Upload
others
View
12
Download
0
Embed Size (px)
Citation preview
Towards Global Earth Observation of Rice
Lam Dao Nguyen, Hoang Phi PhungVNSC/VAST, Vietnam
Pham Duy TienUniversity of An Giang , Vietnam
Sandrine DanielCapgemini, Toulouse, France
Thuy Le Toan Hoa Phan, Alexandre Bouvet, Thierry Koleck
CESBIO, Toulouse, France
o Among EO data, SAR data have been proved efficient for ricemonitoring since late 80’s , but applications have been hamperedby lack of systematic and cost effective data
o Sentinel-1 represents unpreceding opportunity for operational ricemonitoring applications
o R&D Demonstrator projects were urgently needed with the launchof Sentinel-1 in April 2014
The GEORICE Project
Rice is the most critical staple food for more than half of humanity,with the majority in developing world (90% in Asia)
Context of the project
The project focuses on demonstrating the full use of Sentinel-1observations for providing operational rice mapping products
Requirements/feedbacks of potential users in the studyregions are investigated to target the service provisiondevelopment
The demonstration was performed on the Mekong river delta,in Vietnam Extension, adaptation to other regions and countries wereforeseen if successful initial demonstration
Project objectives
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
1. User requirements are well understood (products, accuracy,
timeliness)
2. Systematic EO data acquisitions
3. Freely accessible (or cost effective) EO data
4. Methods to transform the data into products that meet the
users requirements are developed and validated
5. Operational data processing (e.g. cloud processing..)
6. Timeliness of the products dissemination to the users
Requirements for operational applications
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
The Target Users
The main national target users identified during the course of the project in Vietnam are governmental agencies.
These users are also the target of GEOGLAM in which Asia-Rice is a component.
The project focus will not be on the users of private sectors.
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Asia-RiCE Website: http://www.asia-rice.org
Asia-RiCE (Asia Rice Crop Estimation & Monitoring) program led by JAXA with CNES and more than 20 Asian Space Agencies and Ministries of Agriculture with International organizations such as ASEAN/AFSIS, UN/FAO, IRRI, from 2013 . POC: [email protected], [email protected], [email protected]
GEORICE project is conducted in connection with Asia-Rice
1. Main requirements :Rice sown and harvested area, rice cropping density, rice production for statistics at administrative units (province, region, countryRice status, growth anomaly for early qualitative information on future production
2. Other requirements:Rice phenology: to manage irrigation, fertilisation, pesticide, and combined with weather forecast, for disaster mitigationRice sowing date: for planning of irrigation, treatment, harvestRice varieties: for market informationField water status: for irrigation planning, water consumation
1. 1. Understanding and meeting the user needs
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
1. Understanding and meeting the user needs
Main rice monitoring products
Rice area, rice cropping density, growth status
Users Temporal scale Spatial scale
GEOGLAM/Asia Rice Monthly reporting National to global
National agencies Monthly to yearly Commune, district, province, country
Local Agencies NRT (on IT platform) Village, commune, district, province
1. 1. Understanding and meeting the user needs
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
2. Systematic EO data acquisition
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Every 12 days except few gaps6 days with Setinel-1B since 1 Oct 2016
3. Freely accessible Sentinel-1 data
05 May 2017
VHVV
Automatically dowloaded and preprocessed at CESBIO
Big Data challenge
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
1. To understand the S1 response of rice fields with- specific frequency, polarisations , incidence angle - diversity of rice ecosystems over 250x250 km scenes
2. To exploit dense time series of data (12 or 6 days)
3. To develop methods for mapping of rice grown areas, monitoring of phenological stages, growth anomaly
4. To deal with large amount of data (at 20 m resolution)
4. Research needed for method development
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
10/01/2015 VH
Examples of RGB combinations of different dates of Sentinel-1 over rice fields in the An Giang province
Complexity in the S1 SAR response of rice fields
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
GEORICE: ESA-DUE INNOVATORS III
PreprocessingImage inter-calibrationGeocodingMultitemporal/spatial filtering
