112 CHAPTER- V SPATIAL AND TEMPORAL ... - Shodhganga [PDF]

CHAPTER- V

SPATIAL AND TEMPORAL SIGNATURES OF THAR DESERT IN INDIA

The broad level characterization of land cover classes have been discussed in previous chapter. The detailed investigations of variation exists within one of the land cover class, namely arid/semi-arid („Thar‟ desert), carried out using C-band scatterometer data is discussed in this chapter. A portion of hot (“Thar”) desert in India has been assessed using time-series of backscatter response. Major dune types were identified and their temporal trends were investigated. The annual variation of ERS-2 C-band 2000 scatterometer data over this region showed the variability of backscattering coefficient of the order of 11.27 dB (20.08 to -8.71 dB) during one year. These variations were also analysed in terms of scattering due to various factors like soil moisture, vegetation, rainfall and surface roughness. In addition, nine dune types were identified for detailed signature studied.

112

5.1 INTRODUCTION Optical as well as microwave remote sensing data have been successfully used for monitoring and mapping of deserts all over the world. In optical bands, the desert shows higher reflectance whereas, in microwave frequencies, active sensor shows low backscattering coefficient values and passive sensors show high brightness temperature values (Mishra et al, 2002). Among the space-borne SAR systems, C-band data is widely used for various land applications due to availability of ERS-1/2, Radarsat, and Envisat SAR. In addition to this, ERS-1/2 (C-band) and QuikSCAT (Ku-band) scatterometers have also been investigated for land applications (Birrer et al. 1982; Drinkwater et al. 2000 & 2001; Kennett et al. 1989; Kerr and Megagi 1993; Long et al 1994, 1999; Mougin et al. 1995; Prigent et al. 2005; Wagner et al, 1999a, 1999b; Wismann, 1998). Spaceborne scatterometers have provided continuous synoptic microwave coverage of earth for nearly two decades. Stephen et al. (1997) analysed the normalized radar cross section (NRCS) measurement, obtained over Thar Desert in Pakistan between May 1994 and May 1996. Spatial variations in the radar cross section are compared with vegetational and meteorological parameters. Seasonal as well as inter-annual variations are investigated by correlating the radar backscatter with Phenological, meteorological and AVHRR NDVI data. It is demonstrated that valuable information can be delineated from the ERS scatterometer data over arid regions in order to provide data for environmental and especially climatic change studies.

5.2. STUDY AREA The study area is Indian part of the Thar Desert, which is located between 24o 36‘ – 29o 21‘ N Latitude and 69o 32‘ - 75o 26‘ E Longitude (Figure 5.1). The Thar Desert (also known as the ―Great Indian Desert‖) in India is geographically, located in the state of Rajasthan, between the foothills of the Aravalli – ranges in the east and the international border with Pakistan in the west. 113

Dune types

Figure 5.1: Study Area and Dune type (the Thar Desert) It lies mostly in the Indian state of Rajasthan, and extends into the southern portion of Haryana and Punjab and further into the northern region of Gujarat state. The Thar desert is bounded on the northwest by the Sutlej River, on the east by the Aravalli Range, on the south by the salt marsh known as the Rann of Kutch (parts of which are sometimes included in the Thar), and on the west by the Indus River. Depending on the areas included or excluded, the nominal size of the Thar can vary significantly. According to the WWF definition, the Thar has area of 238,700 km². Another source gives the area of 114

446,000 km² that has 805 km length and about 485 km width, out of this 208,110 km² is in India. Of the Indian portion, 61% falls in Rajasthan, 20% in Gujarat and 9% in Punjab and Haryana (combined). The Thar Desert is dominated by the south-west monsoon, which controls both the wind vector and the vegetation cover. The configuration of atmospheric dynamics and sinking air masses in the region inhibit rain in this region despite the fact that considerable precipitable moisture exists in the atmosphere. Minor changes in the atmospheric circulation patterns result in amplified changes in the rainfall, the winds and the Aeolian dynamism. It is an austere area where water is scarce and occurs at great depths, from 30 to 120 m below the ground level. The region is dominated by Aeolian bedforms of different dimensions, including the sand dunes. The thickness of Aeolian cover can range from 1m to 100 m. West-ward, the natural vegetation becomes gradually sparse, cultivation on dune slopes becomes less frequent, and reactivation of the high dunes are more recurrent. Aeolian activity in the Thar Desert is mainly restricted to the period of summer winds associated with the south west monsoon. The north eastern wind of winter months plays only a minor role in Aeolian activity and is largely limited to the northern fringe of the desert. Strong sand and dust shifting winds begin from March onwards when the surface is dry and maximum wind speed (20 km/h or more) is reached at all the meteorological stations during the month of June. May and July are also very windy. Since this is also the period when much of the ground flora is dry, the environment is suitable for Aeolian activities. The wind and the sand dynamics cease with the arrival of monsoon rains (end of June along the eastern margin of the desert, and mid-July in the western part). Higher wind strength and lower rainfall favours erosivity of the wind (Singhvi and Kar 2004).

