Development%20of%20a%20combined%20crop%20and%20climate%20forecasting%20system

1
Development of a combined crop and climate
forecasting system
Tim Wheeler and Andrew Challinor t.r.wheeler_at_rdg.a
c.uk

Crops and Climate Group
2
A combined crop and climateforecasting
systemReport fromFood Crops in a Changing
Climate
3
Linking climate informationto crop models
general circulation model
crop model
At what scale should information pass between
crop and climate models?
4
Development of a combinedcrop / climate
forecasting system
Fully coupled crop-climate simulation
Find spatial scale of weather-crop relationships
Osborne (2004)
Challinor et. al. (2003)
Ensemble methods
Climate change
(Challinor et al, 2005b,c)
(2005c,d)
Crop modelling at the working spatial scale
Hindcasts with observed weather data
Challinor et. al. (2004)
(Challinor et al, 2004)
and reanalysis
(Challinor et al, 2005a)
5
Simple correlations betweenrainfall and crop
yield
Seasonal rainfall and groundnut yields for all
India. Time trend removed. rainfall yield
6
Patterns of seasonal rainfall and yield of
groundnut in India
District level groundnut yields (kg ha-1) Mean of
1966 - 1990 Data source ICRISAT
7
Patterns of seasonal rainfall and yield of
groundnut in India
Sub-divisional level seasonal rainfall (JJAS, cm)
Mean of 1966 - 1990 Data source IITM
8
General Large Area Modelfor Annual Crops (GLAM)

• Aims to combine
• the benefits of more empirical approaches (low
  input data requirements, validity over large
  spatial scales) with
• the benefits of a process-based approach (e.g.
  the potential to capture intra-seasonal
  variability, and so cope with changing climates)
• Uses a Yield Gap Parameter to account for the
  impact of differing nutrient levels, pests,
  diseases, non-optimal management to simulate farm
  yields

Challinor et. al. (2004)
9
Hindcasts of groundnut yield for all India using
GLAM
10
Capturing the effects ofintra-seasonal
variability
1975 Total rainfall 394mm Model 1059 kg/ha Obs
1360 kg/ha
1981 Total rainfall 389mm Model 844 kg/ha Obs
901 kg/ha
11
Using ERA40 reanalysis data

• Gujarat bias correction of climatological mean
  rainfall works well
• Correlation with observed yields 0.49 ? 0.60
• Andhra Pradesh simulated mean yield lt observed,
  variability gtgt observed
• Incorrect seasonal cycle (both mean and
  variability) though Jun and Sept good. This is
  harder to correct.

12
Using probabilistic climate forecasts
Model average
63 ensemble members
713 kg ha-1
Observed
775 kg ha-1
Use of DEMETER multi-model ensemble for groundnut
yield in Gujarat, 1998 from Challinor et al
(2005)
13
Probabilistic forecasting of crop failure

• The number of ensemble members predicting yield
  below a given threshold is an indication of
  probability of occurrence
• Found predictability in crop failure

14
The impact of water and temperature stress at
flowering under climate change
Hadley Centre PRECIS model, A2 (high emission)
scenario
1960-1990
2071-2100
1 no impact
0 max. impact
Groundnut

• Current risk is dominated by water stress in
  the future climate run temperature stress
  dominates in the north.

15
Variety response to temperature stress alone
under climate change
Hadley Centre PRECIS model, A2 (high emission)
scenario 2071-2100 Number of years when the total
number of pods setting is below 50.
Sensitive variety
Tolerant variety
16
An integrated approach to climate impact
assessments

• Crops can modify their own environment
• The water cycle and surface temperatures vary
  according to land use
• Integrate biological and physical modelling
• By working on common spatial scale
• By fully coupling the models

17
Fully coupled crop-climate simulation
Crops growing in HadAM3
18
Fully coupled crop-climate simulation
19
Using satellite estimates of rainfall
TAMSAT Teo Chee-Kiat David Grimes
20
Conclusions

• A combined crop and climate modelling system has
  been developed and tested for the current
  climate.
• It shows skill in seasonal hindcasts and with
  climate ensembles
• It has been used to study crop responses to
  climate change
• Can be fully coupled to a GCM, and driven by
  satellite data