Fieldwork, investigation and geographical data skills

The route to geographical enquiry: seven steps from question to conclusion

Geographical investigation follows a fixed sequence, and examiners expect you to know where you are on it. Step 1: identify the issue, question or problem and turn it into a hypothesis or a set of guiding questions. A topic is recognised through observation, discussion, reading or previous study, and the hypothesis or question is designed to test it. Step 2: define the objectives of the study in specific terms, deciding what data are relevant to the study and how those data can be collected.

Step 3: collect the data. Fieldwork gives you primary data: questionnaires, interviews, mapping, sketching, observation, counts and measurements. Secondary data complement it: census information, the internet, published maps, books, newspapers and magazines, used for comparison or to introduce the topic. Step 4: select and collate the data, deciding which of the figures you have gathered are relevant to the hypothesis and organising them so they can be worked with.

Step 5: present and record the results in appropriate forms, using a variety of maps, graphs and diagrams, each chosen for the data it has to show. Step 6: analyse and interpret the findings in response to the original question, with reference to relevant geographical concepts - the processes and ideas from the content topics. Step 7: draw conclusions and evaluate. State what the data show about the original question and the hypothesis, comment on the strengths and limitations of the data, suggest improvements, and suggest possibilities for further study.

• 1 Identify the issue, question or problem, and design the hypothesis or guiding questions.
• 2 Define the objectives: what data are relevant, and how they will be collected.
• 3 Collect the data: primary fieldwork plus secondary sources.
• 4 Select and collate the data relevant to the hypothesis.
• 5 Present and record the results using maps, graphs and diagrams.
• 6 Analyse and interpret the findings with reference to geographical concepts.
• 7 Draw conclusions, evaluate the strengths and limitations, and suggest further work.

Hypotheses, sampling, sample size, accuracy, reliability and control groups

A hypothesis is a testable statement that the data can support or reject, such as 'the velocity of the river increases downstream from the source'. It names what you will measure and the direction you expect, so that at the end you can say plainly whether the data supported it. An aim states what the investigation sets out to find out, and the objectives break that down into the specific tasks and measurements that will achieve it.

A sample is the part of the whole you actually measure, and it has to stand for the whole. Sample size is the number of measurements, sites or respondents. Too few and a single unusual reading carries the whole conclusion; more samples reduce the effect of any one odd value and make the estimate of the whole more reliable. Decide the number after a trial run, and be able to justify it: 'we sampled every 10 m along the transect because a pilot showed that the vegetation changed over about that distance' is a reason; 'we did ten because we ran out of time' is a limitation to admit, not a justification.

The way you choose the sample matters as much as its size. Random sampling gives every part of the area an equal chance of being chosen, usually from random numbers, which removes the bias of picking the convenient or interesting places. Systematic sampling takes samples at fixed intervals along a line or across a grid, which is quick, repeatable and good for following a change across a site. Stratified sampling divides the area into groups that matter, such as land use types or housing age, and samples each group in proportion, so that a small but important group is not missed. Choosing where to sample by eye as you go is none of these, and it is the commonest reason a school survey produces a pattern that means nothing.

Two words are used loosely in everyday speech and precisely in an investigation. Accuracy is how close a measurement is to the true value: a tape measure held slack, or a noise meter held at a different height each time, makes the readings inaccurate. Reliability is how consistent the results are: repeating the measurement and getting the same answer makes it reliable, so reliability is improved by repeats, by a larger sample, by a standard method and by the same observer using the same equipment throughout. Data can be reliable and still inaccurate if everyone makes the same mistake in the same way, which is why equipment is calibrated and the method is checked.

A control group or control site is the comparison you measure against. If you are investigating the effect of a footpath on vegetation, the control is the same habitat away from the path; if you are surveying the effect of a new leisure centre on a town, the control group might be residents who do not use it. Without a control you cannot say whether the difference you found was caused by the factor you are studying or would have been there anyway. Finally, be alert to bias: leading questions in a questionnaire, only asking the people who look approachable, sampling one side of a street, or working only in the places you can easily reach all push the results in a particular direction.

Fieldwork equipment and how to use it

The syllabus names the equipment you must know how to use, and the list is not exhaustive, so you should also be able to choose equipment relevant to your own fieldwork. The marks come from saying what the instrument measures, how it is used correctly, and what has to be kept the same each time so that the readings can be compared. Read the table below as a set of instructions rather than a list of names.

