¶ Summary
Eoscala (alongside its sister project, Velkscala) focuses primarily on the global historical modelling of economic data over time at a 5-arcminute resolution (4320x2160, WGS84 Equirectangular). These rasters are currently provided at 1000-year intervals from 10000BC to 1AD, at 100-year intervals from 1AD to 1700AD, at 10-year intervals from 1700AD to 1950AD, and at 1-year intervals from 1950AD to 2023AD. Eoscala's data is reliable to 2022AD, and Velkscala's data to 2023AD.
(Eoscala 1.0)
Whilst Eoscala's data is reliable to 2022AD, Velkscala is currently undergoing major methodological changes, and as such, has only been released at a 0.5 version, though it can still be utilised when upscaled.
The latest version of Eoscala is 1.1.
- Eoscala/Velkscala (Stable): https://github.com/Confoederatio/Eoscala-Velkscala
- Eoscala/Velkscala (Production): https://github.com/Confoederatio/Eoscala-Velkscala-Production
¶ Data & Papers.
Eoscala 1.2/Velkscala 0.7 (Better applied scaling)
Eoscala 1.1/Velkscala 0.6 (Refactor, Bugfixes)
Eoscala 1.0/Velkscala 0.5
¶ Decoding and Encoding.
RGBA rasters can be decoded as follows for each pixel: ((r << 24) | (g << 16) | (b << 8) | a) >>> 0, where the resultant integer contains the given cell value (i.e. population/GDP PPP, 100s). RGBA rasters may be encoded via an inverse function: r = (number >> 24) & 0xFF, g = (number >> 16) & 0xFF, b = (number >> 8) & 0xFF, number & 0xFF. Make sure to use these methods when interpreting pixel values.
¶ File Formatting.
Eoscala. Negative years represent BC, and postive years represent AD. 0 is used in place of 1AD.
Eoscala's file directories are divided into the following folders:
./eoscala/economic_activity_rasters- Organically modelled OLS-based proxies for potential economic activity at each time interval../eoscala/gdp_ppp_rasters- Gridmaps reflecting actual GDP PPP per cell, in 100s of FY2000 International Dollars.
Relevant model weighting data for Eoscala ML models, including the base OLS HYDE-SEDAC model, may be found in the ./models/ folder.
Regional subdivisions mentioned in papers may be found as follows:
Their JSON values are currently hardcoded in source, with future plans for modularisation.
¶ Known Errors.
Main article(s): Eoscala Errors, Velkscala Errors
A number of errors, mainly as a result of original coding errors to do with HYDE3.2 as well as Maddison are structurally present in Eoscala. We have done our best to ameliorate most of them, although some of them continue to persist and will be addressed in future updates.
Because of relatively lengthy forms of manual error correction, these original encoding errors have been made available as separate articles.