5.3 MATERIALS AND METHODS The data available from ERS-2 Scatterometer of South Asian region, for year 2000 was used to generate the backscattering images of India. The 115

8.9 km enhanced resolution for ERS-2 C-band data was used. In addition to this, SPOT VEGETATION - NDVI product of same year was used in this study. All the scatterometer and NDVI data were co-registered with the r.m.s value of 0.01 using Arc info software. A portion of the Thar Desert in India was extracted and the backscatter time-series was analysed along with the meteorological and biophysical parameters. Dune map (Singhvi and Kar, 2004) was used to identify the major dune types in this desert region. For identifying the presence of vegetation, the Landsat data (ETM+) was visually interpreted. The NASA Shuttle Radar Topographic Mission (SRTM) DEM (90 m resolution) was also used to measure the height variation in the area. This data (5 deg x 5 deg tiles) was obtained from the USGS ftp site. The SRTM 90m DEM's has a resolution of 90m at the equator. Subsequently, nine locations over six sand dunes type with or without vegetation (Figure 5.1) were identified and signatures statistics were extracted and analysed. A transact profile was also generated from Shahgarh in Western portion to Aravalli (Near Jaipur) in the Eastern portion. Attempts were made to assess the aerodynamic roughness length Z 0, which can be correlated with o. The height above the displacement plane at which the mean wind becomes zero (when extrapolating the logarithmic windspeed profile downward through the surface layer) is the theoretical height that must be determined from the wind-speed profile, although there has been some success at relating this height to the arrangement, spacing, and physical height of individual roughness elements such as trees or houses. Attempt was made to calculate the Z0 using a log-linear model (Prigent et al., 2005). 5.4 RESULTS AND DISCUSSION 5.4.1 Spatio-Temporal Backscatter Pattern The Thar Desert presents picture of contrast in backscatter image in comparison to other land covers in India (Figure 5.2). The monthly averaged o images over Thar Desert and surrounding arid region are shown in Figure 5.2. The Thar Desert is characterized by low o value (-25 to -13 dB), 116

surrounded by (arid region) comparatively high o value (-13 to -9 dB). As discussed by Stephen at al. (1997) the geophysical parameters of the land could also be reflected in the incidence angle diversity of the ERS scatterometer.

JAN

FEB

MAR

APR

MAY

JUN

THAR DESERT 30O44'N 67O48'E

19O57'N 76O16'E

JUL

AUG

SEP

YEAR - 2000 Sigma-0 (dB) -25 - -22 -22 - -18 -18 - -16

OCT

-16 - -15

NOV

DEC

-15 - -14 -14 - -13 -13 - -11 -10 - -9

Figure 5.2: Monthly averaged o over arid area including the Thar Desert in the year 2000.

117

The data in year 2000 over this region showed the variability of o of the order of 11.27 dB (-20.08 to -8.71 dB). This large range of o values observed in desertic terrain is attributed to the changes in dunes roughness (Singh et al. 2006). These variations can be explained in terms of scattering due to various factors like soil moisture, vegetation, rainfall, and surface roughness due to wind. In general, high backscattering was observed in the months of July and August and low backscattering was observed in the months of May and June (Table.5.1 and Figure 5.3). It is observed in (Figure 5.3) that the Min-Max range of observed o over arid area is lowest during the period of December and January and highest in the month of May and June. As given in the table 5.1, observed maximum o value is around

(-9±1) dB during the entire year. In contrast to

monthly maximum o values, the large variation is observed in the minimum values of o from about -17 dB during winter (December and January) period to about -20 dB in the months of May and June i.e. summer. Lower o values observed during the summer period could be the result of the reduced fractional vegetation cover that reduces the effective surface roughness. The high o value (-10 to -9 dB) has also been found over mountainous region, which is due to high topography (Figure 5.4).

