Remote Sensing based indices for agricultural drought assessment

Dr. C.S. Murthy Head, Crop Monitoring & Assessment Division

Remote Sensing Applications Area National Remote Sensing Centre, Hyderabad

Understanding Drought and its management

• Complex nonlinear interactions • Slow process with multiple impact • No single index • Different states adopt different norms

Weather

Soil

Crop

Agricultural drought

Management

Information needs for Drought Management

Short term Management

Long term Management

Monitoring & Assessment

Prediction & Early warning

Agro-advisories Crop damage

assessment

Vulnerability maps

Risk maps Prioritization Impact

monitoring

Different indices • Meteorological • Soil moisture • Surface water • Ground water • Crop related • Socio-economic

Meteorological drought: reduced rainfall

Hydrological drought: reduced surface water

Agricultural drought: reduced soil moisture

Conceptualisation, development, operational services and institutionalisation of a remote sensing application project

2012+

• Institutional Arrangement at Ministry of Agriculture, GOI

• Creation of MNCFC

• Transfer of NADAMS project to MNCFC April 2012 • Enhanced end-use of NADAMS project

NADAMS Project A success story of

Institutionalization

• IRS WiFS based district / subdistrict level assessment • Participation of user departments 1998+

• Use of multiple indices • IRS AWiFS based sub-district level assessment • Decision rules for drought warning & declaration • Enhanced content & frequency of reporting • Institutional participation & Capacity building

2004+

• Conceptualization of NADAMS project at NRSC • Development of Methodology

1988

• Supplementation of WiFS with MODIS • Agricultural area monitoring • Increased number of indices

2002+

2009+

• Operational Services from NRSC • NOAA AVHRR • Regional/district level assessment • Prevalence, intensity and persistence of agricultural

1989 +

• New approach for sowing-period drought assessment • Geospatial products on soil mositure and rainfall • New Indices – Shortwave Angle Slope Index

Institutionalisation phase

Traj

ecto

ry o

f N

AD

AM

S p

roje

ct-

Dev

elo

pm

en

t an

d o

per

atio

nal

ser

vice

s

National Agricultural Drought Assessment and Monitoring Systems (NADAMS)

For NADAMS Drought reports kharif 2012 onwards visit: www.ncfc.gov.in

Wavelength range (µm)

0.62 – 0.67 (red) 0.841 – 0.876 (NIR) Spatial Resolution 250m Swath 2330 kms

IRS-WiFS IRS P6 AWiFS

Spatial resolution

56 metres

Wave lengths 4 bands

(green, red, NIR and MIR)

Swath : 700 kms.

Terra Modis NOAA AVHRR

Spatial resolution

188 metres

Wave lengths 3 bands

(green, red and NIR )

Swath : 700 kms.

Wavelength range (µm) 0.58 – 0.68 (red) 0.725 – 1.1 (NIR) 3.55 – 3.93 (MIR) 10.3 – 11.3 (TIR) 11.5 - 12.5 (TIR) Spatial Resolution 1.1 Km

Data from multiple satellites/sensors NATIONAL STATE DISTRICT

Cell

structure

Leaf

pigments

Chlorophyll

absorption

Water content

Water absorption R

E

F

L

E

C

T

A

N

C

E

(%)

WAVELENGTH (um)

0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6

80

70

60

50

40

30

20

10

0

Spectral response of vegetation

Dominant factor controlling leaf reflectance

NDVI

Ratio of difference and sum of surface reflectance in NIR and red spectral bands NIR can see roughly 8 layers Red – one layer Most successful indicator for describing vegetation Normalisation - reduces the effect of sensor degradation sensitive to changes in vegetation easy to compute and interpret available from most of the sensor systems

Spectral response of vegetation Red – more absorption due to chlorophyll Near Infra red – more reflection due to leaf structure

Normalized Difference Vegetation Index (NDVI)

NIR – Red / NIR+Red Reflected radiation in Near infrared and red bands. NDVI ranges from -1 to +1 Water – negative NDVI Clouds – zero NDVI Vegetation – positive NDVI represents density, vigor

