An Extended VIIRS-like Artificial Nighttime Light Data Reconstruction (1986-2024)

TL;DR

The paper tackles the challenge of generating a long-term, high-quality NPP-VIIRS-like nighttime light dataset for China by bridging DMSP-OLS and VIIRS observations. It introduces EVAL, a two-stage deep learning framework that leverages Landsat surface reflectance and high-resolution impervious-surface masks to overcome intensity underestimation and structural omission, producing a 1986–2024 500 m time series. Quantitative validation shows EVAL outperforms existing VIIRS-like products in pixel- and city-scale accuracy and yields stronger correlations with socioeconomic indicators like GDP and population across multiple periods. The dataset enables robust, long-range analyses of urbanization and development while remaining openly available for researchers and policymakers. This approach provides a scalable path toward unified, radiometrically consistent, long-term NTL records across large regions.

Abstract

Artificial Night-Time Light (NTL) remote sensing is a vital proxy for quantifying the intensity and spatial distribution of human activities. Although the NPP-VIIRS sensor provides high-quality NTL observations, its temporal coverage, which begins in 2012, restricts long-term time-series studies that extend to earlier periods. Current extended VIIRS-like NTL data products suffer from two significant shortcomings: the underestimation of light intensity and the omission of structural details. To overcome these limitations, we present the Extended VIIRS-like Artificial Nighttime Light (EVAL) dataset, a new annual NTL dataset for China spanning from 1986 to 2024. This dataset was generated using a novel two-stage deep learning model designed to address the aforementioned shortcomings. The model first constructs an initial estimate and subsequently refines fine-grained structural details using high-resolution impervious surface data as guidance. Quantitative evaluations demonstrate that EVAL significantly outperforms state-of-the-art products, exhibiting superior temporal consistency and a stronger correlation with socioeconomic indicators.

Figures (7)

• Figure 1: The architecture of our proposed framework, detailing the overall pipeline and its key components. (a) The overall pipeline consists of a construction stage and a refinement stage. The construction stage uses a U-Net-based architecture to generate an initial prediction via our specialized decoder (HFD). The subsequent refinement stage then performs fine-grained adjustments on this prediction using the refiner module (DFR). (b) The detailed structure of the HFD. It is composed of a Structure Residual Fusion (SRF) module, which intelligently incorporates fine-grained details from the encoder's skip connection, and a Multi-scale Aggregator (MA), which captures and adaptively fuses contextual features from varying receptive fields. (c) The detailed structure of the DFR module. It utilizes residual blocks and a Cross-Resolution Local Attention to fine-tune the reconstructed image, guided by high-resolution features.
• Figure 2: The EVAL product for the year 2012. The image has been contrast-stretched using histogram equalization for visualization purposes.
• Figure 3: The scatter plot comparing VIIRS-like NTL data with actual NPP-VIIRS NTL data at the city scale.
• Figure 4: Analysis of prediction results in major urban areas for the year 2012, specifically the megacities of Beijing, Tianjin, and Shanghai. Blue and red denote overestimation and underestimation, respectively, where color intensity is proportional to the magnitude of the deviation.
• Figure 5: Analysis of prediction results in rural areas and along road networks for the year 2012, focusing on Henan and Shandong—two provinces characterized by dense rural settlements. Blue and red denote overestimation and underestimation, respectively, where color intensity is proportional to the magnitude of the deviation.