1 of 34

So Hot Right Now

Data Science at

Queensland Fire and Emergency Services

2 of 34

3 of 34

This Talk

QFES context

2 big challenges for QFES Futures

  • 1 hyperproblem

​

4 of 34

5 of 34

Operational Data

6 of 34

QFES Futures

Decadal outlook of spatial and temporal demand

Evidence based policy

Data driven decision making

7 of 34

8 of 34

QFES Futures

Predictive Investment Modelling System (PIMS)

Demand Forecasting

Scenario Generation

Response Simulation

9 of 34

Challenge 1: App Culture

10 of 34

“How do we quantify the impact of proposed investment on community safety?”

​

​

11 of 34

“How do we quantify the impact of proposed investment on community safety?”

=> minimum viable analysis =>

“How do we quantify the impact of moving Acacia Ridge station to <new_location> on expected response time”

​

12 of 34

13 of 34

Agile Data Science

Thoen, E. (2019). Agile Data Science with R. [online] Edwinth.github.io. Available at: https://edwinth.github.io/ADSwR/ [Accessed 24 Nov. 2019].

14 of 34

“We’re about building a capability not systems.”

15 of 34

Challenge 2:

Geo-Spatial-Temporal Forecasting

16 of 34

H3 Binning (Uber) via {h3jsr}

Source: https://eng.uber.com/h3/

17 of 34

| hex | period | x_1 | x_2 | count |

|-----|--------|-----|-----|-------|

| ... | ... | ... | ... | ... |

​

For a given incident type.

​

18 of 34

Geo-spatial GAM {mgcv}

bam(n_incidents ~ te(lon, lat) +

month +

s(population) +

s(h3Index, bs = "re"),

family = nb(),

...)

​

19 of 34

​

​

20 of 34

21 of 34

“Geo-spatial-temporal statistics has an opening for a software visionary.”

22 of 34

23 of 34

Can we incorporate climate change science into an expectation of future bushfire workload?

​

​

There’s no dataset for that...

24 of 34

First Order Effects

Earlier start to fire season

More days of extreme fire weather

  • High heat
  • Low humidity
  • Wind

Drier Fuel

25 of 34

Second Order Effects

Fire Season now overlaps with:

  • Partner’s fire seasons
  • Flu season

Longer season means more dependent on aircraft

26 of 34

Second Order Effects

Drier climate

  • Less days safe to do hazard reduction burning
  • Water scarcity
  • Change in species distribution

Increased CO2 Fertilises woody plants

  • Faster fuel accumulation

​

27 of 34

Spatial. Temporal. High Dimensional. Non-stationary everything. High-order interactions.

28 of 34

Spatial. Temporal. High Dimensional. Non-stationary everything. High-order interactions. THERE’S BUGGER ALLNOT VERY MUCH DATA.

29 of 34

A period of global instability will bring both great demand and great difficulties for data science.

30 of 34

Your New Role?

More emphasis on:

  • Specifying assumptions transparently
  • Sensitivity testing
  • Exploratory data analysis
  • ?

​

31 of 34

In Summary

Building systems vs capability

H3 + {mgcv} for geo-spatial-temporal

Climate change data science hyperproblem

​

32 of 34

Thankyou...

Emily Smart, Nicholas Tierney, Jason Brown, Nicole Lott, Anthony North, Russell Molaei, Tess Pham, #rstats

​

@milesmcbain

milesmcbain

http://milesmcbain.xyz

33 of 34

GAM(M)S

https://noamross.github.io/gams-in-r-course/

https://www.fromthebottomoftheheap.net/blog/

34 of 34

Street Network Analysis

Padgham, M., Boeing, G., Cooley, D., Tierney, N., Sumner, M., Phan, T. G., & Beare, R. (2019). An Introduction to Software Tools, Data, and Services for Geospatial Analysis of Stroke Services

​

https://richardbeare.github.io/GeospatialStroke/RehabCatchmentAdvanced/Googleway_Mapdeck.html