GEOS2111 Chap.11 Mapping hazard and risk with GIS
Mapping hazard and risk with GIS
The practical stream is not a side channel. Two of the five assessment components come directly out of it and together they are 40 percent of the unit mark, and one of the nine learning outcomes is about spatial analysis technologies specifically.
The stream also does something the lectures cannot: it forces the abstract argument about hazard, exposure and vulnerability into layers that have to be combined at a stated resolution, in a stated projection, with the uncertainty visible.
The unit defines GIS as technology for analysing and displaying geographically referenced data and locates its power in layering, and underneath that all spatial data reduce to two models: vector data as points, lines and polygons each carrying an attribute row, suited to discrete objects, and raster data as a matrix of pixels on a grid, suited to continuous surfaces.
The question being asked selects the model and the model selects which operations are available, which makes the choice analytical rather than technical.
Weeks 2 to 4 build a coastal exposure and vulnerability picture for an allocated river delta by joining tabular climate projections to a country polygon layer, adding a lowland elevation model and a gridded population raster, and running a two-timepoint risk assessment on supplied likelihood and impact matrices.
Weeks 7 to 10 build a seven-part multi-hazard risk map for a town on a volcanic island, converting historical event points into a density surface with a chosen search radius, output resolution and magnitude weighting, then reclassifying that surface into five named hazard zones.
Half the skill is knowing what a layer does not tell you: what unit it was aggregated at, what a median hides, whether two layers share a resolution, and how good the elevation data are. The chapter closes on layout discipline and on what the two assessments actually reward, which is a derived ranking rather than an asserted one.
What this chapter covers
- 01
Why the class makes maps: national resilience and risk index products as the real artefacts
- 02
Vector against raster, and how the question chooses the data model
- 03
Joining tabular data to spatial features on a shared key, and exporting the result
- 04
Digital elevation models: what they represent and what they deliberately exclude
- 05
Gridded population, resolution mismatch, and the overlay problem it creates
- 06
From event points to a density surface: search radius, output resolution and weighting
- 07
Reclassifying meaningless units into five named hazard zones
- 08
Print layouts, locked layers, and the elements without which a map is a picture
- 09
Classification as an argument: three ways to mislead without saying anything false
- 10
What the group presentation and the individual flyer actually reward
Justifying the parameters behind an earthquake hazard zone map
- +1State which events entered the analysis, with the record period and any completeness threshold. A catalogue records detected events and detection improves over time, so including everything back to the earliest entries mixes two sampling regimes.
- +1State whether points were weighted. Counting events equally treats a small tremor and a destructive event as the same evidence; weighting the density surface by magnitude reflects that they are not, and is the option the practical offers. Note that magnitude is logarithmic, so a linear weight already compresses an enormous range.
- +1Justify the search radius. A small radius produces a spiky surface that mostly maps where individual events happened; a large one smooths the pattern into a blur that distinguishes nowhere. Relate the radius to the spatial scale of the structures generating the events, and admit that it was chosen by inspection across a range.
- +1Justify the output resolution and the class breaks together. The resolution has to be fine enough to be meaningful at the scale the map is read at and coarser than the accuracy of the input locations. The density units are not interpretable, so the surface was cut into five classes with reader-facing names; state the rule used and note that a different rule shifts the boundaries.
- +1Say what the map does not show. It is a hazard surface, not a risk surface. It contains no exposure and no vulnerability, so it cannot rank places by expected loss until it is crossed with population and asset layers, and the zones are relative within this island rather than comparable to anywhere else.
Key terms
- Vector data
- Spatial data represented as points, lines and polygons, each feature carrying a row in an attribute table. It suits discrete objects such as towns, rivers and administrative boundaries.
- Raster data
- Spatial data represented as a matrix of pixels on a grid, with a value in each cell. It suits continuous surfaces such as elevation, population density and modelled hazard.
- Attribute join
- The operation that attaches a table of values to spatial features through a field the two share. The table must be imported explicitly as having no geometry, and the key field must match exactly.
- Digital elevation model
- A raster representing bare-ground topography, excluding trees, buildings and other surface objects. That exclusion is the limitation that matters most in a built-up area.
- Coordinate reference system
- The framework that ties coordinates to positions on the Earth. Areas and distances are distorted differently by different systems, so a hazard area computed in the wrong one is simply wrong.
- Graduated symbology
- A vector classification that assigns colours to ranges of a chosen numeric field, using a stated classification mode. The mode and the number of classes are decisions rather than defaults.
- Singleband pseudocolour
- A raster rendering method that assigns colours to value ranges through class breaks the analyst chooses, used in the practicals to resolve elevations below a few metres.
- Kernel density
- An operation that converts point events into a continuous surface, controlled by a search radius, an output resolution and an optional weighting field.
- Reclassification
- The step that converts an uninterpretable continuous surface into a small number of named classes. Without it a density map is a picture rather than a hazard map.
