Fieldwork Investigations

Planning an investigation: the aim, the hypothesis and the pilot study

Every fieldwork investigation begins with a question about the real environment. The aim is a single sentence stating what the investigation sets out to find out, such as 'to find out how plant species richness changes from the strandline to the crest of a sand dune'. The aim fixes where you work, what you measure and how long you spend, so a vague aim such as 'to look at the plants on a dune' cannot be investigated and cannot score.

A hypothesis is a testable prediction. It names the two variables you are measuring and states the direction of the change you expect: 'plant species richness will increase with distance inland from the strandline'. Because it predicts a direction, it can be proved wrong by the data, which is what makes it scientific. A statement such as 'dunes are interesting' can never be tested, and 'distance inland affects plants' gives no direction and no measurable expectation.

Before the main survey, carry out a pilot study: a small trial of the whole method at the site. The pilot tells you whether the quadrat size suits the plants growing there, whether the recording sheet works in wind and rain, how long each site takes, whether the planned route is safe and passable, and roughly how variable the habitat is, which sets the number of samples you need. It also lets you standardise the timing, because if one site is sampled in the morning and another in the late afternoon the light, temperature and animal activity are no longer comparable.

Not every investigation counts organisms in a frame. A survey collects information by systematic observation or by asking people, which is where questionnaires belong. A questionnaire used to investigate how many visitors use a nature reserve, how far they travelled or what they think of the facilities must use closed questions with fixed answers so that the replies can be counted, must be piloted so that leading or ambiguous questions are removed, and must not be handed only to the visitors who look approachable, because that choice biases the sample.

• Aim: one sentence stating what the investigation is trying to find out.
• Hypothesis: a testable prediction that names the independent and dependent variables and the direction of the expected change.
• Pilot study: a small trial of the method, timing and recording sheet before the main survey.
• Survey or questionnaire: used when the data come from people or from systematic observation; closed, piloted questions keep the answers countable.

Variables, repeats, replicates and sample size

The independent variable is the thing you choose or change deliberately: the distance along a transect, the depth of the water, the pH of the soil, or which of two fields you sample. The dependent variable is what you measure, and it depends on the independent variable: the percentage cover of marram grass, the number of mayfly nymphs in a kick sample, the turbidity of the water. State both at the start, because they decide how the graph is drawn and what the conclusion can claim.

Control variables are everything that must be kept the same so the comparison is fair: the size of the quadrat, the time of day, the depth at which the water sample is taken, the season, and even the person estimating cover, because two students will judge percentage cover differently. Some control variables cannot be held constant out of doors, so they are measured instead - air temperature, light intensity and wind speed recorded at every sample point. Recording them lets you show afterwards that conditions were similar at each site, and therefore that they are unlikely to explain the pattern you found.

Repeats are repeat measurements of the same thing, made to reduce the effect of random error: reading the turbidity meter three times at one point and averaging the readings. Replicates are separate samples spread through the habitat: the ten quadrats placed across a field, or the five kick samples taken at intervals along a stream. Both are needed. Replicates show how much the population varies from place to place, and repeats show how much the measurement itself varies. One quadrat and one water sample give single numbers with no idea of the spread, and one unusually high value would then carry the whole conclusion.

Sample size follows from the variability of the habitat and is decided after the pilot study. A uniform mown lawn needs fewer quadrats than patchy dune vegetation to give the same confidence, and the usual practical rule is to keep adding quadrats until extra samples stop changing the running mean much. In an exam answer, a larger number of samples is more representative of the whole area and reduces the influence of any one anomalous value.

Random sampling and systematic sampling

Random sampling means that every part of the study area has an equal chance of being sampled. In practice you divide the area into a grid, use a random number generator or a table of random numbers to give the coordinates of each sample point, and put the quadrat there - even if the spot looks bare, difficult or uninteresting. That is the whole point of it: random placement removes the unconscious bias of choosing the places where your species is easy to see, which is the commonest way a school survey goes wrong.

