Varietal discrimination of Basmati rice in north - west India · IRS-P6 AWiFS: 03-Oct-2010 Covering Haryana and Punjab States. Basmati . Sharbati ... In general, the high yielding

Varietal discrimination of Basmati rice in north-west India

A. N. Singh*, Dharmesh Verma** and M. H. Kalubarme*Global Institute of Land, Water and Environment Management, Lucknow, India

** United Phosphorous Ltd., India

*** BISAG, Gandhinagar, India

What is Basmati Rice?• Basmati is a premium long grained aromatic rice grown in a specific geo-

environment , i.e. NW India and parts of Pakistan for centuries. Documentary evidences show that Basmati has been grown in this area for more than 250 years (Nene, 2003).

• Its high value stems from its unique eating qualities, which includes aroma in both the raw and cooked state, kernel length 7 mm or more, excellent linear elongation on cooking almost double its kernel length, soft and flaky consistency of cooked rice.

• Basmati 370, Taraori Basmati, Type3 and Ranbir Basmati are the Traditional Basmati varieties grown in Punjab, Haryana, western U.P. , Uttarakhand and J&K. These are tall (140-148 cm plant ht.),145-150 d growing period, kernel length varying from 6.9 to 7.3 mm, breadth 1.7 to 1.9 mm, kernel elongation ratio after cooking 1.8 to2.1 .

Evolved Basmati varieties

• Pusa Basmati-1, Pusa-1121, CSR-30 and Pusa-1509 are the varieties evolved (recently bred) Basmati having traditional Basmati varieties as one of the parent. These varieties are being grown in a larger area now due to higher yield.

• Factors favouring aroma in Basmati are cool temp. during flowering and grain development (25 degree C/ 31 degree C night/day temp. during crop maturity), use of organic manures, fertile, light-textured and well-drained soil, direct sowing, etc.

Objectives of study

The study, commissioned by the Agricultural Products Export Development Agency (APEDA), Govt. of India, had the following objectives:• Remote Sensing data based area estimate of traditional and evolved

Basmati varieties in Punjab (21 dist.), Haryana (20), western Uttar Pradesh (26), Uttarakhand (4), Jammu & Kashmir (2), and Himachal Pradesh (2).

• CCE based yield prediction in different districts/regions.• Monthly report on Basmati during its cropping season – growth, crop

condition, biotic and abiotic stresses • Annual change in area under basmati varieties.

Study Area

IRS-P6 AWiFS : 24-SEP-2008 Covering Haryana and Punjab States

IRS-P6 AWiFS : 03-Oct-2010 Covering Haryana and Punjab States

Basmati

Sharbati

An attempt was made to differentiate High yielding riceand basmati varieties based on their phonological stagedifferences which in turn have impact on spectralreflectance on the satellite data.In general, the high yielding rice varieties are sown andtransplanted one month in advance of Basmati varieties inmost of the study areas.The most important reason for late sowing andtransplantation of basmati varieties is that these varietiesshould mature during the cold night periods during lateOctober or first week of November which helps toproduce better aroma in the basmati.

REMOTE SENSING OF CROP PHENOLOGY

Seed to seed duration of commonly grown Basmati varieties

Traditional Basmati• Basmati-386 / Taraori Basmati 150 d• Basmati-370 145 dEvolved Basmati • CSR-30 140 d• Pusa Basmati-1 140 d• Pusa-1121 130-135 d

MethodologyData Used• Multi-temporal IRS AWiFS (56m), LISS-III (23m) digital data for the rice growing period

and Liss-IV (6m) for selected area.• Due to cloud cover, good quality data available were generally after 15 Sept. . AWiFS

with 5 d repeat cycle provided more frequent

Field Data and GPS Measurements• Ground Truth (GT) was collected during fourth week of August to first week of October,

which coincided with flowering to grain formation stage of rice crop. Agronomic data like variety, stage/vigor, and height of the crop canopy, soil exposure were recorded. Minimum size of plot considered was 300 * 300 sq. m. to collect data using GPS.

Crop Cutting Experiments (CCE) for yield estimation in all the states. For example, in Haryana, CCE were conducted in 190 plots covering 10 districts.

Selection of images for varietal study based on crop calender

• In Punjab, Traditional Basmati varieties (Basmati 386) is in flowering stage in last week of October and harvested in 4th week November.

