BioMA workgroup (Multi-model crop yield estimates) Roberto Confalonieri & Marcello Donatelli [email protected] - .

BioMA workgroup

(Multi-model crop yield estimates)

Roberto Confalonieri & Marcello Donatelli

[email protected] - www.robertoconfalonieri.it

E-AGRI kick-off meeting, 24-25 March 2011, VITO (Mol, Belgium)

WP 3

• Multi-model approach to crop yield estimates

• Simulation of the impact of diseases (!!!) and abiotic damages on crop productivity

• Dynamic forcing crop models state variables using exogenous information, i.e., NDVI or NDVI derived leaf area index (only for rice)

• Provide the statistical tool (generating the forecast using simulated and remote sensed data, and historical yield series) with different typologies of simulated information:

same state variables (simulated indicator) simulated using different models

simulated state variables without forcing (to be statistically post-processed together with remote sensing indicators) and the same variables resulting by dynamic forcing of crop models (to be post-processed alone)


WP 3 tasks

• Task 3.1: Ground data collection for BioMA

• Task 3.2: Adaptation of BioMA for multi-model rice monitoring in China

• Task 3.3: BioMA piloting for multi-model rice monitoring and yield forecasting in JIANGHUAI Plain, China

• Task 3.4: Adaptation of BioMA for multi-model wheat monitoring in Morocco

• Task 3.5: BioMA piloting for multi-model wheat monitoring and yield forecasting in Morocco


Task 3.1

Task 3.2

Task 3.3

Task 3.4

Task 3.5

Rice, China Wheat, Morocco

BioMA adaptation

BioMA piloting

Task 3.1 descriptionTask leader: JAAS; partners: JAAS, INRA

• Activity 3.1.1: Identification of the group of cultivars to be calibrated for the BioMA crop models (WARM, CropSyst, WOFOST)

• Activity 3.1.2: Identification of measurable key variables and parameters needed for a robust calibration of the BioMA models

• Activity 3.1.3: Collection of data (i) for each group of cultivar [3.1.1], (ii) for suitable variables [3.1.2], (iii) for different combinations site year

• Activity 3.1.4: Development of a database for the parameterization and calibration activities according to specifications provided by Task 3.2


Task 3.1

Task 3.2

Task 3.3

Task 3.4

Task 3.5


BioMA adaptation

BioMA piloting

Tasks 3.2 & 3.4 descriptionTask leaders: UNIMI, JRC; partners: UNIMI, JRC

• Activity 3.2(4).1: Spatially distributed sensitivity analysis of the BioMA models to identify the most relevant parameters

• Activity 3.2(4).2: Parameters calibration for each model and group of cultivars

• Activity 3.2(4).3: Evaluation of the BioMA models for field-scale simulations for each group of cultivars

• Activity 3.2(4).4: Evaluation of the BioMA models for large-area simulations using official yield statistics


Task 3.1

Task 3.2

Task 3.3

Task 3.4

Task 3.5


BioMA adaptation

BioMA piloting

Tasks 3.3 & 3.5 descriptionTask leaders: UNIMI, JRC; partners: UNIMI, JRC, JAAS, INRA

• Activity 3.3(5).1: Evaluation of the suitability of the BioMA platform for rice/wheat monitoring and yield forecasts in China/Morocco

• Activity 3.3(5).2: Evaluation of the usefulness of the multi-model approach for monitoring and forecasting activities

• Activity 3.3(5).3: Evaluation of possible improvements in monitoring and forecasting capabilities due to the injection in the models of exogenous data (i.e., forcing state variables using NDVI or LAI data)


Task 3.1

Task 3.2

Task 3.3

Task 3.4

Task 3.5


BioMA adaptation

BioMA piloting

Models

• Six modelling solutions will be developed and evaluated within E-AGRI

Rice in China: multi-model simulations with and without forcing the models with RS data

WARM (Confalonieri et al., 2010)

WOFOST (Van Keulen and Wolf, 1986)

CropSyst (Stöckle et al., 2003)

Wheat in Morocco: multi-model simulations

WOFOST

CropSyst


Models calibration (1)

• Model parameters will be calibrated for different groups of rice and wheat varieties with similar morphological and physiological features

