Nexus between Light Pollution and Air Temperature: A Study of Bangladesh

Abstract

Rapid urbanization and population growth have introduced a new phenomenon called "light pollution", which is the result of ungoverned use of artificial lighting. Natural land cover is being replaced by impervious surface to meet the growing demand. As a result, urban areas are getting warmer compared to the surrounding rural areas. Scholarly literature confirms the relationship between light pollution and air temperature. However, the nature of direct relationship is yet to be explored. Using remote sensing and weather station data, this research reveals the nature of relationship between these two variables. The analysis confirms that a significant relationship exists which can be explained by geographically weighted regression (GWR) in a far better way compared to ordinary least square (OLS) regression. According to the GWR, overall 50% change in air temperature is influenced or affected by light pollution, where in urban areas the impact of light pollution on air temperature is higher compared to rural areas. This research also unearths that light pollution is increasing at a dissolute rate, four-fold in 10 years. With this said, considering the nexus between light pollution and air temperature along with its other negative effects the authorities are expected to take measures to reduce light pollution.

Full text

Journal of Bangladesh Institute of Planners ISSN 2075-9363
Vol. 11, 2018 (Printed in June 2020), pp. 1-9, © Bangladesh Institute of Planners
Nexus between Light Pollution and Air Temperature:
A Study of Bangladesh
Aninda Sundar Howlader *
Kazi Saiful Islam**
Abstract
Rapid urbanization and population growth have introduced a new phenomenon called
“light pollution”, which is the result of ungoverned use of artificial lighting. Natural land
cover is being replaced by impervious surface to meet the growing demand. As a result,
urban areas are getting warmer compared to the surrounding rural areas. Scholarly
literature confirms the relationship between light pollution and air temperature.
However, the nature of direct relationship is yet to be explored. Using remote sensing and
weather station data, this research reveals the nature of relationship between these two
variables. The analysis confirms that a significant relationship exists which can be
explained by geographically weighted regression (GWR) in a far better way compared to
ordinary least square (OLS) regression. According to the GWR, overall 50% change in air
temperature is influenced or affected by light pollution, where in urban areas the impact
of light pollution on air temperature is higher compared to rural areas. This research also
unearths that light pollution is increasing at a dissolute rate, four-fold in 10 years. With
this said, considering the nexus between light pollution and air temperature along with its
other negative effects the authorities are expected to take measures to reduce light
pollution.
Introduction
The people inhabit only 5% of the world’s total land area, where more than 50% of the
total world population is living in urban areas. Recently, the developing countries are
experiencing the fastest growth of the urban population compared to the developed
countries (Elvidge et al., 2007a; Elvidge et al., 2007b). As a result, urban areas are
becoming more impervious than ever (Sinha et al., 2016). With rapid urbanization and
economic development, the use of artificial night light is also increasing responding to a
higher extent of human activities at night (Jiang et al., 2017). Nevertheless, poor design
and unregulated use of artificial light are creating a phenomenon called ‘Light Pollution’
(LP) (Gallaway, 2010). LP is the unwanted, unintended and obtrusive aspects of artificial
lighting, which largely results from bad lighting design and has negative impacts on the
environment (Olsen et al., 2014).
Humans have altered the natural land cover (LC) type to meet their various needs from
the beginning of time. With the increasing number of people, LC is changing drastically.
Previous studies have found that different LC types have different relationships with
land surface temperatures (LST) (Dontree, 2010). For example, naturally vegetated areas
have a negative relation with LST whereas, the impervious surface has a positive relation
* Urban Planner, Email: aninda.sh15@gmail.com
** Professor, Urban and Rural Planning Discipline, Khulna University, Khulna-9208, Bangladesh,
Email: saiful_ku@yahoo.com