Normalized Difference Vegetation Index

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

10

1-M

ay-1

10

9-M

ay-1

11

7-M

ay-1

12

5-M

ay-1

10

2-J

un

-11

10

-Ju

n-1

11

8-J

un

-11

26

-Ju

n-1

10

4-J

ul-

11

12

-Ju

l-1

12

0-J

ul-

11

28

-Ju

l-1

10

5-A

ug-

11

13

-Au

g-1

12

1-A

ug-

11

29

-Au

g-1

10

6-S

ep

-11

14

-Se

p-1

12

2-S

ep

-11

30

-Se

p-1

10

8-O

ct-1

11

6-O

ct-1

12

4-O

ct-1

10

1-N

ov-

11

09

-No

v-1

11

7-N

ov-

11

25

-No

v-1

10

3-D

ec-

11

11

-De

c-1

11

9-D

ec-

11

27

-De

c-1

10

4-J

an-1

21

2-J

an-1

22

0-J

an-1

22

8-J

an-1

20

5-F

eb

-12

13

-Fe

b-1

22

1-F

eb

-12

29

-Fe

b-1

20

8-M

ar-1

21

6-M

ar-1

22

4-M

ar-1

20

1-A

pr-

12

09

-Ap

r-1

21

7-A

pr-

12

25

-Ap

r-1

20

3-M

ay-1

21

1-M

ay-1

21

9-M

ay-1

2

ND

VI

NDVI profile over agricultural area

Satellite/ Sensor Indices Relevant Parameter

NOAA AVHRR (1km) NDVI Crop condition

Oceansat 2- OCM (360m) NDVI, ARVI Crop condition

Terra MODIS (500 m) SASI, NDWI Surface wetness/ sown area discrimination

Terra AMSRE (25 km) Soil moisture Surface wetness/ sown area discrimination

INSAT 3A CCD (1 km) NDVI Crop condition

Satellites Sensor Spatial resolution

Temporal resolution Swath

Resourcesat 1 & 2 AWiFS 56 m 5 days 750 km LISS III 23 m 26 days 140 km LISS IV 6 m 48 days 70 km LANDSAT 8 OLI 30 m 16 days 185 km

Some important VI data sets

Coarse resolution data

Moderate resolution data

.

-100

-50

0

50

100

150

200

250

300

.

12/6 19/6 6/6 3/710/7 17/7 24/7 31/7 7/8 14/8 21/8 28/8 4/9 11/9 18/9 25/9

% d

evi

atio

n

NDVI based agricultural drought assessment

0

0.1

0.2

0.3

0.4

0.5

0.6

June July Aug. Sept. Oct. Nov.

Month

ND

VI

Normal delayed season Drought

(1) relative deviation from normal, (2) vegetation Condition Index, (3) in season rate of transformation

Integration with ground data

.

0

10

20

30

40

50

60

70

80

90

100

5 J

un

12

Ju

n

19

Ju

n

26

Ju

n

3 J

ul

10

Ju

l

17

Ju

l

24

Ju

l

31

Ju

l

7 A

ug

14

Au

g

21

Au

g

31

Au

g

11

Se

p

18

Se

p

25

Se

p

30

Se

p

% o

f n

orm

al

Seasonal NDVI profiles for drought assessment

Weekly deviations of rainfall

Weekly progression of sown areas

NDWI/LSWI

Reflectance in 0.9 – 2.5 microns dominated by liquid water absorption

Sensitive to surface wetness/ vegetation moisture

Less affected by atmosphere

Reflectance 1.5-2.5 microns does not saturate till LAI reaches 8

SWIR availability – AWiFS, LISS-III, MODIS, INSAT 3A

MODIS – 3 SWIR bands – 1240 nm, 1640 nm, 2100 nm

LSWI – Land Surface Wetness Index – uses SWIR at 2100 nm – very sensitive to wetness NDWI/LSWI uses Agriculture – crop stress detection, crop yield, classification of succulent crops, surface moisture Forest – Monitoring, burnt area detection

Normalized Difference Wetness Index/ Land Surface Wetness Index

Most commonly adopted index – NDVI a) chlorophyll based index b) plant vigour and density c) easy to compute and interpret d) robust index e) Limitations – soil back ground, saturation, time lag etc.

Commonly used indices for drought assessment

Recently popularized index – NDWI/LSWI a) Plant moisture based index b) NIR and SWIR based c) No saturation issues d) Immediate response e) Sensitive to surface wetness during sowing period

Combination of NDVI and NDWI a) Overcomes limitations of either one b) amplifies anomalies and c) more responsive to ground situation

AWiFS NDVI - Odisha – June to November 2016

Jun 01-15, 2016 Jun 16-30, 2016 Jul 01-15, 2016 Jul 16-31, 2016

Aug 01-15, 2016 Aug 16-31, 2016 Sep 01-15, 2016 Sep 16-30, 2016

Oct 01-15, 2016 Oct 16-30, 2016 Nov 01-15, 2016 Nov 16-30, 2016

Increasing NDVI (greenness) Non-crop area

NDVI anomaly

% dev. from normal (actual NDVI-normal NDVI)/normal NDVI*100

Selection of normal year – average of recent past normal years NDVI is a conservative indicator and hence anomalies are not very high

