Presentation by Professor Peter Atkinson of Lancaster University of Zimbabwe at the One Health for the Real World: zoonoses, ecosystems and wellbeing symposium, London 17-18 March 2016

1. Agent-based modelling as an integra1ve
framework for one-health:
Trypanosomiasis in eastern Zambia
Pete Atkinson (Lancaster)
Simon Alderton, Joanna Kuleszo (Soton)
Neil Anderson, Ewan Macleod, Kathrin Schaten, Sue Welburn (Edinburgh)
Noreen Machilla, Mar1n Simuunza (Zambia), Alex Shaw

3. ESPA DDDAC Project

4. Feedbacks

5. Forecas1ng

6. The Luangwa Valley

7. Zambia Team

8. Lwangwa Valley - Mambwe District

9. The Situa1on in Luangwa Valley
• In terms of the DDDAC project, the focus is how popula1on movements
down (i.e., northwards) into the Luangwa Valley may lead to:
– Increases in popula1on density
– Increases in the area of seSled or farmed land, and reduced natural land
– Proximity to tsetse areas (including proximity to the SLNP)
• With regard to the laSer, we need to think about how people interact
with tsetse dense areas (primarily along the boundary – landscape
ecology)
• People need to enter the tsetse dense areas due to the need to access
ecosystem provisioning services (i.e., resources)
– A good example is the need to get water for caSle and for humans. This may
mean accessing streams by walking through, e.g., tsetse dense woodland and
bush.
– For local, especially poor, communi1es there may be not much choice about
this. There are some wells, but not everywhere.

10. Modelling
• We will explore this with the agent-based model. Lots of assump1ons in
the model, but in principle:
• We can see what happens to disease prevalence in caSle and humans:
– as the land changes historically
– If we change the land further (e.g., more coSon)
– If more people move down the Luangwa Valley
– If we take some other ac1on (e.g., add water boreholes)
– We can adjust people’s behaviour if they get the disease
– And crucially, we can determine which groups of the popula1on are more
likely to be impacted
• So, we are able to connect ES, disease and poverty.

11. Agent-based Modelling as an
Integra1ve Framework
Acknowledgements to Simon Alderton

12. Modelling background
• Epidemiological models are tradi1onally created using dynamic,
compartmentalised approaches…
• Sleeping sickness is represented by the – –
(SIS) model due to the absence of immunity.
Popula1on sizes
Infec1on rate (contact rate)
The theore1cal number of people
in each compartment at a given
1me.
Homogeneous mixing.
Differen1al equa1ons.

13. An Alterna1ve Approach…
• Agent-Based Modelling
“An agent-based model (ABM) is one of a class of
computa1onal models for simula1ng the ac1ons and
interac1ons of autonomous agents (e.g. individuals)
with a view to assessing their effects on the system
as a whole.”
13

14. Example
autonomous agents/
individuals
Fed
40 days old
Female
3 offspring
Mature infecFon
Teneral
1 Day Old
Male
Male
Farming
39 y/o
Female
Going to school
7 y/o
TARGET, MOVE
TO BITE/INFECT

15. Given the sparsely populated nature of the study
site – an ABM seems a reasonable choice
Now to utilise data from the project…

16. Human and animal inputs
• ~ 16,000 humans
– with daily rou1nes based on age/gender/caSle
ownership
– daily rou1ne consists of trips to resources water/
crops etc.
– paths created using A-Star pathfinding algorithm
• ~ 3,000 caSle.
• ~ 11,000 domes1c animals including goats, pigs,
donkeys, cats and dogs.
16

17. ABM Canvas/Background Layers
A. An 11m resoluFon aerial image of the
study area.
B. A 11m m resoluFon land classificaFon
map
C. An 11 m resoluFon image of the
kernel density esFmaFon (KDE)
influence of posi1ve fly count
loca1ons (black)
A B C

18. Tsetse popula1on
• Total popula1on rises ini1ally due to the
teneral injec1on, before reaching
eqilibrium.
• Star1ng pop. had 2:1 female to male
distribu1on (based on model proposed by
Vale & Torr (2005)), accoun1ng for the fact
that at any given 1me a greater propor1on
of the fly popula1on is female, despite
births being 1:1.
• The ABM acted as an integra1ve
framework around which we were able to
test our understanding as experts, use
literature findings and data (and other
modelling approaches).

19. Host infec1ons
• Only 2 new
infec1ons are
observed in hosts
• Reflects the low-
level yet persistent
prevalence of the
disease in the
popula1on

20. Human ac1vity when biSen

21. Discussion
• From a complexity science perspec1ve cau1on is advised!
– Challenging to know if the model is plausible
– U1lity depends on the purpose
– Computa1onally demanding
• Such a rich model has a mul1tude of applica1ons
– Impacts of historical land cover change (RS study)
– Impacts of future possible land cover changes on different tribes
– Other management interven1ons that lie between environment, health and poverty
– Generalisa1on to determine what informa1on is required and what inference is possible in
the inverse sense
• A real tool for real world problems, but only one tool
• Alterna1ve integra1ve and complementary approaches...

22. This work, Dynamic Drivers of Disease in Africa Consor1um, NERC project number
NE-J001570-1, was funded with support from the Ecosystem Services for Poverty
Allevia1on (ESPA) programme. The ESPA programme is funded by the Department
for Interna1onal Development (DFID), the Economic and Social Research Council
(ESRC) and the Natural Environment Research Council (NERC).
Thank you