In-depth practical guides

Compare geographic coverage

Check codes, aggregate membership and territorial versions before joining files or comparing years.

Updated :

Mapping workspace with a map, compass and illustrative coastal image.

AI-generated illustration

Do identically named areas cover the same units?

A label defines neither a boundary nor a population. M49 classifies statistical units; a map layer supplies geometry. Link classification, layer and indicator coverage explicitly.

  1. Identify classification

    Keep code as text, level and version. Retain leading zeros; do not equate M49, ISO and NUTS codes by appearance.

  2. Check membership

    List members and exclusions for each aggregate. Changed membership can change a total while national series remain unchanged.

  3. Handle changes

    Check mergers, splits and boundaries during the period. A correspondence needs direction, validity and multiplicity; one old code can map to several new ones.

  4. Validate joins

    Check duplicates, unmatched units and level. Never add an aggregate and its members together; maps also need coordinate system and layer date.

Reading workshop

An inclusive total must not be added to its members

DEMO territories do not exist. The exercise separates exact aggregation from allocation with unknown weights.

Fictional data and records for learning: none of these results comes from a live service query.

A and B form C: count once

For the same period and count definition, A has 30 and B has 70. They are disjoint and C combines them exactly. A later classification splits B into B-North and B-South without detailed counts.

If the table exceeds the screen, scroll horizontally; arrow keys work when the region is focused.

A and B form C: count once
Fictional unitKnown valueScope relationDecision
Old unit A30Member of C, no overlap with BMay be included in A+B.
Old unit B70Other member of CSame period and definition as A.
Aggregate C100Already includes A and BChoose C alone or A+B; never A+B+C.
B-North and B-SouthAllocation of 70 unknownNew boundaries split BRequest counts or a documented method; do not assume halves.

Which total and allocation are defensible?

Open each proposition to compare its diagnosis. More checks may still be needed.

200 for C by adding all rows

No. 30+70+100 counts the same units twice. These levels are not disjoint.

100 for C; B’s breakdown remains unknown

Yes. A+B aggregation is documented, but transfer from B to the new units lacks appropriate weights.

35 for each half of B

No. Two new codes establish neither equal populations nor equal shares of the variable.

Fictional scope record

Classification: DEMO-old then DEMO-new
Period and definition: identical for A, B and C
A=30; B=70; C includes exactly A+B=100
Selected total: C alone, without its members added
New boundaries: B-North and B-South
Allocation: not calculated, weights unknown

The correspondence table establishes membership here, not transfer weights. Retain direction, validity and multiplicity of each link.

Documentation of the rules, separate from fictional examples

Compare other situations and exercises

Same aggregate name

Two files name a region alike, but one excludes a unit. Compare membership before explaining change.

Split territory

One old area becomes two. A code correspondence alone supplies no weights for allocating an amount.

Growth created by an expanded territory

Fictional example: A has 100 residents, then 130 after absorbing B, which had 30. The total rises 30%, but this example shows no growth within the original boundary. Reconstruct A+B for the starting year when data permits before interpreting the change.

Regional rates need population weights

Two illustrative areas have rates of 10% among 100 people and 20% among 900 people. Their combined rate is (10+180)/(100+900), or 19%, not the simple average of 15%. Denominators must describe the same population and period.

From observation to decision

A 20% fall that hides a 20% rise

In 2024, the published total covers A and B: 100 + 50 = 150 units. In 2025, B leaves the scope; only A remains, with 120 units. Comparing totals mixes changes in the phenomenon with changes in coverage.

Fictional learning case: counts and dates illustrate the method and do not describe a real assignment.

Scope changes the interpretation

Common scale from 0 to 160 fictional units. The first two bars cover different areas.

  • 2024 · A + B150
  • 2025 · A only120
  • 2024 · A only100
  • 2025 · A only120
Case reading grid
ObservationUseful checkNext action
Published totals: 150 then 120(120 − 150) / 150 = −20%.Describe a fall in the published total without calling it a constant-scope change.
A alone: 100 then 120(120 − 100) / 100 = +20%.Report the change for A, provided definitions and units remain the same.
B in 2025: value missingThe 2025 total for A + B cannot be reconstructed.Do not assign zero to B; retain the missing status and seek compatible data.
Your decision — Is a code correspondence table enough?

Explained answer

No. It identifies territories but does not automatically supply historical values. Mergers, splits or coverage changes may need detailed data. Check additivity before summing; rates cannot be reconstructed by adding rates.

Document reference: COG provides geographical vintages and histories. The territories and values in this case are fictional. Insee — Code officiel géographique

Worked exercise

Two areas merge: recalculate the rate

Fictional example: the values and documents teach the method and do not describe a real organisation.

A and B are disjoint and share period and definition. C is exactly their union.

If the table exceeds the screen, scroll horizontally; arrow keys work when the region is focused.

Starting information
AreaCasesPeople
A10100
B90300

Your checks

  1. Check non-overlap and boundary correspondence.
  2. Sum numerators and populations.
  3. Compare the aggregate rate with a simple mean.
Read the answer

Answer and decision

C has 100 cases among 400 people: 25%. The simple mean of 10% and 30% is 20%, giving equal weight to unequal populations. The boundary correspondence permits this aggregation; similar names would not.

