A Study of After-Effects of Kerala Floods Using VIIRS-OLS Nighttime Light Data


Remote sensing is a widely used technique for the study of the Earth’s surface features. Nighttime light imagery forms a useful data set in remote sensing. It is an essential dataset for the study of human activities on Earth and their effects. Such datasets are also found to be of great importance in the study of the atmosphere and other natural processes. A few common examples are the detection and monitoring of city lights, fires, dust storms, volcanoes, gas flares and population/economic geography (Chuvieco 2016). As the name suggests, nighttime light imagery data is collected during the night. It usually contains two types of features: self-illuminating features and moon-illuminated features. Some examples of self-illuminating features include gas flares, forest fires, human-caused disasters, volcanic lava and bioluminescence, whereas moonlight illuminating features include snow cover, sea ice and volcanic ash along with various surface features such as mountains, deserts, rivers and moon glint. Another source of illumination for the clouds can be airglow. It is the luminosity occurring due to chemical reactions in the upper atmosphere. Nighttime visible imaging was initiated by the Defense Meteorological Satellite Program - Operational Linescan System (DMSP-OLS) in the 1960s, which was the only source of visible nighttime images until the launch of Suomi National Polar- Orbiting Partnership - Visible Infrared Imaging Radiometer Suite (SNPP-VIIRS) in 20И. The VIIRS instrument collects visible and infrared imagery and global observations of the land, atmosphere, cryosphere and oceans. In Doll (2008) and Elvidge et al. (2017), a detailed explanation of the generation and usage of the Visible Infrared Imaging Radiometer Suite Day/Night Band (VIIRS-DNB) satellite data is available.

Several studies have highlighted the use of nighttime light data for efficient disaster management. For example, Gillespie et al. (2014) have utilized the nighttime light imagery from the DMSP-OLS to study the effect of the 2004 Indian Ocean mega-tsunami in Indonesia. Their findings were validated based on the extensive aftermath survey results. An empirical relationship between economic expenditure after the tsunami and the nighttime light data was established (Gillespie et al. 2014). Similarly, Zhao et al. (2018) used the National Polar-Orbiting Partnership Visible Infrared Imaging Radiometer Suite Day/Night Band (NPP-VIIRS DNB) daily data to study the effect of three major disasters, namely the earthquake, storms and floods. The percent-of-normal light (PNL) method was employed to assess the damage caused by the disaster. The NPP-VIIRS DNB data one month prior to disaster and ten days after the disaster was averaged to obtain predisaster and postdisaster values. In their study, a longer time period for predisaster was chosen so as to account for the variations of the light intensity due to clouds and other effects, whereas a shorter postdisaster period was chosen to represent a critical period after which postdisaster activity occurs. The nonparametric Mann-Whitney U-Test was employed to test the statistical significance of the results. In an another effort, Wang et al. (2018) studied the power outages in the United States as an aftermath of Hurricane Sandy in 2012 and Hurricane Maria in 2017 using NASA black marble product data that removes cloud-contaminated pixels and corrects for atmospheric, terrain, vegetation, snow, lunar and stray light effects on the VIIRS DNB radiances. Percent-of-normal light technique was used to assess the damage.

Motivated by the above applications of nighttime light data in disaster assessment and aftermath management, the present study aims to estimate flood-related damage in the state of Kerala using VIIRS-DNB scan data. The floods that occurred in the month of August 2018 in Kerala were caused by the unusually high rainfall during the monsoon season. They were one of the worst floods in Kerala in nearly a century, causing property damage of about <1400 billion (US$5.8 billion) and casualty of over 483 human deaths in the state.