- Print layout
- The composition environment in which map frames, legend, scale bar, title, annotation and sources are assembled for export. Layers are locked once each frame's extent is correct.
- Resolution mismatch
- The condition in which two layers being overlaid have different cell sizes, so the combined product cannot be more precise than the coarser of the two.
- Hazard surface
- A map of where a hazard is expected, containing no information about who or what is present. Crossing it with exposure and vulnerability is what turns it into a risk product.
Mapping hazard and risk with GIS FAQ
How much GIS does the unit expect, and do I need experience first?
A good deal, and no. The practicals use open-access software and teach it from the beginning: the first week has no practical class but provides pre-recorded material on basic GIS concepts, and the following weeks build a map from scratch and use it to explore exposure and vulnerability to coastal hazards. That work becomes the group presentation.
From the seventh week the practicals become a seven-part multi-hazard mapping exercise, and that becomes the individual assignment. Between them the two artefacts are 40 percent of the unit mark. What is being assessed is not software fluency for its own sake but spatial reasoning: choosing layers that answer a hazard question, handling elevation data honestly, and producing a map a non-specialist can read and act on.
What is the difference between vector and raster, and why does it matter?
Vector data are points, lines and polygons, each with an attribute row, and they suit discrete objects with properties: towns, rivers, administrative units, individual events. Raster data are a grid of pixels with a value in each cell, and they suit continuous surfaces: elevation, population density, modelled hazard. The distinction matters because it determines which operations are available.
Attribute joins, selections, buffers and overlays belong to the vector world; class breaks, reclassification and density surfaces belong to the raster world. The practicals convert between them in exactly one direction, turning event points into a density raster, and that conversion is where the analytical decisions concentrate.
Choosing the model is therefore a decision about the question you are asking rather than a technical preference.
What are the four questions I should ask of any layer I am given?
At what unit was this aggregated, what does this single value hide, do these layers share a resolution, and how good is the elevation. The practicals assess all four directly. Country-level mapping of climate projections lets opposing regional trends cancel, so a country with severe regional drying can read as unaffected. A median of an ensemble discards the spread, so genuine model disagreement becomes invisible confidence.
Overlaying a fine population raster on a coarser elevation model implies a precision neither has. And exposure estimates are extremely sensitive to vertical error, which is why the unit sets two readings on elevation data quality, one reporting that better data substantially raises estimates of global exposure to coastal flooding. Asking the four questions in your write-up is a mark, not a hedge.
Why is classification described as an argument?
Because two maps of the same data with different class breaks can support opposite conclusions, and neither is lying about the numbers.
Three techniques mislead without stating a falsehood: class breaks chosen to place a boundary either side of a value of interest, a colour ramp whose visual weight does not match the variable so that a mild class reads as alarming, and an extent cropped so that a pattern appears to stop at the map edge.
Each is a defensible design decision in some context, which is exactly why the reasoning has to be stated rather than assumed. The unit's own lava-flow exercise says so explicitly: the class logic is the student's choice but it must be defensible, and the worked logic offered is recency, with recent flows implying very high hazard.
What are the two assessments actually looking for?
A derived conclusion rather than an asserted one, in both cases. The group presentation is built from three practicals on an allocated river delta and is ten minutes with five minutes of questions, and the reasoning being marked runs from map, to defensible generalisation, to mechanism: why did past events become disasters, and who was differentially affected.
The scope discipline is explicit, since ten minutes means one or two documented examples done properly rather than a research report. The individual flyer is a public-audience artefact of four to six pages carrying your own maps, covering hazard background, the vulnerability of the town, and recommended actions for the highest hazard risk.
The word derived matters there: the ranking of hazards has to fall out of a synthesis step, because a hazard with a high zone rating over empty land ranks below a moderate rating over the settlement.
Exam move
Revise this stream as a subject rather than as a series of lab sessions, because that is how it will be examined and how the assignment is assembled. Write the six operations in order on one page, load and inspect, join, symbolise, generate a surface, reclassify and label, compose and export, and put beside each the parameter you have to defend.
Second, make the four data-criticism questions into a checklist you apply to every layer you touch, and write your answers into the assignment text as you go, because reconstructing them later is where students lose marks that were already earned.
Third, keep every project file with its layers intact and export every map as you produce it, with two lines about what it showed; the individual assignment is assembled from seven practical parts and rebuilding part three in the final week is miserable. Fourth, learn the map layout elements as a list and know what is lost if each is missing, since they are worth marks individually and cost nothing to include.
Finally, practise the synthesis move specifically: given several hazard layers and a vulnerability layer, rank the hazards for a named place and say why. That is the seventh part of the practical, it is where the marks concentrate, and it is the whole unit compressed into one map operation.
Working through Mapping hazard and risk with GIS in GEOS2111? Sia is AskSia’s AI Environmental Science tutor — ask any GEOS2111 Mapping hazard and risk with GIS question and get a clear, step-by-step explanation grounded in how GEOS2111 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
