Monitoring Economies Through Space Using Nighttime Lights

World Bank Workshop

Authors

Robert Marty, Data Scientist, DECDI

Sahiti Sarva, Data Scientist, DECDG

Published

October 1, 2025

1 Monitoring Economies from Space Using Nightlights

Presentation Conducted on 1st October 2025: Access the original PowerPoint presentation

Hands On Session Link: Teams Recording

1.1 Outline

Data Sources: Evolution and Access
Core Applications in Economics and Social Science
Economic Monitoring
Crisis Response
New Research using Nightlights
Existing Resources from WB

1.2 Nighttime Lights Data Sources

Two sources of Nightlights are DMSP-OLS & VIIRS. Each has multiple versions of processed data.

1.2.1 Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS)

Agency: US Air Force
Coverage: Daily data from 1992-2013 (2018)
Processed Data: Colorado School of Mines, World Bank – Light Every Night

1.2.2 Visible Infrared Imaging Radiometer Suite (VIIRS)

Agency: NOAA in Partnership with NASA
Coverage: Daily data from 2012-present
Processed Data: Colorado School of Mines, World Bank – Light Every Night, NASA BlackMarble

Note: Processed data facilitates analysis of human-generated light by removing effects of moonlight, applying atmospheric corrections, etc.

1.3 Defense Meteorological Satellite Program (DMSP)

1.3.1 Key Characteristics

Agency: USA Air Force
Time Period: Original series from 1992-2013; extended to 2018
Satellites: Different satellites; requires calibrating across satellites
Resolution: 30 arc second (~1km at equator)
Units: Digital Number, integer from 0 – 63; censoring an issue for bright areas

1.3.2 Processed Data Sources:

Colorado School of Mines (Monthly & Annual)
World Bank – Light Every Night (Daily)

1.3.3 Limitation: Censoring

Bright areas are censored at value 63, losing information about variation in very bright locations.

1.4 Visible Infrared Imaging Radiometer Suite (VIIRS)

1.4.1 Key Characteristics

Agency: NOAA in partnership with NASA
Coverage: Daily Data: 2012-present
Resolution: 500m
Units: Radiance (nanoWatts/sr/cm²)
- How much light energy is received [nanoWatts]
- By a surface per unit area [cm²]
- Per unit solid angle [sr, steradian] → direction light is emitted/detected

1.4.2 Processed Data Sources:

Colorado School of Mines (More commonly used)
NASA BlackMarble (Released in 2021)

1.5 Comparing DMSP-OLS and VIIRS

VIIRS has higher, improved radiometric resolution, and is not saturated, unlike DMSP.

Feature	DMSP	VIIRS
Availability	1992–2013 (2018)	2012–present
Spatial Resolution	30 arc seconds (~1 km)	15 arc seconds (~500 m)
Radiometric Resolution	6-bit	12- or 14-bit
Wavelength Range	0.4–1.1 µm	0.5–0.89 µm
Units of Pixel Values	Relative (0–63 scale)	Radiance (nW cm⁻² sr⁻¹)
Overpass Time	~19:30	~1:30
On-board Calibration	No	Yes
Pixel Saturated	Yes	No

References: Li et al. (2020), Tu et al. (2020)

1.6 VIIRS: Colorado School of Mines vs BlackMarble

1.6.1 Snowfall, Vegetation and Surface Reflectance Corrections in BlackMarble

1.6.1.1 Snowfall Correction

Issue: Snowfall reflects light
BlackMarble Solution: Estimates snow-free lights

1.6.1.2 Vegetation Correction

Issue: Vegetation can hide light; less vegetation in winter can result in higher NTL values
BlackMarble Solution: Corrects for vegetation effects

1.6.1.3 Surface Reflection Correction

Approach: Different approach for accounting for surface reflection (e.g., moon, stars, atmospheric particles) using Bidirectional reflectance distribution function (BRDF)

1.6.1.4 Outlier Removal

Process: In creating monthly/annual data, Black Marble drops daily outliers (also done in Colorado School of Mines V2)

For more info, see The Spatial Edge (Yohan Iddawela)

1.7 VIIRS: Satellite Angle Variations

1.7.1 Colorado School of Mines

Approach: No differentiation; single dataset

1.7.2 Black Marble

Provides multiple datasets for annual data: - Near-nadir [above]: Typically brighter - Off-nadir [at angle]: Captures at an angle, allows greater coverage of a single image - Combined / “All Angle”

1.7.3 Research Findings

Wang et al. (2018): Comparing Dubai and Rome, found that near-nadir was brighter but varied more, while off-nadir was less bright but varied less.

For more info, see The Spatial Edge (Yohan Iddawela) and European Space Imaging

1.8 BlackMarble: Monthly and Annual Variables

1.8.1 Choice Options: (1) Angle; (2) Snowfall

AllAngle_Composite_Snow_Covered
AllAngle_Composite_Snow_Free
NearNadir_Composite_Snow_Covered
NearNadir_Composite_Snow_Free
OffNadir_Composite_Snow_Covered
OffNadir_Composite_Snow_Free

1.8.2 Example: Snowstorm Filomena

Without snow control: Large spike in January 2021
With snow control: No spike in January 2021

For more info, see The Spatial Edge (Yohan Iddawela)

1.9 BlackMarble: Stray Light Limitation

Stray Light: When sunlight hits the satellite - the biggest limitation of BlackMarble

1.9.1 Colorado School of Mines

Approach: Adjusts for stray light contamination

1.9.2 Black Marble

Approach: Drops pixels contaminated by stray light
Impact: Contains many missing values for northern countries from June - August

For more info, see The Spatial Edge (Yohan Iddawela)

1.10 BlackMarble: Data Quality Assessment

Nighttime lights can be impacted by a number of factors, particularly cloud cover.

1.10.1 BlackMarble Quality Flags (`Mandatory_Quality_Flag`)

0: High-quality, Persistent nighttime lights
- Lights that remain stable over time, typically from human-made sources (e.g., city lights)
1: High-quality, Ephemeral nighttime lights
- Temporary or short-lived sources of lights (e.g., wildfires, fishing fleets)
2: Poor-quality, Outlier, potential cloud contamination or other issues
No Data: Removed due to cloud cover

1.10.2 Gap-Filling for Daily Data

Allows using data from a previous date to fill the value
Uses temporally gap-filled NTL value: Gap_Filled_DNB_BRDF-Corrected_NTL

1.11 BlackMarble: Daily Data Variables

1.11.1 Day/Night Band (DNB) with BRDF Correction

BRDF: Bidirectional Reflectance Distribution Function – correction for surface reflectance

DNB_BRDF-Corrected_NTL: NTL value for given day
Gap_Filled_DNB_BRDF-Corrected_NTL: Will use latest high quality pixel within 30 days
Latest_High_Quality_Retrieval: Number of days between day & pixel used

1.12 BlackMarble: Monthly/Annual Data Quality

1.12.1 Number of Observations

NearNadir_Composite_Snow_Free_Num: Number of daily observations used for monthly composite (daily observations with cloud cover or other issues removed)

