3.4, 3.4.1, 3.4.2.5, 4.1, 4.2 - Critical data use and evidenced geographical arguments

Integrating evidence with enquiry

Skills gain geographical meaning when used to investigate a place, process or interaction rather than performed as isolated calculations.

For a fictional question, 'How has riverside redevelopment affected access and flood risk?', a land-use map locates change, comparable water-level readings investigate physical conditions, pedestrian counts describe activity and interviews explore access or belonging. Quantitative data describe amounts and patterns; qualitative data describe meaning and experience. Neither substitutes for the other: a count cannot explain an avoided route, and a vivid quotation cannot estimate a whole population's frequency. A roughly balanced experience of both approaches across the course develops different forms of reasoning, though each particular enquiry uses a justified mixture rather than a fixed quota.

Primary data are collected first-hand for the enquiry, such as your own readings or voluntary interview material. Secondary data were collected or produced by others, such as an existing flood record or census table. Either can be numerical, spatial or qualitative and either can be strong or weak. First-hand collection does not make a narrow convenience sample representative; an authoritative secondary source still needs date, unit, definitions and coverage checks.

Images record selected views; factual text may document a planting count; discursive campaign material argues for an interpretation; creative writing can convey identity or loss. Digital numerical tables, geo-located records and satellite classifications offer further evidence but preserve the limitations of their collection and processing. Crowd-sourced reports and big datasets add scale and speed while posing coverage and verification questions.

Start with a manageable question, define geographical units and periods, choose complementary methods, collect or select evidence, analyse with an appropriate technique, then interpret what it can establish. A river transect might connect measurement, GIS and dispersion to a channel-process question; a place enquiry might connect text, coding and mapped access to identity. Adapt the training design to lawful feasible access around your centre, documenting missing evidence instead of claiming work that has not occurred. Fieldwork generates new knowledge by testing ideas against direct observation and experience, while its selected places, times and protocols delimit that knowledge.

Databases and reproducible ICT

Electronic databases enable selection, comparison and joins across many records, but errors can be amplified unless the stored units and transformations are checked.

A table should have defined fields and a stable unique record identifier. The illustrative structure uses site ID, collection date/time, coordinates and reference system, count duration, pedestrian count and an explicit missing-value flag. A missing count is not zero. A GIS join using site ID can combine a site table with spatial locations; duplicate IDs may multiply rows, and mismatched IDs may omit sites. Check record totals before and after filtering/joining, the geographical boundary and the period represented. A national table cannot automatically resolve variation within one street.

For an original spreadsheet count of 15 pedestrians in three minutes, rate = count ÷ duration = 5 pedestrians/minute. A count of 25 in five minutes gives the same rate. Converting permits a duration comparison but assumes the windows are suitable; it cannot remove unequal times of day. Use explicit units and formulas, keep raw data separate from derived fields and document filters and corrections. Spot-check a result by hand before trusting a full-column formula. Do not silently average text-coded missing entries as zeros.

ICT can produce graphs and statistical calculations efficiently. Choose the graph for the data, check scale and zero baseline, inspect legends and units, and ensure proportional symbols encode area correctly. For statistics, verify paired IDs, tie handling, quartile convention, supplied SD denominator and the chosen significance reference. A default function can implement a different valid convention from the intended task. Save the version, formula and unrounded intermediate results so that another reader can reproduce the transformation.

For ungraded software practice, enter the two original duration/count pairs above, calculate their rates and produce a labelled comparative bar graph. Check both heights are 5 and retain durations in an accompanying note. Then consider whether matched-duration data would answer a different or better question. Written interpretation assesses these checks; a screenshot alone does not demonstrate that the input, formula or population scope was correct.

Crowd sourcing, big data and power

Large or rapidly updated datasets can reveal patterns while systematically excluding people and places whose activities are less digitally visible.

Crowd sourcing obtains contributions from many participants, such as voluntarily geo-located reports of flooded paths. It can identify emerging local problems, but reporting depends on awareness, connectivity, incentives and repeated submissions. More reports can indicate more participation rather than more flooding. Verify time, location, duplication and independent observations; compare against a suitable exposure denominator and other evidence.

Big data refers to datasets whose scale, speed or variety challenge conventional handling, such as location traces from many devices. A million pings are not necessarily a million people: one device can contribute repeatedly, several devices can belong to one person, and some residents have none. A precise coordinate can be technically accurate while the sample of residents is biased. Spatial clustering and repeated units can undermine statistical independence. Compare coverage across groups and periods before mapping totals as population risk or movement.

Quantitative evidence about communities also has ethical and socio-political consequences. Location-time combinations can reveal home, workplace or sensitive visits even if names are removed. Collect only necessary information with an appropriate lawful, transparent basis, minimise precision where suitable, restrict access and assess whether publication can identify or stigmatise groups. Aggregation reduces some risks but is not an automatic guarantee. Permissions for data use are different from consent to identifiable public display.

Who owns the platform, defines a 'risk area', funds the analysis or controls access can determine whose experience influences policy. A sparse reporting map might lead to less maintenance in a poorly connected neighbourhood even if its exposure is high. Document the denominator, coverage and uncertainty, involve relevant perspectives and avoid turning absence of records into absence of need. Combining verified physical observations with community accounts can expose this mismatch without pretending that one dataset represents everybody.

Scale, systems and causal reasoning

A causal explanation connects a process to an outcome through a defensible mechanism while considering other influences, interactions and scale.

In the fictional riverside enquiry, place includes the site's physical characteristics and lived meanings; space concerns distributions and connections, such as which homes connect to an accessible route. Scale identifies the extent and resolution of analysis, from a tunnel entrance to a catchment or global supply chain. Environment includes physical conditions and their interactions with people. A town-wide mean can conceal inaccessible local paths, while a single flood-window observation cannot establish a long-term catchment trend. Match evidence to both temporal and spatial scale.