Embed Size (px)
Citation preview
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
Wor
ld R
ice
Pric
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
Wor
ld R
ice
Pric
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
Wor
ld R
ice
Pric
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
Wor
ld R
ice
Pric
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 crisisBGD Famine
Panicked response from rice importers and exportersdespite stable supply
2008 rice crisis12.1 millionfell bellow the poverty linein 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 medialabhttp://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 systemof Rice-YESplease visitposter byRomuga et al
NASA-PowerCHIRPSLocal Wth
WISEHWSD
VegetationCloud Model(Atema & Ulaby, 1978)
Photosynthesis& Respiration (School of De Wit)
WeatherID453300454742454741454744
Rice-YES LUTID Yield1 58472 57483 51834 42405 2153
ORYZA MAPscape-RICE
MAPscape-RICE
Multi-temporal σ°
LAI
Start of Season (SoS)
SoS2015297201530920153212015333
LAI1.65-3.501.45-1.651.41-1.451.38-1.411.34-1.38
remapping
Yield Raster
Spatial allocation approach of rice yield simulation
150 m resolution30 hr processing time (for entire MRD, VNM)
20 m resolution5 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-RiceUser'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 = 150Overall accuracy 97.3%Kappa index 0.95
Rice area map accuracy computation
Ground Survey Data
Map
Dat
a
Nelson & Setiyono et al. (2014)Remote Sens2016, 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
LAI (
m2 m
-2)
0
1
2
3
4
5
ObservedSimulated
RMSE = 0.43 m2 m-2
NRMSE = 37 % r = 0.85n = 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
LAI (
m2 m
-2)
0
1
2
3
4
5
ObservedSimulated
RMSE = 0.43 m2 m-2
NRMSE = 37 % r = 0.85n = 22
Yield Estimation
Official Yield (t ha-1)
2 3 4 5 6 7 8
Mod
eled
Yie
ld (t
ha-1
)
2
3
4
5
6
7
8
PhilippinesVietnamTamil Nadu, IndiaThailandCambodia
RMSE = 284 kg ha-1
NRMSE = 6.3 %
Country, sitesYear 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
200Km
25Km
±
Max. LAI
1.23 - 3.00
3.01 - 3.50
3.51 - 4.00
4.01 - 4.50
4.51 - 5.50
5.51 - 6.84
Max. LAI
2.41 - 3.05
3.06 - 3.52
3.53 - 4.04
4.05 - 4.54
4.55 - 5.47
5.48 - 6.84
Ha Tay
Thai Binh
Hai Duong
Hanoi
Nam DinhNinh 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 DinhNinh Binh
Haiphong
Ha Nam
Hung Yen
Bac Ninh
Vinh Phuc
106°30'0"E106°0'0"E
LAI EstimationMODIS
PROSAILRadiative 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
MO
DIS
-bas
ed r
ice
yiel
d (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
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
MO
DIS
-bas
ed r
ice
yiel
d (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
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
MO
DIS
-bas
ed r
ice
yiel
d (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
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
MO
DIS
-bas
ed r
ice
yiel
d (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
(a)
Spring
(c)
(b) (d)
RMSE = 0.73 t ha-1
NRMSE = 12 %
SummerRMSE = 0.29 t ha-1
NRMSE = 5 %
SpringRMSE = 0.30 t ha-1
NRMSE = 5 %
SummerRMSE = 0.46 t ha-1
NRMSE = 8 %
Application for Crop Insurance: Supporting Pradhan Mantri Fasal BimaYojana(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
Insurance: Area Yield
Index Insurance
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
1998 2000 2002 2004 2006 2008 2010 2012 2014
Yiel
d (t
/ha)
Year
Simulated Observed
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
1998 2000 2002 2004 2006 2008 2010 2012 2014
Yiel
d (t
/ha)
Year
Simulated Observed
trigger
Average bias = 148 kg/ha
624 kg/ha
no payout
payout
Historical Yield
MODIS+SAR
SAR(ex-ante)
Conclusion
• SAR-based yield estimation system can provide reliable timely sub-national rice information to support food security policy.
• The innovative spatial allocation approach ensures the efficiency of yield data processing by supporting its potential use in an operational setting.
• MODIS+SAR based yield estimates well suited for reconstructing historical yield data in the context of crop insurance programs.