1.12.2 Quality Assessment

NearNadir_Composite_Snow_Free_Quality: - 0: Good-quality – The number of observations used for the composite is larger than 3 - 1: Poor-quality – The number of observations used for the composite is 3 or less

1.13 Trends from 1992 - Present: Combining DMSP-OLS and VIIRS

1.13.1 The Challenge

Issue: Comparing values from DMSP-OLS & VIIRS is an apples-to-oranges comparison. - Units differ: DMSP-OLS uses a “digital number” and VIIRS uses radiance - Data quality differs: For example, differences in resolution

1.13.2 Solutions

Downscale VIIRS: Create “DMSP-like” data
Upscale DMSP-OLS: Create “VIIRS-like” data (relies on using other imagery, such as daytime imagery, for upscaling)

1.13.3 Example Datasets

{Downscale VIIRS}: A harmonized global nighttime light dataset 1992–2018 (Li et al, 2020) [dataset] Updated beyond 2018
{Upscale DMSP}: A global annual simulated VIIRS nighttime light dataset from 1992 to 2023 (Chen et al, 2024) [dataset]

Reference: Li et al., 2020

1.14 Innovations in Nighttime Lights

1.14.1 SDGSAT-1

Developer: Chinese Academy of Sciences - Launch: 2021 - Resolution: Glimmer imager on SDGSAT 1 captures nighttime light data at 10-40 meter resolution

2 Uses of Nightlights in Economics and Social Sciences

Key Insight: Leveraging nighttime lights requires appreciating what does and does not generate lights

Reference: Earth at night (2022)

2.1 Sources of Nightlights

2.1.1 Brightness Ranking (from More Light to Less Light)

This ranking is suggestive/speculative; will differ based on context.

More Light: - Gas Flaring - Ports and Industrial Complexes (large bright clusters) - Manufacturing Units / Industrial Facilities - Airports (runway and terminal lighting) - Large Sports and Event Venues (stadiums with floodlights) - Major Highways and Traffic Corridors - Electric Lighting in Human Settlements (outdoor lighting) - Retail and Trade Activity Areas - Lights on Shipping Vessels - Mining sites and quarries (often bright industrial sites) - Wildfires and Controlled Burns - Fishing Boats - Data Centers and Service Sector Buildings - Rural Villages and Agricultural Areas - Auroras and Atmospheric Phenomena

Less Light

Reference: Earth at night (2022)

2.1.2 Economic Activity Correlation

2.1.2.1 Same Regardless of Economic Output

Data Centers and Service Sector Buildings

2.1.2.2 Differs with Economic Output

Gas Flaring

2.2 Indicators Using Nighttime Lights

Aggregating night lights differently creates different indicators:

Aggregation Type	Meaning
Sum of Lights	Total brightness summed over a region, representing the overall amount of light emitted by all sources combined
Average of Lights	Mean brightness per pixel or unit area, showing the typical light intensity level independent of area size
Centre of Gravity of Lights	Spatial geographic center weighted by brightness values, highlighting where the “center” of activity is located
Threshold of Light	Area above a brightness cutoff, separating lit from unlit or less lit zones

2.3 Average and Sum of Light

2.3.1 Sum vs Mean: Measuring Activity vs Density

‘Sum’ is better for GDP estimates because it’s a stock estimate
Total lights is typically biased towards larger areas
Research about human development, if it needs to be unbiased with area, could choose average light values
Average has been used by Henderson et al., 2018 as a proxy for human development indicators

2.4 Center of Gravity of Light

2.4.1 Movement in the Economic Center of Russia

Application: The central point where light is concentrated in a country was used as a proxy for economic centers (Hande et al., 2025)

Global Analysis: A global analysis of the same was done with DMSP data to show the movement of global economic centers eastward from 1992 to 2009 (Cauwels, Pestalozzi, & Sornette, 2014)

2.5 Nighttime Lights and Urbanization

2.5.1 Threshold Detection of Urban Areas

CIESIN Collaboration: CIESIN in collaboration with the World Bank used a threshold value after aggregating nightlights to determine city limits.

Electrification Studies: Threshold values were also used in studies to identify electrification of areas.

2.6 Use Cases of Nighttime Lights

2.6.1 Primary Applications

Use Case	Description
Proxy for Economic Activity	Estimation and nowcasting of subnational GDP
Rate of Urbanization	Increase in urban area, density and spatial agglomeration
Crisis Response	Post crisis damage assessment, needs assessment
Population Estimation	Post crisis damage assessment, needs assessment and resilience
Oil Production Monitoring	Measurement of oil-based GDP estimates and offshore oil production
Rate of Electrification	Measurement of electricity access and use

3 Nighttime Lights and Economic Activity

3.1 Key Research Findings

3.1.1 GDP Correlation Studies

Henderson et al. (2012): Nighttime lights correlated with levels and changes (long difference) in GDP
Hu and Yao (2019): NTL correlates with GDP. Elasticity of lights to GDP systematically varies by a country’s income level. Can be as high as 2.3 for low-income countries and close to 0 for high income countries
Gibson et al. (2021): VIIRS provides a more accurate proxy to GDP than DMSP-OLS
Martinez (2022): Finds that the elasticity of lights to GDP is larger in authoritarian regimes, suggested overstating annual GDP growth
Mellander et al. (2015): Using firm data in Sweden, shows NTL strongly associated with establishment density and population; weaker association with wages. For some indicators, lights overestimates in urban areas and underestimates in rural areas

3.2 Case Study: Mexican Municipalities

Analysis by Hugo Foster & Marie Lechler (2022)

3.2.1 Key Findings

Border Effects: More luminosity than expected, given population, in key border crossing locations with large cargo infrastructure & manufacturing activities. Less luminosity than expected given population in poorer locations
Supply Chain Nodes: Strong NTL in key supply chain & industry nodes, such as ports & oil/gas facilities
Juarez-El Paso: More nighttime lights around border crossings, transportation terminals and manufacturing sites
State Level: States with more NTL tend to have higher economic output

3.2.2 Important Limitation

“Luminosity data is less likely to be useful where local economies are based around service industries that emit less light compared to manufacturing”

3.3 Subnational GDP and Oil GDP Estimation

3.3.1 World Bank Applications

African Countries: World Bank paper uses Nightlights following Henderson et al., to estimate subnational GDP in African countries
Iran Oil Analysis: The log of nighttime lights from gas flaring locations shows an R-squared value of 0.83 when compared to GDP per capita

3.4 Underestimation of Service Sector & Agriculture

3.4.1 Timing Issues

VIIRS Data Collection: Collects data between 1:30PM UTC and 1:30 AM UTC and regionally this varies on the exact time it detects light.

