Presented by Karl M. Rich (with contributions from Jared Berends, Greg Cooper, Chisoni Mumba, Magda Rich, Helene Lie, Kanar Dizyee and Sirak Bahta) at a training course on systems thinking, participatory modelling and value chains, April 2020.
Pests of safflower_Binomics_Identification_Dr.UPR.pdf
Principles of group model building and spatial group model building
1. Principles of group model building and
spatial group model building
Better lives through livestock
O
K
A
PiS
Training course on Systems Thinking, Participatory Modeling, and Value Chains
Materials prepared and presented by Karl M. Rich (with contributions from Jared Berends, Greg Cooper, Chisoni Mumba,
Magda Rich, Helene Lie, Kanar Dizyee, and Sirak Bahta)
Foresight Modeling & Policy Team, Policies, Institutions, and Livelihoods
International Livestock Research Institute (ILRI)
Version April 2020 (draft)
2. 2
Outline
Overview of group model building (and its cousin, mediated
modeling)
Spatial group model building – how it extends GMB and how it’s
different
Examples
• Urban agriculture (NZ)
• East coast fever (Zambia)
• Value chain upgrading - pigs (Myanmar)
• Aggregation systems in horticulture (Bihar)
qualitative
quantitative
3. 3
Session goals
An emerging understanding of (spatial) group model
building and why we use it
An appreciation of the potential of participatory
processes based on previous work
A desire to learn more on implementing SGMB in
practice!
Picture credit: K.M. Rich 2012 (Maroantsetra, Madagascar)
4. 4
Overview
SD models, given their graphical and intuitive nature, can
be developed in collaboration with a variety of groups
Group model building (GMB) is one way to both obtain
information and parameterize relationships that exist in
the system in question.
5. 5
What is group model building?
Group Model Building or GMB
“focuses on building system
dynamics models with teams in
order to enhance team learning, to
foster consensus and to create
commitment with a resulting
decision.”
Source: Vennix (1996)
Picture credit: K.M. Rich 2019 (Myeik, Myanmar)
6. 6
What is group model building?
A participatory process aimed at:
• Identifying and prioritizing the key problems in the system
• The causes of these problems
• The consequences of these problems
• Feedbacks between consequences and causes
• Development of models from these sessions (qualitative or
quantitative)
Use of SD principles and language (stocks/flows/feedbacks)
to facilitate this discussion.
Involvement of stakeholders in the model building process to
increase the effectiveness and ownership of the final
product
Source:
https://commons.wikimedia.org/wiki/File:Kelly%27s_Kin
dergarten_(1898-11-27).jpg, Public Domain
7. 7
GMB and participatory modeling
GMB is not the only type of participatory modeling
technique that uses systems thinking/dynamics tools.
Mediated modeling (van den Belt 2004) is a similar
concept – it involves a wider range of stakeholders in the
process of model building, rather than a smaller client
group (Antunes et al. 2006).
Mediated modeling has been used primarily in
environmental applications.
8. 8
Why GMB?
Messy problems: Problems are usually complex
and not easily defined.
In complex problems, individuals have a limited
(or narrow) view of the problem (silo thinking).
Our mental models are limited by our individual
ability to process information: role of groups
Difficulties in identifying multiple causes of
complex problems their interconnections
Source: Vennix (1996)
Picture credit: S.Bahta 2019 (New Delhi, India)
14. 14
Client
group has
problem
Is SD
appropriate?
Use
preliminary
model?
Yes
No
Use something
else
Questions to consider:
• Is problem dynamic?
• Short vs. long term effects
• Reference mode of behavior
• Qualitative or quantitative
• Who to involve?
Yes: model based on:
No: start from scratch
Informal
interviews
GMB sessions
Conclusions
Interviews
Documents
Questionnaires/
Workbooks
Source: Adapted from Vennix (1996), figure 4.1, p. 103
Note: questionnaires
both inform and
triangulate our GMB
sessions
A way to “short-cut” the process if time or
resources are scarce, but can reduce ownership
of model and power of participatory process
(Vennix 1996)
15. 15
GMB design issues
• Small (5-12) or large (12+) group?
• Type of participants - homogenous vs
heterogeneous groups?
• Level of model complexity needed: problem
conceptualization vs. full model
development with clients/stakeholders?
Picture credit: K.M. Rich 2019 (Jessore, Bangladesh)
16. 16
Limitations of conventional GMB (1)
Lack of use of value chain/natural resource/LDC settings
(an exception: Lie Ph.D., see Lie and Rich 2016)
Issues of comparability/replication (Scott et al. 2016)
Issues of scaling across contexts
17. 17
Limitations of conventional GMB (2)
GMB sessions do not focus on spatial dynamics
However, the processes that generate change within
systems could have important spatial dimensions
(land use, population dynamics, etc.)
