Remote Sensing based

Crop Yield Monitoring & Forecasting

Tri Setiyono1, Andrew Nelson1, Francesco Holecz2
1International Rice Research Institute (IRRI), Philippines
2sarmap, Switzerland

Expert Meeting on ‘Crop Monitoring for Improved Food Security’

International Rice Research Institute 17 February 2014., Vientiane, Lao PDR

Presentation Outline
• Background
• RIICE Service Platform
•The tools:
– Radar remote-sensing
– Crop growth model
• Integrated RS & crop modeling
• Yield forecasting
•Summary

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR
 Background

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR

Rice in South East & South Asia:

Small holder farmers, vulnerable to climate uncertainties

2011 Thai Flood (Chao Phraya Basin)

 Remote sensing-based Information and Insurance
for Crops in Emerging economies

Project Targets:
7 countries, 5 million farmers, in 5 years.

Reduce the vulnerability of 5 million

rice farmers in Asia and beyond to
flood and drought over the next 5
years.
2

Help Governments and NGOs to

better plan for food crises through
better crop growth monitoring.

Increase efficiency and effectiveness

of crop insurance solutions and turn
it into a viable business also in
emerging markets. www.riice.org

RIICE Service Platform

 RADAR Remote Sensing: MAPScape-Rice

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR

REMOTE SENSING-BASED INFORMATION AND INSURANCE FOR CROPS IN EMERGING ECONOMIES

Rice from optical and radar RS

8
 REMOTE SENSING-BASED INFORMATION AND INSURANCE FOR CROPS IN EMERGING ECONOMIES

Space-born Earth Observation

• Software: MAPscape-Rice (Sarmap SA)
• Radar Data: Cosmo-SkyMed (40 x 40 km, 3 m,
X-band) (Italian Space Agency/e-GEOS)
Radar backscatter

grain filling

harvesting

time

Leyte, Philippines

Rice extent, 1 ha resolution Single season rice area; 3m CSK; June to September 2012
ASAR wide swath © Cosmo-SkyMed data ASI distributed by e-GEOS,
2004 to 2012
Data processed using MAPscape-Rice.

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR
 Leyte, Philippines 12
28
13
21
30
26 Sep
Jul
Aug
Jun2012
2012
2012
Rice field Mid September,
End
Late July,
August,
June,most
the most
mostarea
ofareas
the
most
the
of
is area
athe
phenology
mixture
cropped
areas
is
area in
mixture
are land
landarea
inpreparation
green
maturity
preparation
is in peak
crops
Wet season 2012
Inferred from [blue] and
vegetation
reproductive
[yellow],
(green),
[blue]. maturing
some
Some crop
orare
stage
flowering
are
instill
still in
SAR product. tillering
[dark
[green]
reproductive
[yellow],
fallow green]
tending
stage
andand
[brown]. stage
already
[green].
tosome
still in tillering
maturity
[green]
harvested and
[yellow].
fields
somestage
[brown].
are
[green]. harvested
already
[brown].

© Cosmo-SkyMed data
ASI distributed by e-  Fallow  Land preparation  Vegetative stage  Reproductive stage  Maturity  Harvested
 Fallow  Land GEOS,
preparation processed usingstage
Vegetative  Reproductive stage  Maturity  Harvested
MAPscape-Rice.

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR

Leyte, Philippines
Rice field
phenology
Wet season 2012
Inferred from
SAR product.

© Cosmo-SkyMed data
ASI distributed by e-  Fallow  Land preparation  Vegetative stage  Reproductive stage  Maturity  Harvested
 Land preparation  Vegetative  Reproductive
GEOS, processed
stage using stage  Maturity  Harvested
MAPscape-Rice.

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR
Red River Delta, Vietnam
Rice extent (1 ha), ASAR WS (2003
to 2010).
Seasonal rice, 3m Cosmo-Sykmed
(SM) January to May 2013. Different
colors correspond to the different
dates of start of season (SoS).

Cambodia
Rice extent (1 ha), ASAR WS (2005
to 2010).
Seasonal rice area for short duration
(green) and medium-long duration
(green) in sample area near Takeo,
Cambodia based on 3m Cosmo-
SkyMed SM (September 2012 to
March 2013).

© Cosmo-SkyMed data ASI distributed by

e-GEOS, processed using MAPscape-Rice.
Data processed using MAPscape-Rice.

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR

Tamil Nadu, India

Rice area (15m), ASAR IM Seasonal rice area, Thanjavur, TN, India, Cosmo-Sykmed
(2011), Cauvary Delta, India (Sep 2012 to Feb 2013). Colors = start of season (SoS).
© Cosmo-SkyMed data ASI distributed by e-GEOS, processed using MAPscape-Rice.
Data processed using MAPscape-Rice.
Remote Sensing based Crop Yield Estimation
Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR
 Damage Assessments

Flooding Muang Yang, Thailand, Oct 2013

(image provided by GISTDA)
Oct 2 – blue; Oct 6 – cyan

Taiphon Haiyan (Yolanda),

Leyte Philippines, Nov 2013
Data processed using MAPscape-Rice.
Remote Sensing based Crop© Cosmo-SkyMed
Yield Estimation data ASI
Expert Meeting on ‘Crop Monitoringdistributed by e-GEOS,
for Improved Food Security’ processed
International Rice Research Institute using Lao
17 February 2014., Vientiane, MAPscape-Rice.
PDR

Subang, Indonesia, Feb 2014 Areas flooded on

February 9th are in
blue.

The red box borders

the area where
some floods could
have occurred in
between February
5th and 9th.

