Farmland Drought Assessment Based on the Assimilation of Multitemporal SAR into AquaCrop Model(ID: 10448)
Hao Yang1, Guijun Yang1, Xiuliang Jin1, Stefano Pignatti2, Raffaele Casa3, Simone Pascucci2 , Paolo C. Silvestro3
1NERCITA , Beijing Academy of Agriculture and Forestry Sciences (CHINA) 2CNR-IMAA (ITALY) 3DAFNE, University of Tuscia (ITALY)
Precipitation Anomaly
Soil Anomaly
Crop Anomaly
Driving Factor System Response
Farmland Drought Development
Drought Loss Evaluation
Drought Risk Assessment
Farmland Drought Assessment
Farmland Drought Monitoring
Key Variables Monitoring
Shaanxi (Northwest of China)
Beijing (North China)
Inner Mongolia (Northeast of China)
Experiment Campaign
2 3
1
Project Overview
Biomass retrieval by SAR data
Assimilation of AquaCrop model
Our work
Content
Study Area: Shangkuli Farmland, 3000 hectares typical agricultural land use of Northeast China > 18 hectares, each field
1. Biomass estimation by multi-temporal compact SAR data
SAR data: 5 Radarsat-2, C-band, 8m
5
Sowing Elongation Flowering Filling Ripening
Ground truth data: Crop/Soil/Environment
𝑺𝑺 Sinclair matrix
𝒌𝒌𝒍𝒍, 𝒌𝒌𝒑𝒑Scatteringvectors 𝑪𝑪 Covariance matrix 𝑻𝑻 Coherency matrix
𝐶𝐶 = 𝑘𝑘𝑙𝑙𝑘𝑘𝑙𝑙∗𝑇𝑇
𝑇𝑇 = 𝑘𝑘𝑝𝑝𝑘𝑘𝑝𝑝∗𝑇𝑇
𝑘𝑘𝑙𝑙 = 𝑆𝑆ℎℎ 2𝑆𝑆ℎ𝑣𝑣 𝑆𝑆𝑣𝑣𝑣𝑣𝑇𝑇
𝑘𝑘𝑝𝑝 =12𝑆𝑆ℎℎ + 𝑆𝑆ℎℎ 𝑆𝑆ℎℎ − 𝑆𝑆ℎℎ 2𝑆𝑆ℎ𝑣𝑣 𝑇𝑇
Fully polarization SAR (FP)
H V H V
H V H V
H V H V 𝑺𝑺 =
𝒔𝒔𝒉𝒉𝒉𝒉 𝒔𝒔𝒉𝒉𝒗𝒗𝒔𝒔𝒉𝒉𝒗𝒗 𝒔𝒔𝒗𝒗𝒗𝒗
Scattering matrix Trans
Rec
2 PRF
CP CP
X X
Y Y 𝒌𝒌𝑪𝑪𝑪𝑪 = 𝑪𝑪𝑪𝑪𝑪𝑪
𝑪𝑪𝑪𝑪𝑪𝑪
Scattering vector
rece
ptio
n
reception
V
H
RC transmit reception
reception
V
H
transmit
RC
receive
RC transmit
LC receive π/4 𝐷𝐷𝐶𝐶𝐷𝐷 𝐶𝐶𝑇𝑇𝐶𝐶𝐶𝐶
Mode Transmit Receive 1 Receive 2 𝑘𝑘𝐶𝐶𝐶𝐶
π/4 45° H V 𝑆𝑆ℎℎ + 𝑆𝑆ℎ𝑣𝑣 𝑆𝑆𝑣𝑣𝑣𝑣 + 𝑆𝑆ℎ𝑣𝑣 𝑇𝑇/ 2
DCP RC RC LC 𝑆𝑆𝑅𝑅𝑅𝑅 𝑆𝑆𝑅𝑅𝑅𝑅 𝑇𝑇
CTLR (Hybrid or π/2) RC H V 𝑆𝑆ℎℎ − 𝑖𝑖𝑆𝑆ℎ𝑣𝑣 −𝑖𝑖𝑆𝑆𝑣𝑣𝑣𝑣 + 𝑆𝑆ℎ𝑣𝑣 𝑇𝑇/ 2
Trans
Rec
X
Y
1 PRF
Compact polarization SAR (CP)
C2 matrix of CP
Stokes Matrix
Scattering vector
Compact SAR simulation by fully polarization Radarsat-2
Model-based polarization decomposition method
DBL ODD VOL Total
ODD = Odd Scattering DBL = Double Scattering VOL = Volume Scattering
Freeman decomposition for FP data
Polarization decomposition method for CP SAR
m-delta (Raney, TGRS2007) m-chi (Raney, TGRS2012)
DBL ODD VOL
Freeman
m-chi
m-delta
Single-Pol.