If the case changes

If part of B remains outside C, obtain its counts. Area alone does not automatically provide population weights.

Record to retain

Keep the correspondence table, period, both counts and C calculation.

DOCUMENTED REFERENCE

M49: a hierarchy without double counting

M49 assigns three-digit numeric codes to countries or areas and geographic groupings. ISO letter codes appear alongside them but belong to another system. These are actual mappings from the consulted official table.

001150155250
M49: a hierarchy without double counting
ReferenceValue or formWhat it identifies
001WorldWorld: the global aggregate.
150EuropeEurope: an M49 geographic region, different from the European Union.
155Western EuropeWestern Europe: the M49 subregion containing France.
250FRA / FRFrance: M49 250, ISO three- and two-letter codes FRA and FR.
004AFG / AFAfghanistan: the leading zero is part of the M49 code; preserve it in joins.

Implication for the record

Keep geographic level, the code as text and classification version. World → Europe → Western Europe → France describes nested sets, not four territories to add. For older periods, check territorial and code changes before applying today’s classification.

Can two tables be joined on code 150 alone?

Check the system and level as well. M49 code 150 means the Europe region; an internal code with the same number may mean something else. Name columns explicitly, such as m49_code and geographic_level, and inspect unmatched values before aggregation.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

NUTS 2024: Paris and Île-de-France have different scopes

These codes and levels were checked in the GISCO NUTS 2024 file properties on 22 October 2026. They are statistical units; no population figure is attached here. GISCO’s LEVL_CODE includes the country at level 0.

FRFR1FR10FR101GISCO · NUTS 2024 · 0 / 1 / 2 / 3
NUTS 2024: Paris and Île-de-France have different scopes
ReferenceValue or formWhat it identifies
NUTS 20242024-01-01The classification applies from January 2024. NUTS 2027 is announced for 1 January 2027; do not treat the future version as that of a 2024 series.
FR0France: country, level 0 in this file. NUTS then organises three regional levels, 1, 2 and 3.
FR11Île-de-France at NUTS 1. Keep the code as text, even though it combines letters and digits.
FR102Île-de-France at NUTS 2. Sharing a name with FR1 does not make these records interchangeable in a level-specific join.
FR1013Paris at NUTS 3. Do not assign it a value published for the whole FR10 area; retain each observation’s territorial unit.

Implication for the record

Keep code, level and territorial version alongside the indicator period. The file uses generalized 1:60 million geometry, suitable for this classification check, not parcel boundaries or address analysis.

Are two observations with the same region name comparable?

Not necessarily. Check code, level, territorial edition and actual boundaries. Boundary changes may require a correspondence table without permitting values to be copied between units.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

Geodesic distance: units, ellipsoid and argument order

Official GeographicLib examples provide reproducible inputs and outputs on WGS 84. These documented calculations do not route or geocode your addresses.

Geodesic distance: units, ellipsoid and argument order
ReferenceValue or formWhat it identifies
WGS 84 · a6 378 137.0 mEllipsoid semi-major axis in metres, not point accuracy or a universal degree length.
WGS 84 · 1/f298.257223563Inverse flattening defines the model with a; a sphere differs.
Wellington → SalamancaInverse(-41.32, 174.81, 40.96, -5.50)Inverse takes starting latitude, longitude then ending latitude, longitude, unlike GeoJSON. City coordinates have two decimals.
s1219 959 679.267 mEllipsoidal geodesic distance displayed to a millimetre does not make rounded input positions millimetre-accurate.
Perth · Direct(-32.06, 115.74, 225°, 20 000 000 m)Direct starts near Perth at initial azimuth 225° for 20000 km. Azimuth need not remain constant.
Direct · latitude32.11195529°Positive arrival latitude is north; do not copy the departure sign.
Direct · longitude−63.95925278°Negative arrival longitude is west. Retain order, units and model.

Implication for the record

This is not travel time, road access or terrain-following distance. Euclidean degrees and Web Mercator length are different calculations. Preserve inputs and method; assess positional uncertainty separately.

Do many decimals prove an exact distance between cities?

They express numeric output for rounded inputs, not a unique official point or real-world positioning uncertainty.

Primary documentation

Documentation consulted on .

Record to retain

  • Code, label, level and classification version.
  • Membership, exclusions and validity period.
  • Correspondence direction and any weights.
  • Unmatched units and treatment method.
  • Geographic vintage, weighting populations and results on a constant boundary.

Official references

Related source profiles

Related protocols

Checks before reaching a conclusion

  • Do levels match?
  • Is membership dated?
  • Are multiple relationships explicit?
  • Are aggregates and members counted separately?

Frequently asked questions

Can a code draw a map?

It can join an appropriate layer, but supplies no coordinates or geometry. Check layer licence, vintage and level.

Can all variables share one weight?

Area weights do not automatically suit population or amounts. Justify weights against the variable and available data.

Which NUTS version should a series use?

Eurostat lists NUTS 2024 for data transmission in 2024–2026 and NUTS 2027 from 2027. Also inspect series metadata: a backcast series may use a common geography for earlier years. The observation year alone does not determine its geography.

