crop productivity assessment through remote sensing: radiation-driven and water-driven algorithms,...
TRANSCRIPT
CROP PRODUCTIVITY ASSESSMENT THROUGH
REMOTE SENSING
Christopher M.U. Neale
Isidro Campos
GENERAL FRAMEWORK
• Satellite imagery has been extensively used for crop monitoring and yield
forecasting for over 30 years and plays an important role in a growing number of
operational systems
• The estimation of spatially-distributed biophysical parameters is a good first step
towards the use of remote sensing for crop yield estimation, but more complete,
and complex, approaches have to be developed
• The methodologies developed to date can be described (Rembold et al., 2013)
as:
Regression modelling
Mechanistic and dynamic crop growth models
Assimilation models
Biomass production and partitioning models based on earth
observation (EO) data
GENERAL FRAMEWORK: REGRESSION MODELS
• Data-driven approaches developed with the
base of strong empirical relationships between
biophysical parameters derived from EO and
ground-yield data.
• Necessity of new calibration to be applied at
different location or meteorological conditions
and relative low accuracy for individual plots.
• The use of these approaches for yield
forecasting and global application is limited.
Regression modelling
Mechanistic and dynamic crop growth models
Assimilation models
Biomass production and partitioning models based on EO data
GENERAL FRAMEWORK: ASSIMILATION MODELS
• For these approaches, the crop growth model
inputs and initial conditions that have a large
effect on biophysical parameters are selected.
• Then, those selected inputs and initial conditions
may be ‘‘tuned’’ in order to better fit the estimated
biophysical parameters based on remote sensing
data.
Regression modelling
Mechanistic and dynamic crop growth models
Assimilation models
Biomass production and partitioning models based on EO data
GENERAL FRAMEWORK: BIOMASS PRODUCTION Regression modelling
Mechanistic and dynamic crop growth models
Assimilation models
Biomass production and partitioning models based on EO data
• The driving forces for biomass production are
absorbed photosynthetic active radiation (APAR)
and plant transpiration. Both can be derived
from remote sensing based models
• APAR and Transpiration are related to biomass
production using light use efficiency (LUE) and
water use efficiency (WUE) parameters, with
yield estimated through a yield partitioning factor
• These approaches require the adequate
estimation of both WUE/LUE and the yield
partitioning factors, so previous knowledge about
both factors is needed
(Muñoz-Padilla et al. 2012)
LUE=2.5 g MJ−1 PAR
YPF=0.8
fPAR/APAR
estimates
Crop specific fPAR-VI
relationships and
radiative transfer
models
Yield/Grain
production
ET & water stress
estimates
Hybrid model
based on
ALEXI-DISALEXI
Site/Plot specific
Yield partitioning
Existing models on
Yield partitioning
Biomass
High to Medium resolution RS data: Landsat Constellation
MODIS Constellation, VIIRS, others
Agro-meteorological data Existing ground-networks
and satellite based estimates
OUR PROPOSAL: GENERAL OVERVIEW
Water productivity
Site/Plot specific
LUE
Existing
models on
LUE
Site/Plot specific
WUE
Existing
models on
WUE
Crop coefficient and reference ET: • Reflectance-based crop coefficient models where
vegetation indices are related to ET crop coefficients. Relationships are typically crop specific. Uses shortwave (Visible, NIR) bands of satellite instruments.
Energy balance models: • One layer models examples: empirical models (OLEM),
SEBS, SEBAL, METRIC, SSEBop
• Two-source models, ALEXI-DisALEXI
• Detailed Process models
Energy balance models require the use of both the thermal infrared and the visible/near-infrared bands
Hybrid Methodologies
fPAR/APAR
estimates
Crop specific fPAR-VI
relationships and radiative
transfer models
Yield/Grain
production
ET & water stress
estimates
Hybrid model based on
ALEXI-DISALEXI
Site/Plot specific
Yield partitioning
Existing models on
Yield partitioning
Biomass
Site/Plot specific
WUE/LUE
Existing models on
WUE & LUE
High to Medium resolution EO data: Landsat Constellation
MODIS Constellation
Agro-meteorological data Existing ground-networks
and satellite based estimates
OUR PROPOSAL: GENERAL OVERVIEW
BIOMASS PRODUCTION
• Estimation of Biomass production based on the estimates of crop transpiration
and PAR absorption based on remote sensing models
Biomass=WUE•ΣTr … or … Biomass=LUE•ΣAPAR
o Crop canopy transpiration with the summation over the time period in which the biomass is produced (ΣTr)
o Crop canopy absorbtion of photosyntetic active radiation with the summation over the time period in which the
biomass is produced (ΣAPAR)
o Water use efficiency (WUE)—the slope of the relationship between biomass produced and water transpired by the
crop canopy
o Light use efficiency (LUE)—the slope of the relationship between biomass produced and photosyntetic active
radiation absorbed by the crop canopy
Regression modelling
Mechanistic and dynamic crop growth models
Assimilation models
Biomass production and partitioning models based on EO data
Analyzing the relationship between transpiration & crop biomass
production
Effect of ETo in the Water
Productivity
• Different WP for C4 and C3
crops
• Lower WP under increasing
atmospheric demand
Conclusion:
• Need to normalized for
atmospheric demand
Experimental analysis in
corn (C4) and soybeans (C3) in
eastern Nebraska
Normalization of Water Productivity for
atmospheric demand
Strengths
• Do not depend on crop specie (but
different for C3 and C4 crops)
• Narrow range of variability with respect
to CO2 concentration, phenology and
soil fertility
• Normalized for atmospheric demand
Weaknesses
• Does not consider further reductions of
biomass production depending on
incoming radiation
(Raes et al. 2012)
Analyzing the relationship between transpiration & crop biomass
production
Analyzing the relationship between SAVI (crop transpiration) &
crop production
Experimental analysis in corn (C4) and soybeans (C3) in eastern Nebraska
ESTIMATION OF WUE-LUE
(Sadaña & Pinochet 2014)
Preliminary analysis of LUE based on
HybridMaize & CERES family models
Strengths
• Allows the consideration of the effect of
crop water stress (opportunity to
assimilate RS based stress estimates)
• Differential analysis of carbon fixation
and crop respiration
Weaknesses
• Strong variability depending on crop
type and environment
• Additional effects of the light
compensation and saturation
phenomena
• Difficulty to estimate daily fPAR from
instantaneous measurements-estimates
Wheat
Pea
BIOMASS PRODUCTION
Considering that the driving force for biomass production could
vary depending on the location, environmental conditions and the crop
phenology an integrated analysis of biomass production based on both
WUE and LUE approaches should be considered.