Missing Peak Hours: Could miss peak lighting hours in some countries, especially if service sector is active at night (Bhattarai et al., 2023)

3.4.2 COVID-19 Study in India

Findings: Study conducted to analyze impact of COVID on India’s economy noted that the decrease in the productivity of the service sector wasn’t proportional to the decrease in nightlights (Dasgupta, 2022)

3.4.3 Agricultural Limitations

Weak Agricultural Signal: The weak signature of agricultural light makes NTL a weaker proxy for agricultural GDP in comparison to manufacturing

Reference: The Spatial Edge

4 Nighttime Lights and Crises

4.1 Nighttime Lights and Disasters

4.1.1 Earthquake in Turkey and Syria (2022)

Immediate Impact: Nightlights dropped in some parts of the earthquake impact region 3 days after the earthquake
Recovery Pattern: 2 weeks later, saw an increase in multiple admin regions
Hypothesis: Relief activities caused high intensity lights in the region

4.1.2 Hurricane Michael

Impact: After making landfall as a Category 5 storm, Michael knocked out power for at least 2.5 million customers in the southeastern United States
Visualization: Data visualization showed where lights went out in Panama City, Florida

4.2 Nighttime Lights and War

4.2.1 Conflict Impact Analysis

Gaza: >80% decline in lights in Gaza as early as October 22nd, 2023 when the war started on October 7th, 2023
Sudan: Significant reduction in nightlights before and after the conflict began in the region
Syria: Similar patterns observed during conflict periods

5 Evaluating Impact of Spatially-Explicit Infrastructure and Policies

5.1 Evaluation of New/Upgraded Roads

5.1.1 Methodological Advantages

Spatial Coverage: Nighttime lights are available across time and space; useful for quasi-experimental approaches (not spatially constrained in where to select control group)
Quasi-Experimental Design: Diff-in-diff: Compare treated with yet-to-be-treated units, using NTL as an outcome variable
Heterogeneity Analysis: Explore heterogeneity in results by baseline NTL

5.1.2 Research Examples

Chen et al. (2023): Rural road connectivity and local economic activity: Evidence from Sri Lanka’s iRoad Program. Rural roads with improved connectivity showed higher nighttime luminosity compared to areas without improved connectivity
BenYishay et al. (2018): Evaluation of road upgrades in Palestine. Find positive effects, particularly in areas with higher baseline luminosity
Alder (2025): Chinese roads in India: The effect of transport infrastructure on economic development. India’s network of highways led to gains in economic activity but had unequal effects across regions
Bolivar (2022): Roads illuminate development: Using nighttime luminosity to assess the impact of transport infrastructure. In Bolivia, Paraguay, and Ecuador, municipalities that benefited from paved major roads saw more economic activity compared to control locations

5.2 Evaluation of Rural Electrification Programs

5.2.1 Research Question

To what extent can nighttime lights capture rural electrification programs?

5.2.2 Detection Studies

Min et al. (2013): Detection of rural electrification in DMSP-OLS night lights imagery. Using sample of villages in Senegal and Mali, finds that electrified villages are brighter than unelectrified villages. Appear brighter largely because of the presence of streetlights; correlation of light output with household electricity use was low
Dugoua et al. (2018): Satellite data for the social sciences: measuring rural electrification with night-time lights. Nighttime lights provides an accurate measure of rural electrification in India, but detects electrification less accurately when the supply of power is intermittent
Min and Gaba (2014): Tracking Electrification in Vietnam Using Nighttime Lights. Find a one-point increase in DMSP-OLS DN for every 60-70 additional streetlights and 240-270 electrified homes

5.2.3 Program Evaluation Examples

Burlig and Preonas (2024): Out of the Darkness and into the Light? Development Effects of Rural Electrification. Evaluation of India’s national electrification program, RGGVY
Berthelemy (2024): Using Night-Time Light to Estimate the Impact of Mini-Grid Electrification Projects on Electric Power Consumption: Analytical Background and Case Study in Madagascar. Uses lights to evaluate mini-grid project. Find a positive impact; also find that positive impact could be overestimated by one third if street lighting is not taken into account

6 Combining Lights with Other Metrics

6.1 Poverty Estimation

6.1.1 Machine Learning Approaches

Challenge: Nighttime lights too coarse a measure to be a strong proxy for poverty/wealth estimation alone.

6.1.2 Research Examples

{Satellite imagery}: Jean et al. (2016) and Yeh et al. (2020) combine nighttime lights and daytime imagery to predict poverty measured from DHS in 5 African countries (Jean et al) and across Africa (Yeh et al)
{Satellite imagery and beyond}: Pokhriyal and Jacques (2017) and Marty and Duhaut (2024) use nighttime lights, daytime imagery, and other sources such as CDR data, Facebook data, etc, for poverty estimation

6.2 Downscaling Nighttime Lights

6.2.1 Methodology

Use finer scale data sources that suggest the likelihood of where lights are/aren’t coming from within a NTL pixel.

6.2.2 Common Approach: NDVI

Vegetation Index: One common source is NDVI (vegetation). Lights less likely to come from pixels with high NDVI/vegetation values.

References: Wu et al. (2024), Liu et al. (2022), Guo et al. (2024)

6.3 Combining Nightlights with NO2 and Exports

6.3.1 World Bank Study in Addis Ababa (Ongoing)

Methodology: Combined NO2 and NTL data to estimate its relationship with industrial production in Ethiopia measured through industrial exports data

Key Findings: - In a GMM model, every 1% change in exports caused a 2% change in average nightlights - Nightlights were the second biggest factor influencing exports in XGBoost model

6.4 Combining Nightlights with EVI

6.4.1 Enhanced Vegetation Index Integration

EVI and NDVI: Normalized Difference Vegetation Index and Enhanced Vegetation Index are products from daytime satellite imagery and measure greenness of canopy cover

Agricultural Applications: Studies have suggested the use of EVI alongside NTL to improve agricultural proxies

Status: This is an ongoing research question being explored within the team

7 Popular Global Public Goods Using Nightlights

7.1 Population and Settlement Mapping

7.1.1 Meta’s High Resolution Population Density

Resolution: 30m resolution population density maps

7.1.2 WorldPop Population Density Estimates

Resolution: 100m resolution population maps (annual)

7.1.3 Global Human Settlement Layer

Coverage: Global spatial data on human settlements (10m built up area, 100m population maps)

7.2 Urban Mapping Projects

7.2.1 Global Urban Footprint

Methodology: Combines population counts, settlement points, and nighttime lights to delineate urban and rural areas globally

7.2.2 Global Urban Rural Mapping Project (GRUMP)

Objective: Mapping the physical extent of built-up land worldwide with fine spatial resolution using SAR data

8 Existing and Upcoming Resources

8.1 Current Tools

8.1.1 R and Python Packages

BlackMarbleR: R Package to get data from NASA BlackMarble
BlackMarblePy: Python package to get data from NASA BlackMarble

8.1.2 Data Platforms

Space2Stats: Annual nighttime lights data among other metrics
Light Every Night/High Resolution Electricity Access: Daily NTL dataset from DMSP and VIIRS from 1992-2020 used to create a dataset of global energy access

8.2 Upcoming Resources

Space2Stats - Global Monthly NTL Data at ADM0 to ADM2: Coming soon!

