Rice Crop Growth Monitoring and Yield Prediction with ... Rice Crop Growth Monitoring and Yield Prediction

  • View
    0

  • Download
    0

Embed Size (px)

Text of Rice Crop Growth Monitoring and Yield Prediction with ... Rice Crop Growth Monitoring and Yield...

  • Tri D. Setiyono International Rice Research Institute (IRRI), Philippines

    Rice Crop Growth Monitoring and Yield Prediction with Remote-Sensing and ORYZA Crop Growth Model

  • The need for timely reliable rice production information

    Year

    52 56 60 64 68 72 76 80 84 88 92 96 00 04 08 12 16

    W or

    ld R

    ic e

    Pr ic

    e (U

    SD M

    t-1 )

    0

    200

    400

    600

    800

    1000

    1200

    1400

    1600

    Year

    52 56 60 64 68 72 76 80 84 88 92 96 00 04 08 12 16

    W or

    ld R

    ic e

    Pr ic

    e (U

    SD M

    t-1 )

    0

    200

    400

    600

    800

    1000

    1200

    1400

    1600

    Year

    52 56 60 64 68 72 76 80 84 88 92 96 00 04 08 12 16

    W or

    ld R

    ic e

    Pr ic

    e (U

    SD M

    t-1 )

    0

    200

    400

    600

    800

    1000

    1200

    1400

    1600

    Year

    52 56 60 64 68 72 76 80 84 88 92 96 00 04 08 12 16

    W or

    ld R

    ic e

    Pr ic

    e (U

    SD M

    t-1 )

    0

    200

    400

    600

    800

    1000

    1200

    1400

    1600

    Green Revolution

    Widespread drought (SEA), THA banned rice export

    1974 rice crisis BGD Famine

    Panicked response from rice importers and exporters despite stable supply

    2008 rice crisis 12.1 million fell bellow the poverty line in BGD

  • Insurance helps to keep rice farmers in business

    Smallholder farmers, representing 85% of the world’s farms, face numerous risks to their agricultural production from pest and disease outbreaks, extreme weather events, and market shocks that threaten their household food and income security

  • Photo: ESA, © ESA/ATG medialab http://www.esa.int/spaceinimages/Images/2014/01/Sentinel-1_radar_vision

    Sentinel-1a • Launched 3rd/April/2014 by ESA • Sun-synchronous near-polar orbit, repeat

    cycle of 12 or 24 days • C-Band (5.405 GHz) • Dual polarization: VV+VH • Free and open access to imagery

    Satellite technology can reliably identify losses over large geographies, ensuring farmers receive indemnification more swiftly and enhancing stakeholder trust in crop insurance.

  • SAR+ORYZA Rice Yield Estimation System

    For detailed information on the IT system of Rice-YES please visit poster by Romuga et al

    NASA-Power CHIRPS Local Wth

    WISE HWSD

    Vegetation Cloud Model (Atema & Ulaby, 1978)

    Photosynthesis & Respiration (School of De Wit)

  • WeatherID 453300 454742 454741 454744

    Rice-YES LUTID Yield 1 5847 2 5748 3 5183 4 4240 5 2153

    ORYZA MAPscape-RICE

    MAPscape-RICE

    Multi-temporal σ°

    LAI

    Start of Season (SoS)

    SoS 2015297 2015309 2015321 2015333

    LAI 1.65-3.50 1.45-1.65 1.41-1.45 1.38-1.41 1.34-1.38

    remapping

    Yield Raster

    Spatial allocation approach of rice yield simulation

    150 m resolution 30 hr processing time (for entire MRD, VNM)

    20 m resolution 5 hr processing time (for entire MRD, VNM)

  • Subang, Indonesia, Nov ‘13 – Apr ‘15 SAR-based Rice Area

    Congalton, R.G. 2001. Accuracy assessment and validation of remotely sense and other spatial information. Int. J. Wildland Fire 10: 321–328.

    Rice Non-Rice User's

    Accuracy

    Rice 91 1 98.9%

    Non-Rice 3 55 94.8%

    Producer's Accuracy 96.8% 98.2% 97.3%

    User's Accuracy 96.9% Producer's Accuracy 97.5% n = 150 Overall accuracy 97.3% Kappa index 0.95

    Rice area map accuracy computation

    Ground Survey Data

    M ap

    D at

    a

    Nelson & Setiyono et al. (2014) Remote Sens 2016, 6, 10773- 10812

  • Subang, Indonesia, Nov ‘13 – Apr ‘15 SAR-based Rice Area

  • LAI estimation

    SAR backscatter coefficient (db)

    -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5

    LA I (

    m 2 m

    -2 )

    0

    1

    2

    3

    4

    5

    Observed Simulated

    RMSE = 0.43 m2 m-2

    NRMSE = 37 % r = 0.85 n = 22

    Hirooka et al., 2015. Field Crops Res. 176: 119-122

    SAR backscatter coefficient (db)

    -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5

    LA I (

    m 2 m

    -2 )

    0

    1

    2

    3

    4

    5

    Observed Simulated

    RMSE = 0.43 m2 m-2

    NRMSE = 37 % r = 0.85 n = 22

  • Yield Estimation

    Official Yield (t ha-1)

    2 3 4 5 6 7 8

    M od

    el ed

    Y ie

    ld (t

    h a-

    1 )

    2

    3

    4

    5

    6

    7

    8

    Philippines Vietnam Tamil Nadu, India Thailand Cambodia

    RMSE = 284 kg ha-1

    NRMSE = 6.3 %

    Country, sites Year and

    season

    SAR-based yield estimates,

    t ha-1 Official

    yield/ CCY

    t ha-1

    RMSE

    kg ha-1

    NRMSE

    (%) Min. Mean Max.