Thumb rule: > 20% reduction in NDVI – drought conditions >30% reduction indicate moderate to severe drought conditions

Interpretation of NDVI changes to assess Agricultural drought

• Derivative of NDVI • Substitute to NDVI deviation

VCI = NDVI- NDVImin /(NDVImax –

NDVImin) *100 Range 0-100% 0-40 % drought 40-60 mild to moderate drought 60-100 good

Vegetation Condition Index (VCI)

Interpretation of NDVI changes to assess Agricultural drought

Vegetation Condition Index (VCI)

Computation

September

year NDVI VCI 2000 0.5 8

2001 0.55 50

2002 0.6 92

2004 0.52 25

2005 0.55 50 2006 0.51 17

2007 0.49 0 2008 0.59 83

2009 0.61 100 2010 0.54 42

2011 0.51 17

Mean 0.54

Min 0.49 Max 0.61

diff 0.12

Some critical issues Time series data base (at least 10-12 years) Differences due to cropping pattern, crop calendar to be normalised

VCI = NDVI- NDVImin /(NDVImax – NDVImin)

*100

Interpretation of NDVI changes for drought assesment

Reduction in NDVI is caused by; • Crop moisture stress • Flooding/excess rainfall • Crop and crop stage differences

between the two years under comparison

Despite these limitations, NDVI is still a successful indicator for agricultural drought assessment

Limitations of NDVI Chlorophyll based index – saturates with LAI (=3) Limited capability to detect vegetation water content Over-estimation when the veg. density is less Saturation at peak vegetative phases Conservative index Time lag

Block level crop condition – Anantpur district Comparison between normal year (2005) and drought year (2002)

Tadipatri mandal - Anantpur (Groundnut)

-0.100

-0.050

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

June July August Sep October

ND

VI

2005 2002

Pamidi mandal - Anantpur (Paddy)

-0.050

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

0.400

0.450

June July August Sep October

ND

VI

2005 2002

2002 drought conditions at mandal level in Anantpur district

Mudigubba mandal - Anantpur district

groundnut

-0.100

-0.050

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

0.400

0.450

June July August Sep October

ND

VI

2005 2002

Kanekal mandal - Anantpur district

groundnut + paddy

-0.100

-0.050

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

0.400

June July August Sep October

ND

VI

2005 2002

Crop area affected by drought in kharif 2015, West Bengal

5

17

40

29

8 1

0

10

20

30

40

50

< 10 % 10-20%

20-30%

30-40%

40-50%

< 50 %

Agr

iltu

ral a

rea

Aff

ecte

d

(%)

% of reduction in crop condition

Normal

Mild Moderate

Severe

Normal

Mild Moderate

Severe

10

26

47

15

2 0 0

10

20

30

40

50

< 10 % < 50 %

Agr

iltu

ral a

rea

Aff

ecte

d

(%)

% of reduction in crop condition

Normal

Mild Moderate

Severe

23

39 31

6 1 0

0

10

20

30

40

50

Agr

iltu

ral a

rea

Aff

ecte

d (

%)

% of reduction in crop condition

Purulia district Bankura district West Midnapur district

AWiFS derived crop condition anomalies showing agricultural drought situation in Andhra Pradesh, kharif 2011

Sep 2011 Sep 2010 NDVI anomaly Sep 2011

NDWI anomaly Oct 2011

0.00

0.10

0.20

0.30

0.40

0.50

FN1-SEP FN2-SEP FN1-OCT FN2-OCT Nov

ND

VI

2010 2011

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

F1SEP F2SEP F1OCT F2OCT F1NOV

ND

WI

2011 2010

NDWI - Roddam mandal, Anantpur NDVI - Roddam mandal, Anantpur

164 mandals 394 mandals

562 mandals

Assessment at sub-district level

Kharif 2010 – Agricultural drought assessment, Bihar

AWiFS NDVI Oct. 2010 (Drought year)

AWiFS NDVI Oct. 2008 (Normal year)

1. Pas. Champaran 2. Pur. Champaran 3. Sheohar 4. Sitamarhi 5. Madhubani 6. Supal 7. Ararai 8. Kishanganj 9. Purnia 10. Madhepura 11. Saharsa 12. Darbhanga 13. Muzaffarpur 14. Gopalganj 15. Siwan 16. Saran 17.Vaishali 18. Samastipur 19. Begusarai

20. Khagaria 21. Katihar 22. Bhagalpur 23. Banka 24. Munger 25. Lackeesarai 26. Sheikhpura 27. Nalanda 28. Patna 29. Bhojpur 30. Buxar 31. Bhabua 32. Rhotas 33. Aurangabad 34. Jahanabad 35. Gaya 36. Nawada 37. Jamui

