Rice crop monitoring in the Mekong Delta, Vietnam using radar remote sensing data Nguyen Lam-Dao – GIRS/HCMIRG/VAST, Vietnam Thuy Le-Toan – CESBIO/CNRS, France Phung Hoang-Phi – GIRS/HCMIRG/VAST, Vietnam Asian Conference on Remote Sensing WORKSHOP ON CROP MONITORING AND FOOD SECURITY 22 October 2013, Bali, Indonesia
52
Embed
Rice crop monitoring in the Mekong Delta, Vietnam using radar remote sensing data Nguyen Lam-Dao – GIRS/HCMIRG/VAST, Vietnam Thuy Le-Toan – CESBIO/CNRS,
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Rice crop monitoring in the Mekong Delta, Vietnam
using radar remote sensing data
Nguyen Lam-Dao – GIRS/HCMIRG/VAST, VietnamThuy Le-Toan – CESBIO/CNRS, France
Phung Hoang-Phi – GIRS/HCMIRG/VAST, Vietnam
Asian Conference on Remote Sensing
WORKSHOP ON CROP MONITORING AND FOOD SECURITY
22 October 2013, Bali, Indonesia
2
Contents
1. Introduction
2. Previous research results
3. Ongoing and further works
4. Technical demonstrator site –
the Mekong delta, Vietnam
3
Introduction: Mekong Delta, Vietnam
Optical data
13 provinces and city;
Population: 17.3 M (20% or 1/5);
Area: 40,500 Km2 (12% or 1/8)
Rice production: 23.2 Mton (> 50% or 1/2)
Source: GSO, 2011
4
Rice cropping systems
Main rice-based cropping systems in the MD, Vietnam
Rice crop monitoring using new generation synthetic aperture radar (SAR) imagery (ENVISAT-ASAR data, 2007-2008)
Utilisation of SAR data for rice crop monitoring (ERS2-SAR data, 1997-1998)
Other projects in the Mekong and Red River Delta
9
SAR data used for previous projects
Satellite Years Agency Frequency - Polarisation
Resolution - Swath
Special
ERS-1 1991-2000 ESA C - VV
25 m 100 km
Interferometry (with ERS-2)
JERS 1992-1998 NASDA L-HH 25 m 100 km
Region. mosaic available
ERS-2 1995 ESA
C - VV
25 m 100 km
Interferometry (with ERS-1)
RADARSAT-1 1995 CSA C - HH 10 -100 m 45 - 500 km
Multi-incidence
ENVISAT - ASAR
2002 ESA C - HH/VV/HV 25 - 1000 m 50 - 500 km
Multi-incidence
ALOS - PALSAR
2006-2011 JAXA L - Polarimetric 10 - 100 m 100 - 350
km
Multi-incidence
TerraSAR-X 2007 DLR X-Polarimetric 1 m Interferometry 1 day
RADARSAT 2 2007 CSA C - Polarimetric
< 10 m Multi-incidence
New satellites: COSMO-SkyMed, RISAT-1, ALOS-2 & Sentinel-1 (2013)
10
Objectives
To evaluate the use of SAR data in rice mapping and yield estimation, towards an operational system for rice crop inventory in the Mekong Delta, Vietnam.
24/03/0717/02/2007
11
Rice seasons
Main rice seasons in An Giang province, Mekong Delta
Rice crop Planting Harvesting
English name Local name
Winter Spring Dong Xuan Nov./Dec. Mar./Apr.
Summer Autumn He Thu Apr./May Jul./Aug.
Rainy season Thu Dong (Autumn Winter) Jul./Sep. Oct./Dec.
Mua (Traditional rice) Jul./Sep. Nov./Jan.
