Presenting it
A clean spreadsheet, your chosen variables and your statistical result if you ran one. Settle the numbers before you make any figures.
- At least five different presentation techniques, each chosen for a reason
- Every figure numbered, titled, labelled and integrated into the text
- Any statistical result presented where it is used, not in an appendix
- A locational map that has become a data map
"Most appropriate technique" is a judgement you have to make, and be seen to make.
The marks go to presentation that suits the data type, a sufficient range used effectively, and conventions followed. A spreadsheet's default chart is not a choice.
Use at least five different techniques (our rule), each chosen because it suits that variable. Four bar graphs and a pie chart is appropriate but limited in variety, which keeps a report in the middle band; the top descriptor asks for "a sufficient range … used effectively".
By hand or on a computer: both are fine. A hand-drawn graph or map counts as much as one made on screen if it is accurate, clear and follows the conventions. Scan it, or photograph it flat in good light.
On a computer, use the common tools: Excel, Google Sheets or Logger Pro for graphs, Felt for maps. No graph generators (such as those on geographyfieldwork.com); step 4's cross-section calculator is for checking arithmetic, not for making figures.
No AI-made graphs, maps or diagrams (your teachers' decision, not an IB rule).
Sketches of the form, not finished figures. They show each technique's shape and axis labels, with no values to copy. Build your own figures from your own spreadsheet, full size, with a title, units and a key.
Any variable against distance from source: your workhorse. The line of best fit shows the relationship and r² its strength; both are one tick box in Excel or Sheets (step 6 shows how to read r²).
Site number on the x-axis instead of distance: the commonest error in this IA, and it changes what the graph shows.
How the channel's shape changes, not just its size. Draw every site on the same vertical and horizontal scales (two shown here) so the change is real.
Different scales between sites, which makes them incomparable: on autoscale, these two panels would look almost identical.
Altitude against distance from source, the spine of the investigation. Add a second dataset at each site (here bankfull width as circles) to show in one figure where the sites are, how steeply the river falls and how the channel changes.
A bare profile, which shows only that rivers flow downhill. An exaggerated vertical scale not stated on the figure. Symbols hiding each other where sites are close (sites 5 and 6 are 40 m apart): make them translucent or offset them.
Discharge or velocity shown where it was measured: the technique that most marks out a Geography IA from a set of graphs.
Scaling by radius: set the radius from the square root of the value so the area carries the meaning. It is still a map: it needs all five BOLTS conventions, a credited basemap (see step 2) and a key to symbol sizes.
Categorical data, such as Powers roundness classes: frequency up the y-axis, the classes in order along the x.
Bars for continuous data such as distance downstream, where a scattergraph shows the relationship better.
- Very angular (0)
- Angular (0)
- Sub-angular (2)
- Sub-rounded (15)
- Rounded (12)
- Well rounded (1)
What one sample is made of. The roundness classes in one site's bedload are the only dataset here that fits: every stone falls into exactly one class and the classes add to 100%.
Almost everything else: discharge, velocity and width are not parts of a whole. And your roundness hypothesis is about change between sites, which paired bars show far better than two pies.
The thirty stones from each site as a distribution, not an average: cumulative percentage up the y-axis, stone size along the x, two or three sites per graph. Read the median at 50% and the interquartile range at 25% and 75%: a direct test of downstream fining, and a median per site to set against velocity or add to your long profile.
Using it for anything with one value per site: width, velocity and discharge belong on a scattergraph. More than three curves on one graph. Mixed units: use one unit per variable across your spreadsheet, figures and write-up.

Bed material, bank form, engineering, land use and, as here, the precautions that keep your readings comparable between sites.
Annotations that only name what is in the frame: say something a reader could not see for themselves. Keep each to 10 words or fewer; at 11 words the whole annotation counts towards your limit.
- May 2015: heavy rain, and discharge exceeds the channel
- Flooding on the Asse plain, at the railway and Paléo
- Flood-protection and renaturation scheme, from 2019
- About 4.5 km inside Nyon to be widened and reprofiled
- There, width and depth are a design decision, not distance downstream
- So a site there can sit off the trend, and you can explain why
A process or chain of reasoning you would otherwise spend eighty words on. This one reasons about our own river, the kind of chain Criterion D (Written analysis) rewards.
Decoration. A flow chart must replace prose, and a chain of things a reader already believes replaces nothing.
Everything below is how we suggest you actually do it.
Integrated means next to the sentence it supports
Presentation and analysis must be integrated: no page of graphs, no figures gathered at the end, no separate "Results" section. Each figure sits beside the paragraph that discusses it, and that paragraph refers to it by number.
Appendices are not read as evidence. Anything your argument depends on must be in the report, as a figure.
