Back to the on-screen lesson ·
A GIS stores the world as vector and raster layers; a raster cell covers its width squared, so resolution sets both detail and size.
Paper packet. Every task here also exists on screen, where it is checked automatically; answers written on paper are not assessed by Nydus. When you are back at a device, enter your answers there.
By the end of this lesson you will be able to choose vector or raster layers for a feature and measure ground area from raster cells.
You can read scale, choropleth and symbol maps, and you know coordinates only line up when they share a system. This lesson opens the geographic information system itself: how it stores the world, and why that choice limits what it can show.
| Term | What it means |
|---|---|
| Layer | One set of geographic data of a single kind for an area, such as roads or elevation. |
| Vector | Data stored as points, lines and polygons with exact coordinates. |
| Raster | Data stored as a grid of square cells, each with one value. |
| Cell | One square of a raster, also called a pixel. |
| Resolution | The ground width of one raster cell. |
| Attribute | Information attached to a feature, such as a road's name or a parcel's owner. |
A geographic information system, or GIS, stores the world as stacked layers, each holding one kind of data for the same place.
Another way: picture
Picture tracing paper sheets laid over a base map, one for roads, one for rivers, one for parcels. Lift any sheet and the others stay in place. A GIS works the same way, except the sheets are data files that line up by their coordinates, and a raster sheet is ruled into a checkerboard with a number in every square.
Another way: steps
A layer's geometry answers where; its attributes answer what was recorded there. In a fictional garden inventory, polygon 17 is a boundary, while its ID, land use and inspection date are fields in a row. The ID is a key, not a measured quantity. A separate inspection table can join by that key only after you check missing and duplicate IDs. A one-to-many join can multiply rows and accidentally multiply population totals. Compare row counts and totals before and after a join.
A coordinate reference system (CRS) gives the coordinate numbers their spatial meaning: datum, coordinate axes, units and, for a projected system, projection parameters. A projection trades off properties when flattening the Earth. Choose one suited to the area and measurement: a convenient web display is not automatically an appropriate basis for area or distance. Assigning a missing CRS identifies what the existing numbers mean; transforming converts them. Never guess an unknown system simply to make a layer appear nearby.
Suppose an orchard study provides tree locations, management-plot boundaries and a gridded moisture estimate. First retain trees as points and plots as polygons: the geometries preserve the objects being studied. Next attach plot management records by a verified unique plot ID; check that an unmatched record remains missing rather than becoming zero. Keep moisture as a raster because each cell estimates conditions over part of a surface. Then inspect the source CRS and transform compatible copies into the chosen analysis CRS. Check several surveyed locations before extracting raster values at trees. Finally report both the moisture cell width and tree-position uncertainty. A precise-looking point does not make the moisture estimate precise at that tree. This plan supports a bounded association between management and moisture; it does not by itself prove that management caused it.
Vector data is good for features with sharp edges and identities. Each fire hydrant is a point with an address and inspection date; each street is a line with a name and speed limit; each parcel is a polygon with an owner and assessed value.
The U.S. Census Bureau publishes its TIGER/Line files as vector layers of roads, rivers and boundaries, free for anyone to use; many local maps start from them.
Raster data is good for things that vary continuously, such as elevation, temperature and rainfall, and for images from satellites and aircraft. Every cell has a value, so there are no gaps.
The U.S. Geological Survey's National Land Cover Database classes every thirty-meter cell of the lower forty-eight states as forest, cropland, developed land, water and other types, and has been updated every few years since 2001.
A cell's area is its width squared.
| Cell width | Area of one cell | Cells per square kilometer |
|---|---|---|
| 10 m | 100 m² | 10,000 |
| 20 m | 400 m² | 2,500 |
| 30 m | 900 m² | about 1,111 |
| 50 m | 2,500 m² | 400 |
Halve the width and the area of a cell falls to a quarter, so four times as many cells cover the same ground.