The “where” of the system matters as much as the
“what”, “how”, and “why” Picture credit: Ingrid Kallick (http://www.ikallick.com)
/ Public domain; source
https://upload.wikimedia.org/wikipedia/commons/2/
23/SphericalCow2.gif
18. 18
Spatial group model building: GMB with spatial attributes
Key characteristics
• Grounding problems, causes, and
consequences spatially
• Identifying spatial and temporal changes and
their co-evolution
• Using maps and GIS concepts to facilitate
model and system building through physical
platforms such as LayerStack (and eLayerstack
using Vecta) or other related tools
Picture credit: K.M. Rich 2016 (Lincoln, New Zealand)
19. 19
Spatial group model building: toolkits (1)
LayerStack: an offline, participatory GIS-type facilitation platform (funded by KiwiNet)
Use of plastic acetates as data layers (land use, VC actors, climate, disease patterns,
production characteristics) over a map
Use of variety of consumables (stickers, markers) to denote physical location and
temporal/spatial movement
Improves visualization of system and facilitation of model development
Simple, low-tech, hands-on, easy to store information
22. 22
Spatial group model building: toolkits (2)
“Necessity is the mother of all innovation”
COVID-19 has made face-to-face participatory
processes challenging
“eLayerstack” – an online means of conducting
SGMB with groups online using the web-based
Vecta platform (http://vecta.io)
Same principles of layers and consumables to
“draw” on base maps, but in real-time with
stakeholders
Picture credit: K.M. Rich 2020 (online snapshot)
23. 23
Spatial group model building
Process for model development outlined in Rich,
Rich, and Dizyee (2018)
Eight-step process, but flexible depending on use
for qualitative or quantitative modelling.
In the next presentation, we will demonstrate
how we implement in practice (offline and
online).
Picture credit: K.M. Rich 2019 (Dakar, Senegal)
30. 30
Spatial group model building: process (6)
Source: Rich, Rich, and Dizyee (2018)
Note: spatial co-evolution of models remains an important area of
future research; we’re not there yet, but hope to go in that
direction
31. 31
Example #1: urban agriculture (UA) in Christchurch, NZ
UA has a long tradition in Christchurch (WWII,
Vegetable Campaigns)
Since the earthquakes in 2010 and 2011, Christchurch
has experienced a revival in UA
More complex situation:
High prices of fresh produce
Psychological and emotional impact of the
earthquake
UA as a way to reconnect with the city
Divergence between planners and practitioners: role of
SGMB to articulate key spatial issues and leverage
points
Picture credit: M. Rich 2017 (Christchurch, New Zealand)
36. Production
Number of
market outlets
O1 O14
P1 P12
L1 L9
PO1 PO10
Land for urban
agriculture
Population
LEGEND
IN THE MODEL IN THE MAP
Profits
Demand
Distance to market
-
Growth rate of
market outlets
-
Factors to promote
awareness in UA
+
Actual UA
participants
Number of
market outlets
+
Selling to
consumers
+
+
Community &
consumer awareness
of UA
+
+
Production
Land for urban
agriculture
+
+
+
+
+
+
Population
+
+
Production
Number of
market outlets
O11
O11
O12
O1 O14
P4
P5
P6
P7
P8
P1 P12
L4
L5
L6
L7
L1 L9
PO1 PO10
Land for urban
agriculture
Population
LEGEND
IN THE MODEL IN THE MAP
Profits
Demand
Distance to market
-
Growth rate of
market outlets
-
Factors to promote
awareness in UA
+
Actual UA
participants
Number of
market outlets
+
Selling to
consumers
+
+
Community &
consumer awareness
of UA
+
+
Production
Land for urban
agriculture
+
+
+
+
+
+
Population
+
+
Source: Rich, Rich, and Dizyee (2018)
37. 37
Urban agriculture in Christchurch: insights
Spatial dimension of UA in Christchurch
extremely important – land use patterns vs.
population movement patterns.
An opportunity: How to bring UA products from
producers to consumers?
Model remained qualitative – not parameterized
quantitatively
Picture credit: M. Rich 2017 (Christchurch, New Zealand)
38. 38
Example #2: East coast fever in Zambia
ECF – an important livestock disease in East Africa,
including Zambia.
Recent field work (Mumba 2018) highlighted
importance of ECF relative to other government
priorities (e.g. FMD)
Little known about drivers/context of control and how
this differs across space.
How to identify and quantify impact of interventions
that would both improve communal involvement in the
chain and reduce disease?