Systematic sampling means that the samples follow a fixed pattern. The usual form is a transect: a tape is laid in a straight line across the habitat and a quadrat is placed at regular intervals along it, every 2 m or every 5 paces. The pattern is decided before sampling begins and repeated identically at the next site, so two sites can be compared fairly and another group can repeat the survey from the same instructions and get comparable data.

The question decides which strategy you use. Choose random sampling when you want to estimate the population of a species, or the species richness of an area, and to compare one area with another without bias. Choose systematic sampling when you expect the habitat to change across the site, because a transect actually measures that change: across a sand dune from strandline to crest, from the water's edge up a shingle bank, or from a river bank into woodland. On a transect the distance along the tape becomes the independent variable, and the pattern it reveals is called zonation.

• Random sampling: unbiased and representative of the whole area, so it is used to compare sites or to estimate a population; but it can miss rare or patchily distributed species, it may place samples in places that cannot be reached or worked, and many samples are needed before the estimate is reliable.
• Systematic sampling: quick, easy to repeat and it reveals a gradient or zonation that scattered random quadrats would blur; but a single line can coincide with an unusual feature, it can miss variation running parallel to the tape, and a fixed pattern is unrepresentative if the habitat is patchy rather than graded.
• The strategy must be chosen before the first quadrat is placed and then followed exactly; deciding where to sample as you go is neither random nor systematic.

Quadrats: percentage cover, frequency and counts

A quadrat is a square frame, usually 0.5 m x 0.5 m or 1 m x 1 m, placed on the ground so that the organisms inside it can be recorded. It is the standard tool for sampling plants and other organisms that cannot move away, and it works with either strategy. The same quadrat size must be used throughout, otherwise the figures from different sites are not comparable and the comparison fails.

There are three ways of recording what a quadrat contains. For small, countable plants or sedentary animals, count the number of individuals in the frame. For plants that creep and overlap, such as grass, clover or moss, count percentage cover: use a frame divided into 100 small squares and record how many of them contain the species, so 34 squares occupied means 34% cover. For a species that is scattered and hard to count or estimate, record frequency, the proportion of quadrats in which the species occurs at all: frequency = (number of quadrats containing the species / total number of quadrats) x 100.

Worked example. A student places twelve 1 m x 1 m quadrats on a lawn. Plantain is present in nine of them, so its frequency is 9 / 12 x 100 = 75%. Inside one quadrat, plantain covers 34 of the 100 small squares, so its percentage cover in that quadrat is 34%. The two figures answer different questions: frequency says how widely the species is spread through the lawn, and percentage cover says how much of the ground it occupies. A small creeping plant can therefore have a high frequency and a low percentage cover at the same time.

Every technique has limits, and the limits earn the marks. A quadrat records only the ground it covers, so too few samples can miss a species that really is present; the roots and overhanging leaves of a large plant stand outside the frame; mobile animals such as beetles and spiders run out of the frame before they are counted; and estimating cover by eye is subjective, so the same observer should record every quadrat, or two students should compare estimates until they agree. The frame is also placed by someone walking through the habitat, and trampling changes the community you are trying to measure, so sample lightly and keep to the same route.

Transects and the techniques for sampling animals

A line transect records what lies along a tape: the species touching the tape itself, or the species seen within each section of it. A belt transect records a strip of fixed width laid along the same line, usually divided into square sections that are treated as quadrats, which gives abundance as well as presence. A point quadrat is a frame holding ten pins, or a single pin dropped at fixed intervals along the tape; you record only whether a pin touches each species, which produces fast, repeatable cover data with no judgement about how much leaf lies inside a frame. Quadrats placed at fixed intervals along a tape are the commonest exam answer, and they are the practical version of a belt transect.