118

Table 5.1: o Variability in the Thar Desert (Year 2000) MIN o MAX o RANGE S.N. MONTHS

(dB)

(dB)

(dB)

1

January

-16.72

-9.97

6.745

2

February

-17.61

-10.07

7.531

3

March

-18.98

-9.92

9.06

4

April

-19.65

-9.98

9.67

5

May

-20.08

-9.98

10.1

6

June

-19.94

-9.86

10.08

7

July

-18.17

-8.76

9.406

8

August

-16.94

-8.71

8.233

9

September -17.76

-9.70

8.059

10

October

-17.64

-10.35

7.295

11

November

-18.03

-9.94

8.095

12

December

-17.53

-9.91

7.611

0 MIN MAX

Sigma-0 (dB)

-5 -10 -15 -20 -25

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC

Figure 5.3: Variability of o over the year 2000.

119

0

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

-2 -4

Sigma-0 (dB)

-6 -8 -10

ELEVATION (M) 500 400

-12 -14

300

-16

200

-18 SHAHGARH JAISALMER

JODHPUR

PHALODI

-20 0

71

142

213

285

356

427

498

MERTA DEGANA

570

640

ARAVALLIS

712

100

783

Distance (Km)

Figure 5.4: Transact Profile of o values

5.4.2 Effect of vegetation cover The minimum o observed for sand-dune areas without vegetation cover, in the last week of June, are attributed to dry soil conditions. During this period, lower NDVI values were observed. The increase in o was attributed mostly due to presence of vegetation in the second week of September and roughness arises due to dune size and shape. As mentioned by Stephen et al. (2007), the values of o could be modulated with view direction [incidence angles θ (thetas) and azimuth Φ (phi angles)], where the modulation characteristics reflect the surface geometry. It also varies spatially and reflects the spatial inhomogeneity of the sand surface. High correlation has been observed in o and NDVI (Figure 5.5). As seen in Figure 5.5, in areas of sand dunes with-vegetation cover the minimum o observed in the last week of May, that is -19.5 dB, which is due to lack of soil moisture during summer period. The increase in value of o continues subsequently with the growth of vegetation and corresponding increase in surface roughness.

120

0.3

NDVI

0.35

0.25 0.2

Parabolic dunes 0.15

-12

Sigma-0 (dB)

Linear dunes -14 -16

Transverse dunes

-18

Netwok transitional parabolic dunes Ju l Au g Se p O ct N ov D ec

Ja n Fe b M ar Ap r M ay Ju n

-20

Figure 5.5: o and NDVI response for dunes (with vegetation cover). The low value of o is attributed to dry smooth Soil and sparse vegetation cover. During this period, the NDVI value is also found to be of the lowest order (0.16). The maximum o observed in this area is -12.7 dB mostly due to presence of vegetation cover (NDVI=0.27). A gradual decrease is observed in o values until last week of June. Afterwards, a sudden increase in o is observed for a period of July and August, as the moisture level increases due to arrival of monsoon in this region (Figure 5.6). Onset of monsoon suddenly increases the soil moisture content, which leads to less transmission of energy through the medium and increase in surface scattering. From September onward o decreases due to decrease in soil moisture attributed to percolation of water and high evaporation rate. In the summer season i.e. from the beginning of March, moisture content in the desert region decreases significantly, due to increase in surface temperature resulting in lower o (Figure 5.7).

121

Parabolic-With and Withoutout Veg Sigma-0 Vs Rainfall 20

0

RF - 24 Parabolic - Without Veg Parabolic- with Veg

16

-6

14

-8

12

-10

10

-12

8

-14

6

-16

4

-18

2

-20

0

01 -J 13 an -J 25 an -J 06 an -F 18 eb -F 01 eb -M 13 ar -M 25 ar -M 06 ar -A 18 pr -A 30 pr -A 12 pr -M 24 ay -M 05 ay -J 17 un -J 29 un -J u 11 n -J 23 ul 04 Jul -A 16 ug -A 28 ug -A 09 ug -S 21 ep -S 03 ep -O 15 ct -O 27 ct -O 09 ct -N 21 ov -N 03 ov -D 15 ec -D 27 ec -D ec

Sigma-0 (dB)