9 Main Takeaways

9.1 Sources of Nightlight Data

9.1.1 Two Sources of Nightlights

DMSP-OLS (1992-2013)
VIIRS (2012-present) - Better in multiple ways including granularity

9.1.2 Two Sources of Processed Data

Colorado School of Mines
NASA Black Marble – More recent, provides additional corrections and choice to pick variables
- Daily: Gap filled vs not gap filled
- Monthly/Annually: Angle & snow correction

9.2 Key Questions for Economic Monitoring

9.2.1 Understanding Light Sources

Critical Questions: - What produces light and what doesn’t? - Which source has variability of light and which doesn’t? - Example: Gas Flaring location vs Service sector and agriculture

9.2.2 Spatial Aggregation Control

Analytical Choices: - Do we exclude gas flaring locations? - Do we only include lights in city boundaries? - With nighttime lights, we can control which pixels we use to spatially aggregate

9.3 Thank You

This presentation provides a comprehensive overview of using nighttime lights for economic monitoring and analysis. For more detailed information and hands-on training, please refer to the accompanying training materials and resources.

10 Additional Research Applications

10.1 Urban and Population Analysis

Identify extents of urban areas [link]
Estimate urban population size and density [link]

10.2 Energy and Environmental Applications

Measure electricity use, energy consumption, and GHG emissions [link]
Estimating offshore oil production [link]

10.3 Economic Activity Monitoring

Proxy for economic activity / GDP [link 1] and [link 2]

10.4 Crisis and Impact Assessment

Natural disaster, conflict, etc damages [link 1], [link 2]
Impact of infrastructure / policies (e.g., roads) [link 1], [link 2], [link 3]

10.5 Summary of Key Insights

10.5.1 Nighttime Lights Data Sources

Two sources of raw data: DMSP-OLS and VIIRS. VIIRS goes until present and is better in many ways (e.g., more granular)
Two sources of processed VIIRS data: Colorado School of Mines & Black Marble
- Black Marble is more recent & makes additional corrections (snow, vegetation)
- Black Marble provides additional choice on NTL variable
  - Daily: Gap filled vs not gap filled
  - Monthly/Annually: Angle & snow correction

10.5.2 Applications of Nighttime Lights

Wide use of nighttime lights in social science research, from GDP to urbanization
Key consideration: Think about what does and does not contribute to NTL (e.g., manufacturing vs service sector; agriculture and forestry)
Spatial control: With nighttime lights, we can control which pixels we use to spatially aggregate. Do we exclude gas flaring locations? Do we only include lights in city boundaries?
Data integration: Combining lights with other sources enhances analytical capabilities

