AnalysisMapping regional economic activity from night-time light satellite imagery
Introduction
The need to link human and natural physical systems for a greater understanding of global change requires socio-economic datasets, which can be easily integrated with other sources of environmental data. Socio-economic data are usually provided on a national basis. While this is a convenient administrative unit to collect the data, it may be an inappropriate unit on which to conduct environmental analysis. This fundamental difference in data collection schemes is one of the major obstacles preventing more widespread integration of socio-economic datasets with environmental datasets. Raster datasets also offer the flexibility to perform analysis at more convenient aggregated spatial units such as ecological zones or watersheds.
This paper builds upon previous work to map economic activity and carbon dioxide emissions from night-time light satellite data (Doll et al., 2000) and preliminary assessments of sub-national economic characteristics from space (Doll, 2003). In the former study, the first ever global map of GDP was produced using a country level lit area–GDP relationship. This map had a spatial resolution of 1°. Here, the radiance-calibrated dataset was used in place of the time frequency composite data (Elvidge et al., 1997) employed previously. This facilitates the consideration of the intensity of light recorded at the sensor rather than just a simple areal estimation of light coverage. Linear relationships between radiance and Gross Regional Product (GRP) are observed for a sub-set of countries within the European Union as well as the United States. These relationships exhibit a number of key features, which suggested that the overall relationship is more complex than first thought. It appears that the use of a single coefficient to disaggregate the national GDP statistic may not be accurate for all areas. Bearing these considerations in mind, spatially disaggregated maps of GDP were produced and are analysed here. These results are more detailed than those published in Ecological Economics by Sutton and Costanza (2002). These authors outlined a method of estimating global economic activity from night-time light radiance collected by the Defense Meteorological Satellite Program's Operational Linescan System (DMSP-OLS). The results of their appraisal of the market economy were compared to an assessment of so-called non-market capital (as defined by Sutton and Costanza, 2003) through the use of ecosystem service valuations and a satellite derived land-use map.
In the night-time lights dataset, remote sensing stands uniquely poised to contribute to providing gridded information on socio-economic parameters. This comes at a time when debate as to what are the socio-economic drivers of global change and how these can be best represented is gaining currency among global change research scientists.
Section snippets
Scaling issues
The results presented here provide a further insight into the considerations of using DMSP-OLS night-time light imagery to disaggregate GDP data at the sub-national scale. In the same way that Konarska et al. (2002) highlighted the potential discrepancies in ecosystem services valuation by comparing 1 km IGBP and 30 m NLCD landcover classifications, so too does the scale of analysis play an important role with respect to making valuations of the market economy. In this case, the problem is less
Night-time light satellite data
The Defense Meteorological Satellite Program originated in the mid-1960s with the objective of collecting worldwide cloud cover on a daily basis (Kramer, 1994). The DMSP satellite flies in a sun-synchronous low earth orbit (833 km mean altitude) and makes a night-time pass typically between 20.30 and 21.30 each night (Elvidge et al., 2001). The Operational Linescan sensor was initially used for producing night-time cloud imagery and as such required a high level of sensitivity—some four orders
Sub-national radiance–GRP correlation results
Each European country was tested for at least one NUTS level depending on the availability of data and the NUTS configuration of that country. The US states were compared to regions defined by the US Census Bureau. The results refer to the 1997 set of economic data, as this is co-temporal with the night-time light data. Strong positive correlations between total radiance and GRP were observed for all the countries tested at the finest level of observation. A number of variations were identified
Using relationships to create spatially disaggregated maps
In producing a map of this nature, one must consider how sensible it is create a map at a spatial resolution where the subject of the map starts to loose its intuitive credibility. Addressing socio-economic issues such as economic activity is one such application where such concerns become important.
The main factors in choosing a spatial resolution for the map were based not only on what the map aimed to display but also on the technical specifications of the DMSP-OLS sensor. Firstly, although
Discussion
A key problem with the map results is that there is no independent way of checking whether the disaggregation is correct. The relationships presented in the previous chapter were constructed from the finest scale data available. While it may not be possible to assess the sub-NUTS-3 accuracy of the map, it is still worthwhile to examine how robust the map result is to aggregation. Simulated regions provide an alternative means of aggregating lower levels of data to compare against conventional
Conclusion and prospects
There are a number of improvements to the present methodology which merit investigation. These modifications deal with the way the relationship between radiance and GRP is constructed. In particular, investigating the use of radiance thresholds to identify areas of homogeneous economic activity. While radiance is the key determinant of economic activity, incorporating the prevailing land use into the model could reduce the presence of outliers and facilitate the construction of a single model
Acknowledgements
The authors would like to thank Chris Elvidge at NOAA-NDGC, Boulder Colorado for supplying the satellite data and the anonymous referees' comments. This work was supported by NERC award GT 04/98 188/TS.
References (26)
- et al.
Radiance calibration of DMSP-OLS low-light imaging data of human settlements
Remote Sensing of Environment
(1999) - et al.
Night-time lights of the world: 1994–1995
ISPRS Journal of Photogrammetry and Remote Sensing
(2001) - et al.
The concept of scale and the human dimensions of global change: a survey
Ecological Economics
(2000) - et al.
Evaluating scale dependence of ecosystem service valuation: a comparison of NOAA-AVHRR and Landsat TM datasets
Ecological Economics
(2002) - et al.
Global estimates of market and non-market values derived from nighttime satellite imagery, landcover and ecosystem service valuation
Ecological Economics
(2002) - BEA, 2002. Bureau of Economic Analysis: Gross State Product data. Available from:...
- BEA, 2003. Regional accounts: gross state product. Online Report. Available from:...
- et al.
Understanding the Scale and Resolution Effects in Remote Sensing and GIS
Nighttime images of the Earth from space
Scientific American
(1978)- et al.
Real output decline in transition economies—forget GDP, try power consumption data!
Transition Newsletter
(1995)
Estimating non-population parameters from night-time satellite imagery
Investigating Asian economic activity with night-time light remote sensing data
Night-time imagery as a tool for global mapping of socio-economic parameters and greenhouse gas emissions
Ambio
Cited by (0)
- 1
Tel.: +44 20 7679 2740; fax: +44 20 7380 0453.