User Input- Region of interest- Period of interest
Ground data-Rice/Non rice
samples- Phenological stages
Data AccessSentinel-1 IW, VV&VH
Rice Mapping
Rice map Near Real TimeWhen: After data acquisition
Phenological stages mapWhen: mid to end of season
Estimation of phenological stages
Deviation fromnormal conditions
Maps of anormalyWhen: during the season
Rice map per seasonWhen: mid to end of season
Cropping density mapWhen: end of season
AccuracyAssessment
Processing chain
Products Asia-RiceGEOGLAM
&National Users
User Inputs
National Users
GEORICE Schematic Scheme
In situ data for understanding of the radar backscatter and algorithm development
40 /60 sampling fields have been surveyed in 2015/16 in An Giang at the same dates of S1. Survey by the University of An Giang
In collaboration with the VNSC /STAC) and CESBIO
General:Day of sowingRice varietiesPlanting densityHarvest dateRice Yield
Intensive information:Phenological stageHeightBiomassSoil conditionWater managementUniformity of plant height
VegetativeReproductiveRipening
S2_2017
S1_2017S1_2015
Sentinel-1 backscatter time series of rice fields
VH/VV
1. Better performance of rice mapping methods based on backscatter temporal variation2. Weekly detection of rice phenology3. Better performance of prediction models ( rice production, methane emission)
Expected performance of 6-day vs 12–day time series
VH
Every 6 days
Example of time series of VH/VV and VV backscatter
VH/VV
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
To adapt Rice phenology scale to EO data
1 Seedling, emergence
Early stage
2 2/3 leaves
3 Tillering start Tillering
4 Tillering max
5 Stem elongation Booting Panicleinitiation6 Booting
7 Heading HeadingFlowering8 Flowering
9 Milky stage Grain filling
10 Milky hard stage
11 Maturation Maturation
12 Maturity ->harvest
VEGETATIVE25 to 75 days
REPRODUCTIVE
35 days
RIPENING30 days
TOTAL90 to 140 days
Germination Panicle initiation
End of flowerin
g
Mature grain
Erectophile structure More randomstructure
Vege
tativ
eR
epro
duct
ive
Rip
enin
g
6 stages12 stages
S1 used to monitor rice growth stage
(Thousand ha) 26/10/2015 07/11/2015 19/11/2015 01/12/2015 13/12/2015 25/12/2015 06/01/2016 18/01/2016 30/01/2016 11/02/2016 23/02/2016 06/03/2016
1. Seedling and emergence 4,3 0,4 1,7 1,9 25,6 11,8 1,7 7,4 12,3 5,7 4,6 2,52. Pretillering - Tilleringstart 0,5 3,8 0,8 1,2 1,8 25,3 11,9 3,9 6,0 16,6 6,9 6,73. Tillering 0,6 6,3 0,6 2,3 3,2 31,4 11,9 14,0 7,9 16,3 6,64. End tillering - Panicleinitiation start 0,8 6,9 2,1 2,3 5,7 40,8 54,3 17,4 11,4 0,05. Heading – Flowering 0,5 5,5 2,5 2,1 8,0 11,0 66,2 29,1 17,46. Grain filling 0,6 1,7 2,7 2,1 2,3 14,9 61,0 37,47. Maturation 0,3 0,3 2,6 2,1 3,4 12,7 40,88. Bare soil / flooding 0,2 0,9 9,6 30,8 16,3 0,8 3,2 1,2 0,9 1,6 5,3 10,2Total rice (1-7) 4,8 4,9 9,6 11,2 37,9 47,1 55,8 76,6 102,2 132,0 142,1 111,3
For each province, estimation of area at each phenological stage at each S1 acquisition
BEN TRE
Results in a form familiar to statistics offices
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Winter-Spring 2016
By March 2016:1,39 M ha of rice grown area
The prevision of the Ministry of AgricultureAnd Rural DevelopmentFor Winter-Spring rice1.56 M ha
Seasonal Rice Area and Statistics
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Result validation
In each of the 3 major rice regions in the Mekong delta
100 GPS samples of rice planting areas +
30 other LULC
In total, 413 data pointsT.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Groundsurvey
S1 product
Rice 299 293No-rice in the
season23
114 120Other LULC 91
Total 413
The Mekong Delta: 413 independent check points: 98%.
Error sources:1. The precision of the GPS coordinates2. The selected time interval for rice
mapping method
Rice/non rice validationDetection of rice planting area during the season
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Ground survey phenological stage (1-12)
GEO
RICE
phe
nolo
gica
lsta
ge (1
-6)
Only 1 plot out of 60 was erroneous(98.3%).
Rice phenology validation
Example of validation for 1 S1 acquisition
CanTho workshop, 24-25 October 2016
User workshops for project presentation, result evaluation and training
Dong Thap
Kien Giang
Hau GiangCan Tho
R2 = 0.9785
Estimated area (thousands of ha)
Rice area extent for Summer-Autumn2016 crop in the 13 provinces ofthe Mekong Delta
Comparison GEORICE estimates andAgency statistical data
Rice area statistics by provinces
Winter-Spring 201611/10/2015 to 29/03/2016
Winter-Spring 201506/10/2014 to 23/03/2015
Winter-Spring 201611/10/2015 to 29/03/2016
Decrease of Winter-Spring rice area in 2016 compared to 2015(decrease of 276,000 ha or 16.7% ; i.e. 1.39M ha vs 1.67M ha) caused byshortage of water and saline water intrusion (El Niño effect) .
