Ecological Fieldwork and Data Handling

Sampling a habitat

You cannot count every organism, so you take a sample and use it to describe the whole habitat. The instrument depends on the organism and on the question.

A line transect records what touches the line. A belt transect uses quadrats placed at regular intervals along it, so you see how the community changes with distance.

Avoiding bias

If you place quadrats where the plants look nicest, your results are biased. To sample without bias, place quadrats randomly: use random numbers as coordinates on a grid, or throw the quadrat over your shoulder. For a gradient, use systematic sampling (equal intervals along a transect). A larger sample also makes the result more reliable.

Measuring abiotic factors

Abiotic factors help to explain why organisms live where they do. Variation in them, such as light, moisture or pH, changes the distribution (where a species is found) and abundance (how many there are) of organisms. For example, seaweeds on a shore are zoned by how long each is exposed at low tide. Measure abiotic factors at the same points where you sample the organisms.

Making readings reliable:

• Take several readings at each point and calculate the mean.

• Take readings at the same time of day and at the same depth or height.

• Calibrate the instrument and rinse probes between samples.

• Do not shade the light meter with your own body.

Error, accuracy and precision

• Accuracy: how close a measurement is to the true value.

• Precision: how close repeated measurements are to each other.

• Random error: unpredictable scatter in readings, sometimes high and sometimes low (e.g. reading a scale by eye). Repeating and averaging reduces it.

• Systematic error: every reading is out by the same amount in the same direction (e.g. a thermometer that always reads 2 °C high). Repeating does not remove it; you need to calibrate or change the method.

A set of readings can be precise but inaccurate. Imagine darts that are tightly grouped but far from the bullseye: that is a systematic error.

Estimating population size

Quadrats: density and total population

1. Count the organisms in each quadrat and find the mean number per quadrat.

2. Divide by the quadrat area to get the density (number per m²).

3. Multiply by the total area of the habitat to estimate the population.

To improve reliability: use more quadrats, place them randomly, and repeat on different days.

Mark–recapture (for mobile animals)

1. Capture a sample, mark the animals harmlessly and release them. Count them: $M$.

2. After they have mixed with the population, capture a second sample: $C$ animals.

3. Count how many of these are marked: $R$.

The idea: the fraction of marked animals in the second sample should match the fraction of marked animals in the whole population, so $\frac{R}{C} = \frac{M}{N}$.

Assumptions of the method:

• Marked animals mix evenly with the rest of the population.

• Marking does not harm the animals or change their survival or behaviour (not more visible to predators).

• The marks do not rub off.

• No significant births, deaths, immigration or emigration between the samples.

• The two samples are taken in the same way (equal chance of capture).

If marked animals are harder to recapture (they become "trap-shy"), $R$ is too small, so the estimate $N$ is too high. If marks wash off, $R$ is again too small and $N$ is too high.

Handling and presenting data

• Mean: sum ÷ number of values.

• Range: largest − smallest.

• Standard deviation (SD): measures how spread out the data are around the mean. A small SD means the values are close to the mean; a large SD means they are widely scattered.

Choosing a graph:

• Line graph (or scatter plot) for two continuous variables, e.g. soil pH against distance along a transect.

• Bar chart for categories, e.g. mean beetle number in woodland and grassland.

• Histogram for the frequency distribution of a continuous variable.

Always put the independent variable on the x-axis and the dependent variable on the y-axis, with a title, labelled axes with units, and an even scale.

Designing and evaluating investigations

A good investigation tests a hypothesis: a testable prediction. The null hypothesis states there is no effect or relationship, and the data are used to test it.

Example: Hypothesis: moss density is higher in lower light. Null hypothesis: there is no relationship between light intensity and moss density.

Variables:

• Independent: the one you change or measure as the possible cause (light intensity).

• Dependent: the one you measure as the result (moss density per quadrat).

• Controlled: kept constant (quadrat size, time of year, soil moisture, tree species, slope).

In a woodland, shaded areas might also be damper, so soil moisture is a possible confounding variable: moss may respond to moisture rather than light. Measure soil moisture at every sample point, or control it by comparing points with similar moisture. To test cause, you control other factors or run an experiment.

Ways to improve an investigation:

• More sample points: reduces the effect of chance.

• Random placement: reduces bias.

• Repeat measurements and take means: reduces random error.

• Calibrate instruments: reduces systematic error.

• Measure and control other abiotic factors.

Reading results with standard deviation

Woodland beetles: mean 8.2, SD 1.1. Grassland: mean 5.9, SD 3.4. The woodland values are tightly grouped, while the grassland values are widely spread. With such a large SD, the grassland range overlaps the woodland values, so the difference in means may be due to chance. "Definitely" is too strong. A statistical test (such as a t-test), a larger sample and repeat sampling on different dates would strengthen the conclusion.

Ethics and conservation of fieldwork:

• Use as few traps as needed and check them often so animals do not starve or get eaten.

• Release non-target animals (frogs, shrews) unharmed.

• Avoid damaging the habitat; get permission and follow NPWS rules in protected sites.

• Return organisms to where you found them; do not collect rare or protected species.

Judging scientific claims in the media

Headlines such as "a new study proves…" need checking. Ask:

• Control group? Without a control you cannot tell whether the treatment caused the change. Compare with a group given a placebo.

• Sample size? 12 volunteers is too few. Larger samples reduce the effect of chance.

• Blinding? In a single-blind trial participants do not know which treatment they get; in a double-blind trial the researchers do not know either. This reduces bias and the placebo effect.

• Replication? Other scientists should be able to repeat the work and get similar results.

• Funding and bias? A study funded by the manufacturer of the product has a conflict of interest.

• Language? Real science rarely "proves"; it says "suggests" or "supports".

Peer-reviewed paper vs press release: a peer-reviewed paper is checked by independent experts before it is published, and it gives full methods and data so others can repeat the work. A press release is written by the organisation or company to promote the work: it is not independently checked, and it may exaggerate.

To assess a source, check who wrote it and their expertise, whether it cites the original paper, whether the data and methods are given, who paid for the research, and whether other independent studies agree.

Back to the rocky shore: you would lay a transect from high to low tide, place quadrats at random or regular intervals, count the periwinkles, calculate density, and scale up to the whole shore area. Repeating the survey on different days would make the estimate more reliable.

Summary

• Sample with the right instrument: quadrat (cover, density), transect (change along a gradient), pitfall trap, pooter, nets. Place samples randomly to avoid bias, and take plenty of them.

• Measure abiotic factors (light, temperature, pH, moisture, wind speed) with suitable instruments, take repeat readings and use consistent methods.

• Accuracy is closeness to the true value; precision is closeness of repeats. Random error is reduced by repeating; systematic error needs calibration.

• Quadrat estimate: mean per quadrat → density per m² → × habitat area. Mark–recapture: $N = \frac{M \times C}{R}$, with its assumptions.

• Present data with the right graph (independent variable on the x-axis), and use mean, range and SD. Correlation is not causation, so control variables and state a hypothesis and a null hypothesis.

• Judge claims by control group, sample size, blinding, replication, funding and peer review. Carry out fieldwork ethically.