• Evolved Basmati (Pusa Basmati-1 and PB1121) flowers in 2nd week of October and harvested in 3rd week of November.

• HYVs are harvested in 4th week of September.• In Haryana, transplanting of Basmati varieties is done about 15 days earlier

than Punjab, and accordingly all crop stages.• In Uttar Pradesh, transplanting is similar to Punjab.• In Jammu & Kashmir, only Ranbir Basmati is grown, which is of shorter

duration. • Hence, images of last week of September to 1st week of October and onwards

were selected for analysis.

Steps in IRS LISS-III Digital Data Analysis

• Geo-referencing• Administrative boundary superimposition• Generation of spatial information in GIS environment • Superimposing GPS locations of Basmati and high yielding rice varieties on the

registered LISS-III digital data, • Identification of basmati and high yielding rice varieties on LISS-III digital data, • Supervised classification using MXL classifier with boundary mask approach, • Area estimation under different rice varieties • Generation of spectral vegetation indices like NDVI

DN to Radiance Conversion• Calculation of at-sensor spectral radiance is the fundamental step in

converting image data from multiple sensors and platforms into a physically meaningful common radiometric scale.

• In order to obtain radiometrically comparable apparent spectral radiance data suitable for further processing, the integer digital number (DN) of each band of all images was transferred into real numbers using the spectral calibration data. The calibration was done by following expression of satellite spectral radiance Lλ, (Lillesand et.al; 2000) which is,

Lrad = {[DN/MAX GRAY] * [Lmax - Lmin]} + Lmin

Where, DN = Digital numbers of a pixel, Max grey: Maximum DN possible for a given data. Lmax and Lmin are the maximum and minimum radiance values for band (mWcm-2 Sr-1 µm1).

Generation of Training Signatures and SeparabilityAnalysis• LISS-III and LISS-IV images of selected growth phases of major HYVs and Basmati

varieties ( Last week of September onwards)were used. GPS based training sites were collected for different rice varieties and other land-use classes.

• Five-to-six classes with different developmental stages and percent ground cover having different vigour for each rice variety were identified for training signature generation.

• The training signatures contain multi-band statistics such as mean, standard deviation, and variance-covariance matrix for each class, which is used in supervised classification.

• Spectral separability of basmati rice varieties and other HYVs were generated.• Before using these signature statistics in the supervised classification, the crop

separability was studied by computing the Transformed Divergence for different classes.

0

20

40

60

80

100

120

0.55 0.65 0.81 1.62

DN

Val

ues

Spectral Reflecatnce of Rice Varieties

Basmati-1Basmati-2Basmati-3HYV-1HYV-2Sharbati-1Sharbati-2

Central Wavelength (micro meter)

IRS LISS-III Spectral BandsBand 2-Green: 0.52 – 0.59Band 3-Red: 0.62 – 0.68

Band 3-NIR : 0.76 – 1.55 – 1.70

IRS-P6 LISS-III images of two dates showing differentiation of Basmati from HYVs in part of Karnal district, Haryana

Basmati

24-SEP-2008 31-Aug-2008

High Yielding Varieties

24-September 200831-August 2008

South-Western Part of Karnal District

Traditional Basmati

Evolved Basmati

Sharbati

GPS points for Training sites & CCE

Class Code No. Pixels 5 10 15 20 30 40 50 60 70 75 85 95 110 115River sand 5 4787 83.79 13 0 0 0 0 0 0 0 0 0 3.18 0.02 0Waste Land 10 567 4.59 93.83 0 0 0 0 0 0 0 0 0 1.23 0 0Habitation 15 9545 3.41 42.38 0 0 0 0 0 0 0.05 0 0.17 44.83 1.32 0Water 20 856 0 0 0 68.81 2.57 0 0 0 0 0 0 3.1 16.56 0Water-2 30 1292 0 0.08 0.09 4.49 75.54 0 0.15 0 0 0 0 3.15 16.56 0Basmati-1 40 6647 0 0 0 0 0 23.3 69.8 0 0 0.02 6.71 0.02 0 0Basmati-2 50 3120 0 0 0 0 0 1.15 77.98 0.03 0.38 2.08 16.44 0.06 0 0Sharbati-1 60 768 0 0 0 0 0 0 8.98 59.77 17.32 11.07 2.86 0 0 0Sharbati-2 70 577 0 0 0 0 0 0 0.35 14.73 83.36 0 0.17 1.39 0 0HYV-1 75 2991 0 0 0 0 0 0.33 4.55 0.74 0.3 62.29 31.16 0 0 0HYV-2 85 1296 0 0.08 0 0 0 0 7.87 0 0.46 5.25 85.65 0.31 0 0Fallow land 95 4461 3.12 0.81 5.04 0 0 0 0.87 0 3.18 0 0.11 78.7 0.99 5.87waterlogged 110 3358 0.12 1.43 0.03 0.09 4.59 0 8.7 0 0 0 0.12 5.48 79.09 0Other Veg. 115 584 0 0 0 0 0 0 1.2 0.17 0.34 0.15 0.35 5.6 0 92.29