    Cambodia, Takeo 2014 MWS 2.74 3.13 3.51 3.03, 339 242, 634 8.0, 19

    Vietnam, Soc Trang 2014 Summer 4.56 5.39 5.77 5.64, 5.45 459, 414 8.1, 7.6

    Philippines, Nueva Ecija 2013-14 Dry 6.77 7.60 8.21 7.21, 7.71 980, 1,029 13.6, 13.4

    Philippines, Nueva Ecija 2014 Wet 3.96 5.30 5.95 5.31/ 5.77 480, 521 8.9, 8.7

  • Added value products: MODIS + SAR

    MODIS

    Sentinel-1

    Start and Peak of Season PhenoRice

    LAI at POS PROSAIL &

    Logistic function

    Area MAPscape - RICE

    Ancillary

    Y ield Simulation

    R ice - YES &

    ORYZA -

    Estimated Yield

    Historical yield estimates for Area-based Yield Insurance

  • Red River Delta, VNM, 2016 Spring & Summer

    200 Km

    25 Km

    ±

    Max. LAI

    1.2 3 -

    3. 00

    3.0 1 -

    3. 50

    3.5 1 -

    4. 00

    4.0 1 -

    4. 50

    4.5 1 -

    5. 50

    5.5 1 -

    6. 84

    Max. LAI

    2.4 1 -

    3. 05

    3.0 6 -

    3. 52

    3.5 3 -

    4. 04

    4.0 5 -

    4. 54

    4.5 5 -

    5. 47

    5.4 8 -

    6. 84

    Ha Tay

    Thai Binh

    Hai Duong

    Hanoi

    Nam Dinh Ninh Binh

    Haiphong

    Ha Nam

    Hung Yen

    Bac Ninh

    Vinh Phuc

    106°30'0"E106°0'0"E

    21°0'0"N

    20°30'0"N

    20°0'0"N

    109°0'0"E

    109°0'0"E

    105°0'0"E

    105°0'0"E

    20°0'0"N

    15°0'0"N

    10°0'0"N

    Ha Tay

    Thai Binh

    Hai Duong

    Hanoi

    Nam Dinh Ninh Binh

    Haiphong

    Ha Nam

    Hung Yen

    Bac Ninh

    Vinh Phuc

    106°30'0"E106°0'0"E

  • LAI Estimation MODIS

    PROSAIL Radiative Transfer

    PhenoRice

    LAI Estimates

    Logistic LAI Model

  • Yield Estimation (MODIS)

    Sentinel-1-based rice yield (t ha-1)

    3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

    M O

    D IS

    -b as

    ed r

    ic e

    yi el

    d (t

    h a-

    1 )

    3.0

    3.5

    4.0

    4.5

    5.0

    5.5

    6.0

    6.5

    7.0

    7.5

    8.0

    Sentinel-1-based rice yield (t ha-1)

    3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

    M O

    D IS

    -b as

    ed r

    ic e

    yi el

    d (t

    h a-

    1 )

    3.0

    3.5

    4.0

    4.5

    5.0

    5.5

    6.0

    6.5

    7.0

    7.5

    8.0

    Reported rice yield (t ha-1)

    3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

    M O

    D IS

    -b as

    ed r

    ic e

    yi el

    d (t

    h a-

    1 )

    3.0

    3.5

    4.0

    4.5

    5.0

    5.5

    6.0

    6.5

    7.0

    7.5

    8.0

    Reported rice yield (t ha-1)

    3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

    M O

    D IS

    -b as

    ed r

    ic e

    yi el

    d (t

    h a-

    1 )

    3.0

    3.5

    4.0

    4.5

    5.0

    5.5

    6.0

    6.5

    7.0

    7.5

    8.0

    (a)

    Spring

    (c)

    (b) (d)

    RMSE = 0.73 t ha-1

    NRMSE = 12 %

    Summer RMSE = 0.29 t ha-1

    NRMSE = 5 %

    Spring RMSE = 0.30 t ha-1

    NRMSE = 5 %

    Summer RMSE = 0.46 t ha-1

    NRMSE = 8 %

  • Application for Crop Insurance: Supporting Pradhan Mantri Fasal Bima Yojana(PMFBY) insurance program in Tamil Nadu, India

    Bi-weekly information on crop growth status, derived by SAR-based Rice Monitoring System

    Observation on anomaly in rice cultivation helps to guide decision on Preventive Sowing/ On Account Payment Cover

    End-of-season yield estimation at village level can guide Crop Cutting Experiments (“Smart sampling”)

  • Rice Yield Map of Vayalore Village of Kodavasal Block, Tiruvarur District, Tamil Nadu

  • Application for Crop