The Egyptian Journal of Remote Sensing and Space Sciences (2015) 18, 99–106

HO ST E D BYNational Authority for Remote Sensing and Space Sciences

The Egyptian Journal of Remote Sensing and Space

Sciences

www.elsevier.com/locate/ejrswww.sciencedirect.com

RESEARCH PAPER

Flood monitoring and damage assessment using

water indices: A case study of Pakistan flood-2012

* Corresponding author.

Peer review under responsibility of National Authority for Remote

Sensing and Space Sciences.

http://dx.doi.org/10.1016/j.ejrs.2015.03.0031110-9823 � 2015 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier B.V.This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Akhtar Ali Memon a,*, Sher Muhammad b, Said Rahman a, Mateeul Haq a

a Pakistan Space & Upper Atmosphere Research Commission (SUPARCO), Pakistanb Key Laboratory of Tibetan Environmental Changes and Land Surface Processes, Institute of Tibetan Plateau Research,

Chinese Academy of Sciences, China

Received 7 August 2014; revised 4 February 2015; accepted 22 March 2015Available online 16 April 2015

KEYWORDS

Water Index;

Flood-2012;

Pakistan floods;

Damage assessment;

MODIS;

Landsat ETM+

Abstract This paper uses Normalized Difference Water Index (NDWI) of McFeeters (1996),

Water Index (WI) introduced by Rogers and Kearney (2004), referred to as Red and Short

Wave Infra-Red (RSWIR) and WI suggested as the best by Ji et al. (2009), referred to as Green

and Short Wave Infra-Red (GSWIR) for delineating and mapping of surface water using

MODIS (Terra) near real time images during 2012 floods in Pakistan. The results from above

indices have been compared with Landsat ETM+ classified images aiming to assess the accuracy

of the indices. Accuracy assessment has been performed using spatial statistical techniques and

found NDWI, RSWIR and GSWIR with kappa coefficient (j) of 46.66%, 70.80% and 60.61%

respectively. It has been observed using statistical analysis and visual interpretation (expert knowl-

edge gained by past experience) that the NDWI and GSWIR have tendencies to underestimate and

overestimate respectively the inundated area. Keeping in view the above facts, RSWIR has proved

to be the best of the three indices. In addition, assessment of the damages has been carried out

considering accumulated flood extent obtained from RSWIR. The information derived proved to

be essential and valuable for disaster management plan and rehabilitation.� 2015 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier

B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/

by-nc-nd/4.0/).

1. Introduction

Pakistan is a flood-prone country with a history of widespreadand repeated flooding that causes loss of lives, substantial

damage to property, infrastructure, loss of agricultural cropsand land (Sardar et al., 2008). It has two dominant types offloods i.e. riverine and flash. The riverine flood, mainly

originating from glacier melt as well as extreme weather events,is predictable which in result allows sufficient time to reactwhereas the converse is true for flash floods. Flash floods,originating from extreme weather events, have relatively less

duration but intensity and impacts are severe. These floodsnormally occur during the South Asian monsoon periodbetween July and September.

The country has been continuously experiencing majorfloods since 2001. The major flood in 2010 inundated an areaof 70,238 km2 and affected around 884,715 homes (Haq,

2010). In 2011, another major flood caused an inundation of

100 A.A. Memon et al.

21,200 km2, affected 5.88 million people, 1500 km of road net-work, 382 km of railway tracks, 498 km2 of forests and16,440 km2 of agricultural land (Haq et al., 2012). In this year,

moderate scale rains have been observed, starting from Augustand intensified in September, resulting in huge economiclosses, destruction of ecological sources, food shortage and

starvation of thousands of people. During the first half ofSeptember, widespread rains were experienced throughoutthe country, which severely affected the areas of Punjab,

Sindh and Balochistan provinces. During 06–11 September,Jacobabad, Larkana, Sukkur, Thatta and Tharparkar inSindh received heavy rains resulting in flood. In addition, rainsin Balochistan not only brought disaster in Balochistan but

also the districts of Sindh received devastating flash floodsdue to hill torrents in the Suleman range. In Punjab, RahimYar Khan, Sahiwal, Okara, Jehlum, Mandi Bahaudin, DG