Fieldwork techniques and designing a data collection sheet

An environmental quality survey scores a set of features at points along a route against a fixed scale, usually 1 to 5, and adds or averages the scores to give an environmental quality index for each point. The features are chosen to be things you can see and judge: litter, graffiti, dog mess, noise, traffic, the condition of buildings, the amount of greenery and the upkeep of open space. A bi-polar survey does the same thing differently: each feature is scored on a scale between two opposite descriptions, such as 'quiet 3 2 1 0 1 2 3 noisy', and the scores are plotted as a pair of bars on either side of a central axis for each site. The result is a shape that can be compared at a glance between sites. Both surveys depend on judgement, so the scale must be defined before the survey starts - what exactly counts as a 1 and as a 5 - and either the same observer scores every site or two observers agree a score together.

Questionnaires and interviews collect data from people. A questionnaire uses closed questions with fixed answers so that the replies can be counted, and a set number of respondents chosen by a stated method; an interview is face to face, allows follow-up questions and gives richer answers, but fewer people can be covered and the interviewer's manner can influence what is said. Neither should use leading questions such as 'do you agree that the town centre is too crowded?', and a pilot run with a few respondents will show up questions that are ambiguous or that nobody can answer.

Counts and surveys of the built environment are systematic. A traffic count records the number and type of vehicles, with their direction, at a fixed point for a fixed period, at the same times on the same day of the week, avoiding school holidays and unusual days. A pedestrian count works the same way for people, classified by direction and, if relevant, by age or gender. A land use survey records the use of every building or plot along a street or inside a grid square, using a coded key such as R for residential, S for shop, O for office, so that the data can be plotted on a map and the pattern of the CBD or the urban fringe can be seen. A transect is a line across a site along which measurements are taken at fixed intervals, and it is the standard technique for any change across a distance: down a beach, across a river valley, from the edge of a wood into a field.

Some techniques are about recording what you see. A field sketch shows the key features of a view with labels, and it is better than a photograph when the view is too complex to photograph or when you need to show what matters rather than what is in the way; it should have a title, a direction and simple labels. A photograph should be taken from a marked point and a stated direction so that a later visit can be compared with it, and labelled with the location, the direction of view and the date. Delimiting the sphere of influence of a settlement or the boundary of its CBD can be done with a questionnaire asking where people have travelled from, or by walking a street until the land use changes from shops and offices to housing, and marking where that happens. A roundness index is one of the indices used on material rather than people: the longest, intermediate and shortest axes of a pebble or stone are measured, and the index is calculated from them, so that a pebble close to a sphere scores close to 1 while a long, angular pebble scores lower. The same idea of a calculated index runs through fieldwork: an environmental quality index, a river discharge, a percentage of a total.

A data collection sheet is designed before you leave, and it decides how good your data can be. Give it a title, and space for the date, the time, the weather and the name of the recorder. Put one sample point on each row and one variable in each column, with the unit written in the column heading, so the sheet can be filled in quickly without writing sentences. Leave a column for notes, where the reason for an unusual reading can be recorded at the time - a quadrat that landed on a path, a count interrupted by a delivery lorry. Use a key or a code for categories, keep the sheet on a clipboard with a pencil, and record the raw readings only, working out means and percentages later so that an arithmetic error can be traced. Pilot the sheet: a sheet that has no space for a reading you need, or that cannot be read in wind and rain, wastes the whole day.

• Environmental quality survey: features scored against a defined scale and added to give an index.
• Bi-polar survey: each feature scored between two opposite descriptions and plotted as bars on either side of a central axis.
• Questionnaires: closed, piloted questions and a stated method of choosing respondents; interviews cover fewer people in more depth.
• Counts: traffic, pedestrians and land use, taken at fixed points for fixed periods, at the same times on the same day of the week.
• Transects: measurements at fixed intervals along a line, used wherever something changes across a distance.
• Field sketches and photographs: labelled, dated, and taken from a stated point and direction so they can be repeated.

Presenting and analysing the data

Present the results so that the pattern is visible, and choose the form for the data rather than for decoration. A table comes first: the raw readings with their units, then the processed values such as means, percentages and totals beside them, with a title and a clear column heading for every variable. Then a graph or map chosen to show what the data are doing - a line graph for a trend over distance or time, a bar graph for a comparison between places, a scatter graph for the relationship between two variables, a pie chart or divided bar for the parts of a total, a choropleth or flow line map for a spatial pattern. Label the axes with their units, use a scale that suits the data, and annotate the graph with the names of the places and the highest and lowest values so that it can be read without the text.

