What is the difference between a screening map and a risk assessment?

A screening map flags where a hazard might exist using available regional data; a risk assessment quantifies likelihood and consequence for a specific site, usually requiring field data, modelling, and a licensed professional's judgement.

What specific triggers mean a GIS risk map needs specialist review?

Any output that informs life safety, building permits, engineering design, insurance pricing, regulatory submissions, or legal/public communication. Also when the data scale is coarser than the decision, or local conditions contradict the regional model.

How do hazard, susceptibility, and regulatory layers differ?

A hazard layer records a process or past event; a susceptibility layer is a model of where a process could occur; a regulatory zone is a legally defined boundary (e.g. a FEMA flood zone). They carry very different authority and must never be merged silently.

What should a GIS analyst hand to the reviewing specialist?

Source layers with provenance (origin, scale, date, CRS), the exact overlay logic and any scoring, the assumptions and thresholds used, and an explicit statement of what the map does and does not decide.

When GIS Risk Maps Need Specialist Review

Short answer

A GIS risk map needs specialist review the moment it stops describing where to look and starts implying what to do about life safety, permitting, engineering, insurance, or legal exposure. Desktop GIS is excellent at screening — overlaying authoritative hazard layers to flag candidate problems early. It cannot substitute for a licensed engineer, engineering geologist, or hydrologist making a site-specific call, because the regional data behind a screening map almost never resolves the local conditions a real decision depends on. The skill is recognising the line and referring before it is crossed.

Screening versus assessment: not a difference of polish

The dangerous confusion is treating a tidy overlay as an assessment. The two answer different questions.

A screening map asks "across this area, where might hazard X be a concern, given the data we can assemble at our desks?" It uses regional or national layers, often at coarse scale, and is appropriate for route comparison, exclusion mapping, due-diligence triage, and prioritising fieldwork.
A risk assessment asks "at this specific location, what is the likelihood and consequence of hazard X, and what design or decision follows?" It combines hazard and exposure and vulnerability, usually needs field data and modelling, and carries professional liability.

Hazard and risk are themselves distinct. A hazard is a potentially damaging process (a landslide-prone slope, a flood-prone reach). Risk is hazard combined with what is exposed and how vulnerable it is — a steep unstable slope in empty wilderness is high hazard but low risk; the same slope above a school is high risk. A GIS overlay that colours slopes by steepness has mapped a hazard proxy, not risk, and saying otherwise oversells it.

Three layer types that must never be silently merged

Risk maps routinely combine layers of very different authority. Keeping them visually and logically separate is a core discipline.

Observed/inventory layers — recorded events: a landslide inventory, historical flood extents, mapped fault traces. These are evidence, bounded by where someone looked.
Susceptibility / model layers — predictions of where a process could occur, from statistical or physical models (e.g. a landslide susceptibility raster). These carry model uncertainty and depend entirely on input quality.
Regulatory zones — legally defined boundaries with statutory force, such as FEMA Special Flood Hazard Areas under the National Flood Insurance Program, or seismic design zones in a building code. These are not opinions; they have legal consequences and must not be redrawn or blended into a derived score.

When a single "risk" raster fuses a regulatory flood zone with an informal susceptibility model, the reader cannot tell which pixels carry legal weight and which are a modelling guess. That is the most common way GIS risk products mislead.

Concrete triggers for specialist referral

Refer when any of these are true:

The decision affects life safety — occupancy, evacuation, emergency planning.
It feeds a permit or regulatory submission — building consent, environmental licence, planning approval.
It becomes engineering design input — foundations, slope stabilisation, drainage sizing, setback distances.
It informs insurance or financial pricing of hazard exposure.
It will be communicated publicly or used in litigation, where precision and defensibility are scrutinised.
The data scale is coarser than the decision. A 1:50,000 regional susceptibility layer cannot resolve whether one parcel is safe; using it as if it could is a category error.
Local conditions contradict the regional model — recent earthworks, drainage changes, observed seepage, or fresh cracking that the dataset predates.
The map's date is stale relative to a dynamic hazard (post-wildfire debris-flow risk, coastal change, recent seismicity).

Several of these can hold at once; any single one is enough.

Worked overlay: doing the screening honestly

Suppose you screen a proposed corridor for landslide and flood concern. A defensible workflow:

Normalise everything to one projected CRS and note each layer's native scale and date. ogr2ogr -t_srs EPSG:25831 ... and gdalwarp -t_srs EPSG:25831 .... Mixing a 1:25,000 inventory with a 30 m global susceptibility raster is fine if labelled; it is misleading if not.
Keep the three layer types in separate groups — inventory, susceptibility, regulatory — never collapsed into one symbol.
Derive slope from the best available DEM (gdaldem slope dem.tif slope.tif) as a hazard proxy, and treat thresholds (e.g. >25°) as screening flags, not verdicts.
Produce a flag, not a score where you cannot defend the weights. A weighted-overlay "risk index" implies a calibration you usually do not have; an honest output is "intersects regulatory flood zone / within 50 m of mapped landslide / slope > 30°," with each criterion traceable.
Annotate explicitly what the map decides (where to investigate) and what it does not (whether a site is safe to build).

Common pitfalls and why they happen

Calling a screening overlay a "risk assessment." The word inflates authority the data cannot support, and readers act on it.
Blending regulatory and model layers so legal zones and guesses share a colour ramp — usually because a single composite raster is easier to symbolise.
Ignoring scale. Coarse data zoomed in looks precise; the pixels do not become more accurate.
Stale data on dynamic hazards. A pre-fire debris-flow map or pre-development drainage map is quietly wrong.
Suppressing uncertainty for a clean deliverable. A polished map reads as more certain than the inputs justify, and confidence notes get dropped to look tidy.

QA and validation

Provenance audit: every layer carries origin, scale, date, and CRS before it enters the map.
Cross-check against authoritative sources (FEMA flood maps, USGS landslide resources, national geological survey hazard data) rather than relying on a single derived layer.
Re-run thresholds at ±one class to test how sensitive the flagged area is; large swings mean the result is fragile and must be hedged.
Write the review boundary into the map legend — a one-line statement of what triggers escalation to a specialist.

Bathyl perspective

We build terrain-risk products as decision support, with the referral line drawn on the map itself. The value of a good screening layer is early clarity and a clean handoff: provenance documented, layer types separated, uncertainty visible, and an explicit note of when a licensed specialist must take over. A screening map that knows its own limits is far more useful than one that quietly pretends to be an assessment.