• Ideally, LUE and WUE approaches should provide similar estimates under non-
limiting conditions, this is, non-limiting radiation and water availability. Thus both
approaches can be integrated by using assimilation procedures.
• Under limiting conditions, the most restrictive approach could provide the better
approach.
fPAR/APAR
estimates
Crop specific fPAR-VI
relationships and radiative
transfer models
Yield/Grain
production
ET & water stress
estimates
Hybrid model based on
ALEXI-DISALEXI
Site/Plot specific
Yield partitioning
Existing models on
Yield partitioning
Biomass
Site/Plot specific
WUE/LUE
Existing models on
WUE & LUE
High to Medium resolution EO data: Landsat Constellation
MODIS Constellation
Agro-meteorological data Existing ground-networks
and satellite based estimates
OUR PROPOSAL: GENERAL OVERVIEW
Yield/Grain
production (Y)
ET & water stress
estimates
Based on soil water balance
remote sensing driven and hybrid
model
Site specific
Harvest index (HI)
Existing models on
Yield partitioning and
experimental data
Biomass (B)
Normalized water
productivity (WP*)
Normalization of WUE for
atmospheric demand
High to Medium resolution EO data: Landsat Constellation
MODIS Constellation
Agro-meteorological data Existing ground-networks
and satellite based estimates
GENERAL OVERVIEW: SIMPLIFIED APPROACH
Y=B*HI
• Theoretical approach to Biomass production at canopy scale and plant
transpiration or transpiration coefficient (Kt) (Steduto et al. 2007)
Ktmin=0 (for bare soil) and Ktmax≈mid season Kcb (Raes et al. 2012)
• Well established relationship between SAVI and plant transpiration
This rrelationship is generally established in terms of basal crop coefficient (Kcb),
Kcb=0.15 for bare soil and Kcb=max for SAVI max. These approaches may
overestimate Biomass production for low coverage canopies.
But, González-Dugo and Mateos (2008) proposed a non-linear relationship with
Kcbmin=0 and Kcbmax=mid season Kcb
Analyzing the relationship between SAVI & crop biomass
production
K=0.47; K´=0.55
K=0.55; K´=0.57
Re-analyzing the analytical approach to convert VI in crop
coefficients for irrigation management.
• Non-linear relationships
for both crops
• General good agreement
with moderate differences for
minimum SAVI values
• Need to consider the role of
bare soil in ET rates also in the
absence of plant development
Yield/Grain
production (Y)
ET & water stress
estimates
Based on soil water balance
remote sensing driven and hybrid
model
Site specific
Harvest index (HI)
Existing models on
Yield partitioning and
experimental data
Biomass (B)
Normalized water
productivity (WP*)
Normalization of WUE for
atmospheric demand
High to Medium resolution EO data: Landsat Constellation
MODIS Constellation
Agro-meteorological data Existing ground-networks
and satellite based estimates
GENERAL OVERVIEW: SIMPLIFIED APPROACH
Y=B*HI
Analyzing the relationship between SAVI (crop transpiration) &
crop production
Experimental analysis in corn (C4) and soybeans (C3) in eastern Nebraska
Analyzing the relationship between SAVI & crop production
Experimental values of
Harvest Index
for corn and Nebraska
• Mean values around 0.5 for
both crops and irrigation
managements
• Strong variability depending on
meteorological conditions
• Difficulty to determine
empirically because of the
dynamic of biomass
accumulation
FINAL COMMENTS
• Use of remote sensing to estimate crop biomass appears straightforward for the
case of maize and soybeans. Needs to be further examined for multiple crops
• Remote sensing allows for spatial estimates of biomass and yield
• With concurrent estimates of ET, crop water productivity can be estimated spatially
• For the NENA region, downscaled values of ET and biomass would be used