolution can be defined as the size of a pixel in a raster layer (though this definition may vary). For example, a raster layer with a spatial resolution of 25 km will have everything within each 25 km x 25 km area represented by one pixel. 

A larger spatial resolution means that a larger area is represented by one pixel. The spatial resolution of a raster layer is described as ‘coarse’ or ‘large’ when the pixel size is larger. When the pixel size is smaller, the spatial resolution is described as ‘fine’ or ‘small’. For example, a layer with a spatial resolution of 100 km is much more coarse than a layer with a spatial resolution of 5 km.

Downscaling

There are multiple methods of downscaling to increase grid cell resolution, two common types of downscaling are dynamical and statistical.

  • Dynamical downscaling: A regional climate model is placed within a global climate model, informing the GCM local scale climate interactions. 
  • Statistical downscaling: A statistical relationship is defined between large-scale atmospheric variables and small-scale variables. 

The data needed to carry out each method of downscaling is different, which may limit the type of downscaling that is used for a certain application. Each method of downscaling has uncertainties. CHELSA high-resolution climate projections were downscaled using statistical downscaling methods. Zhang and Wang's high-resolution climate projections were downscaled using dynamical downscaling

Future Climate Scenarios

What is a CMIP and how does it help us understand future climate?

Coupled Model Intercomparison Project (CMIP) is a project of the World Climate Research Programme that designs and distributes global climate model simulations. CMIP compares and analyzes climate models, developing a group of endorsed models called CMIP-Endorsed Model Intercomparison Projects (MIPs). MIPs are included in phases CMIP; CMIP5 and CMIP6 being the most recent. The MIPs from CMIP5 and CMIP6 have contributed to the assessment of future climate in the International Panel on Climate Change (IPCC) Annual Reports (ARs).  

Projections from CMIP are included in IPCC annual reports, which are used internationally to inform governments about the state of climate change knowledge. IPCC Annual Report 5 (2014) used CMIP5 models and IPCC Annual Report 6 (2023) used CMIP6 models. 

Scenarios are an important part of climate change research. Scenarios help researchers understand the potential effects of near-term and long-term decisions as well as explore possible future climate conditions. Having a common set of scenarios allows researchers from across different disciplines to compare their projections of future conditions.

Representative Concentration Pathways (RCPs) are scenarios that were used in developing model projections for CMIP5. RCPs represent possible changes in climate for different futures. RCPs quantify the amount of radiative forcing that will be experienced up to 2100. A larger radiative forcing equates to a larger change in global mean temperature. RCPs were used to develop climate model projections for CMIP5 in IPCC AR5. The four RCPs are 8.5, 6, 4.5, and 2.6. RCP values indicate the radiative forcing levels by 2100.

Shared socioeconomic pathways (SSPs) are a set of scenarios that are based on five possible narratives of the future; sustainable development, regional rivalry, inequality, fossil-fueled development, and middle-of-the-road development. Possible energy and land use developments and associated greenhouse gas and air pollutant emissions are incorporated within SSPs.

SSPs were incorporated in the set of earth system models of the sixth coupled model intercomparison project (CMIP6). The scenarios used in CMIP6 are a combination of the five future narratives (SSPs 1, 2, 3, 4, or 5) and Representative Concentration Pathways (RCPs).  The four ‘top priority scenarios’ to be considered are SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5.

Future Climate Scenarios

Shared Socio-economic Pathways

SSP5-8.5: High emissions scenario, with development based on fossil fuel. 

SSP3-7.0: Medium to high emissions scenario under a regional rivalry scenario. Includes substantial land use change, high emissions, and high social vulnerability. 

SSP1-2.6: Low emissions scenario, under a sustainable narrative. Includes land use change, low vulnerability, and low challenges for mitigation. 

Representative Concentration Pathways

RCP4.5: Represents a future in which radiative forcing levels of 4.5 W/m2 are reached by 2100

RCP8.5: Represents a future in which radiative forcing levels of 8.5 W/m2 are reached by 2100

For complete descriptions of SSP scenarios, please see O'Neill, B. C., Tebaldi, C., van Vuuren, D. P., Eyring, V., Friedlingstein, P., Hurtt, G., Knutti, R., Kriegler, E., Lamarque, J.-F., Lowe, J., Meehl, G. A., Moss, R., Riahi, K., and Sanderson, B. M.: The Scenario Model Intercomparison Project (ScenarioMIP) for CMIP6, Geosci. Model Dev., 9, 3461–3482, https://doi.org/10.5194/gmd-9-3461-2016, 2016. 

For complete descriptions of RCPs, please see the following resources:

van Vuuren, D.P., Edmonds, J., Kainuma, M. et al. The representative concentration pathways: an overview. Climatic Change 109, 5 (2011). https://doi.org/10.1007/s10584-011-0148-z

Australian Government Department of the Environment and Energy, Coastal Adapt and National Climate Change Adaptation Research Facility. What are the RCPs?. Coast Adapt, n.d. 15-117-NCCARFINFOGRAPHICS-01-UPLOADED-WEB27Feb.pdf

Scenarios Utilized in High-Resolution Downscaled Climate Projections for American Samoa

Zhang and Wang et al. 2016 dynamically downscaled 12 CMIP5 model projections for two future scenarios

  • RCP4.5
  • RCP8.5

CHELSA CMIP5 statistically downscaled 36 and 37 CMIP5 model projections for the following scenarios:

  • RCP4.5 
  • RCP8.5

CHELSA CMIP6 statistical downscaled 6 CMIP6 model projections for the following scenarios:

  • SSP1-2.6
  • SSP3-7.0
  • SSP5-8.5

These great future climate resources were compiled and developed by Danielle Hall as part of a 2 year long study. A publication detailing all methods will be available soon. Until then a repository where all the layers for American Samoa can be found is accessible at: https://github.com/DaHall15/AmericanSamoa_FutureRainfallData.git.

For more information on Danielle's work please see her website here.