2 Journal of Bangladesh Institute of Planners, Vol. 11, 2018
with LST (USGS, 2016; Zeng et al., 2010). In response to LC change caused by rapid
urbanization, LST is changing dramatically (Islam and Islam, 2013) and widening the gap
between urban and rural temperature (Zhou and Wang, 2011). This is threatening urban
habitats by raising mortality rate, increasing energy consumption and carbon emission,
hence fueling global warming. All of which are damaging both the sustainability and
livability of cities (Li et al., 2018; Sidiqui et al., 2016). Recent studies on continuous air
temperature mapping have identified that LST and LC are significantly related to air
temperature. Air temperature (ArT) measured at weather stations cannot depict the
spatial variation of ArT over a large area because of the limited number. As a solution to
this problem, Huang et al., (2017) propose a new method for estimating ArT. He used
multiple linear regression considering LST, NDVI, and elevation as predictors (Huang et
al., 2017). However, Zhang et al., (2018) used simple linear regression between LST and
ArT to generate high-resolution spatial ArT. Moreover, Good (2015), used a dynamic
multiple linear regression model to generate high-resolution spatial ArT for Europe.
Bangladesh was declared as the lower-middle income country by the World Bank in 2014
and is currently working on achieving the middle-income status by 2021 (The World
Bank, 2016). It is a small country with a large population and 36.5% of them live in urban
areas making these urban centers one of the most densely populated areas of the world.
Urban population is still increasing at more than 3.3% per annum (ibid). Moreover, 62.4%
of its total population has access to electricity.
Evidently, light pollution occurs in areas with higher population density, LC type with a
higher percentage of impervious surface and higher LST. Moreover, ArT is related to the
LC type and LST. Therefore, there might be a significant relationship exists between
night light or light pollution and air temperature. As no prior research has been done
regarding this issue, this paper aims to investigate how LP is affecting the ArT in
Bangladesh.
Materials and Method
Data and Study Area
This research is on Bangladesh. However, as LP does not occur all over the country, only
the light polluted areas of Bangladesh until 2013 is considered in this research. Moreover,
2003 and 2013 is selected as the study period for this research to observe the change.
Night light data is acquired from the Defense Meteorological Satellite Programs’
Operational Lines can System (DMSP-OLS) sensor. This data provides a yearly average
cloud-free night light data globally. Data from F14, F15, and F18 satellites are collected.
For night light data of 2003, an average of F14 and F15 is computed and for 2013 data,
only F18 data is used. After acquiring data, it is processed for further analysis. This
includes projection, clipping, and demarcation of light polluted area. Where pixels of
DMSP-OLS night light image containing positive value is identified as a light polluted
area. Moreover, combining both years of data, the light polluted area was identified
(Figure 1 B).
ArT is estimated using MODIS MOD11A1 daily Land Surface Temperature (LST) and
MOD13A3 monthly vegetation indices data product of Terra constellation and ArT data
from 34 weather stations of Bangladesh Meteorological Department (BMD) (Figure 1 A).

Nexus between Light Pollution and Air Temperature: A Study of Bangladesh 3
Multiple linear regression model is used in this research to estimate air temperature (eq.
1), where LST and NDVI are considered as the explanatory variable and air temperature
as a dependent variable.


     

       
Where, 
 = Air Temperature,  = Land Surface Temperature,  = Normalized
Difference Vegetation Index, = Intercept,   = Regression coefficients and

=
Random error.
Using data collected from weather stations, LST and NDVI data and the regression
parameters   are calculated. Applying the parameters to the satellite image,
annual average continuous air temperature of 2003 and 2013 is calculated.
Source: Author
Figure 1: Locations of BMD Weather Stations (A) and Light Polluted Area (B).
Change Detection
Discriminant function is used to identify the change of LP and ArT data between 2003
and 2013. This function provides a probability of change of DN value between two
different periods for each pixel. The output value of the discriminant function ranges
from 0 to 1, where a value close to 0 means a lesser probability of change and vice-versa.
Among three types of changes, combined change function is used, which provides the
probability of change (increase or decrease) of value over the change of time.
Results and Discussion
Light Pollution
DMSP-OLS night-light data shows that Bangladesh has experienced a rapid increase in
light pollution between the year 2013 and 2003 (Figure 2). Applying the threshold value

