A history reconstructed time series (1992-2011) of annual global NPP-VIIRS-like nighttime light data through a super-resolution U-Net model

The dataset provides an extended time series of Version 2 annual NPP-VIIRS nighttime light (NTL) composites, enabling long-term urban and socioeconomic monitoring from 1992 to 2024.
It consists of three components:
(1) Simulated NPP-VIIRS-like NTL data for 1992–2011, generated from stable DMSP-OLS NTL observations using a super-resolution U-Net deep learning model that integrates Landsat-derived NDVI to enhance spatial detail and temporal consistency;
(2) Annual NPP-VIIRS NTL Version 2.0 data for 2012–2021; and
(3) Annual NPP-VIIRS NTL Version 2.2 data for 2022–2024.
Both the V2.0 and V2.2 NPP-VIIRS products are the average_masked versions, and all pixel values less than zero have been set to zero.
The original NPP-VIIRS V2.0 and V2.2 data (2012–2024) were obtained from the Earth Observation Group at the Colorado School of Mines (https://eogdata.mines.edu/products/vnl/).
The dataset are 15 arc second grids, spanning -180 to 180 degrees longitude and -65 to 75 degrees latitude.Each zip file contains an NTL image in geotiff format. The filename of each image is "SRUNet_NPP_VIIRS_V2_Like_yyyy", where "yyyy" is the year of the image.The simulated NPP-VIIRS NTL data converted from the DMSP-OLS NTL data in 2012 and 2013 are also uploaded, which named "simulated_NPP_VIIRS_V2_yyyy".Based on the super-resolution deep learning technology, the dataset using DMSP-OLS NTL data and Landsat NDVI to reconstruct the NPP-VIIRS NTL data in the early years (1992-2011). The dataset could be easily updated by downloading the version 2 annual NPP-VIIRS NTL data in later years from the academic sector at the Colorado School of Mines (https://eogdata.mines.edu/products/vnl/). The dataset has good spatial pattern and temporal consistency, and is highly consistent with NPP-VIIRS NTL data. It provide a longer time period compared to existing products.Compared with the annual NPP-VIIRS NTL data in 2012, the dataset shows a good consistency at the pixel, city, province and nation levels with R2 of 0.617, 0.747, 0.874, and 0.964, respectively, while the R2 of the NPP-VIIRS-like NTL product provide by Chen (2020) is 0.512, 0.694, 0.86, and 0.966, respectively.