Is it possible to make them comparable?

This depends on the indicator and the available data. Total and ratio are not always gathered in the same way; the rule must preserve the meaning of the numerator and the denominator.

TOOLS / REGISTERS

Information to record

Check duplicates, unmatched units and level. Never add an aggregate and its members together; maps also need coordinate system and layer date.

  • Access date and time zone
  • Exact URL
  • Dataset / series code
  • Exact name
  • Dimensions / filters
  • Reference period
  • Version / revision
  • Relationship / link direction
  • Method / definition
  • Limit / uncertainty
  • Decision / next action
  • Review owner
  • Next review

Apply the method to specific cases

Proposed exercise: this protocol is not a performed service test.

Record observations

DOCUMENTED REFERENCE

Population density: check the area in the denominator

Four countries and one observation year: this table preserves the 2023 WDI values retrieved on 23 October 2026. It supports checking a division before comparing territories. Population figures are midyear estimates, not counts taken on the retrieval date.

Lake Saimaa around Varpasalo island, Finland: a shipping lane crosses the ice between islands. Landsat 8 OLI image acquired on 8 April 2024.

Lake Saimaa around Varpasalo island, Finland: a shipping lane crosses the ice between islands. Landsat 8 OLI image acquired on 8 April 2024.

Real satellite image · Michala Garrison / NASA Earth Observatory · USGS Landsat · Source · Reuse

Image by Michala Garrison, NASA Earth Observatory, using USGS Landsat data. Resized and converted to WebP without cropping. Illustrates inland waters; no area measurement or NASA endorsement of this site.

Population density: check the area in the denominator
ReferenceValue or formWhat it identifies
SP.POP.TOTLPPopulation is the numerator. Its definition and coverage must match the area used.
AG.LND.TOTL.K2LLand area excludes inland waters and is expressed in km².
AG.SRF.TOTL.K2SSurface area includes inland waters and some coastal waterways. It cannot automatically replace land area.
EN.POP.DNSTP / LPublished density uses land area. The reader below repeats the division using values from the same snapshot.
WDI observations · 2023
Country / ISO3Population · peopleLand · km²Surface area · km²Published density · people/km² of land
France · FRA68372286538950606410126.862020595603
Germany · DEU83287273349430357680238.351810090719
Spain · ESP48352528499697.4505975.9896.7636173412149
Finland · FIN558391130396033848018.3705454665088
Published densities · people/km² of land
FRA126.86DEU238.35ESP96.76FIN18.37

Recalculate for a country

Finland · 2023 · Check the definition before interpreting the result.

Population ÷ land area
18.370545
Population ÷ surface area · alternative calculation
16.497019
Absolute difference from published density
0.000000

Local calculation from the 2023 snapshot. Results rounded to six decimals; chart rounded to two decimals.

Source update : · Retrieved :

Data: World Bank, WDI; upstream sources: FAO, UN, national statistical offices and Eurostat depending on the indicator. CC BY 4.0. Values preserved; table, chart and calculations added here.

Implication for the record

A national average describes neither where people live nor neighbourhood density. Do not derive water area by subtracting two series until their territorial coverage, conventions and revisions have been checked. The alternative ratio using surface area is a sensitivity demonstration, not a new official indicator.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

Area: convert m², hectares and km² without losing decimals

Changing an area unit does not change the territory measured. A kilometre is one thousand metres, so a square kilometre contains one thousand times one thousand square metres. The table and converter make that square explicit.

1 km × 1 km = 1 km²1 000 m × 1 000 m = 1 000 000 m²1 km² = 100 ha
Area: convert m², hectares and km² without losing decimals
ReferenceValue or formWhat it identifies
1 km²1 000 000 m²The factor 1,000 applies to each length: multiply km² by 1,000,000 to obtain m².
1 ha10 000 m²A hectare, symbol ha, is exactly the area of a square with 100 m sides.
1 km²100 haOne square kilometre contains exactly 100 hectares.
2.5 ha25 000 m² = 0.025 km²Arithmetic example: 2.5 ha equals 25,000 m² and 0.025 km². No real parcel is represented.

Convert an area

Enter a non-negative number, up to 40 characters. Decimal point or comma; no thousands separator or exponent. Results preserve the entered decimals without binary rounding.

Exact unit conversion.

m²
25000
ha
2.5
km²
0.025

Implication for the record

Conversion factors are exact but do not improve the precision of the original measurement. Unit conversion does not calculate polygon area, correct a projection or validate a cadastral boundary.

Primary documentation

Documentation consulted on .

DEEPER READING / PRACTICE

Check that the areas are comparable.

Two series describe territories whose contours or categories have changed. The comparison must specify the geography used and the validity period of the codes.

Updated on .

Priority checks

Name, code, geographical level and millesime. Examine contour changes and correspondence tables before grouping or approaching values.

Interpretation pitfalls

An identical name can indicate a different scope depending on the date. An aggregation of territories does not automatically retain the meaning of a ratio.

Working record to retain

The correspondence table documents initial territories, rule of approximation and loss of information. Non-comparable cases are kept separately.