-60

-40

-20

0

Jam

ui

Gay

a

Jahan

a…

Bu

xar

Aurang…

Samasti…

Bh

ojp

ur

Nal

and

a

Pat

na

Bhagalp…

% d

evia

tio

n f

rom

n

orm

al

Rainfall deficiency

0

50

100

Jam

ui

Gay

a

Jahan

a…

Bu

xar

Aurang…

Samasti…

Bh

ojp

ur

Nal

and

a

Pat

na

Bhagal…%

of

no

rmal

Crop sown area status

Crop areas affected by agricultural drought situation are showing lower NDVI compared to normal, in kharif 2010 in Bihar state.

Agricultural drought assessment

(based on multiple indices; NDVI, NDWI, SASI)

Moderate (103) Severe (55)

Moderate (103) Severe (55)

Satellite derived Area Favourable for Crop Sowing/Crop Sown Area (AFCS) , Lakh ha.

Kharif Nromal

Area

AFCS AFCS Unfavourable

Jul-10 Aug-10 area

37 19 23 14

Satellite based agricultural drought assessment

kharif 2010, West Bengal

Crop condition assessment Rajkot district, Gujarat Kharif 2016

+ 10% 10-20% 20-30% 30-40% 40-50% < 50%

Deviation

September October November

NDVI Anomaly – September 2016

Normal Normal Mild drought Mild – moderate Moderate to severe

Kerala 2016 Drought

PERCENTAGE OF AREA in different drought classes - SEPTEMBER

S.No Districts C1 C2 C3 C4 C5

1 Alappuzha 19.57 0.04 0.10 0.05 80.23

2 Ernakulam 36.42 0.16 0.10 0.01 63.31

3 Idukki 17.23 0.34 0.17 0.00 82.25

4 Kannur 86.78 0.07 0.05 0.01 13.09

5 Kasaragod 41.25 0.22 0.14 0.00 58.38

6 Kollam 55.96 0.05 0.10 0.01 43.87

7 Kottayam 16.33 0.04 0.06 0.03 83.54

8 Kozhikode 72.66 0.16 0.07 0.00 27.11

9 Malappuram 87.52 0.29 0.08 0.00 12.11

10 Palakkad 63.08 0.19 0.12 0.02 36.59

11 Pattanamthitta 47.75 0.12 0.07 0.01 52.06

12 Thiruvananthapuram 37.84 0.20 0.04 0.01 61.91

13 Thrissur 59.60 0.17 0.15 0.01 40.07

14 Wayanad 76.12 0.07 0.05 0.00 23.75

0

100

200

300

400

500

-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

09

Jun

25

Jun

11

Jul

27

Jul

12

Au

g

28

Au

g

13

Se

p

29

Se

p

So

wn

are

a 0

00’

ha

SA

SI

Weeks

SASI Sown area

Response of SASI to crop sown area

Features SASI value

Dry vegetation low negative

Moist veg. high negative

Shortwave Angle Slope Index (SASI)

βSWIR1 = cos-1 [ (a2 + b2 - c2) /

(2*a*b) ]

Slope = (SWIR2 − NIR)

SASI = βSWIR1 * Slope (radians)

where a, b and c are Euclidian

distances between vertices NIR

and SWIR1, SWIR1 and SWIR2,

and NIR and SWIR2, respectively

Chronological synchronization between (a) Decrease in SASI (b) Increase in rainfall (c) Increase in sown area

NADAMS project Conceptually and computationally simple procedures to discriminate the crop sowing favorable areas at state level