4 Journal of Bangladesh Institute of Planners, Vol. 11, 2018
to declare light pollution (Butt, 2012), it is found that in 2003 only 7.1 percent of the total
area of Bangladesh was light polluted, and in 2013, the amount of area dramatically
increased to 25.4 percent (Table 1).
Table 1: Percentage of the light polluted area in 2003 and 2013.
Year
2003
2013
Not Polluted
92.9
74.6
Polluted
7.1
25.4
Low
5.1
21.1
Moderate
1.2
2.7
High
0.4
1.0
Extreme
0.3
0.6
Furthermore, using the natural breaks method, light polluted areas are categorized into
three categories (low, moderate, high, and extreme). Majority of the light polluted area
fall into low pollution category. However, in 2013 the area under this category increased
more than four-fold compared to 2003. Remaining categories doubled in size by 2013.
Dhaka is the most affected urban area followed by Chittagong, Khulna, Sylhet, Rajshahi,
Jessore, Rangpur, and other urban centers. In 2003, only urban centers were featured
with light pollution. Nonetheless, in 2013 light pollution spread out into the suburban
areas, indicating significant development of the human activity in these areas.
Source: Author
Figure 2: Light polluted areas of Bangladesh in 2003 and 2013.

Nexus between Light Pollution and Air Temperature: A Study of Bangladesh 5
Air Temperature
Air temperature maps are generated using equation 1 (Figure 3). GWR is not an option in
this case as the number of data points (34 weather stations) available for this regression is
not suitable for GWR. According to which, in 2013 the minimum annual average
temperature increased up to 25.5°C compared to 24.6°C in 2003. However, the maximum
annual average temperature remained the same (27°C). Evidently, the temperature in
Rajshahi Division, river, haaor and hill tracks areas are less than the rest of the country.
However, hotspots are found in both maps which are mainly the urban areas having the
highest estimated temperature. Hotspots can be easily identified in the 2003 map for
Dhaka and Chittagong. However, in 2013 the number of hotspots increased compared to
2003 whereas, the hotspots of 2003 increased in size. In 2013, the suburbs of Dhaka
featured increased temperature compared to the surrounding rural areas. This
phenomenon confirms that the urban areas, where natural land cover is less are prone to
have a higher temperature (Islam and Islam, 2013).
Source: Author
Figure 3: Air temperature of Bangladesh in 2003 and 2013.
Nexus between Air Temperature and Night Light
Earlier studies characterized urban areas as light polluted areas (Butt, 2012; Elvidge et al.,
1997; Khorram et al., 2014). Other studies also characterized them as warmer areas
(Dontree, 2010; Sidiqui et al., 2016; Zhou and Wang, 2011). In Bangladesh, areas that have
experienced an increase in LP in 2013 compared to 2003 also experienced an increase in
ArT. To identify the nexus between ArT and LP, estimated discriminant function of
combined change is used in the following analysis. Table 2 suggests that data of both
variables are significantly clustered as depicted by positive Z-score and significant p-
value. Following this result, cluster and outlier analysis is performed on both data sets.

6 Journal of Bangladesh Institute of Planners, Vol. 11, 2018
The analysis shows that clusters exist in urban areas in both data sets. The cluster in
Dhaka is significantly visible than other urban areas (Figure 4).
Table 2: Global Moran’s I Summary
Air Temperature
Night Light
Moran's Index:
0.4
0.1
z-score:
1399.8
489.8
p-value:
0
0
However, to understand the relationship between ArT and NL, both (ordinary least
square regression) OLS and geographically weighted regression (GWR) is used, where
the ArT is considered as the dependent variable and LP as an explanatory variable. The
result shows that GWR outperforms OLS. AICc for GWR is considerably lower compared
to OLS while Adjusted R2 also shows significant improvement for GWR. Thus, the
authors preferred GWR to explain the relationship between the variables. Moreover, the
database contains 67,000 data points which are comparatively more suitable for GWR.
Source: Author
Figure 4: Cluster and outlier analysis maps.
GWR was run using adaptive kernel. However, the model summary reveals that only 50
percent of air temperature change can be explained by the change in light pollution. The
spatial variation of the relationship between ArT and NL (the coefficient of explanatory
variable) is shown in Figure 5. The dark green areas on the map suggest that the relation
between NL and ArT is relatively higher compared to the brown areas. Meaning, in
green areas any change in NL will incur higher impact on ArT and vice versa.