Khan and Rajanpur experienced extensive and continuousdownpour from 08 to 12 September. The available daily rain-fall data of 8 districts for the months of August and September

are given in Table 1(a) and (b).Estimation of Land surface water (area), which includes

rivers, lakes, wetlands and flood inundated areas, plays an

important role in the evaluation of the water cycle. With globalclimate change and the growing influence of human activities,river changes (Xu, 2004; Xia and Zhang, 2008), shrinkage ofwetlands, glacier melting (Chen et al., 2007), flooding, and

other spatial and temporal distribution changes in surfacewater resources are increasingly highlighted. A variety of stud-ies have been focused on impact assessment to flood disaster

on different groups of people and communities (e.g.,Schmuck-Widmann, 1996; Rashid, 2000; Rasid and Haider,2003) and found the poor people to be hardest hit (IPPC,

2001).Modern techniques of remote sensing provide tremendous

potential for monitoring and managing dynamic changes in

large surface water bodies; extracting hydrological parameters,and modeling the water balance (El Bastawesy, 2015;El-Gamily et al., 2010). McFeeters (1996) proposed theNormalized Difference Water Index (NDWI) using Near

Infrared (NIR) and green channels of Landsat that can delin-eate and enhance open water features. Xu (2006) modified theNDWI proposed by McFeeters (1996) by substituting the mid-

dle infrared band for NIR band. The NDWI defined as (refl(0.86 lm) � refl (1.24 lm))/(refl (0.86 lm) + refl (1.24 lm)),was introduced by Gao (1996) for extracting vegetation liquid

water. Rogers and Kearney (2004) introduced NDWI usingred and Short Wave Infra-Red (SWIR) channels of LandsatTM. Xiao et al. (2005) defined a Land Surface Water Index(LSWI) using NIR and SWIR bands of MODIS. Ouma and

Tateishi (2006) proposed a new Water Index (WI) for shorelineedge extraction using a logical combination of Tasseled CapWetness (TCW) index and NDWI using Landsat Thematic

Mapper (TM) and Enhanced Thematic Mapper Plus(ETM+) data. Wu et al. (2008) introduced a new combinatedmethod for extraction of water body information using

ETM+. The method is based on combining the spectralrelationship between different bands with the TCW indexand removing other objects from the mask image using maxi-

mum likelihood classification methods. Li and Zhou (2009)presented a method for Mountain area Enhanced Water-bodyIdentification (MEWI) through the Normalized DifferenceVegetation Index (NDVI) and red band of MODIS. Ji et al.

(2009) used spectral data to simulate the satellite sensors dataand calculated the simulated NDWI in various forms and con-cluded that the calculated NDWI from (Green � SWIR)/

(Green + SWIR), where the SWIR is a region from 1.2 to1.8 lm, has the most stable threshold. Ling et al. (2010) usedthe Combined Index of NDVI and NIR for Water Body

Identification (CIWI) aiming to extract coastal wetlands waterinformation using Landsat TM images. Lu et al. (2011) pro-posed an integrated water body mapping method through

combination of difference between NDVI and NDWI(NDVI–NDWI) with slope and NIR band using HJ-1A/Bsatellite images.

In this paper, we have selected three water indices

proposed by McFeeters (1996), Rogers and Kearney (2004)and Ji et al. (2009). The methods selected in this study aredue to their generality in delineating and enhancing surface

water features. The high temporal resolution of MODISimages used in this study played a major role in monitoringthe 2012 floods in Pakistan. The paper recommends the best

of the three mentioned indices for mapping flood inundatedareas and to assess the damage caused by the 2012 floods inthe country.

2. Study area

The study area includes 20 adjacent districts of the

Baluchistan, Khyber Pakhtunkhwa, Punjab and Sindhprovinces of Pakistan ranging from 26�5013.900 to 33�1302.2100in latitude and 67�6052.6800 to 72�0015.8600 in longitude(Fig. 1). On average the mountainous area of the study area

receives 200–960 mm rain annually and rest of the area, whichis mostly plain, receives 180–460 mm rain annually.