A cell can only hold one value. A thirty-meter cell that is half road and half lawn gets one class, perhaps "developed". A creek narrower than a cell may disappear.
Coarse rasters suit broad patterns across states; fine rasters suit a neighborhood. No resolution is right for every question.
To measure an area on a raster, count the cells in a class and multiply by the area of one cell. $2000$ forest cells at thirty meters cover $2000 \times 900 = 1800000$ square meters, or $1.8$ square kilometers.
The answer is only as good as the classification: every misclassed cell adds or subtracts nine hundred square meters.
Every vector feature carries a row in an attribute table. A county polygon might hold its name, population, area and median income. Choropleth maps from the earlier lessons are drawn by coloring polygons by one attribute.
Joining a table of statistics to a layer by a shared code, such as a county's federal code, is one of the most common GIS tasks.
A city's water utility stores pipes as lines and valves as points, because it needs exact locations and attributes. A climate scientist stores temperature as a raster, because it is measured or modeled everywhere.
Many projects use both, and a GIS can convert between them, with some loss each way.
Checking an answer. An area in square meters should be much larger than the cell count when cells are more than a meter wide. A finer raster always has more cells.
Squaring the width is allowed because each cell is a square with equal sides. Multiplying cells by cell area is allowed because every cell in a raster has the same size.
Multiplying by four when the width halves is allowed because the count doubles both across and down, and two times two is four.
Most American counties keep a GIS of parcels, zoning, roads, addresses and utilities. Emergency dispatch uses the address layer to route ambulances; the assessor uses parcels to send tax bills; planners overlay zoning with flood zones.
Many counties publish these layers on public map viewers, so residents can look up their own property.
A layer without its story is a risk. Metadata records where the data came from, when it was collected, its coordinate system and datum, its resolution or scale, and what each attribute means. The Federal Geographic Data Committee sets a metadata standard that federal agencies follow for the data they publish.
Before overlaying two layers, read their metadata: a parcel layer from last year and a land cover raster from a decade ago may disagree because the land changed, not because either is wrong. A layer that cannot say when and how it was made should be used with care, and its gaps named in any map built from it.
The most common slip is taking the cell width as its area, forgetting to square it. Another is forgetting that halving the width quadruples the cells, not doubles them.
A third is mixing square meters and square kilometers; a square kilometer is a million square meters, not a thousand. A fourth is treating a raster's classes as exact when each cell may be mixed.
Layers are the raw material of every GIS analysis. The next lessons combine them: overlay finds where layers coincide, and buffers find what lies within a distance of a feature.
The U.S. Geological Survey and partner agencies produce the National Land Cover Database, a raster that classes every thirty-meter cell of the lower forty-eight states into about sixteen types, from open water and deciduous forest to high intensity developed land. Editions from 2001 onward let analysts compare how land cover has changed.
Planners use it to track suburban growth, foresters to measure forest loss after fires, and hydrologists to estimate how much rain will run off paved surfaces. Because each cell covers nine hundred square meters, counting cells turns directly into acres and square miles.
The thirty-meter cells come from Landsat satellite images. They are fine enough to show a new subdivision, but a single house, a narrow creek or a row of street trees may vanish into a cell classed as something else, which is why local studies often add finer data.
Almost every American county keeps a vector layer of property parcels, with a polygon for every lot and an attribute table of owners, acreage and assessed values. The assessor uses it to send tax bills; buyers and title companies use it to check boundaries.
Laid over parcels are other layers: zoning polygons that say what may be built, flood zone polygons from FEMA, road lines, address points and utility lines. A planner considering a new school can ask which parcels are large enough, zoned for public use, outside the floodplain and near a main road.
That question only works because every layer shares a coordinate system and datum. When one layer is off by a datum shift of a few dozen meters, parcels seem to spill into roads, and the analysis gives wrong answers.
It is natural to read a raster's resolution, say thirty meters, as the area each cell covers. But resolution is the width of a cell, and the ground inside it is the width squared, nine hundred square meters.