Dual-Pol.
Quad-Pol.
Ground photos in different Day After Sowing
The same growth pattern was found before crop ripening between m-chi-Dbl and crop biomass.
Evolution of measured dry biomass Evolution of m-chi-Dbl
The relationship between m-chi-Dbl and biomass
3 fields (red) in ripening stage
its vegetation water content become low
Conclusion
Propose a method of estimating crop biomass by CP SAR
Demonstrate the importance of polarimetry in crop monitoring by
camparing with traditional single-pol and dua-pol
CP revealed great potential in crop quantitative monitoring when
considering it can reach the level of Quad-pol and it has more imaging
width and less data volume
Motivation • Improving Water Use Efficiency (WUE)
• preventing farmland drought • saving water resource
• Yield loss assessment due to drought • boosting crop production in arid regions
Objective • Yield and WUE estimation in regional scale by combining:
• Remote sensing • AquaCrop model • Data assimilation (PSO)
𝑊𝑊𝑊𝑊𝑊𝑊 =𝑌𝑌𝑖𝑖𝑌𝑌𝑌𝑌𝑌𝑌∑𝑊𝑊𝑇𝑇
2. WUE estimation by AquaCrop and data assimilation
1. Experiment and data
2. AquaCrop Model
3. Global sensitivity
4. CC or Biomass retrieval
5. PSO assimilation
PSO Assimilation
Simulated Biomass
Input (Climate,soil …)
Output (Yield…)
AquaCrop Model
Sensitivity analysis
Retrieved Biomass
Method
Time Growth stages
05/03/2014 Regreening
29/03/2014 Jointing
22/04/2014 Heading
16/05/2014 Filling
09/06/2014 Mature
(1). Experiment Campaign
• Satellite data: – Radarsat-2 SAR – HJ-1 CCD
• Ground data: – Crop – Soil – Climate…
Agronomic management
(variety; sowing dates…)
Soil (hydraulic properties; fertility)
Weather data AQUACROP
• AquaCrop is new Crop Model to simulate yield response to water (Steduto 2009);
• Suited for predicting crop productivity, water requirement, and WUE under water-limiting conditions
(2). AquaCrop Model
Climate data
Management data Soil data
Crops data
• Data collection in Yangling site (China); • AquaCrop have been localized for wheat in China in our previous research (PLOS, 2014)
*.CLI file *.IRR file *.ETo file *.CRO file *.SOL file …
• Objective: • determine the most important
crop parameters • reduce the used parameters
number • Method:
• Extended Fourier Amplitude Sensitivity Test (EFAST)
• assess the contribution of different crop parameters to model output
(3). Global sensitivity analysis for AquaCrop
• 2 indicator: – First order sensitivity index (FOSI) – Total sensitivity index (TSI)
• 40 crop parameters in AquaCrop – ±10% – ±30% – ±50% fluctuations
• 4 variables in AquaCrop output – Static variables : Yield, Maximum
dry biomass – Dynamic variables : Canopy Cover,
Dry biomass
TSI results for yield FOSI result for yield
TSI results for time series dry biomass FOSI results for time series dry biomass
Output variable Parameter range Sensitive parameters
Yield 10% fluctuation cc, polmn,wp, stbio, hi, psto
30% fluctuation cc, wp, remd, hi, stbio, rootdep, mcc, polmn, cgc
50% fluctuation cc, polmx, wp, hi, stbio, mcc, remd, cdc, pstoshp
Maximum dry
biomass
10% fluctuation wp, cc, stbio, rootdep, polmn, mcc, psto
30% fluctuation wp, cc, stbio, mcc
50% fluctuation wp, cc, stbio, mcc
Sensitivity analysis results (TSI) under different parameter variation ranges