First “live” test of Layerstack and (qualitative) SGMB in
the field
Picture credit: K.M. Rich 2016 (Monze, Zambia)
43. 43
East coast fever in Zambia: insights
Drivers of ECF have distinct spatial patterns
• Competition between land
• Market differences (external – Lundazi vs. local –
Monze)
• Cultural norms against mixing animals at dip
tanks based on social status/class
• Variations in herding practices
Spatial differences highlight the need for
developing locally relevant, fit-for-purpose control
strategies.
Picture credit: K.M. Rich 2016 (Monze, Zambia)
44. 44
Example #3: upgrading pig value chains in Myanmar
Project at a glance…
Focus: Pro-poor interventions to upgrade pork and rice value chains in Tanintharyi
region, Myanmar
Duration: 5 years, from November 2017 to October 2022
Client: MFAT, Partnership for International Development (PfID) fund
Contract value: NZ$4.1 million.
Partners
World Vision (WV): Contract holder, personnel, resources and logistical support for
project implementation and monitoring
Vision Fund (VF): Micro-finance for households and farmers, new financial products to co-
finance value-adding ventures
Lincoln University & ILRI: Value chain research and design, steer project implementation
and monitoring, impact assessment
Source: Slide courtesy of Berends and Esnard (2020)
45. 45
Process
Three field visits to Myanmar:
• Five pig SGMB workshops (avg. 13
participants and 50% female) with
farmers, brokers, slaughterhouse
owners, and wholesalers
• Five rice SGMB workshops (avg. 14
participants and 40% female) with
paddy farmers, millers, wholesalers
• Six Reference Group workshops (avg. 6
participants) with government officers,
NGO staff, and lead farmers
• Two Project Advisory Committee(PAC)
meetings to review results and decide
on interventions
Source: Slide courtesy of Berends and Esnard (2020)
46. 46
Tools and outputs: Layerstack for VC dynamics
•
•
•HHHH
•
•
•
•
•
•
•HHHH
•
•
•
•
Layer 2: Input, service and product
flows in the pig value chain
Layer 1: Livelihood zones
Source: Slide courtesy of Berends and Esnard (2020); picture credit J. Berends (2019)
47. 47
Tools and outputs: causes and consequences
• Value chain problems prioritised and then explored by developing reference nodes, and cause and consequence maps
• Goal of identifying causal relationships that determine dynamic behaviours in the chain
Source: Slide courtesy of Berends and Esnard (2020); picture credit J. Berends (2019)
48. 48
Tools and outputs: concept model
• Based on common themes and critical feedback loops from cause and consequence mapping, develop concept model
that contains feedback loops and structure which determine dynamic behaviour in the chain
Source: slide courtesy of Berends and Esnard (2020); picture credit J. Berends (2019)
49. 49
Tools and outputs: modules for scenario analysis
• Concept model then divided into
modules
• Each module structure is further
developed in Stella Architect (SD
software package) and
parametrised
• Modules are then connected
through material flows and
information flows to form a
functioning baseline model
Source: Slide courtesy of Berends and Esnard (2020)
50. 50
Tools and outputs: scenarios
Model scenarios
• Baseline: No project interventions
• Scenario 1: Project interventions cover all
pig producers in target villages
• Scenario 2: Project establishes Producer
Groups (PGs) and targets PG members for
interventions
• Scenario 3: PGs are upgraded to Producer
Organizations (POs) with the institutional
arrangements to support ongoing capacity
investments
Interventions (within scenarios)
• Microfinance loans
• Good Animal Husbandry Practices (Animal
Health Workers and biosecurity)
• Training on hybrid pig production and
commercial pig feed
• Artificial insemination
• Combination of interventions
Source: Slide courtesy of Berends and Esnard (2020)
51. 51
Key findings
Source: Slide courtesy of Berends and Esnard (2020); photo credit J. Berends (2019)
• Establish PGs with a mix of hybrid Farrow-to-Finish (high profits) and Wean-to-Finish (moderate profits) farming systems that can
collectively supply slaughterhouses with a consistent high-quality fattener
• To sustain investments in hybrid breeds a rank order for project interventions is
recommended:
1. Improved credit facilities (high priority)
2. Good Animal Husbandry Practices (high priority)
3. Training and the introduction of commercial pig feed (medium priority)
4. Artificial insemination (low priority)
• Interaction effects: individual activities are negative or barely positive but 1+2+3 = 47% increase in profits
• Focus on functional PGs: Institutional arrangements that reward small-scale farmers in proportion to their patronage and investment
delivered higher reinvestment in PGs and larger profits for members
• Co-investment between PG/PO members and a strategic partner in a hygienic slaughterhouse facility is a high-impact intervention that
widens and deepens the medium and long-term results of the project.
