Input data set: Historical gridded Gross Domestic Product (GDP)



Protocol relation: Protocol
Data Type: Socio-economic
Simulation rounds: ISIMIP2a
Description:

Annual gridded GDP data for 1861-2016.

Scenarios: historical
Variables: gdp
Specifications

These data are based on the gridded GDP data originally created for ISIMIP2b. The years 2000-2016 are rescaled, with the aim to better match measured national GDP in these years.

Rescaling for ISIMIP2a extension:
In the original downscaling, 2000 was the last year where the input national time series were based on measured GDP. From 2010 onwards, SSP2-based national time-series were used as input. As SSP2 projections and actual developments differ somewhat for certain countries (e.g. CHN,IND), the years 2000-2016 were rescaled based on updated national time series until 2016 (also part of the ISIMIP2a extension).

Planned update:
It is planned to provide a new version of the gridded GDP data with the ISIMIP3 input data. This will be based on an improved version of the original downscaling algorithm, and will feature an updated transition between historical and SSP-based years.

Data source

Based on the methodology described here: http://www.cger.nies.go.jp/gcp/population-and-gdp.html and the ISIMIP2b ssp2 gridded population data.

See also:
Murakami, D. and Yamagata, Y. (2019) Estimation of gridded population and GDP scenarios with spatially explicit statistical downscaling, Sustainability 2019, 11(7), 2106; https://doi.org/10.3390/su11072106

These data have been generated by interpolating the data described here: http://doi.org/10.5880/pik.2017.007 from decadal to annual time steps and aggregating from 5arcmin to a 0.5deg grid.

Details of the rescaling method:
The rescaling is performed at the original resolution of 5 arcmin. As 2016 was not part of the originally downscaled years, a GDP grid for 2016 was obtained through linear interpolation between 2010 and 2020. Country- and year-specific rescaling factors were computed for 2010 and 2016 from measured national GDP time series (ISIMIP2a extended national data) and the corresponding SSP projections. When the rescaling factor was not available for 2016 due to missing GDP data, it was calculated from the most recent available year, or ultimately set to unity for a small number of countries.
The rescaling factors were then applied to the (downscaled) 2010 and (interpolated) 2016 GDP grids. The full 2000-2016 data is obtained via linearly interpolating between (2000)-(2010 rescaled)-(2016 rescaled), and then aggregating to the 0.5deg ISIMIP grid.
Combining this with previously provided data for 1861-2000 yields the full historical data for 1861-2016.

Caveats

- Discretization error: The original downscaling was performed at 5 arcmin. At the resolution of 0.5deg provided here, there will be larger inaccuracies when assigning grid cells to countries. This is especially relevant for small countries (e.g. island states) and/or countries with a lot of coastline (e.g. JPN). It is therefore not guaranteed that the sum of all grid cells belonging to a country at certain resolution will exactly match the national GDP.
- There are also a number of coastal cells that are marked as land in the 0.5deg land/sea mask, but fall outside of the country masks used in the rescaling. As the total GDP contained in these cells is small, they were assigned a value of zero for the rescaled years.
- Long-term trend only: As the years between the 10-year time steps of the original downscaling are based on linear interpolation, they do not accurately follow year-to-year variations in national GDP (e.g. the financial crisis).

Download Instructions

For ISIMIP participants, these files are available for download on the DKRZ server using the path
ISIMIP2a: /work/bb0820/ISIMIP/ISIMIP2a/InputData/landuse_humaninfluences/gdp/

Currently they are also available via an rsync facility (described here: https://www.isimip.org/gettingstarted/data-access/#for-external-non-participant-users), for which the password will be provided upon request (isimip-data@pik-potsdam.de).