12 Sample rice fields in Cho Moi (An Giang)
Methods – Sample rice fields
13
Ground data collection at Cho Moi district
Guidelines for ground data collection for rice monitoring experiments using radar data (Thuy Le Toan)
14
SAR data used (2007, 10-11)
ENVISAT ASAR data (2007): - Band: C - Wavelength: 5.6 cm - Polarisation: HH&VV - Resolution: 30 m (APP)
TerraSAR-X data (2010-11):
- Band: X
- Wavelength: 3.1 cm
- Polarisation: HH&VV- Resolution: 3 m (SM)
No. Sensor-ModeDate of image
No.Sensor-Mode
Date of image No. Sensor-Mode Date of image
1 ASAR APP 13-Jan-07 11 ASAR APP 22-Feb-08 21 TSX SM 31-Jan-11
2 ASAR APP 17-Feb-07 12 TSX SM 19-Aug-10 22 TSX SM 11-Feb-11
3 ASAR APP 24-Mar-07 13 TSX SM 30-Aug-10 23 TSX SM 22-Feb-11
4 ASAR APP 28-Apr-07 14 TSX SM 10-Sep-10 24 TSX SM 16-Mar-11
5 ASAR APP 2-Jun-07 15 TSX SM 24-Oct-10 25 TSX SM 27-Mar-11
6 ASAR APP 07-Jul-07 16 TSX SM 04-Nov-10 26 TSX SM 07-Apr-11
7 ASAR APP 15-Sep-07 17 TSX SM 15-Nov-10 27 TSX SM 29-Apr-11
8 ASAR APP 20-Oct-07 18 TSX SM 26-Nov-10 28 TSX SM 10-May-11
9 ASAR APP 24-Nov-07 19 TSX SM 18-Dec-10 29 TSX SM 01-Jun-11
10 ASAR APP 29-Dec-07 20 TSX SM 29-Dec-10
15Temporal variation of average polarization ratio HH/VV of ASAR APP
(left) and TSX SM (right) for sample rice fields in Cho Moi
Results – Rice backscatter analysis (2007, 10-11)
-4.0-2.00.02.04.06.08.0
10.012.0
Dec-06 Feb-07 Apr-07 Jun-07 Aug-07 Oct-07 Dec-07
HH
/VV
(dB)
-4.0-2.00.02.04.06.08.0
10.012.0
Aug-10 Oct-10 Dec-10 Feb-11 Apr-11 Jun-11
HH/V
V (d
B)
16
Results – Rice backscatter analysis of TSX SM (2010-2011)
O charts of HH (UL), VV (UR) & HH/VV (LL) data
-30.0
-25.0
-20.0
-15.0
-10.0
-5.0
0.0
0 10 20 30 40 50 60 70 80 90 100 110
Số ngày sau sạ/cấy
Hệ
số tá
n xạ
ngư
ợc (
dB)
TĐ2010_CM
ĐX2011_CM
ĐX2011_TL
-30.0
-25.0
-20.0
-15.0
-10.0
-5.0
0.0
0 10 20 30 40 50 60 70 80 90 100 110
Số ngày sau sạ/cấy
Hệ
số tá
n xạ
ngư
ợc (
dB)
TĐ2010_CM
ĐX2011_CM
ĐX2011_TL
-6.0
-4.0
-2.0
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
0 10 20 30 40 50 60 70 80 90 100 110
Số ngày sau sạ/cấy
HH
/VV
(dB
)
TĐ2010_CM
ĐX2011_CM
ĐX2011_TL
Source: Thuy Le Toan et. al., 1997
17 HH&VV ratio of land use / land cover classes
Results – LU backscatter analysis of TSX SM (2010-2011)
-4.0
-2.0
0.0
2.0
4.0
6.0
8.0
10.0
12.0
26/12/2010 25/01/2011 24/02/2011 26/03/2011
Thời gian
Hệ
số tá
n xạ
ngư
ợc (
dB)
. Cây hàng năm 1
Cây hàng năm 2
Nông thôn 1
Nông thôn 2
Lúa 1
Lúa 2
Sông 1
Sông 2
Đô thị 1
Đô thị 2
Results – Rice backscatter analysis of ASAR APP (2007)Effect of water / no water
Where: YRa: rice yield (kg/m2), Ra1 : polarisation ratio of first date image,
Ra2 : polarisation ratio of second date image,
Ra3 : polarisation ratio of third date image,
Ra4 : polarisation ratio of four date image,
Ra5 : polarisation ratio of five date image, r2 : the coefficient of determination,
sey : the standard error for the y estimate.