3. Methodology

Five images of MODIS onboard the TERRA satellite acquiredin 0.5 km resolution through NASA’s website ‘‘http://

rapidfire.sci.gsfc.nasa.gov/realtime/’’, have been used as amain input to map flood inundated areas. Images have beengeo-referenced using the header file allied with the data. The

study area has been extracted from the images using Districtadministrative boundaries. Cloud shadow was frequentlyidentified as water, by most water indices. So careful removal

of cloud shadow using visual interpretation has beencarried out. Cloud masking has been performed using NIR(band 2 of MODIS) by clipping the image that has reflectanceabove 0.7 as the spectral reflectance of cloud is greater than 0.7

in NIR region.Flood mapping has been carried out using NDWI, RSWIR

and GSWIR as shown in Eqs. 1–3 respectively by setting the

threshold from 0 to 1. Two images of Landsat ETM+, havingpath 152, row 041, of September 17, 2012 and October 03,2012, have been acquired for accuracy assessment of the results

obtained through mentioned indices. A standard supervisedmaximum likelihood classification of Landsat ETM+ imageshas been performed. Validation of the classification has been

performed by visual interpretation of Landsat ETM+ imagesusing false color 543 composite. Contingency matrix and itsallied statistics (i.e. overall accuracy, producer’s accuracy,user’s accuracy and kappa coefficient (j)), reported at

Congalton (1991), have been performed for accuracy

Table 1 (a) Daily rainfall (mm) data for the month of August 2012 (b) Daily rainfall (mm) data for the month of September 2012.

Districts 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

(a)

Dera Ismail Khan 0 0 0 0 49 0 0 0 0 0 16 0 0 0 0 1 0 0 0 0 0 0 0 0 0 5 63 0 0 0 0

Bhakkar 0 0 0 0 21 0 0 0 0 0 22 0 0 0 0 0 0 0 0 0 0 0 16 0 0 2 0 0 5 0 0

Dera Ghazi Khan 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 1 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0

Rahim Yar Khan 0 0 0 0 0 0 14 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Dadu 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Jacobabad 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Larkana 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Sukkur 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0

(b)

Districts 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Dera Ismail Khan 0 0 0 49 17 0 0 0 13 8 0 0 0 0 0 0 0 34 0 0 0 0 0 0 0 0 0 0 0 0

Bhakkar 0 0 0 0 56 0 3 0 15 20 0 0 0 0 0 7 54 0 0 0 0 0 0 0 0 0 0 0 0 0

Dera Ghazi Khan 0 0 0 1 8 2 0 5 35 76 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

Rahim Yar Khan 0 0 0 0 0 9 0 6 102 32 96 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Dadu 0 0 0 0 0 0 82 0 2 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Jacobabad 0 0 0 0 0 0 24 0 9 – 143 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Larkana 0 0 0 0 0 1 42 10 47 58 58 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Sukkur 0 0 0 0 0 0 1 1 4 164 36 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

‘–’ data not available.

Floodmonito

ringanddamageassessm

entusin

gwater

indices

101

Figure 2 Inundated areas obtained through mentioned indices on different dates.

Table 2 Accuracy assessment of NDWI using error matrix.

Values show the area in km2.

NDWI results Landsat ETM+ Total

Water Others

MODIS

Water 390 80 470

Others 185 331 516

Total 575 411 986

Table 3 Accuracy assessment of RSWIR using error matrix.

Values show the area in km2.

RSWIR results Landsat ETM+ Total

Water Others

MODIS

Water 505 75 580

Others 65 341 406

Total 570 416 986

Figure 1 Location of study area.

102 A.A. Memon et al.

assessment. After analyzing the best Water Index resultingflood water class, obtained from the corresponding index,

has been converted into a shapefile and used as an input toassess damage for various land use/land cover categories.

NDWI ¼ GREEN�NIR

GREENþNIRð1Þ

RSWIR ¼ RED� SWIR

REDþ SWIRð2Þ

Table 4 Accuracy assessment of GSWIR using error matrix.

Values show the area in km2.

GSWIR results Landsat ETM+ Total

Water Others

MODIS

Water 510 115 625

Others 70 291 361

Total 580 406 986

Figure 3 Results of statistica

Figure 4 Extent of cumulative damage

Flood monitoring and damage assessment using water indices 103

GSWIR ¼ GREEN� SWIR

GREENþ SWIRð3Þ

4. Results and discussions

In order to evaluate the performance of the mentioned indices

we have compared them with classified images of LandsatETM+ through statistical techniques mentioned before.

l analysis of error matrix.

assessment on September 17, 2013.