--- title: "Monitoring Economies Through Space Using Nighttime Lights" subtitle: "World Bank Workshop" authors: - "Robert Marty, Data Scientist, DECDI" - "Sahiti Sarva, Data Scientist, DECDG" date: "October 1 2025" organization: "World Bank" --- # Monitoring Economies from Space Using Nightlights > **Presentation Conducted on 1st October 2025**: [Access the original PowerPoint presentation](https://worldbankgroup.sharepoint.com.mcas.ms/:p:/r/teams/DevelopmentDataPartnershipCommunity-WBGroup/_layouts/15/Doc.aspx?sourcedoc=%7BCB01FC45-80A0-41FD-A2F2-25B5EAE0FF49%7D&file=Monitoring%20Economies%20Through%20Space%20Using%20Nighttime%20Lights.pptx&action=edit&mobileredirect=true) **Hands On Session Link**: [Teams Recording](https://worldbankgroup-my.sharepoint.com/:v:/g/personal/ltsegaye_worldbank_org/ETvpjaz3HhJFingCe7T2SQoBz6JmALghWIY4pdKBG-kD-Q?nav=eyJyZWZlcnJhbEluZm8iOnsicmVmZXJyYWxBcHAiOiJTdHJlYW1XZWJBcHAiLCJyZWZlcnJhbFZpZXciOiJTaGFyZURpYWxvZy1MaW5rIiwicmVmZXJyYWxBcHBQbGF0Zm9ybSI6IldlYiIsInJlZmVycmFsTW9kZSI6InZpZXcifX0%3D&e=HSxhmk) ## Outline - **Data Sources**: Evolution and Access - **Core Applications** in Economics and Social Science - **Economic Monitoring** - **Crisis Response** - **New Research** using Nightlights - **Existing Resources** from WB --- ## Nighttime Lights Data Sources Two sources of Nightlights are **DMSP-OLS** & **VIIRS**. Each has multiple versions of processed data. ### Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS) - **Agency**: US Air Force - **Coverage**: Daily data from 1992-2013 (2018) - **Processed Data**: Colorado School of Mines, World Bank – Light Every Night ### Visible Infrared Imaging Radiometer Suite (VIIRS) - **Agency**: NOAA in Partnership with NASA - **Coverage**: Daily data from 2012-present - **Processed Data**: Colorado School of Mines, World Bank – Light Every Night, NASA BlackMarble > **Note**: Processed data facilitates analysis of human-generated light by removing effects of moonlight, applying atmospheric corrections, etc. --- ## Defense Meteorological Satellite Program (DMSP) ### Key Characteristics - **Agency**: USA Air Force - **Time Period**: Original series from 1992-2013; extended to 2018 - **Satellites**: Different satellites; requires calibrating across satellites - **Resolution**: 30 arc second (~1km at equator) - **Units**: Digital Number, integer from 0 – 63; censoring an issue for bright areas ### Processed Data Sources: - Colorado School of Mines (Monthly & Annual) - World Bank – Light Every Night (Daily) ### Limitation: Censoring Bright areas are censored at value 63, losing information about variation in very bright locations. --- ## Visible Infrared Imaging Radiometer Suite (VIIRS) ### Key Characteristics - **Agency**: NOAA in partnership with NASA - **Coverage**: Daily Data: 2012-present - **Resolution**: 500m - **Units**: Radiance (nanoWatts/sr/cm²) - How much light energy is received [nanoWatts] - By a surface per unit area [cm²] - Per unit solid angle [sr, steradian] → direction light is emitted/detected ### Processed Data Sources: - Colorado School of Mines (More commonly used) - NASA BlackMarble (Released in 2021) --- ## Comparing DMSP-OLS and VIIRS VIIRS has higher, improved radiometric resolution, and is not saturated, unlike DMSP. | Feature | DMSP | VIIRS | |---------|------|-------| | **Availability** | 1992–2013 (2018) | 2012–present | | **Spatial Resolution** | 30 arc seconds (~1 km) | 15 arc seconds (~500 m) | | **Radiometric Resolution** | 6-bit | 12- or 14-bit | | **Wavelength Range** | 0.4–1.1 µm | 0.5–0.89 µm | | **Units of Pixel Values** | Relative (0–63 scale) | Radiance (nW cm⁻² sr⁻¹) | | **Overpass Time** | ~19:30 | ~1:30 | | **On-board Calibration** | No | Yes | | **Pixel Saturated** | Yes | No | *References: [Li et al. (2020)](https://www.nature.com/articles/s41597-020-0510-y), [Tu et al. (2020)](https://www.tandfonline.com/doi/full/10.1080/01431161.2020.1731935)* --- ## VIIRS: Colorado School of Mines vs BlackMarble ### Snowfall, Vegetation and Surface Reflectance Corrections in BlackMarble #### Snowfall Correction - **Issue**: Snowfall reflects light - **BlackMarble Solution**: Estimates snow-free lights #### Vegetation Correction - **Issue**: Vegetation can hide light; less vegetation in winter can result in higher NTL values - **BlackMarble Solution**: Corrects for vegetation effects #### Surface Reflection Correction - **Approach**: Different approach for accounting for surface reflection (e.g., moon, stars, atmospheric particles) using Bidirectional reflectance distribution function (BRDF) #### Outlier Removal - **Process**: In creating monthly/annual data, Black Marble drops daily outliers (also done in Colorado School of Mines V2) *For more info, see [The Spatial Edge (Yohan Iddawela)](https://www.spatialedge.co/p/not-all-nightlight-datasets-are-the)* --- ## VIIRS: Satellite Angle Variations ### Colorado School of Mines - **Approach**: No differentiation; single dataset ### Black Marble Provides multiple datasets for annual data: - **Near-nadir [above]**: Typically brighter - **Off-nadir [at angle]**: Captures at an angle, allows greater coverage of a single image - **Combined / "All Angle"** ### Research Findings [Wang et al. (2018)](https://ieeexplore.ieee.org/abstract/document/9779217): Comparing Dubai and Rome, found that near-nadir was brighter but varied more, while off-nadir was less bright but varied less. *For more info, see [The Spatial Edge (Yohan Iddawela)](https://www.spatialedge.co/p/not-all-nightlight-datasets-are-the) and [European Space Imaging](https://www.euspaceimaging.com/blog/2024/01/31/what-is-ona-off-nadir-angle-in-satellite-imagery/)* --- ## BlackMarble: Monthly and Annual Variables ### Choice Options: (1) Angle; (2) Snowfall - `AllAngle_Composite_Snow_Covered` - `AllAngle_Composite_Snow_Free` - `NearNadir_Composite_Snow_Covered` - `NearNadir_Composite_Snow_Free` - `OffNadir_Composite_Snow_Covered` - `OffNadir_Composite_Snow_Free` ### Example: Snowstorm Filomena - **Without snow control**: Large spike in January 2021 - **With snow control**: No spike in January 2021 *For more info, see [The Spatial Edge (Yohan Iddawela)](https://www.spatialedge.co/p/not-all-nightlight-datasets-are-the)* --- ## BlackMarble: Stray Light Limitation **Stray Light**: When sunlight hits the satellite - the biggest limitation of BlackMarble ### Colorado School of Mines - **Approach**: Adjusts for stray light contamination ### Black Marble - **Approach**: Drops pixels contaminated by stray light - **Impact**: Contains many missing values for northern countries from June - August *For more info, see [The Spatial Edge (Yohan Iddawela)](https://www.spatialedge.co/p/not-all-nightlight-datasets-are-the)* --- ## BlackMarble: Data Quality Assessment Nighttime lights can be impacted by a number of factors, particularly cloud cover. ### BlackMarble Quality Flags (`Mandatory_Quality_Flag`) - **0**: High-quality, Persistent nighttime lights - Lights that remain stable over time, typically from human-made sources (e.g., city lights) - **1**: High-quality, Ephemeral nighttime lights - Temporary or short-lived sources of lights (e.g., wildfires, fishing fleets) - **2**: Poor-quality, Outlier, potential cloud contamination or other issues - **No Data**: Removed due to cloud cover ### Gap-Filling for Daily Data - Allows using data from a previous date to fill the value - Uses temporally gap-filled NTL value: `Gap_Filled_DNB_BRDF-Corrected_NTL` --- ## BlackMarble: Daily Data Variables ### Day/Night Band (DNB) with BRDF Correction **BRDF**: Bidirectional Reflectance Distribution Function – correction for surface reflectance - **`DNB_BRDF-Corrected_NTL`**: NTL value for given day - **`Gap_Filled_DNB_BRDF-Corrected_NTL`**: Will use latest high quality pixel within 30 days - **`Latest_High_Quality_Retrieval`**: Number of days between day & pixel used --- ## BlackMarble: Monthly/Annual Data Quality ### Number of Observations **`NearNadir_Composite_Snow_Free_Num`**: Number of daily observations used for monthly composite (daily observations with cloud cover or other issues removed) ### Quality Assessment **`NearNadir_Composite_Snow_Free_Quality`**: - **0**: Good-quality – The number of observations used for the composite is larger than 3 - **1**: Poor-quality – The number of observations used for the composite is 3 or less --- ## Trends from 1992 - Present: Combining DMSP-OLS and VIIRS ### The Challenge **Issue**: Comparing values from DMSP-OLS & VIIRS is an apples-to-oranges comparison. - **Units differ**: DMSP-OLS uses a "digital number" and VIIRS uses radiance - **Data quality differs**: For example, differences in resolution ### Solutions - **Downscale VIIRS**: Create "DMSP-like" data - **Upscale DMSP-OLS**: Create "VIIRS-like" data (relies on using other imagery, such as daytime imagery, for upscaling) ### Example Datasets - **{Downscale VIIRS}**: [A harmonized global nighttime light dataset 1992–2018](https://www.nature.com/articles/s41597-020-0510-y) (Li et al, 2020) [[dataset](https://figshare.com/articles/dataset/Harmonization_of_DMSP_and_VIIRS_nighttime_light_data_from_1992-2018_at_the_global_scale/9828827/8)] *Updated beyond 2018* - **{Upscale DMSP}**: [A global annual simulated VIIRS nighttime light dataset from 1992 to 2023](https://www.nature.com/articles/s41597-024-04228-6) (Chen et al, 2024) [[dataset](https://figshare.com/articles/dataset/A_history_reconstructed_time_series_1992-2011_of_annual_global_NPP-VIIRS-V2-like_nighttime_light_data_through_Super-resolution_U-Net_model/22262545/8)] *Reference: [Li et al., 2020](https://www.nature.com/articles/s41597-020-0510-y)* --- ## Innovations in Nighttime Lights ### SDGSAT-1 **Developer**: Chinese Academy of Sciences - **Launch**: 2021 - **Resolution**: Glimmer imager on SDGSAT 1 captures nighttime light data at 10-40 meter resolution --- # Uses of Nightlights in Economics and Social Sciences > **Key Insight**: Leveraging nighttime lights requires appreciating what does and does not generate lights *Reference: [Earth at night (2022)](https://www.nasa.gov/wp-content/uploads/2019/11/earth_at_night_508.pdf)* --- ## Sources of Nightlights ### Brightness Ranking (from More Light to Less Light) *This ranking is suggestive/speculative; will differ based on context.