Analysis means more than describing. Identify and describe the relationships between two or more sets of data: does environmental quality rise as the distance from the town centre increases, does pedestrian flow fall away from the main shopping street, does river velocity increase downstream? Then explain the relationship using geographical ideas - the processes from the content topics - and use the data as evidence, quoting figures with their units. Trends over time are read from a line graph or a sequence of maps, and a line of best fit drawn through the points shows the overall direction of the change even when individual years are irregular. Where a scatter graph is used, the line of best fit shows whether the relationship is positive, negative or absent, and any point far from the line is an anomaly worth explaining rather than ignoring.

A line of best fit also allows prediction. Interpolation means reading a value from inside the range of the data you have, which is reasonably safe. Extrapolation means extending the line beyond the data, which is only a prediction: a trend that has risen for five years need not keep rising for another five. Always state which of the two you are doing, and say how confident the prediction can be.

Finally, look for the weaknesses and limitations in the statistical presentation itself. Check whether the vertical axis starts at zero or has been broken, because both exaggerate change. Check that class intervals are equal where a histogram is used. Check that percentages are calculated from a sensible total, that the units are given, and that a mean is not quoted without the range, because two sites with the same mean can be completely different. A sample of twelve respondents cannot support a claim about a whole town, and a survey taken on one afternoon cannot describe a normal week.

Mathematical skills: number, indices, significant figures, proportion and averages

Number, area and scale underpin everything else. You need to add, subtract, multiply and divide without a calculator error, to work out the area of a rectangle or a cross-section from length and width, and to convert between map and ground distances using a scale. The commonest error is a units error: keep the units attached to the numbers as you work, and convert to the same unit before you calculate.

Standard notation uses powers of ten to write very large or very small numbers compactly, and you must be able to use both positive and negative indices. A positive index means the number of places the decimal point moves to the right: 10 to the power 3 is 1 000, so 4.2 x 10 to the power 3 is 4 200. A negative index means the number of places it moves to the left: 10 to the power minus 3 is 0.001, so 5 x 10 to the power minus 3 is 0.005. This matters for figures that would otherwise be unwieldy, such as a national population in the millions or a concentration of a pollutant in milligrams per litre.

Significant figures are the digits that carry real information. Round your answers to a sensible number of significant figures - three is usual - so that the figure does not claim more precision than the measurement has. If a tape was read to the nearest centimetre, quoting a velocity to six decimal places is meaningless. When you round, round only at the end of the calculation, not at each step.

Proportion, ratio, magnitude and frequency are the language of comparison. A proportion is a part of a whole, often written as a fraction, a decimal or a percentage; a ratio compares one quantity with another, as in a gradient of 1:25 or a ratio of three cars to one bus; magnitude is the size of a value, so an order of magnitude is a tenfold difference; and frequency is the number of times a value occurs, which is why a frequency table counts how many readings fall in each class. If you record the same reading many times, its frequency is high.

The mean, the mode and the median are the three averages, and the range measures spread. The mean is the total of all the values divided by how many there are. The mode is the value that occurs most often. The median is the middle value when the values are put in order, which is why the data must be sorted first. The range is the highest value minus the lowest. Worked example: a student records five noise readings of 62, 65, 71, 68 and 64 decibels. The mean is 330 / 5 = 66 dB, the median is 65 dB because the ordered values are 62, 64, 65, 68, 71, and the range is 71 minus 62 = 9 dB. Report the mean with the range, because the range shows how variable the readings were: a mean of 66 with a range of 9 describes a very different street from a mean of 66 with a range of 40.

Decimals, fractions, percentages and ratios should be used interchangeably: 0.25, one quarter, 25 per cent and the ratio 1:4 all describe the same thing. To convert a fraction to a percentage, multiply by 100. To work out the angle of a sector on a pie chart, multiply the percentage by 3.6, because the whole circle is 360 degrees. Finally, draw and justify conclusions from numerical and statistical data: state the finding, give the figures that support it, say how strong the relationship is and how much of the data it is based on, and say what it means geographically.

GIS and image skills: photographs, satellite images, text and bias

Photographs come in three forms and each one answers a different question. A ground-level photograph is taken at eye level and shows a small area in detail - the state of a pavement, the materials in a sea wall, the type of housing on a street - but it cannot show the wider pattern. A vertical aerial photograph is taken looking straight down, so it looks like a map and can be used to measure distance and to map land use, but the features are unfamiliar from above and relief is hard to see. An oblique aerial photograph is taken at an angle, which shows the shape of the land and the sides of buildings and is easier to interpret, but the scale changes across the image, so the foreground is much larger than the background and distances cannot simply be measured from it.