104 A.A. Memon et al.

Also visual interpretation, using expert knowledge, of MODISimages with 721 composite has been performed for the pur-pose. In order to preserve normalization for better comparison

we have set the threshold from 0 to 1 while applying waterindices to MODIS images. It has been observed that varyingthe threshold to the best values could impact the efficiency

of the individual method, also reported by McFeeters (1996),Xu (2006), Ouma and Tateishi (2006) and Wu et al. (2008),but gives relatively same accuracy. This might be due to poor

atmospheric correction of images. Note that the EarthObserving System Data and Information System (EOSDIS)provides MODIS images that are atmospherically correctedbut due to diurnal variations in atmospheric parameters it

would not be possible to purely normalize the images byatmospheric correction.

Fig. 2 shows inundated areas obtained on different dates

through mentioned water indices. It reveals that the variationsin inundated areas follow similar trends for RSWIR andGSWIR while the converse is true for NDWI. This suggests

that the NDWI maps roughly the inundated area. It alsoshows that NDWI underestimates the inundated area. The val-idation shown as error matrices (Tables 2–4), is obtained

through MODIS and Landsat ETM+ images of 17September 2012 and 03 October 2012. The rasters obtainedfrom the mentioned indices applied on MODIS images andclassification on Landsat ETM+ were converted to vector

files. The vector files were then intersected to get the resultsas shown in Tables 2–4. Rows in these tables show the inun-dated area derived from MODIS while the columns represent

the inundated area derived from Landsat ETM+ images.The row total shows water and other land use/land cover areain square kilometers derived from MODIS, similarly for

Table 5 Detailed statistics of flood affected land use/land cover ca

Province District Accumulated

inundated area

(km2)

Villa

affec

(Cou

BALOCHISTAN BOLAN 736.24 39.0

BALOCHISTAN JAFARABAD 1144.69 214.

BALOCHISTAN JHAL MAGSI 581.29 24.0

BALOCHISTAN LORALAI 53.97 2.00

BALOCHISTAN NASIRABAD 725.46 35.0

KHYBER

PAKHTUNKHWA

D I KHAN 328.65 27.0

PUNJAB BHAKKAR 27.34 7.00

PUNJAB DERA GHAZI

KHAN

527.11 78.0

PUNJAB Layyah 201.93 57.0

PUNJAB MUZAFFARGARH 821.33 197.

PUNJAB RAHIM YAR

KHAN

500.24 108.

PUNJAB RAJANPUR 1038.30 213.

SINDH DADU 413.98 102.

SINDH GHOTKI 494.56 148.

SINDH JACOBABAD 1557.67 573.

SINDH KASHMORE 1185.42 291.

SINDH LARKANA 250.32 75.0

SINDH QAMBAR

SHADAD KOT

1390.14 332.

SINDH SHIKARPUR 682.92 184.

SINDH SUKKUR 495.80 107.

Landsat ETM+ via column. Each cell value shows the agree-ment/disagreement of MODIS with Landsat ETM+ results.Tables 2–4 and Fig. 3 show underestimation of 18.26% for

NDWI and overestimation of 1.75% and 7.76% for RSWIRand GSWIR respectively, while j (it is a measure of overallagreement of a matrix) for NDWI, RSWIR and GSWIR is

measured to be 46.66%, 70.80% and 60.61% respectively.Overall accuracy, producer’s accuracy (it is a measure ofhow well a certain area is classified), user’s accuracy (it is a

measure of how much the results, obtained through producer’saccuracy, are reliable) and kappa coefficient of NDWI,RSWIR and GSWIR for water and other land use/cover(Fig. 3) clearly depict in detail that RSWIR is relatively more

accurate in delineating water bodies.The vegetation has relatively high reflectance in NIR

region, so the NDWI could not take water under vegetation

into account and ultimately underestimate the inundated area.Most of the study area is vegetated, that is why NDWI has sig-nificantly underestimated the inundated area. Vegetation has

reflectance of around 17%, 10% and 20% for green, red andSWIR regions respectively of electromagnetic spectrum. Thedifference in reflectance between green and SWIR is very small

that could vanish in chaotic atmospheric conditions duringfloods and make GSWIR to take vegetation into account (asmentioned above) which results in overestimation in theinundated area. This could not be the case for RSWIR because

red and SWIR have large differences regarding reflectance.