Features SASI value

Dry soil highly positive

Wet soil low positive

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

09

Ju

ne

17

Ju

ne

25

Ju

ne

3 J

uly

11

Ju

ly

19

Ju

ly

27

Ju

ly

04

Au

g

12

Au

g

20

Au

g

28

Au

g

05

Se

p

13

Se

p

21

Se

p

29

Se

p

SASI

val

ue

s

2002 2006 2009

Seasonal dynamics of SASI

Before crop sowing

Crop growing and

maturity

Commencement of crop sowing

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.40

1-M

ay-1

10

9-M

ay-1

11

7-M

ay-1

12

5-M

ay-1

10

2-J

un

-11

10

-Ju

n-1

11

8-J

un

-11

26

-Ju

n-1

10

4-J

ul-

11

12

-Ju

l-1

12

0-J

ul-

11

28

-Ju

l-1

10

5-A

ug-

11

13

-Au

g-1

12

1-A

ug-

11

29

-Au

g-1

10

6-S

ep-1

11

4-S

ep-1

12

2-S

ep-1

13

0-S

ep-1

10

8-O

ct-1

11

6-O

ct-1

12

4-O

ct-1

10

1-N

ov-

11

09

-No

v-1

11

7-N

ov-

11

25

-No

v-1

10

3-D

ec-1

11

1-D

ec-1

11

9-D

ec-1

12

7-D

ec-1

10

4-J

an-1

21

2-J

an-1

22

0-J

an-1

22

8-J

an-1

20

5-F

eb-1

21

3-F

eb-1

22

1-F

eb-1

22

9-F

eb-1

20

8-M

ar-1

21

6-M

ar-1

22

4-M

ar-1

20

1-A

pr-

12

09

-Ap

r-1

21

7-A

pr-

12

25

-Ap

r-1

20

3-M

ay-1

21

1-M

ay-1

21

9-M

ay-1

2

SASI

Seasonal SASI profile

Area Favourable for Crop Sowing/Crop Sowan Area (AFCS)

Geospatial product on Area Favourable for

Crop Sowing (AFCS) using multi-criteria approach

Soil Texture

Rice area mask

Kharif area mask

SASI Modelled soil moisture

Input datasets

September August

July June

State Kharif normal

AFCS M ha. Unfavorable area

state June July Aug Sep

Andhra Pradesh 7.8 2.0 6.8 6.9 7.3 0.4

Bihar 3.7 0.7 3.6 3.7 3.7 0

Chhattisgarh 4.8 3.2 4.8 4.8 4.8 0

Gujarat 8.7 1.3 5.0 5.8 8.1 0.6

Haryana 2.8 0.6 1.6 2.8 2.8 0

Jharkhand 2.5 0.3 2.4 2.5 2.5 0

Karnataka 7.5 3.5 6.0 6.0 7.0 0.5

Madhya Pradesh 10.4 0.7 9.7 10.3 10.4 0

Maharashtra 14.0 5.5 13.2 13.8 13.8 0.2

Odisha 6.3 3.9 6.1 6.2 6.3 0

Rajasthan 14.3 0.2 4.4 11.7 13.6 0.8

Tamil nadu 2.4 1.1 1.8 2.0 2.0 0.4

Uttar Pradesh 9.3 2.8 8.7 9.2 9.3 0

Sub-Total 94.5 25.8 74.2 85.7 91.7 2.9

All India 108.6 34.2 87.0 97.7 105.5 3.1

Area Favourable for Crop Sowing (AFCS) derived from SASI and water balance methodology, Kharif 2012

4_11 JUNE 12_18 JUNE 19_25 JUNE 26_2 JULY 3_9 JULY 10_16 JULY

Tracking the early season drought conditions of 2014 using LPRM Soil Moisture datasets

17_23 JULY 24_30 JULY 31_06 AUGUST 17_23 JULY

Soil moisture deviations from normal in 2014

0.000.050.100.150.200.250.300.350.40

01

jun

05

Ju

n

09

Ju

n

13

Ju

n

17

Ju

n

21

Ju

n

25

Ju

n

29

Ju

n

03

Ju

l

07

Ju

l

11

Ju

l

15

Ju

l

19

Ju

l

23

Ju

l

27

Ju

l

31

Ju

l

04

Au

g

08

Au

g

12

Au

g

Soil

mo

istu

re (

m3

/m3

)

LPRM_SM 2014 LPRM_SM 2013

Drought frequency in the sowing period

Field Data Collection using Geo-ICT

Observation

Transmission

Information

Decision

Action

Mobile app. from NRSC/ISRO

that allows users to share,

access and upload natural

resources information on a near

real time basis, with Bhuvan

serving as the platform

Crowd sourcing approach with

open source tools like Open

layers, PHP, Geoserver and

Mapserver, etc. for

visualization and uploading

Immense use for agricultural

information collection/analysis

Provision to upload the

information through internet or

customized mobile which will

be geo-tagged for

visualization through Bhuvan

Portal

Geo-tagged in-season field

data enables developing a

repository of agriculture/crop

related data

Successful and On-going applications of FDC • Crop mapping • Pest/disease surveillance • Crop Insurance • Crop damage assessment enumerations • Disaster • Drought impact enumeration.

Conclusions

Satellite indices have the potential to capture drought conditions

Satellite data – free access, analysis/computations simplified

Strengthen the drought monitoring and declaration systems

Selection of suitable satellite data - AWiFS is preferable

States – maintain NDVI, LSWI data, effective utilisation

State RS Centres - capability and expertise, support to State Depts.

National Centres – provide expertise, training etc.

Need for institutional mechanism Thank you