Transcript crop and water resource modelling - START

T2.32: Climate Outlooks and Agent-Based Simulation of Adaptation in Africa
Richard Washington (School of Geography and the Environment, University of Oxford), Mark New (SoGE), Thomas E. Downing (Stockholm Environment Institute), Mike Bithell (Department of Geography, Cambridge University), Alex Haxeltine (Tyndall Centre for Climate Change Research)
Bruce Hewitson (University of Cape Town), Chris Reason (UCT), Roland Schulze (Natal University), Coleen Vogel (University of Witwatersrand), Emma Archer (IRI and UCT), Edmund Chattoe (Department of Sociology, University of Oxford), Gina Ziervogel, SoGE Oxford and SEI), Sukaina Bharwani (SEI) Matt Swann (SoGE Oxford)
Agent Based Social Simulation
Fieldwork in Southern Africa
Introduction
STAKEHOLDER ENGAGEMENT AND VULNERABILITY
ASSESSMENT
We approach climate change adaptation as a learning
process, in that the development of adaptive capacity
to climate forecasts on shorter time-scales enhances
adaptive capacity on all time-scales, extending from
seasonal and interannual to decadal and beyond.
Climate change and climate variability are closely
linked in the operation of the climate system. Some of
the largest impacts of climate change may arise
through the superposition of more intense forms of
existing modes of variability on an underlying global
warming trend.
In previous research:
Stakeholders were interviewed on their use of climate
information
Potential impacts of climate change on agriculture, water and
vulnerability of small farmers was documented
The case study was selected: Mangondi village in Limpopo
Province
See Archer, E. 2002 Identifying Underserved End-User Groups
in the Provision of Climate Information, Bulletin of the American
Meteorological Society (Submitted).
STAKEHOLDER INTERVIEWS AND FIELD WORK IN SOUTH
AFRICA
It is proposed that adaptation strategies geared to
cope with large climate anomalies would embrace a
large proportion of the envelope of change expected
from long term climate change. This idea has become
‘conventional wisdom’, endorsed, for example by
WMO CLIPS and the World Bank. However, it has not
been tested in a rigorous manner.
Requirements from fieldwork:
Construction of adaptation-decision making scenarios. The
suite of socio-economic futures constructed for greenhouse gas
emissions (the SRES scenarios) will be supplemented by
southern African development scenarios
Map of social networks among farmers that govern
dissemination of information
AGENT-BASED SOCIAL SIMULATION
Objectives:
oTranslate stream of climate outlooks into
forecasts
oTrace dissemination and use of forecasts
oModel decision-making among smallholder
farmers
Design based on:
oPrototype
oReview by project team and additional experts
oDiscussions with local stakeholders
Model used in two modes:
oOptimum forecast mode, using ‘perfect’ forecasts
and responses
o‘Realist’ mode degrades optimum chain, with
imperfect, probabilistic forecasts. Decision-making
reflects social processes and perception
This project will develop an innovative modelling
framework that integrates social responses to climate
events and climate predictions on a continuum of timescales, thereby enabling the exploration of adaptation
as a learning process.
The methodology will be applied using a southern
African case study because Africa is arguably the
continent most vulnerable to climate change, and
southern Africa is an area where seasonal climate
prediction is already operational (Washington and
Downing, 1999).
Clockwise from above: mixed crops/fallow in Mangondi
village; Emma Archer, Gina Ziervogel and Tom Downing
are shown maize crops by a local farmer; Gina Ziervogel
and a local farmer.
Climate Outlooks
Summer sea surface temperature correlations with Mangondi region rainfall, from the Hadley
Centre climate model (HadCM3), forced with greenhouse gas increases of 1% per annum.
Growth of stakeholder trust when poor forecasts
damage the trust level. As the number of failed wet
year forecasts increases, the mean level of trust
starts to decline. Note the long timescale over
which this takes place. The curves shown are
means over 500 climate sequences. Scatter about
the mean is of the same order as the mean itself.
AGENT-BASED
SOCIAL
SIMULATION
These figures show the mean correlations between JFM Nino3 and southern African JFM
rainfall for active (left) and inactive (right) 30 year ENSO periods. The 30 year periods are
obtained from all the climate change runs of HadCM3 used in the study, and the correlation
coefficients for active and inactive periods are averaged at each gridbox. There are six
active and nine inactive periods. During active ENSO periods (similar to the last 30 years
in the observed record), El Nino events are associated with negative rainfall anomalies in
southern Africa, but this link weakens when ENSO is less active (comparable to 1941-70 in
the observed record). The implications for seasonal forecasting and decadal-scale planning
are significant. El Nino is not always as powerful a predictor of southern African rainfall as
it is currently. This highlights the need for adaptation strategies on a continuum of
timescales.