To deconstruct a photograph, work through it in a fixed order. Identify the type of photograph and, for an oblique view, the direction it is looking. Use the foreground, middle ground and background to organise what you see, then identify and label the features: the buildings and their materials, the transport, the vegetation, the water, the industrial or commercial activity. Then say what the photograph shows about a process or a pattern, such as the growth of a settlement along a road, the evidence of erosion at the base of a cliff, or the contrast in housing quality between two parts of a city. Finish with what the photograph cannot tell you, because it is a single view at a single moment and may have been taken to make a particular point.

Satellite images cover huge areas and can be repeated over time, which makes them the best evidence for change: the growth of a city, the retreat of a glacier, the spread of deforestation, the spread of a desert. True colour images show what the eye would see; false colour images give the different wavelengths their own colours so that features such as healthy vegetation or water become easy to distinguish, which means you must read the key before interpreting one. Their limits are the resolution, which decides whether a feature can be seen at all, cloud cover, the cost of high-quality imagery, and the fact that an image records shape, colour and extent but rarely the use of a building or the name of a crop.

Cartoons, pictures and diagrams carry a point of view. For a cartoon, identify the issue, the symbols and labels used, and the attitude of the person who drew it - it is a comment on a situation, not evidence about it. For a diagram, check what it shows, whether it is a cross-section, a plan or a flow diagram, and whether the labels explain the process or only name the parts. Written text must be read the same way. Ask who wrote it, who it was written for, whether it is fact or opinion, what has been left out, and what the choice of words is doing: 'a controversial development' and 'a long-awaited new housing estate' can describe the same thing. Recognising that bias is a skill in its own right, and a question that asks you to compare two sources is testing exactly that.

A GIS (geographical information system) is a computer system that stores geographical data in layers, one layer for each kind of information - roads, rivers, land use, population, flood risk, school catchments - so that the layers can be displayed, combined and analysed together. You use GIS to identify trends, patterns and issues and to solve problems. Typical tasks are measuring a distance or an area accurately, counting how many people live within a given distance of a site, finding all the places that meet several conditions at once, and producing a map of the result. A practical example is choosing a site for a new facility: overlay the flood risk layer, the road network, the existing facilities and the population layer, and query the system for areas that are above the flood line, within a short distance of a main road and more than a set distance from a competing facility.

The benefits are that large amounts of data can be stored and updated, layers can be combined in ways that would take weeks by hand, patterns and relationships that are not obvious in a table become visible on a map, results can be shown at different scales, and a decision can be justified by showing the evidence behind it. The limitations are that GIS needs expensive software, a computer, and someone trained to use it, so it is not available to everyone; the output is only as good as the data put in, and data can be out of date, incomplete or mapped at a resolution too coarse for the question; the map that comes out looks precise and convincing even when the data behind it are weak; and it cannot record what people think, only what is already in the database.

• Ground-level photograph: detailed view of one small area. Vertical aerial photograph: looks like a map and can be measured. Oblique aerial photograph: shows relief and buildings but the scale changes across the image.
• Satellite images are best for change over large areas and long periods; read the key first, because false colour does not show what the eye would see.
• For any image or text, ask who produced it, for whom, and what is left out - that is where the bias shows.
• GIS stores data in layers, which can be combined and queried to find patterns and to choose between options.
• GIS benefits: fast analysis of large data sets and clear maps for decisions. GIS limits: cost and training, and output only as good as the data.

Drawing conclusions and evaluating your investigation

A conclusion answers the original question and returns to the hypothesis. State plainly whether the data support it, reject it or support it only in part, and give the evidence: the figures, the places and the direction of the change. If the relationship is strong in one part of the data and absent in another, say so rather than smoothing it over. A conclusion that says 'the hypothesis was correct' with no figures cannot be credited.

The evaluation is a judgement about the method, and it should work through the same headings every time. How big was the sample, and was it large enough? How was it chosen, and could the choice have biased the results? Were the measurements accurate, and was the equipment calibrated? Were the readings repeated, and did they agree? Were the variables that should have been controlled actually controlled - the time of day, the weather, the season, the observer, the distance between sites? Did the technique actually measure what the investigation set out to measure? Was there a control site or a control group to compare against?

Then give improvements that would change the result rather than advice to be more careful: more sample points or a larger sample of respondents, a different sampling method, a longer transect, repeat readings at each site, a second observer to check a subjective score, sampling at the same point in the day and the week, repeating the survey in a different season, calibrating the equipment, or measuring an additional variable to test the explanation. Each improvement should be tied to the weakness it fixes. Finish with further work: what you would investigate next, and why it would help - a different site for comparison, a longer time period, or a second technique on the same question. A good evaluation ends with a judgement about how far the conclusion can be trusted, and that judgement is what the highest marks are given for.