Remote Sensing based Crop Yield Monitoring & Forecasting · 2015. 5. 26. · ME (kg ha -1 ) 42 y = 0.96 x + 0.168 r = 0.94 Performance of ORYZA2000 in simulating rice yield Experimental

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

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

Background



Rice in South East & South Asia: Small holder farmers, vulnerable to climate uncertainties

2011 Thai Flood (Chao Phraya Basin)

Project Targets:

7 countries, 5 million farmers, in 5 years.

2

3

Reduce the vulnerability of 5 million rice farmers in Asia and beyond to flood and drought over the next 5 years.

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.

1

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

www.riice.org

RIICE Service Platform

RADAR Remote Sensing: MAPScape-Rice



8

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

Rice from optical and radar RS

time

Rad

ar

backscatt

er

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)

grain filling

harvesting



Rice extent, 1 ha resolution

ASAR wide swath

2004 to 2012

Single season rice area; 3m CSK; June to September 2012 © Cosmo-SkyMed data ASI distributed by e-GEOS,

Data processed using MAPscape-Rice.

Leyte, Philippines



Leyte, Philippines

© Cosmo-SkyMed data

ASI distributed by e-

GEOS, processed using

MAPscape-Rice.

Fallow Land preparation Vegetative stage Reproductive stage Maturity Harvested

Late June, most areas are in land preparation [blue]. Some are still fallow [brown].

26 Jun 2012 12 Jul 2012

Mid July, the area is a mixture land preparation [blue] and crop in tillering stage [green].

End July, most of the cropped area is in peak vegetation or flowering [dark green] and some still in tillering stage [green].

28 Jul 2012

Rice field

phenology

Wet season 2012

Inferred from

SAR product.

13 Aug 2012 21 Sep 2012 30 Sep 2012

Mid August, most of the cropped area is in reproductive stage [green] tending to maturity [yellow].

Mid September, most areas are in maturity [yellow], some are still in reproductive stage [green] and some are already harvested [brown].

End September, the area is a mixture green crops (green), maturing [yellow], and already harvested fields [brown].





Rice field

phenology

Wet season 2012

Inferred from

SAR product.

Leyte, Philippines

© Cosmo-SkyMed data

ASI distributed by e-

GEOS, processed using

MAPscape-Rice.





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).

Red River Delta, Vietnam

Cambodia

© Cosmo-SkyMed data ASI distributed by

e-GEOS, processed using MAPscape-Rice.

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).



Tamil Nadu, India


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

Seasonal rice area, Thanjavur, TN, India, Cosmo-Sykmed

(Sep 2012 to Feb 2013). Colors = start of season (SoS).

Rice area (15m), ASAR IM

(2011), Cauvary Delta, India



Damage Assessments

Data processed using MAPscape-Rice. © Cosmo-SkyMed data ASI

distributed by e-GEOS, processed

using MAPscape-Rice.

Taiphon Haiyan (Yolanda),

Leyte Philippines, Nov 2013

Flooding Muang Yang, Thailand, Oct 2013

(image provided by GISTDA)

Oct 2 – blue; Oct 6 – cyan

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.

Subang, Indonesia, Feb 2014

Crop Growth Model: ORYZA2000



Potential, water-limited, and/or nitrogen-limited

conditions

Lowland & upland/aerobic rice

Weather, irrigation, nitrogen fertilizer, general

management, variety characteristics, soil

properties

ORYZA2000: a rice growth simulation model

Weather Data

• Daily Time Step

• Variables:

•Solar Radiation

•Min Temperature

•Max Temperature

•Relative Humidity

•Wind Speed

•Precipitation

ORYZA2000 Rice Crop Growth Simulation Model

Observed yield (Mg ha-1

)

0 2 4 6 8 10 12

Sim

ula

ted

yie

ld (

Mg

ha

-1)

0

2

4

6

8

10

12

Rasht, Iran (2007)

Ludhiana, India (2000-2002)

Shizukuishi, Japan (1998-2000)

Jakenan, Indonesia (1995-2000)

Coimbatore, India (2008-2009)

Ludhiana, India (2008-2009)

Amiri & Rezaet, 2009 (World App. Sci J 6:1113-1122)

Aurora 2006 (Agric. Water. Manag. 83: 51-57)

Bannayan et al., 2007 (Field Crops Res. 93:237-251)

Boling et al., 2007 (Agric. System 92:115-139)

Soundharajan et al., 2009 (Paddy Water Environ. 7: 135-149)

Sudhir-Yadav et al., 2010 (Field Crops Res. 122:104-117)

RMSE (kg ha-1

)

578

379

1262

582

219

726

RMSE (kg ha-1)754

ME (kg ha-1)42

y = 0.96 x + 0.168r = 0.94

Performance of

ORYZA2000 in

simulating rice

yield

Experimental plots

Potential, water-, or

nitrogen-limited

environments

ORYZA2000

Rice Crop Growth

Simulation Model

Irrigated Rice Yield (2009 Dry Season) Source: BAS Simulated (ORYZA2000)

BAS - Philippines Bureau of Agricultural Statistics

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)

Crop Growth Model: ORYZA2000

Integrated RS & Crop Growth Model



Rice Yield Simulation

with SoS & LAI from SAR

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

www.riice.org Leyte West, Philippines

Preliminary accuracy assessment of RS-based rice yield estimation

Yield accuracy at barangay level

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

ARBY CCE RS estimate 702

Amahit 2.96 1.94 Accuracy (%)

Cuta 3.79 4.32 85

Liloan 5.96 5.04

Matica-a 5.14 5.69

Sabang Ba-o 4.99 4.94

Leyte West, Philippines, Wet Season 2012

www.riice.org

National

Province

Muncipal

Barangay


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

Municipality Yield trigger

at 95%

(ton/ha)

Yield (ton/ha)

ARBY RS estimate

Barugo 3.62 5.12 5.63

Ormoc City 3.74 5.31 5.56

Agreement between ARBY and RS yield estimates

No payout in WS 2012

*ARBY – Area Based Yield Crop Insurance Policy

National

Province

Muncipal

Barangay

Leyte West, Philippines, Wet Season 2012

www.riice.org


Crop Yield Forecasting



Duration of Forecast (days)

90 80 70 60 50 40 30 20 10 0

Yie

ld (

Mg

ha

-1)

3.0

3.5

4.0

4.5

5.0

5.5

6.0

6.5

7.0

Time of Forecast (day of year)

180 190 200 210 220 230 240 250 260 270 280

max.

75 pct.

25 pct.

min.

Median

Mean

- Climate Information

- Assimilation of secondary data (e.g. RS)

Forecast without RS

= 6.2 Mg/ha

LAI estimate from RS

55 days before harvest

Corrects the forecast &

get closer to actual yield

of 5.2 Mg/ha

Unc

erta

inty

planting anthesis harvestTime

Unc

erta

inty


model uncertainty

climate uncertainty

Unc

erta

inty


model uncertainty

climate uncertainty

Unc

erta

inty


Hansen et al. 2006

Clim. Res 33: 27-41

Reducing uncertainties in yield forecast Integrated RS & Crop Growth Model

- RS -> in-season crop status

- CSM -> weather + crop status = yield

• 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

Summary



Documents

Remote Sensing based Crop Yield Monitoring & Forecasting · 2015. 5. 26. · ME (kg ha -1 ) 42 y = 0.96 x + 0.168 r = 0.94 Performance of ORYZA2000 in simulating rice yield Experimental