For the present study, we use the DNB sensor data of VIIRS instrument (VIIRS- DNB) provided by the Earth Observations Group (EOG) at NOAA/NCEI. Table I4.l summarizes some key specifications of the sensor. The VIIRS-DNB data is mainly of three types: stable lights, radiance calibrated and the average digital number based on the detection frequency, radiance and digital number, respectively. These differences make the datasets usable in several applications, such as the study of urban extent, population change over time, socio-economic activity, greenhouse

TABLE 14.1

VIIRS-DNB Sensor Specification


22 spectral bands from 412 nm to 12 pm

Nadir Resolution

400 m


3.000 km (max)

Average Data Rate

7.674.000 bps

Average Power

319 Watts

gas emissions, light pollution and disaster management. In this work, the study region consists of the state of Kerala, a coastal state located in the southwestern part of India.


The methodology comprises three steps: preprocessing of data, processing and postprocessing analysis.

14.3.1 Preprocessing

When a satellite records data, some phenomena such as stray lights, reflected moonlight from clouds, snow cover, topographic variations and the position of the pixels in the swath substantially affect the recorded radiance value. To account for such disturbances, lunar radiance removal, cloud clearing and edge-of-swath pixel removal are often used. To preprocess the data for the present work, we use the summed cloud cover (SCC) recorded by the satellite and the average-swath technique that considers the aggregated value of swath from 32 different zones on each side of the nadir via a scan-angle-dependent aggregation strategy.

14.3.2 Processing

To enable rapid detection of the flood affected areas in the state of Kerala, we have chosen the percent of normal light (PNL) technique, which is based on the realization that the flood-affected areas will have direct impact on the urban lights. As the nighttime light data captures these lights, they are used to develop the PNL method. To detect the change, the radiance values before and after the disaster are compared by taking their ratio. If the ratio is less than 1, the area is classified as affected. The PNL technique, due to its simplicity and availability of relevant dataset, is often preferred for rapid damage assessment over other methods, such as the false color composite. Flood Proxy Maps (FPM) and estimation using economic parameters.

To implement the PNL technique, VIIRS-DNB scan data for the months of July, August and September was analyzed. The radiance values were first thresholded to ensure a correct calculation of the PNL values, as for very small radiance values.

background noise can result in extremely high or extremely low values of PNL. Thus, Radpre and Rad , values below 0.3 nW.cm -.sr~' have been thresholded. The accrued monthly data was then averaged to generate the composite radiance values. As the floods occurred in the month of August, the data for the month of August was taken to compute postdisaster radiance values. The data for the months of July and September are averaged and taken as predisaster radiance values. Using the above values, the PNL image for the study region was generated by taking the ratio of predisaster radiance values and postdisaster radiance values as

14.3.3 Postprocessing

After obtaining the PNL image, a simple criterion of PNL value less than l is employed to demarcate the affected areas. District-level sum of lights statistics is then generated using the composite radiance values. Sum of lights is defined as the sum of all the radiance values in the district. The obtained values are plotted on a histogram to observe the drop in radiance values after the disaster. The districts with the large drop in radiance values were identified as worst affected.

A flow-chart in Figure 14.1 describes the above methodology of flood damage assessment in the state of Kerala.

Flow-chart of the proposed methodology

FIGURE 14.1 Flow-chart of the proposed methodology.

(a) The PNL values of Kerala state and (b) the 2018 flood-affected areas

FIGURE 14.2 (a) The PNL values of Kerala state and (b) the 2018 flood-affected areas

(PNL < 1).


In this section, we present the results. The image of PNL values for the study region is shown in Figure I4.2a. Areas w'ith low PNL values (less than 1) suggest that the region is highly affected by the floods, as the lights coming from these areas after the flood have decreased by a larger amount. The extent of damage by floods are represented in a color-coded image as shown in Figure 14.2b, whereas a summary of PNL values is provided in Table 14.2.

The predisaster and postdisaster images were used to calculate district-wise statistics for Kerala. The predisaster and postdisaster radiance values for each district were summed and represented in Figure 14.3a and Figure 14.3b, respectively. Using the radiance-sum value for each district, a graph (Figure 14.4) was generated to highlight the change in the predisaster and postdisaster radiance-sum of lights for each district.

