Layer architecture

Separate layers make the model auditable.

LikelyBird treats survey quality, detectability, environmental context, range plausibility and species modelling as distinct components. That makes the modelling pipeline easier to test, explain and improve.

Layer map

SIL

Survey Information Layer
Checklist evidence, local sampling pressure and access context used for audit and modelling.

CQL

Checklist Quality Layer
Observer, effort and checklist quality context.

WDL

Weather Detectability Layer
Weather effects on detection and reporting.

TDL

Temporal Detectability Layer
Time, season and solar timing effects.

ECL

Environmental Covariate Layer
Terrain, water, vegetation, fire and land-cover covariates.

RPL

Range Prior Layer
Species-level ecological plausibility as a soft prior.

Not a single quality score

The project architecture avoids collapsing survey quality, detectability and environment into one hidden number. Each layer should be inspectable on its own, then joined into modelling tables with a known meaning.

That separation matters because a weak checklist, poor detection conditions and unsuitable habitat are different problems. They should influence the model in different ways.

Access and sampling context

Human geography helps explain where birding happens.

Bird observations are shaped by where people can travel, where they live and which places are commonly visited. These covariates help LikelyBird account for uneven sampling opportunity; they are not treated as simple evidence of habitat preference.

Population and settlement

Population density and settlement context help describe where potential observers are concentrated.

Roads and access

Road and track context helps represent where sites are more readily reached and repeatedly surveyed.

Remoteness

Distance and remoteness measures help reveal the uneven availability of places for birding.

Night lights

Night-time light data provides an additional broad proxy for human activity and settlement intensity.

Local sampling pressure

Site history and nearby checklist activity help distinguish well-observed areas from places with little survey coverage.

How it is used

These variables sit with survey information and audit context, helping interpret observation bias alongside effort and detectability.

ECL status

Environmental covariates are a container layer.

ECL is the Environmental Covariate Layer. Several component products are already complete or join-ready, while land cover remains to do.

Partial / container

ECL - Environmental Covariate Layer

Complete / join-ready

ECL: Terrain

Complete / join-ready

ECL: Hydrology / DEA water

Complete / join-ready

ECL: Geofabric streams / waterbodies

To do

ECL: Land cover

Complete / join-ready

ECL: Vegetation / NVIS

Complete / join-ready

ECL: Fire history