4.1. Damage assessment

One of the most important applications of water indices inPakistan is the mapping and monitoring of flood events.

tegories based on MODIS data.

ges

ted

nt)

Affected

railway

(km)

Affected

roads (km)

Affected

agriculture

(km2)

Forest

affected

(km2)

0 2.26 64.07 0.00 0.00

00 11.07 332.14 295.64 0.00

0 0.00 39.69 62.09 0.00

0.00 0.00 0.00 0.00

0 7.58 61.60 330.26 29.42

0 0.00 19.20 5.27 0.00

0.00 0.00 20.07 0.00

0 0.00 29.51 18.88 35.04

0 0.00 0.00 20.15 0.00

00 0.00 11.91 104.46 134.24

00 0.00 10.41 165.68 38.83

00 37.44 160.14 200.44 93.02

00 2.01 95.31 115.03 9.04

00 4.22 74.07 160.09 123.92

00 12.63 377.43 294.87 0.00

00 4.26 90.24 350.99 192.94

0 3.03 26.74 43.69 70.08

00 0.00 172.12 304.16 0.00

00 11.50 119.34 220.45 46.82

00 13.24 37.89 234.05 141.16

Flood monitoring and damage assessment using water indices 105

During flood events, policy makers and disaster managementauthorities require rapid identification and extent of floodingat a reasonable scale aiming at minimizing damage and

vulnerability. The accumulated flood extent (Fig. 4) of13,157 km2 has been observed on September 17, 2012 withsevere rainfall recorded from 06 to 11 September. It has been

observed that inundation in the low lying areas in upperSindh is greater than other provinces despite the low amountof rainfall measurements. This is due to flash floods in

the catchment areas of Baluchistan province that flowsinto Sindh province causing inundation in Jacobabad,Shahdadkot, Kashmore and Ghotki districts. Damagesassessed include agriculture, villages, roads, forest and railway

in order of severity from high to moderate. All the provinceswere affected by floods but assessed damages and losses inSindh were relatively severe followed by Baluchistan, Punjab

and Khyber Pukhtoonkhwa. Table 5 shows the accumulatedinundated areas with damage assessment of land cover/landuse in various districts.

5. Conclusion

Due to high temporal resolution of MODIS it is used widely to

map flood events. But its low spatial resolution questions onaccuracy of delineating flood extent. In this regard, one ofthe aims of the study is to check the accuracy of MODIS data

with respect to LANDSAT in mapping flood events in thecountry without taking the difference in spatial resolution intoaccount. It is estimated that the MODIS data could help tomap the inundated area with an accuracy of 73–81% depend-

ing on the mentioned indices used. Further, the comparativeanalysis of NDWI, RSWIR and GSWIR using spatial sta-tistical techniques and visual interpretation suggests RSWIR

to be more efficient for delineating water bodies during floods.Visual interpretation using past experience provides qualitativeresults and statistical analysis presents quantitatively. The

flood was caused by heavy monsoon rains mostly in theEastern part of Baluchistan and flowed toward upper Sindhprovince during early September 2012. It inundated 22 districts

of Pakistan with maximum cumulative extent of 13,157 km2.The study also provides detailed information about theaffected land use/land cover in the country as mentioned inTable 5.

Acknowledgements

A part of the work has been carried out for the NationalDisaster Management Authority (NDMA), Pakistan andSindh Irrigation Department, Pakistan. We acknowledge the

use of Rapid Response imagery from the Land AtmosphereNear-real time capability for EOS (LANCE) system operatedby the NASA/GSFC/Earth Science Data and Information sys-

tem (ESDIS) with funding provided by NASA/HQ. We aregrateful to Mr. Rahmatullah Jilani of SUPARCO for encoura-ging and giving us moral support to carry out the study.We thank Dr. Muhammad Shafiq from Earth Sciences

Directorate of SUPARCO for giving useful suggestions toimprove the manuscript.

References

Chen, X.Q., Cui, P., Li, Y., Yang, Z., Qi, Y.Q., 2007. Changes in

glacial lakes and glaciers of post-1986 in the Poiqu River basin,

Nyalam, Xizang (Tibet). Geomorphology 88 (3), 298–311.

Congalton, R.G., 1991. A review of assessing the accuracy of

classifications of remotely sensed data. Remote Sens. Environ. 37

(1), 35–46.

El Bastawesy, M., Gabr, S., Mohamed, Ihab, 2015. Assessment of

hydrological changes in the Nile River due to the construction of

Renaissance Dam in Ethiopia. Egypt. J. Remote Sens. Space Sci.

18 (1), 65–75.