* **More Light:** - Gas Flaring - Ports and Industrial Complexes (large bright clusters) - Manufacturing Units / Industrial Facilities - Airports (runway and terminal lighting) - Large Sports and Event Venues (stadiums with floodlights) - Major Highways and Traffic Corridors - Electric Lighting in Human Settlements (outdoor lighting) - Retail and Trade Activity Areas - Lights on Shipping Vessels - Mining sites and quarries (often bright industrial sites) - Wildfires and Controlled Burns - Fishing Boats - Data Centers and Service Sector Buildings - Rural Villages and Agricultural Areas - Auroras and Atmospheric Phenomena **Less Light** *Reference: [Earth at night (2022)](https://www.nasa.gov/wp-content/uploads/2019/11/earth_at_night_508.pdf)* ### Economic Activity Correlation #### Same Regardless of Economic Output - Data Centers and Service Sector Buildings #### Differs with Economic Output - Gas Flaring --- ## Indicators Using Nighttime Lights Aggregating night lights differently creates different indicators: | **Aggregation Type** | **Meaning** | |---------------------|-------------| | **Sum of Lights** | Total brightness summed over a region, representing the overall amount of light emitted by all sources combined | | **Average of Lights** | Mean brightness per pixel or unit area, showing the typical light intensity level independent of area size | | **Centre of Gravity of Lights** | Spatial geographic center weighted by brightness values, highlighting where the "center" of activity is located | | **Threshold of Light** | Area above a brightness cutoff, separating lit from unlit or less lit zones | --- ## Average and Sum of Light ### Sum vs Mean: Measuring Activity vs Density - **'Sum'** is better for GDP estimates because it's a stock estimate - **Total lights** is typically biased towards larger areas - **Research about human development**, if it needs to be unbiased with area, could choose average light values - **Average** has been used by Henderson et al., 2018 as a proxy for human development indicators --- ## Center of Gravity of Light ### Movement in the Economic Center of Russia **Application**: The central point where light is concentrated in a country was used as a proxy for economic centers ([Hande et al., 2025](https://xkdr.org/paper/shedding-light-on-the-russia-ukraine-war)) **Global Analysis**: A global analysis of the same was done with DMSP data to show the movement of global economic centers eastward from 1992 to 2009 ([Cauwels, Pestalozzi, & Sornette, 2014](https://app.amanote.com/v4.5.5/research/note-taking?resourceId=w6yiAnQBKQvf0BhiwH82)) --- ## Nighttime Lights and Urbanization ### Threshold Detection of Urban Areas **CIESIN Collaboration**: [CIESIN in collaboration with the World Bank](https://openknowledge.worldbank.org/server/api/core/bitstreams/adf744b4-8f0d-5f61-a72e-ac1f7d2ae99d/content) used a threshold value after aggregating nightlights to determine city limits. **Electrification Studies**: Threshold values were also used in studies to identify [electrification of areas](https://blogs.worldbank.org/en/energy/using-night-lights-map-electrical-grid-infrastructure). --- ## Use Cases of Nighttime Lights ### Primary Applications | **Use Case** | **Description** | |--------------|-----------------| | **Proxy for Economic Activity** | Estimation and nowcasting of subnational GDP | | **Rate of Urbanization** | Increase in urban area, density and spatial agglomeration | | **Crisis Response** | Post crisis damage assessment, needs assessment | | **Population Estimation** | Post crisis damage assessment, needs assessment and resilience | | **Oil Production Monitoring** | Measurement of oil-based GDP estimates and offshore oil production | | **Rate of Electrification** | Measurement of electricity access and use | --- # Nighttime Lights and Economic Activity ## Key Research Findings ### GDP Correlation Studies - **[Henderson et al. (2012)](https://pubs.aeaweb.org/doi/pdfplus/10.1257/aer.102.2.994)**: Nighttime lights correlated with levels and changes (long difference) in GDP - **[Hu and Yao (2019)](https://www.econ2.jhu.edu/people/hu/paper_HUandYAO.pdf)**: NTL correlates with GDP. Elasticity of lights to GDP systematically varies by a country's income level. Can be as high as 2.3 for low-income countries and close to 0 for high income countries - **[Gibson et al. (2021)](https://www.sciencedirect.com/science/article/pii/S0304387820301772)**: VIIRS provides a more accurate proxy to GDP than DMSP-OLS - **[Martinez (2022)](https://www.journals.uchicago.edu/doi/10.1086/720458?)**: Finds that the elasticity of lights to GDP is larger in authoritarian regimes, suggested overstating annual GDP growth - **[Mellander et al. (2015)](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0139779)**: Using firm data in Sweden, shows NTL strongly associated with establishment density and population; weaker association with wages. For some indicators, lights overestimates in urban areas and underestimates in rural areas --- ## Case Study: Mexican Municipalities **Analysis by [Hugo Foster & Marie Lechler (2022)](https://www.spglobal.com/market-intelligence/en/news-insights/research/how-nighttime-lights-illuminate-economic-activity)** ### Key Findings - **Border Effects**: More luminosity than expected, given population, in key border crossing locations with large cargo infrastructure & manufacturing activities. Less luminosity than expected given population in poorer locations - **Supply Chain Nodes**: Strong NTL in key supply chain & industry nodes, such as ports & oil/gas facilities - **Juarez-El Paso**: More nighttime lights around border crossings, transportation terminals and manufacturing sites - **State Level**: States with more NTL tend to have higher economic output ### Important Limitation > "Luminosity data is less likely to be useful where local economies are based around service industries that emit less light compared to manufacturing" --- ## Subnational GDP and Oil GDP Estimation ### World Bank Applications - **African Countries**: World Bank paper uses Nightlights following Henderson et al., to estimate subnational GDP in African countries - **Iran Oil Analysis**: The log of nighttime lights from gas flaring locations shows an R-squared value of 0.83 when compared to GDP per capita --- ## Underestimation of Service Sector & Agriculture ### Timing Issues **VIIRS Data Collection**: Collects data between 1:30PM UTC and 1:30 AM UTC and regionally this varies on the exact time it detects light. **Missing Peak Hours**: Could miss peak lighting hours in some countries, especially if service sector is active at night (Bhattarai et al., 2023) ### COVID-19 Study in India **Findings**: Study conducted to analyze impact of COVID on India's economy noted that the decrease in the productivity of the service sector wasn't proportional to the decrease in nightlights (Dasgupta, 2022) ### Agricultural Limitations **Weak Agricultural Signal**: The weak signature of agricultural light makes NTL a weaker proxy for agricultural GDP in comparison to manufacturing *Reference: [The Spatial Edge](https://www.spatialedge.co/p/some-surprising-facts-about-nightlights)* --- # Nighttime Lights and Crises ## Nighttime Lights and Disasters ### Earthquake in Turkey and Syria (2022) - **Immediate Impact**: Nightlights dropped in some parts of the earthquake impact region 3 days after the earthquake - **Recovery Pattern**: 2 weeks later, saw an increase in multiple admin regions - **Hypothesis**: Relief activities caused high intensity lights in the region ### Hurricane Michael - **Impact**: After making landfall as a Category 5 storm, Michael knocked out power for at least 2.5 million customers in the southeastern United States - **Visualization**: Data visualization showed where lights went out in Panama City, Florida --- ## Nighttime Lights and War ### Conflict Impact Analysis - **Gaza**: >80% decline in lights in Gaza as early as October 22nd, 2023 when the war started on October 7th, 2023 - **Sudan**: Significant reduction in nightlights before and after the conflict began in the region - **Syria**: Similar patterns observed during conflict periods --- # Evaluating Impact of Spatially-Explicit Infrastructure and Policies ## Evaluation of New/Upgraded Roads ### Methodological Advantages - **Spatial Coverage**: Nighttime lights are available across time and space; useful for quasi-experimental approaches (not spatially constrained in where to select control group) - **Quasi-Experimental Design**: Diff-in-diff: Compare treated with yet-to-be-treated units, using NTL as an outcome variable - **Heterogeneity Analysis**: Explore heterogeneity in results by baseline NTL ### Research Examples - **[Chen et al. (2023)](https://www.sciencedirect.com/science/article/pii/S0967070X23002603)**: Rural road connectivity and local economic activity: Evidence from Sri Lanka's iRoad Program. Rural roads with improved connectivity showed higher nighttime luminosity compared to areas without improved connectivity - **[BenYishay et al. (2018)](https://docs.aiddata.org/ad4/pdfs/West_Bank_Infrastructure_GIE_Report.pdf)**: Evaluation of road upgrades in Palestine. Find positive effects, particularly in areas with higher baseline luminosity - **[Alder (2025)](https://www.sciencedirect.com/science/article/pii/S0022199625000972)**: Chinese roads in India: The effect of transport infrastructure on economic development. India's network of highways led to gains in economic activity but had unequal effects across regions - **[Bolivar (2022)](https://scioteca.caf.com/bitstream/handle/123456789/1955/Roads%20illuminate%20development%20Using%20nightlight%20luminosity%20to%20assess%20the%20impact%20of%20transport%20infraestructure.pdf?sequence=4&isAllowed=y)**: Roads illuminate development: Using nighttime luminosity to assess the impact of transport infrastructure. In Bolivia, Paraguay, and Ecuador, municipalities that benefited from paved major roads saw more economic activity compared to control locations --- ## Evaluation of Rural Electrification Programs ### Research Question **To what extent can nighttime lights capture rural electrification programs?