Appendix 2 3, 4; 9GE0/04 - Designing methods, sampling and safe fieldwork

Methods that address the question

A method is useful when the information it produces can answer a defined part of the geographical enquiry.

Begin with the concept, identify a suitable indicator, and decide exactly what will be observed. A count of people entering a park measures movement during a defined period. An interview can reveal how a participant experiences access to the park. Neither directly measures all aspects of the park's social value. Combining relevant quantitative and qualitative evidence can investigate both a pattern and people's explanations of it; collecting more kinds of data without a purpose does not improve an enquiry.

Distinguish accuracy, closeness to the true value, from precision, how closely repeated measurements agree. Resolution is the smallest increment the instrument can distinguish. A miscalibrated instrument can give closely clustered readings that are consistently wrong; displaying more decimal places does not ensure accuracy. An environmental score can be reliably applied by observers but still be a weak indicator of an abstract concept such as community identity.

Describe the operational procedure sufficiently clearly that another collector could follow it: observation unit, equipment, units, categories, location rule, duration, repeat procedure and recording method. Explain the choice in terms of its expected evidence and limitations. For example, a fixed observation line avoids collectors making different judgements about whether someone “used the area.” A limitation is not a reason to conceal the method; it sets the boundary of the later conclusion.

These are general training principles and a repertoire of methods. For your actual investigation, make and justify choices relevant to your own question. Collaboration in methodological planning is allowed, but a lesson cannot supply a bespoke method that removes your ownership of the investigation.

Sampling frames and selection

Sampling selects a feasible subset from a clearly defined population or study area, so the selection process shapes what the evidence can represent.

In a fictional training frame, 20 of 50 mapped segments are residential. For a proportional stratified target of 15 segments, the residential allocation is:

Random selection within each stratum preserves a clear selection rule. Equal allocation can instead be useful when comparing small and large groups, but it does not reproduce their population proportions: explain that choice and avoid treating an unweighted average as the population average.

Map the boundary and record inaccessible or unsafe units. Safety takes priority over a sampling plan; use a pre-agreed safe replacement rule where appropriate and disclose the exclusion. Quietly moving every selected site to the easiest roadside location turns the actual sample into a different one. Random selection does not justify entering unsafe or unauthorised sites.

Timing, frequency and independent observations

A spatially balanced sample can still be misleading if it observes only one unrepresentative moment.

Timing determines which processes or groups are observable: tide state affects a coastal profile's surveyed extent, preceding rainfall affects infiltration, and time of day changes pedestrian populations. State dates, start and end times and relevant conditions. Repeat observations across justified periods when the question asks about change or variability. Comparable observations need equivalent definitions and durations; convert unequal counting periods to rates only when that comparison is defensible.