From the graph in Figure 14.4, it is observed that the postdisaster values of the sum of lights are less than those for the predisaster values. According to ground surveys

TABLE 14.2 Summary of Results

% area affected


Number of affected pixels (out of 179,310)


Mean PNL value


Range of PNL values


District-wise sum of lights in Kerala; (a) pre-disaster sum of lights and (b) post-disaster sum of lights

FIGURE 14.3 District-wise sum of lights in Kerala; (a) pre-disaster sum of lights and (b) post-disaster sum of lights.

Bar plots for pre-disaster and post-disaster radiance-sum of lights for the districts of Kerala

FIGURE 14.4 Bar plots for pre-disaster and post-disaster radiance-sum of lights for the districts of Kerala.

carried out by the Indian government and various agencies (Kerala Floods 2018), the worst affected areas of the state were Wayanad. Pathanamthitta, Ernakulam, Thrissur, Malappuram, Kozhikode, Kannur, Palakkad and Alappuzha. It can be observed from Figure 14.4 that there is a significant drop in the radiance values for these districts except for Wayanad. Due to insufficient radiance values even before the disaster, the effect of flood measured through the sum of lights is not well represented for Wayanad district.

In summary, in this chapter, we have used the concept of percent of normal light to identify disaster-affected areas in the state of Kerala through the nighttime light image. We further calculated the sum of lights for each district and plotted the results. We have not involved any economic parameters in the analysis as the analysis is carried out for a short duration of time around the disaster. An economic parameter will not be a suitable indicator for such a duration. A significant drop was observed in the values of the postdisaster sum of lights. The districts with the most drop were identified as severely affected. These results were compared with the ground reports and were found to be valid. Thus, the proposed analysis is found to be useful for rapid and efficient flood damage assessment in an area.

Processing system: The system used to generate the results has an Intel i7 8th gen CPU unit along with a NVIDIA GeForce GTX 1050 Ti (4 GB) GPU. The system has a DDR4 RAM of 16 gigabytes. We use ArcGIS (vl0.2.2) suite of geospatial processing software for data processing.


Chuvieco. E. 2016. Fundamentals of Satellite Remote Sensing: An Environmental Approach. Boca Raton, FL: CRC Press, Taylor & Francis Group.

Doll. C.N.H. 2008. CIESIN thematic guide to night-time light remote sensing and its applications. New York, NY: Center for International Earth Science Information Network of Columbia University. Retrieved from http://sedac.ciesin.columbia.edu/tg/ (Accessed on 27 November 2019).

Elvidge, C.D., Baugh. K.. Zhizhin, M.. Hsu. F.-C. and Ghosh. T. 2017. Supporting international efforts for detecting illegal fishing and GAS flaring using VIIRS. 2017 IEEE International Geoscience and Remote Sensing Symposium (IGARSS). Doi: 10.1109/ igarss.2017.8127580.

Gillespie. T.W.. Frankenberg, E., Chum. K.F. and Thomas. D. 2014. Night-time lights time series of tsunami damage, recovery, and economic metrics in Sumatra, Indonesia. Remote Sensing Letters. 5(3), 286-294. Doi: 10.1080/2150704x.2014.900205.

Kerala Floods. 2018. These are the worst affected areas, stay clear of them. Mirror Now News (website), https://www.timesnownews.com/mirror-now/in-focus/article/kerala-floods- 2018-these-are-the-worst-affected-areas-stay-clear-of-them/271873.

Wang. Z„ Roman. M.O., Sun, Q., Molthan. A. L.. Schultz, L. A. and Kalb, V. L. 2018. Monitoring disaster-related power outages using NASA black marble nighttime light product. ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. XLII-3, 1853-1856. Doi: 10.5194/ isprs-archives-xlii-3-1853-2018.

Zhao. X.. Yu. B„ Liu. Y.. Yao. S.. Lian. T„ Chen, L.....Wu, J. 2018. NPP-VIIRS DNB

daily data in natural disaster assessment: Evidence from selected case studies. Remote Sensing. 10(10), 1526. Doi: 10.3390/rsl0101526.

< Prev   CONTENTS   Source   Next >