CSK images provided

by e-geos.
CSK images
processed by sarmap.
 Crop Growth Model: ORYZA2000

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR

ORYZA2000 Rice Crop Growth Simulation Model

ORYZA2000: a rice growth simulation model

 Potential, water-limited, and/or nitrogen-limited
conditions
 Lowland & upland/aerobic rice
 Weather, irrigation, nitrogen fertilizer, general
management, variety characteristics, soil
properties

Weather Data
• Daily Time Step
• Variables:
•Solar Radiation
•Min Temperature
•Max Temperature
•Relative Humidity
•Wind Speed
•Precipitation
ORYZA2000 12
Rice Crop Growth
Simulation Model 10 RMSE (kg ha-1)

Simulated yield (Mg ha )

754

-1
Performance of 8 ME (kg ha-1)
42
ORYZA2000 in
simulating rice y = 0.96 x + 0.168
6
yield r = 0.94

Experimental plots
Potential, water-, or 4
nitrogen-limited
environments
2

0
0 2 4 6 8 10 12
-1
Observed yield (Mg ha )
RMSE (kg ha-1)
578 Rasht, Iran (2007) Amiri & Rezaet, 2009 (World App. Sci J 6:1113-1122)
379 Ludhiana, India (2000-2002) Aurora 2006 (Agric. Water. Manag. 83: 51-57)
1262 Shizukuishi, Japan (1998-2000) Bannayan et al., 2007 (Field Crops Res. 93:237-251)
582 Jakenan, Indonesia (1995-2000) Boling et al., 2007 (Agric. System 92:115-139)
219 Coimbatore, India (2008-2009) Soundharajan et al., 2009 (Paddy Water Environ. 7: 135-149)
726 Ludhiana, India (2008-2009) Sudhir-Yadav et al., 2010 (Field Crops Res. 122:104-117)

Irrigated Rice Yield (2009 Dry Season)

Source: BAS Simulated (ORYZA2000)

BAS - Philippines Bureau of Agricultural Statistics

 Crop Growth Model: ORYZA2000

Yield trends in long-term continuous rice cropping

• Actual measured yield in

dry seasons, 1992 to 2010
• 6.5 to 9.5 tons/ha
• Measured yield depends
on weather and
management

• Potential yield is the maximum plausible yield of a variety under certain climatic
conditions
• Potential yield is obtained using crop simulation models (ORYZA2000)
• Potential yield varies according to weather

Slide: Courtesy of
Roland Buresh (IRRI)

Integrated RS & Crop Growth Model

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR
 Leyte West, Philippines www.riice.org
Rice Yield Simulation
with SoS & LAI from SAR

© Cosmo-SkyMed data ASI distributed by e-GEOS, processed using MAPscape-Rice

Integrated RS & Crop Growth Model

Leyte West, Philippines, Wet Season 2012

www.riice.org
Yield accuracy at barangay level
Preliminary accuracy assessment of RS-based rice yield estimation

Yield (ton/ha) RMSE (kg/ha)

Barangay
ARBY CCE RS estimate 702
Amahit 2.96 1.94 Accuracy (%)
Cuta 3.79 4.32 85 National

Liloan 5.96 5.04 Province

Matica-a 5.14 5.69 Muncipal

Sabang Ba-o 4.99 4.94 Barangay

 Integrated RS & Crop Growth Model

Leyte West, Philippines, Wet Season 2012

www.riice.org
Insurance & RS-based Yield Estimates
Comparison of ARBY* yield and Remote Sensing yield for 2012 WS

Yield trigger Yield (ton/ha)

Municipality at 95%
(ton/ha) ARBY RS estimate

Barugo 3.62 5.12 5.63

Ormoc City 3.74 5.31 5.56 National

Agreement between ARBY and RS yield estimates Province

No payout in WS 2012 Muncipal

Barangay
*ARBY – Area Based Yield Crop Insurance Policy

Crop Yield Forecasting

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
International Rice Research Institute 17 February 2014., Vientiane, Lao PDR
 model uncertainty

Uncertainty

Uncertainty
climate uncertainty

Integrated RS & Crop Growth

Reducing uncertainties Model
in yield forecast
- RS -> in-season
Climate crop status
Information
planting anthesis harvest planting anthesis harvest
- CSM -> weather
Assimilation + crop status
of secondary data=(e.g.
yieldRS) Time Time

model uncertainty

Uncertainty

Uncertainty
Duration of Forecast (days)
climate uncertainty

90 80 70 60 50 40 30 20 10 0
7.0 planting anthesis harvest planting anthesis harvest
Time Time

Median Hansen et al. 2006

6.5 max. Mean Clim. Res 33: 27-41
75 pct.
6.0
Yield (Mg ha )
-1

25 pct. Forecast without RS

5.5 = 6.2 Mg/ha
min.
5.0

4.5
LAI estimate from RS
4.0 55 days before harvest
Corrects the forecast &
3.5 get closer to actual yield
of 5.2 Mg/ha
3.0
180 190 200 210 220 230 240 250 260 270 280

Time of Forecast (day of year)

Summary

• Incorporation of remote sensing products (SAR) into

crop model is essential:
• Reduces non-remote sensing parameters, e.g. SAR derived LAI
capture the response of rice plants to environmental conditions over
large areas
• Includes rice phenology to initialize the model
• Improves yield estimation for actual yields, allowing reliable
application such as in a crop insurance system
• National partners involvements are crucial:
• The only way to sustain, promote and validate operational crop
monitoring system
• Terrestrial data collection for calibration and validation & provision of
knowledge on rice types and practices that are essential for product
generation

Remote Sensing based Crop Yield Estimation

Expert Meeting on ‘Crop Monitoring for Improved Food Security’
17 Institute
International Rice Research February 2014., Vientiane, Lao PDR