Aligning Pixel Values of DMSP and VIIRS Nighttime Light Images to Evaluate Urban Dynamics

Kang Wu, Xiaonan WangRemote 2019 · DOI

TL;DR

This paper introduces a novel method to bridge the gap between two generations of nighttime light (NTL) satellite data (DMSP and VIIRS) by calibrating older DMSP data and simulating DMSP-like data from newer VIIRS data. This allows for a more continuous and consistent analysis of urban dynamics over a longer time period, showing improved correlations with economic indicators like GDP.

What Problem Does This Paper Address?

Nighttime light (NTL) satellite imagery is a valuable proxy for understanding urban dynamics, economic activity, and population changes due to its wide coverage and temporal resolution. However, a significant challenge arises from the differences between the DMSP (Defense Meteorological Satellite Program) and VIIRS (Visible Infrared Imaging Radiometer Suite) NTL datasets. These differences, particularly in pixel value ranges and radiometric resolution, make it difficult to directly combine or compare data from these two sources across their overlapping and sequential timeframes. This discontinuity hinders long-term trend analysis and the comprehensive study of urban evolution. This paper aims to overcome this by creating a unified, continuous NTL dataset that bridges the DMSP and VIIRS eras.

Methods

The research proposes a two-pronged approach:

  1. DMSP Calibration (1992-2013): To address the lack of on-board calibration in DMSP NTL, the authors employed a robust regression algorithm using a quadratic polynomial model. This method iteratively fits a regression model to the data, identifies and removes outliers based on predicted vs. actual pixel values, and refines the model until a stable calibration is achieved. The year 2003 DMSP data was used as a base image for this calibration.
  2. DMSP Simulation from VIIRS (2012-2018): To extend the DMSP-like data beyond its operational period and into the VIIRS era, a simulation model was developed. This model combines a power function to capture the non-linear relationship between VIIRS and DMSP pixel values with a Gaussian low-pass filter to account for the spatial resolution difference. A thresholding step is also applied to constrain the simulated values. The parameters for this simulation model were optimized through a sensitive analysis to minimize the Root Mean Square Error (RMSE) between simulated and original DMSP data.

Datasets

  • DMSP/OLS NTL: Annual composites from 1992 to 2013 were used.
  • Suomi NPP/VIIRS NTL: Monthly composites from April 2012 to August 2018 were used.
  • Administrative Boundaries: For Beijing and Yiwu.
  • Socioeconomic Statistics:
    • City-level GDP for Beijing (1992-2017).
  • District-level GDP for Beijing (2005-2017).
  • City-level GDP for Yiwu (1999-2017).

Key Insights and Ideas

  • Bridging the Data Gap is Crucial: The fundamental insight is that the differences between DMSP and VIIRS NTL are not insurmountable and can be addressed through sophisticated calibration and simulation techniques, enabling a more robust long-term analysis.
  • Quadratic Regression for DMSP Calibration: Using a quadratic polynomial for DMSP calibration, rather than a simple linear model, likely captures more complex relationships in the sensor's output over time and across different sensors.
  • Power Function + Gaussian Filter for Simulation: The combination of a power function (for intensity transformation) and a Gaussian filter (for spatial smoothing/aggregation) is a clever way to mimic the characteristics of DMSP data from the higher-resolution VIIRS data.
  • Improved Correlation with GDP: The simulated DMSP data shows a stronger correlation with GDP than the raw VIIRS data, suggesting that the simulation successfully captures the "DMSP-like" characteristics that are more directly linked to economic activity in the studied regions.

Key Results

  • Enhanced Correlation: The Pearson correlation coefficients between simulated DMSP and original DMSP annual composites in 2012 and 2013 were significantly higher (0.937-0.945) compared to the correlation between VIIRS and original DMSP (0.637-0.658).
  • Reduced Error: The RMSE between simulated DMSP and original DMSP was substantially lower (9.244-9.387) than between VIIRS and original DMSP (34.804-35.526).
  • Improved Urban Dynamics Representation: In Beijing, the correlation between the simulated DMSP NTL and GDP from 1992-2017 reached 0.965, outperforming a direct combination of original DMSP and VIIRS data (0.935).
  • District-Level Improvements: For 11 out of 16 districts in Beijing, the correlation between NTL and GDP was enhanced after using the simulated DMSP data.

Limitations and Future Work

  • DMSP Value Range Limitation: The simulated DMSP NTL values are capped at 50, while the original DMSP NTL can reach 63. The authors suggest developing methods to quantify light intensity in the higher range (51-63) for more precise simulations.
  • Discontinuous SOL: The Sum of Light (SOL) of calibrated DMSP and combined NTL was not perfectly continuous across all years. Future work could explore methods to improve this continuity.
  • Potential for Other Applications: The authors suggest that the methodology could be extended to study other socio-economic problems, such as carbon emission, potentially by integrating with other data sources like NDVI.

Summary updated

Abstract

The brightness of pixels in nighttime light images (NTL) has been regarded as the proxy of the urban dynamics. However, the great difference between the pixel values of NTL from the Defense Meteorological Satellite Program’s Operational Linescan System (DMSP/OLS) and the Suomi National Polar-orbiting Partnership satellite’s Visible Infrared Imaging Radiometer Suite (Suomi NPP/VIIRS) poses obstacles to analyze economic and social development with NTL in a continuous temporal sequence. This research proposes a methodology to align the pixel values of both NTL by calibrating annual DMSP images between the years 1992–2013 with a robust regression algorithm with a quadratic polynomial regression model and simulating annual DMSP images with VIIRS images between years 2012 and 2018 with a model consisting of a power function and a Gaussian low pass filter. As a result, DMSP annual images between years 1992–2018 can be produced. Case study of Beijing and Yiwu are conducted and evaluated with local gross domestic product (GDP). Compared with the values of DMSP and VIIRS annual composites, the Pearson correlation coefficients of DMSP and simulated DMSP annual composites in 2012 and in 2013 increase significantly, while the root mean square error (RMSE) decrease evidently. In addition, the correlation of the sum of light of NTL and local GDP is enhanced with a simulation process. These results demonstrate the feasibility of the proposed method in narrowing the gap between DMSP and VIIRS NTL in pixel values.
BibTeX
@inproceedings{wu2019aligning,
  author = {Kang Wu and Xiaonan Wang},
  title = {{Aligning Pixel Values of DMSP and VIIRS Nighttime Light Images to Evaluate Urban Dynamics}},
  booktitle = {Remote},
  year = {2019},
  doi = {10.3390/rs11121463},
}
Figure