A distribution map of estimated rice yield in AW 2010 crop at Cho Moi district using five-date polarisation ratios:• SM 30/08/2010 (8)• SM 10/09/2010 (19)• SM 24/10/2010 (63)• SM 04/11/2010 (74)• SM 15/11/2010 (85)
AW 2010 Crop, Case 1
31
YRa = -0.0422*Ra1 + 0.0068*Ra2 + 0.0969*Ra3 + 0.4918
r2 = 0.781, sey = 0.16 ton/ha where YRa : rice yield (kg/m2),
Ra1 : polarisation ratio of first date image, Ra2 : polarisation ratio of second date image, Ra3 : polarisation ratio of third date image, r2 : the coefficient of determination, sey : the standard error for the y estimate.
A distribution map of estimated rice yield in AW 2010 crop at Cho Moi district using there-date polarisation ratios:• SM 10/09/2010 (19)• SM 24/10/2010 (63)• SM 15/11/2010 (85)
AW 2010 Crop, Case 7
Percentage error by commune derived from two cases
Commune name
Estimatedproduction
(Ton)
Agencydata(Ton)
PercentageError(%)
Long Kien 4215 5880 -28.3
My Luong town 2069 2204 -6.1
Long Giang 5968 5940 0.5
My An 2449 1659 47.6
Kien Thanh 7297 7800 -6.5
Long Dien B 5832 5490 6.2
Tan My 4493 4680 -4.0
Long Dien A 4662 5292 -11.9
Cho Moi town 228 342 -33.3
Total 37212 39287 -5.3
Commune name
Estimatedproduction
(Ton)
Agencydata(Ton)
Percentageerror(%)
Long Kien 4214 5880 -28.3
My Luong town 2081 2204 -5.6
Long Giang 5992 5940 0.9
My An 2461 1659 48.3
Kien Thanh 7331 7800 -6.0
Long Dien B 5820 5490 6.0
Tan My 4521 4680 -3.4
Long Dien A 4655 5292 -12.0
Cho Moi town 229 342 -33.0
Total 37303 39287 -5.0
Case 1 (five-date data) Case 7 (three-date data)
Multiple linear regression analysis were performed using
LINEST function:
Rice yield estimation using ASAR APP (2007)
Rice cropr2
HH VV HH/VV
WS 2007 0.575 0.661 0.675
SA 2007 0.653 0.328 0.833
The regression equations between in situ rice yield and polarisation
ratios of sampling fields at Cho Moi district in WS & SA 2007 crop using
SA crop: YRa = 0.072 Ra1 – 0.017 Ra2 – 0.002 Ra3 + 0.503
r2 = 0.833, sey = 0.11 ton/ha
Rice yield estimation (ASAR APP, 2007)
YRa : rice yield (kg/m2),Ra1 : polarisation ratio of first date image,Ra2 : polarisation ratio of second date image,Ra3 : polarisation ratio of third date image,r2 : the coefficient of determination,sey : the standard error for the y estimate.
Summer Autumn 2007 rice crop
Rice yield map in Cho Moi (ASAR APP, 2007)
3.2
Rice cropStatistical
data(Ton)
Estimated Production
(Ton)
Percentage error (%)
WS 2007 131,595 106,128 -19.4
SA 2007 79,256 81,820 3.2
36
What we learned from previous researches
The radar backscattering behaviour of rice is much different from that of the traditional rice plant and other LULC classes;
The temporal changes of radar backscattering of HH and VV are different during rice growing stages;
HH/VV ratio of the single-date Envisat-ASAR APP and TerraSAR-X SM image acquired in the middle period of the crop season is a good rice classifier;
The results using ASAR APP and TSX SM data acquired at a single date have provided a high accuracy of planted rice areas, and three acquisition dates are sufficient to mapping cropping systems during a year (triple crops);
The study also pointed out that at least three-date SAR data (TerraSAR-X SM, Envisat-ASAR APP) can be used to estimate the rice yield and finally rice production by using statistical model (multi linear regression analysis).
