GHSL Datasets information

Read below to know about the progress of the new update.

Tests with new independent reference data show that GHSL R2022A matches or outperforms other data sources for accuracy in epochs 2018 and 2020 and matches or outperforms also all the other single epochs (1975, 1990, 2000, and 2015) included in the previous release GHSL R2019. The accuracy of the time series and its change rates, however, are lower especially in the rural domain. According to the JRC internal tests, the anomaly is expected to introduce a positive bias in predicted change rates of built-up surfaces and built-up volumes after the year 2000. The positive bias is especially remarkable in the rural domain as set by the GHS-SMOD R2022A. Previous GHSL R2019 was affected by larger omission errors in rural areas compared to R2022A, therefore generating an underestimation of the change rates in the rural domain.

Thus, the use of the GHSL Data Package 2022 (GHS P2022) is currently not recommended for supporting multi-temporal studies and indicators including built-up surfaces, built-up volumes, and population, especially if stratified by GHS-SMOD grid class. Applications relying on data for epochs 2018 and 2020 are not affected.

A new multi-temporal model was developed and is currently being processed at global level. Due to the complexity of the problem, this took longer than anticipated. Once the GHS_BUILT R2023A release is finished, the downstream multi-temporal products including the population grids and the degree of urbanisation grids will be updated (GHS-BUILT-S, GHS-BUILT-V, GHS-POP, GHS-SMOD, and GHS-BUILT-LAU2STAT). A new GHSL R2023A data release is expected for April 2023. Versions of the GHSL data in WGS84 and GHS-DUC will be made available only based on GHSL R2023A.

We thank all GHSL users for your continued support and apologize for the inconvenience.