Add hotspots visualisation layer

[tomalrussell]

To add as a raster layer (set of layers). Absolute minimum would be current/baseline/aggregated total risk (EAD+EAEL - expected annual damages plus expected annual economic losses).

TODO:

• [ ] define parameter space in discussion with @itrcrisks (do we need multiple epochs/climate scenarios, breakdown by hazard, infrastructure sector, EAD, EAEL?)
• [ ] create layers from asset-level risk results
• [ ] ingest to backend
• [ ] visualise on frontend
• [ ] document process for updating

[tomalrussell]

For each infrastructure sector / hazard scenario combination, we could aggregate values for EAD and EAEL (both min/mean/max)

• sector: transport, power, water
• subsector: airport_polygon_areas, port_polygon_areas, rail_edges, rail_nodes, roads_edges, roads_nodes, electricity_network_v3.1_edges, electricity_network_v3.1_nodes, irrigation_assets_NIC_edges, irrigation_assets_NIC_nodes, pipelines_NWC_edges, potable_facilities_NWC_nodes, waste_water_facilities_NWC_nodes
• hazard: coastal, cyclone, fluvial, surface
• rcp: baseline, 4.5, 8.5 (some 2.6)
• epoch: 2010, 2050 (some 2070, 2080, 2100)

As a first suggestion, this could be the level of detail to calculate and visualise:

• variables: EAD, EAEL (both mean), plus "total risk"
• sector: transport, power, water, plus "all sectors"
• hazard: coastal, cyclone, fluvial, surface, plus "all flooding"
• rcp: baseline, 4.5, 8.5
• epoch: 2010, 2050

Spatial resolution is open for discussion. Suggest a 1km grid in projected space (Jamaica Metric Grid - EPSG:3448).

Smoothing or kernel density estimation is also up for discussion. Suggest start by looking at the outputs with no smoothing, as this is a relatively coarse grid and direct aggregation is simple to interpret/explain.

[tomalrussell]

After some discussion today, the next proposal is to do something smaller, but with an additional focus on pure exposure (distribution of value/use alone), where the first three variables are independent of hazards, and then risk is aggregated for present/baseline conditions only:

• variables: "total value", "economic use", "population use", "total risk", EAD, EAEL
• hazard: "none", cyclone, "all flooding"

Resolution 1km and no smoothing is fine to start, will need visual examination for legibility and check that e.g. small concentrations along the coast show up well.

[eatyourgreens]

After some discussion today, the next proposal is to do something smaller, but with an additional focus on pure exposure (distribution of value/use alone), where the first three variables are independent of hazards, and then risk is aggregated for present/baseline conditions only:

• variables: "total value", "economic use", "population use", "total risk", EAD, EAEL
• hazard: "none", cyclone, "all flooding"

Resolution 1km and no smoothing is fine to start, will need visual examination for legibility and check that e.g. small concentrations along the coast show up well.

I've got a branch in #39 that mocks up how this might work in the frontend. Variable is chosen from a select menu, hazard from radio buttons. I've still got menus for epoch and RCP but those can be removed.

Variable + hazard combination maps to a raster layer with legend and tooltips, like the existing hazard raster layers.