37
Ongoing and further works
Integrated system of remote sensing, GIS and mathematical model for assessing climate change in Southern Vietnam (08/2013-02/2016, MOST – National level)
Nghiên cứu xây dựng hệ thống tích hợp viễn thám, GIS và mô hình toán trong đánh giá biến đổi khí hậu khu vực phía Nam Việt Nam.
Utilisation of satellite imagery VNREDSat-1 and equivalent for monitoring agricultural land cover/land use of the Mekong Delta, Vietnam in the context of socio-economic development and climate change (07/2013-12/2015, MOST – National level)
Ứng dụng ảnh vệ tinh VNREDSat-1 và tương đương nghiên cứu giám sát hiện trạng sử dụng đất nông nghiệp khu vực Tây Nam bộ phục vụ phát triển kinh tế - xã hội và ứng phó với biến đổi khí hậu.
Rice crop monitoring in the Mekong Delta, Vietnam (07/2013-06/2015, SAFE project);
.
38
Asia RiCETechnical Demonstrator Site – Mekong Delta, VN
Technical/Implementation Agency: Vietnam Academy of Science and Technology (VAST), Space Technology Institute (STI), and Ho Chi Minh Institute of Resources Geography (HCMIRG)
Links to Existing Agricultural Authorities: Department of Agriculture and Rural Development (DARD) in An Giang Province
Summary of Expected Outputs and Benefits:
From the results of the Asia-RiCE project, remote sensing methods in more accurate and reliable are expected to apply for monitoring of rice crop in practice. Accuracy of rice area and rice production estimates is improved. Such methods will be used to combine with current in-situ methods in order to support agricultural managers and planers at local to national level.
Resources Required: capacity building program and multi-dimension RS data
42
SAFE Prototype
Title of the Prototype: RICE CROP MONITORING IN THE MEKONG DELTA, VIETNAM
Executing Agency: HoChiMinh City Institute of Resources Geography (HCMIRG) – VAST
End-User Agency: Department of Agriculture and Rural Development of An Giang province
Purpose of the Prototype
Monitoring of rice cropping area and rice growth;
Estimating rice yield and production.
Expected Outputs of the Prototype
Rice distribution maps of the Vietnam’s Mekong Delta using SAR and optical data;
Rice yield estimation map of An Giang province for one district using SAR data.
43
SAFE Prototype Time Period and Milestones for Prototype Implementation
Time Period: From July 2013 to June 2015
Mid-Term Report#1 (6th Month): Collecting and analysing data.
Mid-Term Report#2 (12nd Month): Developing processor of rice mapping and establishing rice yield estimation model.
Final Report (24th Month): Completion of the Prototype.
44
SAFE Prototype Work Plan of the Prototype Activity
Establish distribution map of rice area
Collecting and analysing data;
Developing methods for mapping rice area using SAR data (ALOS PALSAR, ALOS-2, Cosmo-Skymed, TerraSAR-X, RADATSAT-2, RISAT-1, etc.) and optical data (VNREDSat-1, Landsat 8, MODIS, SPOT VGT, etc.);
Developing crop calendar using high-frequent revisit data (MODIS);
Assessing rice mapping methods;
Establishing GIS database.
Rice yield prediction model
Surveying and measuring field data of rice parameters;
Analysing radar data to establish relationships between rice parameters (such as biomass) and backscattering coefficients;
Establishing rice yield estimation models such as regression models relating rice yield to a combination of backscattering coefficients or to optical indicator (NDVI);
Estimating rice yield harvested according to crops;