• Different models would need different measured data (parameters, state variables, driving variables, etc.) for a rigorous calibration, e.g., CropSyst needs specific leaf area (SLA) at emergence, WARM needs SLA at emergence and at mid-tillering, WOFOST would need SLA periodically (about 10-12 times during the crop cycle length)

• Different production levels (e.g., potential, water limited) would need different data for the calibration activities



• The first steps in BioMA adaptation for rice in China and wheat in Morocco will be:

the identification of groups of similar varieties, of their spatial distribution, and of information on their management[JAAS and INRA will be in charge for this]

definition of protocols for data collection to allow the creation of databases with measured data suitable for the calibration of the parameters of all the models[UNIMI will provide the protocols]



• The second step is to understand which are the most relevant parameters for each model (those on which to concentrate the effort during the calibration) under the conditions explored

• This will be carried out by:

running spatially distributed sensitivity analysis

comparing the spatial patterns of

the groups of cultivars and

the sensitivity analysis results


Example of spatiallydistributed sensitivity analysis


Example of spatiallydistributed sensitivity analysis


Models evaluation

• Once key parameters will be set to values derived from measurements, the others will be calibrated using automatic optimization algorithms

• Calibrated parameters will be validated using independent sets of field observations

• Models performances for yield estimates will be evaluated at administrative level using historical yield series

by directly comparing simulated and official yields

by comparing post-processed (statistical tool) simulated results and official yields (cross-validation)

• Performances of the different models (with and without dynamic forcing for rice) will be evaluated using different evaluation metrics


Piloting BioMA

• BioMA can be easily deployed:

no installation is required

a data-layer component allows for interfacing BioMA with whatever typology of database (Oracle, Access, …) and with whatever typology of database structure

• Integrated tools for zonation and automatic calibration are available in BioMA, thus allowing people in China and Morocco to autonomously refine calibrations in case further observations become available

• Tools for data display (maps, different typology of charts, etc.) are integrated, showing also agroclimatic indices, thus favouring the work of the analyst


Activity planning

GANTT diagram for the whole WP 3


ID Activities2011 2012 2013

Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar

1 Ground data collection for BioMA

2Adaptation of BioMA for multi-model monitoring in JIANGHUAI Plain, China

3BioMA piloting for multi-model rice monitoring and yield forecasting in JIANGHUAI Plain, China

4Adaptation of BioMA for multi model wheat monitoring in Morocco

5BioMA piloting for multi-model wheat monitoring and yield forecasting in Morocco

3.1

3.2

3.3

3.4

3.5

Task2014

Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar

Activity planning

GANTT diagram for Task 3.1


ID WP 3.12011 2012 2013 2014

Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar

1

2

3

4

Identification of the cv to be calibrated for BioMA crop models (WARM, WOFOST, CropSyst)

Identification of the parameters needed to calibrate BioMA crop models

Collection of data for each group of cv

Organization of data as required by WP 3.2








3.1

3.2

3.3

3.4

3.5

Task2014


Activity planning

GANTT diagram for Tasks 3.2 and 3.4









3.1

3.2

3.3

3.4

3.5

Task2014


ID WP 3.2 – WP 3.42011 2012 2013 2014


1

2

3

4

Sensitivity analysis of the BioMA models for rice/wheat to identify most important parameters

Calibration of parameters for each cv and crop model

Evaluation of BioMA models for field-scale simulation of rice/wheat

Evaluation of BioMA models for large area simulation of rice/wheat

Activity planning

GANTT diagram for Tasks 3.3 and 3.5









3.1

3.2

3.3

3.4

3.5

Task2014


ID WP 3.3 – WP 3.52011 2012 2013 2014


1

2

3

Testing suitability of BioMA platform for rice/wheat monitoring and yield forecasting

Evaluating the usefulness of the multi-model approach

Evaluating possible improvement of the system via forcing crop models with remote sensed data

…Thank you for your kind attention

BioMA workgroup (Multi-model crop yield estimates) Roberto Confalonieri & Marcello Donatelli [email protected] - .

Documents

bioma task

morocco task

china task

belgium task

tasks task

bioma models activity

adaptation of bioma

description task leaders