Where the animals cannot be counted from a frame, traps and nets are used. A pitfall trap is a smooth-sided container sunk so that its rim is level with the soil surface; ground-living invertebrates such as beetles, woodlice and ants walk over the edge, fall in and cannot climb out. A pooter is a small pot with two tubes: you suck on the mouth tube, which draws the animal up the other tube into the pot, so small, delicate animals such as aphids and small flies can be taken from a leaf unharmed. A sweep net is swept through tall grass or across a bush in a standard way - the same number of sweeps, at the same height, at the same speed - so that samples from different places are comparable, and it catches the insects living on the vegetation: beetles, bugs, grasshoppers and moths.

Kick sampling is the freshwater equivalent of the sweep net. You stand in the stream with a net held just downstream of your feet and kick the river bed in front of you for a fixed time, so that the invertebrates dislodged from the stones are washed into the net. The animals caught are identified with a key, and the result is used as a measure of water quality: a sample containing mayfly and stonefly nymphs, caddis larvae and a wide range of other groups indicates cleaner, well-oxygenated water, while a sample containing only worms, leeches, water lice and bloodworms suggests organic pollution. The fixed kicking time or distance is the control variable that makes one sample comparable with the next.

Each technique measures something slightly different, and a good answer chooses one and says why: pitfall traps for ground-living beetles and other surface invertebrates, a pooter for insects on one plant, a sweep net for the insect community of a meadow, kick sampling for the invertebrates of a stream. Traps and nets can also be used to estimate the size of a mobile population by mark-release-recapture: animals are caught, marked in a way that does not harm them, released, and the proportion of marked individuals in a later catch is used to estimate the total population, which is the usual approach where individuals cannot be counted directly.

• Pitfall trap: catches ground-living invertebrates; measures activity near the trap, not the exact population size.
• Pooter: catches small, delicate insects from leaves and stems without harming them.
• Sweep net: a standard number of sweeps through vegetation; used to compare insect communities between sites.
• Kick sampling: a fixed period of kicking in a stream; the invertebrates collected indicate water quality.
• Mark-release-recapture: estimates the size of a mobile population from the proportion of marked animals in a second catch.
• Line transect records along a line; belt transect records a strip of fixed width; a point quadrat records only whether a pin touches the species.

Testing the soil and water, and sampling from a distance

Many investigations measure the abiotic conditions rather than the organisms. In the field, water is tested for temperature, pH, turbidity, nitrate concentration and dissolved oxygen, and soil is tested for pH and for nutrient content. Temperature and pH are read with a calibrated meter or probe; turbidity with a turbidity meter or a Secchi disk, lowered until it can no longer be seen, with the disappearance depth recorded; nitrate and other nutrients with a test kit whose colour is compared with a chart; dissolved oxygen with a calibrated oxygen meter, which is more reliable in the field than a chemical test kit because the reading is not affected by the titration being rushed. Every reading must be taken at the same depth, at the same time of day and at the same distance from any discharge, because oxygen and turbidity change through the day as the water warms and plants photosynthesise.

These measurements support the ecological data. Low dissolved oxygen with a high nitrate concentration points to organic pollution, because the bacteria breaking down the sewage or the run-off from farmland use up the oxygen, and only pollution-tolerant invertebrates survive; high turbidity cuts light from submerged plants and can carry pathogens as well as sediment. Aerial photography and drones sample in a completely different way. A drone flown on fixed flight lines at the same height can map habitat boundaries, the extent of a forest clearance, an algae bloom, a spread of invasive water hyacinth or a nesting colony that could not be approached on the ground, and images from different years can be compared to show change.

Automated sampling replaces the observer with equipment that records on its own. A data logger left in a stream or a soil pit records temperature, pH or dissolved oxygen every few minutes, day and night, through a whole season, which no group of students could do; camera traps photograph passing mammals; acoustic recorders store bat calls and birdsong; and light traps switch on at dusk to catch night-flying insects. Because the equipment does not tire and makes no judgement about where to stop, automated sampling reduces observer bias and covers far more time and space than a handheld survey.