• Potential negative impacts for smaller farmers if disease outbreak occurs during upgrading. Improved loan product, subsidize introduction of
GAHPs, keep funds in reserve to cover loan defaults.
52. 52
Example #4: Aggregation systems for horticulture in Bihar
People in Bihar consume less than half of the global
recommendation of 400 grams/capita/day (FAO and
WHO, 2014)
Consumers dependent upon nutritionally vulnerable
markets (i.e. traditional, small and often rural) likely to
face the greatest challenges to F&V access and
affordability.
Do aggregation systems like LOOP (a program of Digital
Green) improve availability/accessibility for poorer, more
remote HHs? Can they be made more nutritionally-
sensitive? Are there trade-offs in doing so?
Picture credit: K.M. Rich 2019 (Muzzafapour, India)
Source: Slide courtesy of Cooper et al. (2020)
53. 53
The LOOP aggregation scheme
1. F&V aggregation
from farmers
2. Aggregator
sells F&V at
market
3. Aggregator
collects money
and receipts
4. Returns revenues
and receipts to farmers
LOOP
LOOP: a mobile app-based aggregation service that has collected
and sold the F&V supplies of over 28,000 farmers in Bihar, India
Key farmer-facing
benefits:
Cut transport costs
(1.5 Rs/kg 0.5-1
Rs/kg)
Market access
Increased bargaining
power
Time-savings
BUT …
The combination of
lower transport costs
and access to higher
capacity vehicles has
contributed to
aggregation pathways
clustering around
large urban markets
(occasionally
bypassing smaller
rural markets)
Source: Slide courtesy of Cooper et al. (2020)
54. 54
Approach
Example: the total number of farmers
registered to LOOP in Koilwar block, Bihar
Spatial group model building (SGMB): involving
stakeholders in model conceptualisation,
formulation, analysis, evaluation and decision-
making (Mumba et al. 2017); using the participatory
GIS tool ‘LayerStack’ (Rich et al. 2018)
LOOP dashboard data: real-time market transaction
data covering LOOP supply quantities, F&V types,
prices and associated meta-data
Household survey data: 360 farming household
surveys on production and marketing habits
Source: Slide courtesy of Cooper et al. (2020)
55. 55
Output timeseries
Reference Extension Quota
Cold storage Consumer demand
LOOP farmers LOOP profits
LOOP sales Small market
F&V retail
purchases per
customer
Source: Slide courtesy of Cooper et al. (2020)
57. 57
Wider trade-offs
-1
0
1
2
3
4
LOOP extension
-1
0
1
2
3
4
Small market
quota
-1
0
1
2
3
4
Cold storage
-1
0
1
2
3
4
Retail demand growth
Reference
baseline
Outcome
relative
to
reference
Outcome
relative
to
reference
Source: Slide courtesy of Cooper et al. (2020)
58. 58
Implications
Aggregation systems: real potential to improve the availability
and affordability of F&V in small, rural markets.
However, nutrition-facing benefits may come at the expense of
producer-facing financial outcomes.
Likewise, changes in the wider enabling environment may
compound these trade-offs (e.g. cold storage stabilising prices
in smaller markets)
Picture credit: K.M. Rich 2019 (Muzzafapour, India)
Source: Slide courtesy of Cooper et al. (2020)
59. 59
References
Antunes, P., Santos, R., & Videira, N. (2006). Participatory decision making for sustainable development—the use of mediated modelling techniques. Land Use
Policy, 23(1), 44-52.
Berends, J., Rich, K.M., & Lyne, M.C. (2020). A pro-poor approach to upgrade value chains in Tanintharyi region of Myanmar. Oral presentation for the 3rd Asia-Pacific
System Dynamics Society Conference, Brisbane, Australia, 4 February 2020.
Cooper, G.S., Rich, K.M., Shankar, B., Rana, V., Ratna, N., Kadiyala, S., Alam, D. & Nadagouda, S.B. (in review).Identifying ‘win-win-win’ futures from inequitable value
chain trade-offs: a system dynamics approach. Submitted to Agricultural Systems.
Lie, H., Rich, K.M., & Burkart, S. (2017). Participatory system dynamics modelling for dairy value chain development in Nicaragua. Development in Practice 27 (6), 785-
800.
Lie, H., Rich, K.M., van der Hoek, R., & Dizyee, K. (2018). Quantifying and evaluating policy options for inclusive dairy value chain development in Nicaragua: A system
dynamics approach. Agricultural Systems 164, 193-222.