El-Gamily, I.H., Selim, G., Hermas, E.A., 2010. Wireless mobile

field-based GIS science and technology for crisis management

process: a case study of a fire event, Cairo, Egypt. Egypt. J. Remote

Sens. Space Sci. 13 (1), 21–29.

Gao, B.C., 1996. NDWI––normalized difference water index for

remote sensing of vegetation liquid water from space. Remote Sens.

Environ. 58 (3), 257–266.

Haq, M., Said, R., Rahmatullah, J., Mohammad, S., 2010. Pakistan –

flood 2010 monitoring using MODIS data. In: The 9th

International Workshop of the CAS-TWAS-W Forum (2010

CTWF) on Climate and Environmental Changes and Challenges

for Developing Countries, November 17–20, 2010, Beijing, China.

Haq, M., Akhtar, M., Muhammad, S., Paras, S., Rahmatullah, J.,

2012. Techniques of remote sensing and GIS for flood monitoring

and damage assessment: a case study of Sindh province, Pakistan.

Egypt. J. Remote Sens. Space Sci. 15, 135–141.

Houghton, J.T., 2000. The IPCC Report 2001. In: The Solar Cycle

and Terrestrial Climate, Solar and Space weather, vol. 463, pp. 255.

Ji, L., Zhang, L., Wylie, B., 2009. Analysis of dynamic thresholds for

the normalized difference water index. Photogramm. Eng. Remote

Sens. 75 (11), 1307–1317.

Li, B., Zhou, X.Z., 2009. Research on extraction method of river

water-body in mountain area of Western Sichuan based on MODIS

data. Geogr. Geoinf. Sci. 2, 017.

Ling, C., Zhang, F.Q., Lin, H., 2010. Research on extracting coastal

wetlands water information using composition ciwi water index.

Resour. Environ. Yangtze Basin 2, 007.

Lu, S., Wu, B., Yan, N., Wang, H., 2011. Water body mapping

method with HJ-1A/B satellite imagery. Int. J. Appl. Earth Obs.

Geoinf. 13 (3), 428–434.

McFeeters, S.K., 1996. The use of the normalized difference water

index (NDWI) in the delineation of open water features. Int. J.

Remote Sens. 17 (7), 1425–1432.

Ouma, Y.O., Tateishi, R., 2006. A water index for rapid mapping of

shoreline changes of five East African Rift Valley lakes: an

empirical analysis using Landsat TM and ETM+ data. Int. J.

Remote Sens. 27 (15), 3153–3181.

Rashid, S.F., 2000. The urban poor in Dhaka city: their struggles and

coping strategies during the floods of 1998. Disasters 24 (3), 240–

253.

Rasid, H., Haider, W., 2003. Floodplain residents’ preferences for

water level management options in flood control projects in

Bangladesh. Nat. Hazards 28 (1), 101–129.

Rogers, A.S., Kearney, M.S., 2004. Reducing signature variability in

unmixing coastal marsh thematic mapper scenes using spectral

indices. Int. J. Remote Sens. 25 (12), 2317–2335.

Sardar, M.S., Tahir, M.A., Zafar, M.I., 2008. Poverty in riverine areas:

vulnerabilities, social gaps and flood damages. Pak. J. Life Soc. Sci

6 (1), 25–31.

Schmuck-Widmann, H., 1996. Living with the Floods: Survival

Strategies of Char-Dwellers in Bangladesh. LN-Vertrieb.

Wu, W.Y., Shen, X.H., Zou, L.J., Lu, S.L., Zhang, G.F., 2008. An

integrated method for water body detection and delineation using

Landsat ETM+ data. Bull. Sci. Technol. 24, 252–259.

106 A.A. Memon et al.

Xia, Jun, Zhang, Yongyong, 2008. Water security in North China and

countermeasure to climate change and human activity. Phys.

Chem. Earth Parts A/B/C 33 (5), 359–363.

Xiao, X., Boles, S., Liu, J., Zhuang, D., Frolking, S., Li, C., Moore III,

B., 2005. Mapping paddy rice agriculture in Southern China using

multi-temporal MODIS images. Remote Sens. Environ. 95 (4),

480–492.

Xu, J., 2004. A study of anthropogenic seasonal rivers in China.

Catena 55 (1), 17–32.

Xu, H., 2006. Modification of normalised difference water index

(NDWI) to enhance open water features in remotely sensed

imagery. Int. J. Remote Sens. 27 (14), 3025–3033.