Nexus between Light Pollution and Air Temperature: A Study of Bangladesh 7
Interestingly, the concentration of higher coefficients is observed mainly in urban areas.
In other words, the more LP occurs in an urban area, the warmer the area is.
Source: Author
Figure 5: Spatial variation in the relationship between the change of light pollution and
the change of air temperature.
Conclusion
As lights generate heat, it is most likely to feel warmer where light pollution is higher.
Moreover, light pollution also indicates the concentration of impervious surface, man-
made objects, and higher human activity. As vegetation is less or absent in impervious
surface, these areas are warmer than its surroundings. Therefore, increasing population
density in urban areas will result in the brighter night sky and warmer temperature.
However, light pollution alone cannot explain the total change in air temperature. Other
factors such as land cover type, wind, humidity, solar radiation might have their own
impact on it. As there is a lack of availability of high-resolution climatic data, they were
not considered in this research. Moreover, land cover classification of large areas (e.g.
entire Bangladesh) using remotely sensed data products (e.g. Landsat) is quite
challenging for many reasons. Though machine-learning algorithms have proven to be
competent in image classification with higher accuracy, they have their own limitations
too. However, this research has successfully achieved its purpose in identifying the nexus

8 Journal of Bangladesh Institute of Planners, Vol. 11, 2018
between light pollution and air temperature, which is quite significant and considering
the rate at which light pollution is increasing in Bangladesh, we should now consider it
with due importance. The planning community and the researchers should work
together to devise measures to limit light pollution in our country before it is too late.
Countries like USA, UK, Canada, Italy, Germany, Chez Republic, and Singapore have
taken planning interventions to reduce light pollution in their countries. And we should
not fall behind.
References
Butt, M. J. 2012. 'Estimation of Light Pollution Using Satellite Remote Sensing and Geographic
Information System Techniques', GIScience & Remote Sensing, vol. 49, no. 4, pp. 609–621.
Dontree, S. 2010. ‘Relation of Land Surface Temperature (LST) and Land Use/Land Cover (LULC)
from Remotely Sensed Data in Chiang Mai – Lamphun Basin’, paper presented in the
SEAGA conference 2010, at Hanoi, Vietnam, 23 - 26 November, 2010.
Elvidge, C. D., Baugh, K. E., Kihn, E. A., Kroehl, H. W., and Davis, E. R. 1997. ‘Mapping City Lights
with Nighttime Data from The DMSP Operational Linescan System’, Photogrammetric
Engineering and Remote Sensing, vil. 63, no. 6, pp. 727-734.
Elvidge, C. D., Cinzano, P., Pettit, D. R., Arvesen, J., Sutton, P., Small, C., … Ebener, S. 2007. ‘The
Nightsat Mission Concept’, International Journal of Remote Sensing, vol. 28, no. 12, pp.
2645–2670.
Elvidge, C. D., Safran, J., Tuttle, B., Sutton, P., Cinzano, P., Pettit, D., … Small, C. 2007. ‘Potential for
Global Mapping of Development Via a Nightsat Mission’, GeoJournal, vol. 69, no. 1–2, pp.
45–53.
Eckstein, D., Künzel, V., and Schäfer, L. 2017. Global climate risk index 2018: Who suffers most
from Extreme weather events? Weather-related loss events in 2016 and 1997 to 2016.
Germanwatch Nord-Süd Initiative eV.
Gallaway, T. 2010. ‘On light pollution, passive pleasures, and the instrumental value of beauty’,
Journal of Economic Issues, vol. 44, no. 1, pp. 71-88.
Good, E. 2015. ‘Daily Minimum and Maximum Surface Air Temperatures from Geostationary
Satellite Data: Satellite Min and Max Air Temperatures’, Journal of Geophysical Research:
Atmospheres, vol. 120, no. 6, pp. 2306–2324.
Huang, F., Ma, W., Wang, B., Hu, Z., Ma, Y., Sun, G., … Lin, Y. 2017. ‘Air Temperature Estimation
with MODIS Data Over the Northern Tibetan Plateau’, Advances in Atmospheric Sciences,
vol. 34, no. 5, pp. 650–662.
Islam, M. S. and Islam, K. S. 2013. ‘Application of Thermal Infrared Remote Sensing to Explore the
Relationship Between Land Use-Land Cover Changes and Urban Heat Island Effect: A Case
Study of Khulna City’, Journal of Bangladesh Institute of Planners, vol. 6, pp. 12.
Khorram, A., Yusefi, M. and Fardad, M. 2014. ‘Assessment of Light Pollution in Bojnord City Using
Remote Sensing Data’, International Journal of Environmental Health Engineering, vol. 3,
no. 1, pp. 19.
Li, H., Zhou, Y., Li, X., Meng, L., Wang, X., Wu, S., and Sodoudi, S. 2018. ‘A New Method to
Quantify Surface Urban Heat Island Intensity’, Science of The Total Environment, vol. 624,
pp. 262–272.
Sidiqui, P., Huete, A. and Devadas, R. 2016. ‘Spatio-Temporal Mapping and Monitoring of Urban
Heat Island Patterns Over Sydney, Australia Using MODIS and Landsat-8’, paper presented