Mumba, C., Skjerve, E., Rich, M., & Rich, K.M. (2017). Application of System Dynamics and Participatory Spatial Group Model Building in Animal Health – A Case Study of
East Coast Fever Interventions in Lundazi and Monze Districts of Zambia. PLOS One, http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0189878.
Rich, K.M., Rich, M., & Dizyee, K. (2018). Participatory system approaches for urban and peri-urban agriculture planning: the role of system dynamics and spatial group
model building. Agricultural Systems 160, 110-123.
Scott, R. J., Cavana, R. Y., & Cameron, D. (2016). Recent evidence on the effectiveness of group model building. European Journal of Operational Research, 249(3), 908-
918.
Vennix, J. A. M. (1996). Group Model Building. Facilitating Team Learning Using System Dynamics. New York: Wiley & Sons
moving it to the heart of livestock agendas and investments and driving technical and transformational interventions so women can achieve better lives through livestock
Maputo declaration- 10% of public resources to agriculture
moving it to the heart of livestock agendas and investments and driving technical and transformational interventions so women can achieve better lives through livestock
Maputo declaration- 10% of public resources to agriculture
Maputo declaration- 10% of public resources to agriculture
Maputo declaration- 10% of public resources to agriculture
Maputo declaration- 10% of public resources to agriculture
Maputo declaration- 10% of public resources to agriculture
Maputo declaration- 10% of public resources to agriculture
LMP, GLAD, TASSL and ADGG in particular
moving it to the heart of livestock agendas and investments and driving technical and transformational interventions so women can achieve better lives through livestock
LMP, GLAD, TASSL and ADGG in particular
Maputo declaration- 10% of public resources to agriculture
LMP, GLAD, TASSL and ADGG in particular
LMP, GLAD, TASSL and ADGG in particular
LMP, GLAD, TASSL and ADGG in particular
LMP, GLAD, TASSL and ADGG in particular
LMP, GLAD, TASSL and ADGG in particular
LMP, GLAD, TASSL and ADGG in particular
LMP, GLAD, TASSL and ADGG in particular
LMP, GLAD, TASSL and ADGG in particular
LMP, GLAD, TASSL and ADGG in particular
Maputo declaration- 10% of public resources to agriculture
Maputo declaration- 10% of public resources to agriculture
First introduce LOOP farmer membership and total LOOP sales trends. Note how extension leads to ~4 times more farmers than the baseline, whilst having to send 20% of produce to Market B (smaller market) limits the attractiveness of LOOP membership relative to non-loop.
The system is less sensitive to the external scenarios (i.e. cold storage and demand); not the same feedback magnitudes/effect on LOOP membership and production
Market B Quota scenario leads to LOOP profits falling by 1/3 relative to the reference run by October 2021 (lower prices and higher wastage rates in Market B). However, positive implications for the availability and affordability of F&V in Market B, with a ~12% increase in cumulative purchases over the reference scenario.
Interesting, LOOP extension on its own may have negative implications for the avail and affordability of F&V in smaller markets (i.e. this scenario may not actually be nutritionally sensitive); non-LOOP farmers that previously supplied the smaller market are now able to access the larger market through LOOP (essentially diverting supplies away from the smaller market, making supplies less available and more expensive).
How do these runs plot on to the trade-off axes? (next slide)
NOTE: the jagged cumulative profit lines are caused by farmers investing in F&V land and higher yields.
First set up the idea of the trade-off space: where does each scenario land on the trade-off space between LOOP farmer profits (x-axis) and F&V purchases (proxy for availability and affordability) in Market B (small market)? The reference mode sits in the middle…
And the four scenarios fall within the four quadrants
Most noticeably, sending 20% of all LOOP supplies to Market B leads a significant improvement in availability and affordability, but also the steepest decline in LOOP profits. Likewise cold storage, where the reduction in waste and dampening of prices helps to improve avail and affordability, but reduce revenues and profits
Where can we go from here?
How do we arrive at the win-win space for consumer nutrition and producer livelihoods? Is it a combination of the one-at-a-time runs here? Can we run Monte Carlo like simulations to understand the interactions between the scenarios and internal drivers (not plotted here due to time/space limits).
Compare trade-offs from internal and external interventions.
We’re also able to visualise some of the other trade-offs across the wider value chain. e.g. forcibly increasing LOOP supplies to smaller markets may reduce LOOP return on investments (i.e. only able to sell smaller quantities in smaller market, losing out on recouping LOOP transport costs which are Rs/kg sold) and reduce the attractiveness of LOOP to farmers.
- And, whilst cold storage in the small market may help to reduce retail prices below the reference (and increase F&V avail and affordability), the attractiveness of LOOP supply