Make figures big enough to read on screen. At least half the page width for a graph, a full page for a map or for cross-sections placed on a map (our rule). A figure the examiner has to zoom into is evidence they may not see.
No tables of data in the body. Data reaches the reader as graphs and maps; raw data and calculations go in an appendix, referred to where it matters ("full data in Appendix 1"). A short table in words, summarising methods or comparing sites, is fine.
Plot against distance from source, never site number: the most common single mistake in this investigation.
Site numbers are evenly spaced; the sites are not. There is 2.3 km between the first two and 40 m between sites 5 and 6, yet a site-number axis draws both gaps the same. The question asks about distance from the source, so a graph against site number cannot answer it.
- Border
- Orientation
- Legend
- Title
- Scale bar
Every map in the report, locational or data, gets all five.
Annotate, don't just label
A label names what is there: riffle, bankfull level, bridge. Anyone could have written it.
An annotation says something about it: what it shows, why it matters, what it did to your data. It shows understanding, and it turns a figure into evidence.
Annotate photographs, sketches and diagrams (our advice). Labels are enough where naming is the job, on a sketch map or a cross-section; a photograph with only labels is a wasted figure.
A photograph of a site with three arrows: bridge, bankfull level, pool.
Correct, but nothing the reader could not see.
The same photograph, three notes: Bridge abutments fix the width here: expect an anomaly. Bankfull line hard to judge: banks vegetated. Pool, not riffle: slack velocity, unrepresentative of this reach.
Each note shows a judgement about the data.
An annotation of ten words or fewer is outside the word limit (IB rule), so five short notes on one photograph are all free.
At eleven words the whole annotation counts. So does text in a box beside a figure, and a caption that describes the figure. Titles and source lines do not count.
Cross-sections and the long profile
Channel cross-sections: plot depth against distance across the channel, with the water surface as a line at zero.
Use the same scales at every site, so the sections are directly comparable and the channel visibly grows downstream. Autoscaled sections all look the same shape, which hides the change you are investigating.
The long profile (optional): altitude against distance from source. Use it as a base: place proportional symbols, bars or annotations for a second variable at each site along it, so one figure shows location, gradient and the variable together.
The sites fall from 711 m to 413 m over 8.9 km: at true scale the profile is almost flat, so the vertical axis must be exaggerated.
State it on the figure, "vertical scale exaggerated ×20", and keep it the same on every profile.
Located data, on a map
Your data has coordinates, so some of it belongs on a map. Proportional symbols, a circle at each site sized by discharge, velocity or bedload, show magnitude at located points. For a steady trend downstream a scattergraph is clearer: use each for what it does best.
In Felt: load your sheet, size the symbols by a variable, set the colours and legend, choose a basemap, and export the view as an image with the legend.
A circle four times the radius looks sixteen times the size. Make area proportional to the value, check your largest and smallest symbols against the ratio of the values, and say in the caption that symbols are scaled by area.
- Border
- Orientation
- Legend
- Title
- Scale bar
All five, plus the basemap credited and a key to the symbol sizes.
Share the Google Sheet with "Anyone with the link", then paste the link into Upload Anything → From URL. A private sheet fails silently.
The sheet needs latitude and longitude columns in decimal form, `46.409795`, not `46° 24' 35.26" N`. Right-click a point on map.geo.admin.ch and take the WGS 84 (lat/lon) pair.
What the 400 words are actually for
Not describing your figures: the reader can see them. Spend the words on decisions: why this technique for this variable, why these scales, what you did about a site left off a graph, why Spearman's rank, and what its result means for the hypothesis.
"Figure 6 is a bar chart showing the average bedload size at each of the nine sites, measured in centimetres, with site number along the bottom."
Twenty-six words a glance at the figure supplies.
"Bedload was plotted as a scattergraph against distance rather than a bar chart by site, because the hypothesis concerns a continuous change with distance and bars imply discrete categories. Roundness, being an ordinal Powers class, is shown as paired bars instead."
Two decisions, each with its reason.
Put a test result in the analysis where you use it, not in an appendix. Give the coefficient, whether it is significant, and what that means for your null hypothesis from step 1.
It can react to a figure you have already made: you describe it and say what it should show, and ask whether it does.
It cannot choose your techniques or make your figures. Choosing the right technique for each variable is what Criterion C (Quality and treatment of information collected) awards.
Ready for step 6?
0 of 18Read your figures as a stranger would. Any figure you would have to explain out loud is not finished.
Step 6 is the written analysis: the largest criterion and the largest share of the word count.
Anything marked “our river” is specific to our fieldwork on the River Asse at Nyon. Photographs are the author’s own.