** ### Detection Studies - **[Min et al. (2013)](https://www.tandfonline.com/doi/abs/10.1080/01431161.2013.833358)**: Detection of rural electrification in DMSP-OLS night lights imagery. Using sample of villages in Senegal and Mali, finds that electrified villages are brighter than unelectrified villages. Appear brighter largely because of the presence of streetlights; correlation of light output with household electricity use was low - **[Dugoua et al. (2018)](https://www.tandfonline.com/doi/full/10.1080/01431161.2017.1420936?utm_source=chatgpt.com)**: Satellite data for the social sciences: measuring rural electrification with night-time lights. Nighttime lights provides an accurate measure of rural electrification in India, but detects electrification less accurately when the supply of power is intermittent - **[Min and Gaba (2014)](https://www.mdpi.com/2072-4292/6/10/9511)**: Tracking Electrification in Vietnam Using Nighttime Lights. Find a one-point increase in DMSP-OLS DN for every 60-70 additional streetlights and 240-270 electrified homes ### Program Evaluation Examples - **[Burlig and Preonas (2024)](https://www.journals.uchicago.edu/doi/epdf/10.1086/730204)**: Out of the Darkness and into the Light? Development Effects of Rural Electrification. Evaluation of India's national electrification program, RGGVY - **[Berthelemy (2024)](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4812767)**: Using Night-Time Light to Estimate the Impact of Mini-Grid Electrification Projects on Electric Power Consumption: Analytical Background and Case Study in Madagascar. Uses lights to evaluate mini-grid project. Find a positive impact; also find that positive impact could be overestimated by one third if street lighting is not taken into account --- # Combining Lights with Other Metrics ## Poverty Estimation ### Machine Learning Approaches **Challenge**: Nighttime lights too coarse a measure to be a strong proxy for poverty/wealth estimation alone. ### Research Examples - **{Satellite imagery}**: [Jean et al. (2016)](https://www.science.org/doi/epdf/10.1126/science.aaf7894) and [Yeh et al. (2020)](https://www.nature.com/articles/s41467-020-16185-w) combine nighttime lights and daytime imagery to predict poverty measured from DHS in 5 African countries (Jean et al) and across Africa (Yeh et al) - **{Satellite imagery and beyond}**: [Pokhriyal and Jacques (2017)](https://www.pnas.org/doi/10.1073/pnas.1700319114) and [Marty and Duhaut (2024)](https://www.nature.com/articles/s41598-023-49564-6) use nighttime lights, daytime imagery, and other sources such as CDR data, Facebook data, etc, for poverty estimation --- ## Downscaling Nighttime Lights ### Methodology Use finer scale data sources that suggest the likelihood of where lights are/aren't coming from within a NTL pixel. ### Common Approach: NDVI **Vegetation Index**: One common source is NDVI (vegetation). Lights less likely to come from pixels with high NDVI/vegetation values. **References**: [Wu et al. (2024)](https://ieeexplore.ieee.org/document/10707291), [Liu et al. (2022)](https://www.mdpi.com/2072-4292/14/24/6400), [Guo et al. (2024)](https://www.sciencedirect.com/science/article/pii/S1569843224002784) --- ## Combining Nightlights with NO2 and Exports ### World Bank Study in Addis Ababa (Ongoing) **Methodology**: Combined NO2 and NTL data to estimate its relationship with industrial production in Ethiopia measured through industrial exports data **Key Findings**: - In a GMM model, every 1% change in exports caused a 2% change in average nightlights - Nightlights were the second biggest factor influencing exports in XGBoost model --- ## Combining Nightlights with EVI ### Enhanced Vegetation Index Integration **EVI and NDVI**: Normalized Difference Vegetation Index and Enhanced Vegetation Index are products from daytime satellite imagery and measure greenness of canopy cover **Agricultural Applications**: Studies have suggested the use of EVI alongside NTL to improve agricultural proxies **Status**: This is an ongoing research question being explored within the team --- # Popular Global Public Goods Using Nightlights ## Population and Settlement Mapping ### Meta's High Resolution Population Density - **Resolution**: 30m resolution population density maps ### WorldPop Population Density Estimates - **Resolution**: 100m resolution population maps (annual) ### Global Human Settlement Layer - **Coverage**: Global spatial data on human settlements (10m built up area, 100m population maps) ## Urban Mapping Projects ### Global Urban Footprint - **Methodology**: Combines population counts, settlement points, and nighttime lights to delineate urban and rural areas globally ### Global Urban Rural Mapping Project (GRUMP) - **Objective**: Mapping the physical extent of built-up land worldwide with fine spatial resolution using SAR data --- # Existing and Upcoming Resources ## Current Tools ### R and Python Packages - **BlackMarbleR**: R Package to get data from NASA BlackMarble - **BlackMarblePy**: Python package to get data from NASA BlackMarble ### Data Platforms - **Space2Stats**: Annual nighttime lights data among other metrics - **Light Every Night/High Resolution Electricity Access**: Daily NTL dataset from DMSP and VIIRS from 1992-2020 used to create a dataset of global energy access ## Upcoming Resources - **Space2Stats - Global Monthly NTL Data at ADM0 to ADM2**: Coming soon! --- # Main Takeaways ## Sources of Nightlight Data ### Two Sources of Nightlights - **DMSP-OLS** (1992-2013) - **VIIRS** (2012-present) - Better in multiple ways including granularity ### Two Sources of Processed Data - **Colorado School of Mines** - **NASA Black Marble** – More recent, provides additional corrections and choice to pick variables - **Daily**: Gap filled vs not gap filled - **Monthly/Annually**: Angle & snow correction --- ## Key Questions for Economic Monitoring ### Understanding Light Sources **Critical Questions**: - What produces light and what doesn't? - Which source has variability of light and which doesn't? - Example: Gas Flaring location vs Service sector and agriculture ### Spatial Aggregation Control **Analytical Choices**: - Do we exclude gas flaring locations? - Do we only include lights in city boundaries? - With nighttime lights, we can control which pixels we use to spatially aggregate --- ## Thank You *This presentation provides a comprehensive overview of using nighttime lights for economic monitoring and analysis. For more detailed information and hands-on training, please refer to the accompanying training materials and resources.* --- # Additional Research Applications ## Urban and Population Analysis - Identify extents of urban areas [[link](https://www.sciencedirect.com/science/article/abs/pii/S0034425797000461)] - Estimate urban population size and density [[link](https://www.tandfonline.com/doi/full/10.1080/01431160802430693)] ## Energy and Environmental Applications - Measure electricity use, energy consumption, and GHG emissions [[link](https://www.tandfonline.com/doi/full/10.1080/01431160903261005)] - Estimating offshore oil production [[link](https://www.sciencedirect.com/science/article/abs/pii/S0924271620301453)] ## Economic Activity Monitoring - Proxy for economic activity / GDP [[link 1](https://www.aeaweb.org/articles?id=10.1257/aer.102.2.994)] and [[link 2](https://www.pnas.org/doi/10.1073/pnas.1017031108#sec-5)] ## Crisis and Impact Assessment - Natural disaster, conflict, etc damages [[link 1](https://www.sciencedirect.com/science/article/abs/pii/S0143622819308525)], [[link 2](https://www.mdpi.com/2072-4292/10/6/858)] - Impact of infrastructure / policies (e.g., roads) [[link 1](https://www.aiddata.org/publications/evaluation-of-usaid-west-bank-gaza-infrastructure-needs-program)], [[link 2](https://docs.iza.org/dp12018.pdf)], [[link 3](https://documents.worldbank.org/en/publication/documents-reports/documentdetail/099332404062230683/idu073a7158605532046490b712098aed9008539)] ## Summary of Key Insights ### Nighttime Lights Data Sources - **Two sources of raw data**: DMSP-OLS and VIIRS. VIIRS goes until present and is better in many ways (e.g., more granular) - **Two sources of processed VIIRS data**: Colorado School of Mines & Black Marble - Black Marble is more recent & makes additional corrections (snow, vegetation) - Black Marble provides additional choice on NTL variable - Daily: Gap filled vs not gap filled - Monthly/Annually: Angle & snow correction ### Applications of Nighttime Lights - **Wide use** of nighttime lights in social science research, from GDP to urbanization - **Key consideration**: Think about what does and does not contribute to NTL (e.g., manufacturing vs service sector; agriculture and forestry) - **Spatial control**: With nighttime lights, we can control which pixels we use to spatially aggregate. Do we exclude gas flaring locations? Do we only include lights in city boundaries? - **Data integration**: Combining lights with other sources enhances analytical capabilities