-4

18

Rainfall (inch)

-2

Figure 5.6: o v/s Mean-24 hrs Rainfall (Parabolic Dunes) Parabolic-With and Withoutout Veg Sigma-0 Vs Temp. 0

40

-2

35

-4 30 25

-8 -10

20

-12

15

Temp. (deg C)

Sigma-0 (dB)

-6

-14 10 -16

Parabolic - Without Veg Parabolic- with Veg T Mean

-18

0

01 -J a 19 n -J a 06 n -F e 24 b -F e 13 b -M a 31 r -M a 18 r -A p 06 r -M ay 24 -M a 11 y -J u 29 n -J un 17 -J 04 ul -A u 22 g -A u 09 g -S e 27 p -S e 15 p -O c 03 t -N o 21 v -N o 09 v -D e 27 c -D ec

-20

5

Figure 5.7: o v/s Mean Temperature (Parabolic Dunes) 5.4.3 Backscatter Response to Dune types From Figure 5.8 it is evident that the overall range of o follows a cyclic pattern for all the features. Due to physical scattering mechanism (surface 122

roughness and dielectric properties as well as volume scattering from vegetation) attributing to variability in o, it is possible to discriminate dune types. -9

Parabolic - Without Veg Sand streaks

-11 Linear- Without Veg

Sigma-0 (dB)

-13

Transverse-Without Veg Star Dune type-1

-15 Netwok transitional parabolic-without Veg-2

-17

Network Sinuous Dunes Barchans and Barchanoids

-19

Major obstacle dunes

01 -J 19 an -J 06 an -F 24 eb -F 13 eb -M 31 ar -M 18 ar -A 06 pr -M 24 ay -M 11 ay -J 29 un -J u 17 n -J 04 ul -A 22 ug -A 09 ug -S 27 ep -S e 15 p -O 03 ct -N 21 ov -N 09 ov -D 27 ec -D ec

-21

Figure 5.8: Temporal trend of radar Backscatter for major Dune Types The discrimination of different sand dunes is best possible in the month of June (Figure 5.9). Various types of Dunes can be classified in the month of June and July comparing their backscatter values (Figure 5.9). Variability of backscatter coefficient, corresponding to different Dune types in the month of June, is 5.75 dB (-11.75 to -17.5 dB) due to dry Soil and no rainfall condition, whereas it is 2.75 dB (-10.5 to -13.25 dB) in August due to rain and sparse vegetation. Higher o (-14 to -13 dB) range (blue colour in Figure 5.2) may also be due to presence of some paleochannel loaded with moisture which needs to be validated. The order of o of different dune types studied (in increasing order) is as under: Network transitional parabolic > Major obstacle > Network Sinuous > Parabolic > Barchans and Barchanoids > Sand streaks > Linear > Transverse > Star > Megabarchanoids

123

-9

Parabolic Dune - Without Veg

-10

Parabolic Dune - With Veg

Sigma-0 (dB)

-11

Sand streaks

-12

-15

Netwok transitional parabolic Dune - With Veg Netwok transitional parabolic Dune - Without Veg Netwok transitional parabolic Dune - bare soil (sand) Network Sinuous Dunes

-16

Barchans and Barchanoids

-17

Major obstacle Dune

-13 -14

-18 0

June 1

July 2

Aug 3

Months→ 4

Figure 5.9: Separability of different dune types in the months of June, July and August.

5.5 HIGHLIGHTS OF THIS CHAPTER C-band radar time-series backscatter response for the year 2000 covering the Thar Desert region was studied. The data showed the variability of o of the order of 11.27 dB (-20.08 to -8.71 dB). Major dune types were identified and their temporal trends were also studied. In general, high backscattering was observed in the months of July and August and low backscattering was observed in the months of May and June. Over Thar Desert region lower values of o were observed and found to vary gradually from the summer to the monsoon, and winter seasons. The sharp change in moisture level in desert region is also reflected clearly from o. It can also be concluded that the dune type separation using o is best possible in the month of June. Time series analysis using scatterometer data over many years (decades) may give better insight into the formation of the dunes, ―dune shift‖ and spread of desertification. The dune spacing using SRTM-DEM and its relation with o in the Thar can also be attempted in future. The study indicated the potential of C-band 124

scatterometer data for monitoring temporal variability for modelling and monitoring desert ecosystem.

125