Nexus between Light Pollution and Air Temperature: A Study of Bangladesh 9
in the 4th International Workshop on Earth Observation and Remote Sensing Applications
(EORSA) 2016, pp. 217–221, at Guangzhou, China, 4-6 July, 2016.
Sinha, P., Verma, N. K. and Ayele, E. 2016. ‘Urban Built-Up Area Extraction and Change Detection
of Adama Municipal Area Using Time-Series Landsat Images’, International Journal of
Advanced Remote Sensing and GIS, vol. 5, no. 1, pp. 1886–1895.
USGS, 2016. ‘Understanding Land Surface Temperature Dynamics’, https://www.youtube.com/
watch?v=UUQAX-aPbbE&t=80s, retrieved on April 4, 2018.
The World Bank, 2016. ‘Helping Bangladesh Reach Middle Income Country Status’,
http://www.worldbank.org/en/news/feature/2016/04/07/World_Bank_Group_s_New_
Country_Partnership_Framework_helps_Bangladesh_Reach_Middle_Income_Country_Stat
us, retrieved on April 2, 2018.
Zeng, Y., Huang, W., Zhan, F., Zhang, H. and Liu, H. 2010. ‘Study on The Urban Heat Island
Effects and Its Relationship with Surface Biophysical Characteristics Using MODIS
Imageries’, Geo-Spatial Information Science, vol. 13, no. 1, pp. 1–7.
Zhang, H., Zhang, F., Zhang, G., Ma, Y., Yang, K. and Ye, M. 2018. ‘Daily Air Temperature
Estimation on Glacier Surfaces in The Tibetan Plateau Using MODIS LST Data’, Journal of
Glaciology, vol. 64, no. 243, pp. 132–147.
Zhou, X. and Wang, Y. C. 2011. ‘Dynamics of Land Surface Temperature in Response to Land-
Use/Cover Change: Dynamics of Land Surface Temperature’, Geographical Research, vol.
49, no. 1, pp. 23–36.