The same slip shows up when resolution changes: halving the width does not double the cells, it multiplies them by four, and the file grows to match.
A raster has thirty-meter cells. Square the width.
$30^2 = 900\ \text{m}^2$
Width times height.
Name the slip of doubling.
$60\ \text{m}^2$
Width plus width.
Count cells per square kilometer.
$1000000 \div 900 \approx 1111$
A million square meters.
Say what each cell holds.
$\text{one value}$
For its whole square.
A thirty-meter land cover raster has $5000$ forest cells. Recall the area of one cell.
$900\ \text{m}^2$
Thirty squared.
Multiply by the forest cells.
$5000 \times 900 = 4500000\ \text{m}^2$
Square meters.
Convert to square kilometers.
$4500000 \div 1000000 = 4.5$
A million per square kilometer.
Convert to acres for a landowner.
$4.5 \times 247 \approx 1112$
About 247 acres per square kilometer.
Say what could make it wrong.
$\text{misclassed cells}$
Each adds or removes 900 square meters.
A square area $3$ km on a side has thirty-meter cells. Convert the side.
$3000\ \text{m}$
Meters.
Count the cells across.
$3000 \div 30 = 100$
One row.
Count all the cells.
$100^2 = 10000$
Rows times columns.
Halve the cells to fifteen meters and count across.
$3000 \div 15 = 200$
Twice as many.
Count all the finer cells.
$200^2 = 40000$
Four times as many.
Say the trade-off.
$\text{more detail, larger file}$
Four times the storage.
Find the area of one cell.
$20^2 = 400\ \text{m}^2$
Width squared.
Divide a square kilometer by it.
$1000000 \div 400 = 2500$
Cells.
Compare with ten-meter cells.
A raster layer has square cells $10$ m wide. What area of ground does one cell cover, in square meters?
Complete the worked solution: a square study area $1.5$ km on a side is covered by a raster with cells $30$ m wide. Find the cells across, the total cells, and the total if the cell width is halved.
Find the cells across.
$\text{side in meters} \div \text{cell width} =$ c
One row of the grid.
Find the total cells.
$\text{cells across squared} =$ t
Rows times columns.
Find the total with half-width cells.
$\text{four times as many} =$ f
Twice across, twice down.
Say what that does to the file.
$\text{about four times larger}$
One value per cell.
Match each feature to the kind of layer that best stores it.
| a point layer | a line layer | a polygon layer | a raster layer | |
|---|---|---|---|---|
| fire hydrants | ||||
| streets | ||||
| property parcels | ||||
| ground elevation |
A square kilometer is $1000000$ square meters. Fill in how many raster cells cover it when the cells are 10 m, 20 m and 50 m wide.
| cells | |
|---|---|
| 10 m cells per square kilometer | |
| 20 m cells per square kilometer | |
| 50 m cells per square kilometer |
A raster has cells $10$ m wide. Write the ground area, in square meters, covered by $x$ cells.
Answer:
In a land cover raster with 30 m cells, $3000$ cells are classed as forest. How much forest is that, in square kilometers?
Answer: unit: m2 / km2
The National Land Cover Database maps the United States in 30 m cells. Suppose a county covers $900$ square kilometers. About how many cells cover it?
Answer: cells
Mark the supported observation, data-link plan and spatial-comparison limit in this fictional garden-study audit. The city supplies 4 proposed gardens, parcel polygons with unique IDs, a land-use table keyed by those IDs and a 10 m elevation grid. Parcels use a local projected CRS in meters; the grid has another known CRS.
This task has no paper form; do it on a device.
Lesson test: one question per skill, one attempt each, no hints. Your answers are checked when you submit.
A raster has cells $30$ m wide. Write the ground area, in square meters, covered by $x$ cells.
Answer:
You can work with map layers. Explain why halving a raster's cell width makes four times as many cells.
26. Your turn: how many twenty-meter cells cover a square kilometer?, step 3
$10000$
Four times as many.