Official report by Vietnam MARD 2017:The severe drought and salinity intrusion strongly affected 11 of the 13 provinces in the MRD. Rice areas affected by drought and salinity intrusion rapidly increased from 139,000 ha in mid March 2016 to 224,552 ha by mid April 2016.
Impacts of El Nino on rice area
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Winter-Spring 201611/10/2015 to 29/03/2016
Winter-Spring 201701/10/2016 to 30/03/2017
Increase of Winter-Spring rice area in 2017 compared to 2016, 2015( More fields planted with rice and conversion of aquaculture. Among causes:shortage of rice production in 2016 and increase of rice price in 2016)
However:Rice yield was expected to be lower than average, due to lingering salinity andperiods of excess rains and floods during the growing season(FAO, country profile, August 2017)
Winter-Spring 201611/10/2015 to 29/03/2016
Winter-Spring 201506/10/2014 to 23/03/2015
Impacts of El Nino on rice area
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Mapping at country scale
Winter-Spring Rice 2016
Mapping at country scale
Thai Binh
Winter-Spring Rice 2016
Mapping at country scale
Quy Nhon
Phu Yen
Winter-Spring Rice 2016
Quy Nhon
Winter-Spring Rice 2016
An Giang
Winter-Spring Rice 2016
Winter-Spring rice in Cambodia 2016
Mapping at country scale
Thuy Le Toan, French-Thai Seminar on Enhancing Agrigriculture Management Using Space Technology, Bangkok, 21 June 2016
20/01/201601/02/201613/02/201625/02/201608/03/201620/03/201601/04/201613/04/201625/04/201607/05/201619/05/201631/05/201612/06/2016
Rice map in Suphan Buri province, 12/06/2016
S1 data
Tech4Earth
Application Store
Services
Big Data Architecture for Data Processing
Data Lake
1 0 1 1
0 0 1 1 1 0 1
0 1 0 0 1 1 0 1
1 0 1 0 1 1 0 0
1 0 1 1
GEORICE application
Service Platform: geospatial handling, data management
Processing engines for analysis in near real time
Data Hub, to collect, ingest and archive data
External data
Satellite data
User data
CollectSentinel-1 images
Users Community
5. Operational data processing
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Tech4Earth
Application Store
Services
Big Data Architecture for Data Processing
Data Lake
1 0 1 1
0 0 1 1 1 0 1
0 1 0 0 1 1 0 1
1 0 1 0 1 1 0 0
1 0 1 1
GEORICE application
Service Platform: geospatial handling, data management
Processing engines for analysis in near real time
Data Hub, to collect, ingest and archive data
External data
Satellite data
User data
CollectSentinel-1 images
Users Community
GEORICE implementation
Building Analysis Ready Data (ARD) for Sentinel-1 on Datacube – Application on Rice monitoring
37
S1Data
CenterS1 data images
ARD pre-
processingAnalysisReadyData
https://peps.cnes.fr
Ground projected image (GRD).SAFE format
Python scriptS1toARD.py
S1 images ready to use in Datacube
First result of rice mapping on VN Data Cube, March 6, 2018
Implementation in Vietnam DataCube
o Rice monitoringo Forest monitoring
Beyond the GEORICE demonstrator project
1. User issues
Convince and involve the ‘real’ users (Institutions, Ministries..) with contribution/endorsement from national monitoring systems2. Promote coordination of supported projects Promote development and cross validation on different sites,
different countries Promote coordination of supported projects (e.g. via Asia-Rice,
GEOGLAM)
Strategic Issues to bring R&D results into operation
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
GEORICE proposed extension
• GEORICE: proof-of-conceptOne 41 000 km² region: the Mekong Delta, VietnamRice/Non rice maps:
1. near-real time after S1 acquisition 2. map per rice season (3 seasons per year)3. annual synthesis (cropping intensity)
Anomaly maps: interannual variationPhenological stage maps
• GEORICE-Extension: demonstration of larger scaledeployment (2018-2019)
o Extension to 1/3/4/5 full countries of South-East Asiao Seasonal Rice/Non rice mapso Rice growth status for selected regions
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
Vietnam + Lao PDR + Cambodia + Thailand + Myanmar1 940 000 km²297 tiles
Collaboration needed in the region to achieve the objective
We are a whole community to make it happen !
T.Le Toan LC/LUC –Int Regional Science Meeting-28-30 May 2018
SALAMAT!