1 Monitoring Economies from Space Using Nightlights

1.1 Outline

1.2 Nighttime Lights Data Sources

1.2.1 Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS)

1.2.2 Visible Infrared Imaging Radiometer Suite (VIIRS)

1.3 Defense Meteorological Satellite Program (DMSP)

1.3.1 Key Characteristics

1.3.2 Processed Data Sources:

1.3.3 Limitation: Censoring

1.4 Visible Infrared Imaging Radiometer Suite (VIIRS)

1.4.1 Key Characteristics

1.4.2 Processed Data Sources:

1.5 Comparing DMSP-OLS and VIIRS

1.6 VIIRS: Colorado School of Mines vs BlackMarble

1.6.1 Snowfall, Vegetation and Surface Reflectance Corrections in BlackMarble

1.6.1.1 Snowfall Correction

1.6.1.2 Vegetation Correction

1.6.1.3 Surface Reflection Correction

1.6.1.4 Outlier Removal

1.7 VIIRS: Satellite Angle Variations

1.7.1 Colorado School of Mines

1.7.2 Black Marble

1.7.3 Research Findings

1.8 BlackMarble: Monthly and Annual Variables

1.8.1 Choice Options: (1) Angle; (2) Snowfall

1.8.2 Example: Snowstorm Filomena

1.9 BlackMarble: Stray Light Limitation

1.9.1 Colorado School of Mines

1.9.2 Black Marble

1.10 BlackMarble: Data Quality Assessment

1.10.1 BlackMarble Quality Flags (Mandatory_Quality_Flag)

1.10.2 Gap-Filling for Daily Data

1.11 BlackMarble: Daily Data Variables

1.11.1 Day/Night Band (DNB) with BRDF Correction

1.12 BlackMarble: Monthly/Annual Data Quality

1.12.1 Number of Observations

1.12.2 Quality Assessment

1.13 Trends from 1992 - Present: Combining DMSP-OLS and VIIRS

1.13.1 The Challenge

1.13.2 Solutions

1.13.3 Example Datasets

1.14 Innovations in Nighttime Lights

1.14.1 SDGSAT-1

2 Uses of Nightlights in Economics and Social Sciences

2.1 Sources of Nightlights

2.1.1 Brightness Ranking (from More Light to Less Light)

2.1.2 Economic Activity Correlation

2.1.2.1 Same Regardless of Economic Output

2.1.2.2 Differs with Economic Output

2.2 Indicators Using Nighttime Lights

2.3 Average and Sum of Light

2.3.1 Sum vs Mean: Measuring Activity vs Density

2.4 Center of Gravity of Light

2.4.1 Movement in the Economic Center of Russia

2.5 Nighttime Lights and Urbanization

2.5.1 Threshold Detection of Urban Areas

2.6 Use Cases of Nighttime Lights

2.6.1 Primary Applications

3 Nighttime Lights and Economic Activity

3.1 Key Research Findings

3.1.1 GDP Correlation Studies

3.2 Case Study: Mexican Municipalities

3.2.1 Key Findings

3.2.2 Important Limitation

3.3 Subnational GDP and Oil GDP Estimation

3.3.1 World Bank Applications

3.4 Underestimation of Service Sector & Agriculture

3.4.1 Timing Issues

3.4.2 COVID-19 Study in India

3.4.3 Agricultural Limitations

4 Nighttime Lights and Crises

4.1 Nighttime Lights and Disasters

4.1.1 Earthquake in Turkey and Syria (2022)

4.1.2 Hurricane Michael

4.2 Nighttime Lights and War

4.2.1 Conflict Impact Analysis

5 Evaluating Impact of Spatially-Explicit Infrastructure and Policies

5.1 Evaluation of New/Upgraded Roads

5.1.1 Methodological Advantages

5.1.2 Research Examples

1.10.1 BlackMarble Quality Flags (`Mandatory_Quality_Flag`)