.jpeg)
Project Hyggja / Awa · Causal Systems
A living flood model
with people inside it.
A digital twin of the Manawatū River environment combined with a multi-agent crisis simulator to test how warnings, roads, shelters, barriers and human behaviour alter flood outcomes.
Research question
What would have happened if we acted differently?
Traditional hazard maps describe where water may go. Project Hyggja adds people, decisions, infrastructure and time.
System architecture
A physical twin and a social twin.
The combined environment allows interventions to be compared against the same baseline event.
River, terrain and infrastructure
River level, flood extent, terrain, roads, bridges, centres, barriers and location-specific exposure.
People and decisions
Households, movement, warning response, evacuation behaviour, delay, congestion, assistance needs and shelter capacity.
Actions and timing
Early or late warning, road closure, evacuation centre activation, sandbag barriers and resource placement.
Consequences
Exposure minutes, peak exposed population, evacuation, emergency demand, damage proxy and intervention trade-offs.
Counterfactual engine
Compare interventions against one baseline.
The system is designed to distinguish an intervention that sounds sensible from one that actually improves measured outcomes inside the simulated event.
Baseline run
Simulate the event without the proposed intervention.
Treatment run
Change one or more actions while holding the event assumptions constant.
Outcome comparison
Measure whether exposure, evacuation, emergency pressure or damage improves.
Sensitivity analysis
Test whether the result survives changes in timing, agent behaviour and uncertain physical inputs.
Why Ashhurst and the Manawatū
Begin where the system can be understood.
A geographically bounded river system, known communities and historical flood context provide a tractable proving ground before any wider deployment.
Grounded assumptions
Local roads, settlements, flood history and community behaviour can be reviewed by people who know the area.
Physical evidence
River observations, topography, weather and historical events create a basis for calibration.
Architecture beyond one river
The goal is to test a causal digital-twin method that could later transfer to other hazards and communities.
Current programme status
V0 established the comparison architecture.
The early prototype can run baseline and intervention scenarios and report comparative metrics. It is not yet an operational emergency-management system.
Stage 01 · Simulation skeleton — complete
Physical hazard state, agents, interventions and scenario outputs can be generated.
Stage 02 · Historical calibration
Connect the physical model to historical Manawatū events, terrain and observed river conditions.
Stage 03 · Behaviour validation
Review agent assumptions with emergency-management and community expertise.
Stage 04 · Causal evaluation
Test intervention effects, robustness, confounding and competing explanations.
Stage 05 · Controlled decision support
Evaluate limited operational use with qualified agencies and explicit human oversight.
Research collaboration
Flood intelligence needs local and technical scrutiny.
Pythology is interested in discussions with hydrologists, emergency-management specialists, councils, infrastructure operators, behavioural researchers and communities with relevant flood knowledge.
Project Hyggja / Awa is an exploratory research simulation. It is not a certified flood forecast, evacuation system or emergency-management authority. Outputs require calibration, domain review and independent validation before any operational use.