Results: (1) EFAST is OK (2) TSI > FOSI (3) Four variables show consistency for FOSI, but show difference for TSI;
Canopy cover and biomass estimation by
• A new combined VI was developed to estimate Canopy cover and Biomass by SAR and Optical data
Radarsat-2 HJ-1
05/03/2014 04/03/2014
29/03/2014 07/04/2014
22/04/2014 29/04/2014
16/05/2014 20/05/2014
(4). CC and Biomass retrieval from RS
Note: n=80 for modeling, n=40 for validation
(5). PSO assimilation
Method: Particle Swarm Optimization (PSO) State variables: CC or Biomass Parameters:
Result (1)
CC and Biomass estimation results by PSO method
Canopy Cover Biomass
It shows good consistency between the predicted and the measured for CC and Biomass
Result (2)
Yield WUE
Yield and WUE estimation results by PSO method (biomass as the state variable in assimilation)
Result (3) Yield WUE
Yield and WUE mapping in Yangling by PSO method (biomass as the state variable in assimilation)
Conclusion • PSO method which assimilates the Remote Sensing observation and AquaCrop
model can be used to estimate the yield and WUE in regional scale; • Biomass is more suitable than CC as a state variable in PSO assimilation;
– The estimation result of yield based on Biomass (R2=0.42,RMSE=0.81ton/ha, nRMSE=17.05%) is better than that based on CC (R2=0.31, RMSE=0.94ton/ha and nRMSE=19.79%);
– WUE result based on Biomass is also better;
• This study provide a method for monitoring the yield and WUE in regional scale by combining the AquaCrop model and RS observation;
• This study provide a guideline for improving the irrigation management of winter wheat in arid regions.
Cooperation
Cooperation
1. Hao Yang, Zengyuan Li, Erxue Chen, Chunjiang Zhao, Guijun Yang, Raffaele Casa, Stefano Pignatti, Qi Feng. Temporal Polarimetric Behavior of Oilseed Rape (Brassica napus L.) at C-Band for Early Season Sowing Date Monitoring, Remote sensing, 2014,6(11):10375-10394.
2. Hao Yang, Erxue Chen, Zengyuan Li, Chunjiang Zhao, Guijun Yang, Stefano Pignatti, Raffaele Casa, Lei Zhao. Wheat lodging monitoring using polarimetric index from RADARSAT-2 data, International Journal of Applied Earth Observation and Geoinformation,2015,34(1):157-166.
3. Hao Yang, Chunjiang Zhao, Guijun Yang,Zengyuan Li, Erxue Chen, Lin Yuan, Xiaodong Yang, Xingang Xu. Agricultural crop harvest progress monitoring by fully polarimetric synthetic aperture radar imagery. Journal of Applied Remote Sensing, 2015, 9(1), 09607, 1-11.
4. Assessing water use efficiency in winter wheat by using the AquaCrop model with remote sensing data. Agricultural Water Management. (Under Review)
5. A new optical and radar combination vegetation index for estimating LAI and biomass of winter wheat using HJ and RADARSAT-2 data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. (Under Review)
6. Estimation of water use efficiency for winter wheat based on multi-source remote sensing data and AquaCrop model using particle swarm optimization algorithm. Remote Sensing of Environment. (Under Review)
Co-Papers
Acknowledge
Our work was supported by Dragon III, also in part supported by the Chinese National Science and Technology Support Program under grants 2012BAH29B00 and by the Chinese State Key Basic Project under grants 2013CB733404.
Thanks for your attention!