INTEGRATED FRAMEWORK AND OUTPUT
INDICATORS
The three elements of the project are brought
together with a consistent user interface for easy
access. Outputs include
Time series of climate outlooks
Characteristics of adaptation pathway scenarios
A vulnerability profile
Indicators such as the water poverty index being
developed for DFID by the University of Natal.
DETAILED
CLIMATE
OUTLOOKS
CROP AND
WATER
RESOURCE
IMPACT
MODELLING
6
5.5
5
4.5
9 per. Mov. Avg. (CON)
4
9 per. Mov. Avg. (B1a)
3.5
9 per. Mov. Avg. (B2a)
9 per. Mov. Avg. (A2a)
3
9 per. Mov. Avg. (A1f )
2.5
2
19
60
19
64
19
68
19
72
19
76
19
80
19
84
19
88
19
92
19
96
20
00
20
04
20
08
20
12
20
16
20
20
20
24
20
28
20
32
20
36
20
40
20
44
20
48
20
52
20
56
20
60
20
64
20
68
20
72
20
76
20
80
20
84
20
88
20
92
20
96
Rain (mm/day)
PREPARATION OF BASELINE
CLIMATE DATA AND CLIMATE
OUTLOOKS ACROSS A
RANGE OF PLANNING
HORIZONS
Development of streams of
climate data that provide
probabilistic climate outlooks on
three time scales
oInterannual (associated
with seasonal forecasting)
oMulti-decadal
oLong-term climate change
The streams will drive resource
models and agent-based
simulation on the three time
scales
A range of model integrations
will be used, including UKMO,
incorporating both dynamic and
empirical downscaling of
ensemble model runs
Multi-decadal variability and
climate change (2099)
timescales will be sourced from
the IPCC data distribution centre
and the LINK Project at the
University of East Anglia Climatic
Research Unit.
The potential impact of forecast adoption on a
community for which rainfall fluctuations may lead
to regular food shortages has been modelled. The
figure shows the cost (in tonnes of grain) that a
farmer would have to bear as a result of years
where stored food levels drop to zero. Plotted are
mean cumulative costs as a function of forecast
skill, with standard error generated from 500
simulated climate sequences, for a household of
eight people farming a single hectare field for 50
years. Typical normal grain yield for this region is
1 tonne per hectare. The dotted horizontal line and
blue cross show the no forecast case. In red is the
case where the incorrect forecasts are always as
poor as possible, and in green those where forecast
is incorrect by two terciles at most 10% of the time.
Early results suggest that where the rainfall does
not fall in the forecast tercile more than 60% of the
time, farmers may be worse off using the forecast
than ignoring it.
PROTOTYPE AGENT BASED MODEL
The team has built a prototype model that will address:
oWays in which information is passed between
model elements, including farmers, chiefs,
extension agents and meteorological offices
oDesign of the user interface
oAspects of the core model
Outputs include figures 1 and 2 (right), detailing the
possible effect of forecast implementation on costs to
farmers (figure 1) and levels of trust in the forecasts
(figure 2).
The prototype model is being extended, and is
currently being developed to include a more
sophisticated crop model, more detailed climate inputs,
and further economic and demographic components.
Northern Province Area Average JFM precipitation (9 year moving average). HadCM3 control run and climate
change runs.
CROP AND WATER RESOURCE MODELLING
The principal impacts of climatic variations for
smallholder farmers are effects on agriculture and
water resources
The FAO model ‘CropWat’ has been adapted to run
together with the agent-based simulation, providing
information on a variety of crops
STAKEHOLDER WORKSHOP AND
PARTICIPATORY EXERCISES IN SOUTH
AFRICA
The integrated framework of climate outlooks,
ABSM and a user interface will be reviewed and
assessed in South Africa
The assessment will include:
oParticipatory exercise in the selected village
oWorkshop within the regional SARCOF
meeting to show the approach and elicit
comments from experts
oExtended training and demonstration within
Wits University to discuss ways to take the
methodology forward.
References:
Archer, E.R.M. 2002 Identifying Underserved End User Groups in the Provision of Climate
Information: Bulletin of the American Meteorological Society (submitted)
Washington, R. and Downing, T.E. 1999: Seasonal Forecasting of African Rainfall:
Geographical Journal, 165 pp255-274
Richard Washington
School of Geography and the
Environment
Mansfield Road
Oxford, OX1 3TB, UK
richard.washington@geograp
hy.ox.ac.uk
Mark New
School of Geography and the
Environment
Mansfield Road
Oxford, OX1 3TB, UK
[email protected].
uk
Thomas E. Downing
Stockholm Environment
Institute, Oxford Office
10B Littlegate Street
Oxford OX1 1QT, UK
[email protected]
Mike Bithell
Department of Geography
Downing Place
Cambridge CB2 3EN
UK
[email protected]
k
Alex Haxeltine
The Tyndall Centre for
Climate Change Research
School of Environmental
Sciences
University of East Anglia
Norwich, Norfolk, NR4 7TJ,
UK
[email protected]