Average accuracy = 82.81 % Overall accuracy = 91.58 % KAPPA COEFFICIENT = 0.9220

Percent Pixels Classified by Code

Confusion Matrix

Separability Measure: Transformed DivergenceAverage Separability: 1994636Minimum Separability: 1509900Maximum Separability: 2.000000Minimum Separability: Sharbati-1 and Sharbati-2

Class River Sand Waste Land Habitation Water-1 Water-2 Basmati-1 Basmati-2 Sharbati-1 Sharbati-2 HYV-1 HYV-2 Fallow Land Waterlogged StreamWaste Land 1.999966Habitation 2.000000 2.000000Water-1 2.000000 2.000000 2.000000Water-2 2.000000 2.000000 2.000000 2.000000

Basmati-1 2.000000 2.000000 2.000000 2.000000 2.000000Basmati-2 2.000000 2.000000 2.000000 2.000000 2.000000 1.981544Sharbati-1 2.000000 2.000000 2.000000 2.000000 2.000000 2.000000 1.999990Sharbati-2 2.000000 2.000000 2.000000 2.000000 2.000000 2.000000 2.000000 1.509900

HYV-1 2.000000 2.000000 2.000000 2.000000 2.000000 1.999995 2.000000 2.000000 2.000000HYV-2 2.000000 2.000000 2.000000 2.000000 2.000000 1.999778 1.929037 2.000000 2.000000 1.950382

Fallow Land 2.000000 2.000000 1.999963 2.000000 2.000000 2.000000 2.000000 2.000000 2.000000 2.000000 2.000000Waterlogged 2.000000 2.000000 2.000000 2.000000 1.999815 2.000000 2.000000 2.000000 2.000000 2.000000 2.000000 2.000000

Stream 2.000000 1.999966 2.000000 2.000000 2.000000 1.999966 2.000000 2.000000 2.000000 2.000000 2.000000 1.992781 2.000000Other Veg 2.000000 2.000000 1.999966 2.000000 2.000000 2.000000 1.998234 1.999993 1.999996 2.000000 1.996616 1.999999 2.000000 1.999986

Signature Separability using Transformed Divergence

KURUKSHETRA

KARNAL

KAITHAL

NDVI Image of IRS LISS-III data of 24-Sep-2008 covering Karnal Kaithal & Kurukshetra

Basmati and HYV variety Classification using NDVI Thresholding

Crop Yield

• Crop yield data collected from CCE and Agriculture department from high yielding and basmati growing states in India, along with agro-meteorological data and Spectral Vegetation Index like Normalized Difference Vegetation Index (NDVI) was analyzed for developing zonal Agromet-Spectral –Yield models using multiple regression analysis.

Pusa Basmati: NDVI Vs. Yield (Panipat, Haryana)

Yield = 25.27*NDVI + 30.57R2 = 0.95

3839404142434445464748

0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65NDVI

Yie

ld (

q/ha

)

YIELD (q/ha) Linear (YIELD (q/ha))

Pusa Basmati: NDVI Vs. Yield (Karnal, Haryana)

Yield= 45.15*NDVI + 25.23R2 = 0.91

404244464850525456

0.35 0.40 0.45 0.50 0.55 0.60 0.65NDVI

Yie

ld (

q/h

a)

Yield (q/ha) Linear (Yield (q/ha))

Relationship between CCE Yield and NDVI values of Basmati Variety in Karnal & Panipat Districts

Sharbati Rice: NDVI Vs. Yield (Karnal, Haryana)

Yield = 45.66*NDVI + 34.61R2 = 0.84

525354555657585960

0.35 0.40 0.45 0.50 0.55 0.60NDVI

Yie

ld (

q/ha

)