The limits matter as much as the benefits. A drone sees shape and colour but cannot identify most species from the air, so it maps extent rather than biodiversity; its work depends on calm, clear weather and on permission to fly, and flying low can disturb birds at nest sites. Automated equipment is expensive, loggers flood, batteries go flat, lenses are obscured by mud and data are lost silently because nobody was there to notice; camera traps are triggered by moving vegetation and miss animals that pass quickly. Instruments must also be calibrated and their readings cross-checked by hand, because a probe reading 0.3 units low will give a neat, convincing and completely wrong data set.

From data to judgement: recording, errors, risk and evaluation

Record the data as you collect it, on a prepared sheet with a column for each variable and a row for each sample point, and write the raw readings down with their units before working anything out from them. The processed values - means, percentages, rates - are calculated later, so that an arithmetic slip can be found and corrected. Present the data so that the pattern is visible: a table of raw and processed values first, then a graph with the independent variable on the horizontal axis, such as distance along the transect, and the dependent variable on the vertical axis. Where several species are recorded at each point, a kite diagram shows the proportions, with the bars widening where a species is abundant and pinching to a point where it is absent, so the zonation along the transect can be read at a glance. Always give the mean and the range, because two sites with the same mean can have very different spread.

Anomalies are values that do not fit the pattern of the other readings: a cover of 80% among readings of 5% to 10%, or a stream temperature far above the rest of the sample. Do not delete them. Check the record first for a recording or transcription error, then look for a real reason, such as a quadrat that landed on a dung patch or beside a drainage outlet, or a probe read before it had settled. If there is a reason, keep the value, report the reason and discuss whether it should be included in the mean; if it is clearly a copying mistake, cross it out, say so, and explain the effect it had on the result. Sources of error come in two kinds. Random error scatters readings either side of the true value, through estimating cover by eye, miscounting a moving animal or reading a turbidity tube against the light, and it is reduced by repeats and by a larger sample. Systematic error pushes every reading the same way - an uncalibrated pH meter, a quadrat frame that is not square, samples always taken from the sunny bank - and repeats cannot remove it; only checking the equipment, comparing observers and changing the sampling plan can.

A risk assessment names the hazards of the site and sets out what will be done about each one before anyone sets off. Standing in cold, fast or deep water, steep and slippery banks, falling branches, sun and heat, broken glass or discarded needles at an urban site, ticks and biting insects, heavy rain and lightning, and working alone or out of sight of the group are the usual fieldwork hazards. Each control measure must be specific to its hazard: sample only shallow water in pairs, with one person steadying the other on the bank; wear wellington boots, work gloves and long sleeves where there is litter or dense undergrowth; carry a first-aid kit and know the nearest road access for an ambulance; tell a responsible adult the route and the return time and carry a charged phone or a whistle. 'Be careful' is never a control measure.

Finally, evaluate the investigation. Judge how reliable and how valid the conclusion is: how many replicates and repeats were taken, whether the sample was representative of the whole site, whether the control variables really were controlled, whether the technique suited the organism, and whether the pattern could have another explanation. Then suggest improvements that would actually change the result: more replicates or more transects, a larger quadrat, sampling at the same point in the tidal or daily cycle, repeating the survey in a different season, calibrating the meter, using a second observer to check cover estimates, or extending the range of the independent variable. A good evaluation ends with a judgement - the data are enough to show a pattern along that line, but not enough to claim it for the whole habitat - and that judgement is what the highest marks are given for.

• Record raw readings with units as you go; process them afterwards so an arithmetic error can be traced.
• Table first, then a graph with the independent variable on the horizontal axis; use a kite diagram where several species vary along a transect.
• Anomaly: check the record, look for a real reason, then keep or discard it and justify the decision.
• Random error scatters readings and is reduced by repeats; systematic error shifts every reading and is found by checking equipment and method.
• Risk assessment: name the hazard, then give the specific control measure that deals with it.
• Evaluation: how many, how representative, how controlled, how suitable - then suggest improvements that would change the result, and judge the conclusion.