Yield (q/ha) Linear (Yield (q/ha))

Pusa-1121: NDVI Vs. Yield (Muzaffarnagar, UP)

Yield = 16.15*NDVI + 41.34R2 = 0.67

44

46

48

50

52

54

56

0.35 0.40 0.45 0.50 0.55 0.60 0.65NDVI

Yie

ld (

q/h

a)

Yield (q/ha) Linear (Yield (q/ha))

Relationship between CCE Yield and NDVI values of Sharbati & Basmati Varieties

Crop Cutting Experiments (CCE) for yield estimation

Crop Cutting Experiments were conducted usingstandard procedures in the study area for assessment ofyield.

The CCE derived yield was averaged for the districtand a conversion factor used for offsetting the moisturecontent of the grain for estimation of district levelproduction.

In Haryana State, Crop Cutting Experiments were conducted in 190 plots covering 10 districts. Based on CCE data, the range of productivity of different Basmati varieties computed is given in Table-1.

Normalized Difference Vegetation Index (NDVI) of Basmati and high yielding rice varieties of aparticular administrative district / tehsil for 10 crop seasons were generated

Meteorological data like rainfall, Tmax, Tmin, Relative Humidity (RH %), sunshine Hours etc. of previous10-years have been collected form IMD for a particular Met Station.

The Basmati yield data at district/tehsil-level was also collected from the Department of Agriculture of thesame periods.

Agro-meteorological yield models were generated using this data set and using the current seasons metdata.

These Agro-meteorological yield models can be used for predicting the basmati yields well in advance ofthe harvesting period.

The crop condition term was also be incorporated into the yield models to take into account the yieldreduction due yield reducing factors.

Agro-met-spectral Yield Models

2200

2400

2600

2800

3000

3200

3400

RICE

YIEL

D (kg

/ha)

1994 1996 1998 2000 2002 2004 2006 2008

YEAROBSERVED PREDICTED

Rice Yields in Karnal District

An attempt was made to predict the Basmati rice yields using the Agro-meteorogical yield model. Theobserved and model predicted Evolved basmati yields for Karnal, Panipat, Kurukshetra districts aregiven in the following figure.

Area in '000 ha

Pusa-1121 PB-1 CSR-30 Sharbati Sugandha

1 Punjab 2780.40 4.50 622.70 28.12 53.24 49.30 -

2 Haryana 1081.70 424.65 94.95 74.48 7.70 -

3 Uttar Pradesh 1550.00 18.05 289.85 59.00 - 112.90 70.35

4 Uttarakhand 143.00 8.20 7.00 1.70 - 33.40 3.30

5 Jammu & Kashmir 96.00 34.00 8.50 - - - -

6 Himachal Pradesh 72.88 2.85 - - 40.17 -

7 Delhi 1.505723.98 64.75 1357.05 183.77 127.72 243.47 73.65

Non-Notified

State-wise area under Basmati and other varieties

Total

S. No. State Total RiceBasmati-386, 370, Type-3

Evolved Basmati

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

AREA

(000

, HA)

YEAR

Aromatic Indica Basmati Rice Acreage, Production and Export During Last One Decade

Area Production Export (000, tons) Value (million USD)0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

AREA

(000

, HA)

YEAR

Aromatic Indica Basmati Rice Acreage, Production and Export During Last One Decade

Area Production Export (000, tons) Value (million USD)

Conclusion

• Traditional and Evolved Basmati varieties were separable by proper selection of data base on phenology and analysis of IRS Liss-3 data. AWiFS data with 5d repeat cycle used in conjuction with Liss-3 proved helpful wherever data gap existed.

• Within the two groups of Basmati, ground survey based fraction of diff. varieties was used to arrive at percent area under different varieties.

• Crop cutting based yield was used to calculate the production of different varieties. NDVI and NDWI based yield model was also developed and used to validate the field data.

• Abiotic stress, like flooding in Punjab in 2010 was also studied using RS data, which helped in flood damage assessment and it`s effect on final yield.

• Both RS, detailed ground information and expert knowledge are needed to get information on varietal discrimination and production. Accuracy of production data was compared with market arrivals in different states by AIREA and the user organization.

• The area under Evolved Basmati is increasing due to higher yield and Traditional varieties decreasing in recent years.

SALAMAT Thank You