When environmental scientists head into the field to collect water samples from a river, measure air quality across a city, or assess soil contamination at an industrial site, they face a critical decision: where exactly should they collect their samples? With thousands of possible locations and limited time, budget, and personnel, the choice of sampling method directly determines whether the collected data can accurately represent the larger environment. Sampling methods fall into two broad categories – probability sampling and non-probability sampling – and understanding when to use each is essential for producing reliable environmental data.
Table of Contents
What is sampling in environmental studies?
Sampling is the process of selecting a subset of locations, time points, or specimens from a larger environmental area to draw conclusions about the whole system. Since it is impossible to analyse every drop of water in a lake or every particle of soil in a field, scientists rely on carefully selected samples to represent the environment as accurately as possible. The quality of the entire research project depends on how well the sampling strategy captures the true variability of the system being studied.
According to the U.S. Environmental Protection Agency, selecting the right sampling design is a critical step in the data quality objectives process, influencing every subsequent analysis and decision. The two fundamental approaches – probability and non-probability sampling – differ in how samples are selected and what conclusions can be drawn from the results.
Probability sampling methods
Probability sampling methods ensure that every unit in the study area has a known, non-zero chance of being selected. This randomness is what makes the results statistically valid, allowing researchers to generalise findings from a small number of samples to the entire study area. These methods reduce bias and are preferred whenever regulatory compliance or scientific publication is the goal.
Simple random sampling
Simple random sampling gives every possible sampling location an equal chance of being selected. In practice, an environmental scientist studying water quality in a lake would divide the lake into a grid, assign numbers to each grid cell, and then use a random number generator to pick specific points for sample collection.
This method works best when the study area is relatively homogeneous – meaning environmental conditions don’t vary dramatically from one location to another. Monitoring background air pollution levels in a rural area or assessing general water quality in a well-mixed reservoir are good examples. The EPA recommends simple random sampling when there is little prior information about the area and no major contamination patterns are expected.
The main advantage is its simplicity and the elimination of selection bias. However, it can be inefficient for large or heterogeneous environments because randomly chosen points might cluster in one area, leaving other zones underrepresented.
Systematic sampling
Systematic sampling involves selecting sampling points at regular, predetermined intervals across the study area. A researcher might overlay a grid on a map and collect samples at every intersection point, or take water samples from a river at equal distance intervals downstream.
The process starts by randomly selecting one initial point and then spacing all subsequent points at fixed intervals – whether spatial (every 500 metres along a transect) or temporal (every six hours at a monitoring station). The EPA notes that systematic sampling can be used for estimating means, delineating boundaries, finding hot spots, and estimating spatial or temporal patterns. It is particularly useful for pilot studies and exploratory investigations.
One important caution: systematic sampling can produce biased results if the sampling pattern accidentally aligns with a cyclical pattern in the environment. For example, collecting air samples every Monday morning would be problematic if a nearby factory always runs its heaviest operations on Mondays. Researchers must verify that their sampling intervals do not coincide with any periodic environmental variations.
Stratified sampling
Stratified sampling addresses a key limitation of the previous two methods: they treat the entire study area as uniform, which rarely matches reality. In stratified sampling, the study area is divided into distinct sub-areas (strata) based on known characteristics, and samples are collected independently within each stratum.
Consider monitoring heavy metal contamination in a watershed that includes urban zones, agricultural land, and natural forests. Each land use type likely has different contamination sources and levels. Stratified sampling ensures that data adequately represents all three environments rather than being dominated by whichever zone happens to be largest. Marine biologists studying coral reefs commonly create strata based on depth zones or reef health conditions for the same reason.
According to the EPA’s guidance on sampling design, stratified sampling is recommended when the target area is heterogeneous, when rare groups need sufficient representation, or when sampling costs differ across parts of the study area. This method can also reduce the total number of samples needed compared to fully random approaches, making it more cost-effective for complex environments.
Within stratified sampling, scientists can allocate effort in two ways. Proportional allocation assigns more samples to larger strata, producing better estimates of overall conditions. Equal allocation distributes the same number of samples to every stratum, ensuring that smaller but ecologically significant zones – such as wetlands covering a tiny fraction of a watershed – receive adequate attention.
Non-probability sampling methods
Non-probability sampling methods do not rely on random selection, meaning not every location or unit has a known chance of being included in the sample. While this limits the ability to make statistically rigorous generalisations, these methods are practical and sometimes the only feasible option when real-world constraints come into play. Budget limitations, inaccessible terrain, time pressure, or the need for preliminary data can all make non-probability approaches a reasonable choice.
Convenience sampling
Convenience sampling involves collecting samples from locations that are easiest to access rather than following a random protocol. In environmental work, this is more common than it might initially seem. A researcher studying river water quality may find that most of the river runs through private property or hazardous terrain, and ends up sampling only from publicly accessible bridges, parks, and boat ramps.
This method is fast, inexpensive, and requires minimal planning. As highlighted by research published in Cambridge University Press, convenience sampling is less costly and quicker than other approaches and can be useful for developing hypotheses for more rigorous future studies. It is particularly appropriate in emergency response situations – such as after a chemical spill – where speed matters more than statistical precision.
The biggest drawback is sampling bias. Convenient locations may not reflect typical environmental conditions. Publicly accessible riverbanks, for instance, might be cleaner (or dirtier) than the sections people can’t easily reach. Results from convenience samples cannot be reliably generalised to the entire study area, so this method is best reserved for preliminary assessments, screening studies, or situations where more rigorous approaches are genuinely impossible.
Quota sampling
Quota sampling involves setting specific targets (quotas) for different categories within the study area and then filling those quotas through non-random selection. It resembles stratified sampling in structure but lacks the random selection step within each group.
For example, if an environmental consultant is assessing urban air quality and knows the city is roughly 40% residential, 30% commercial, 20% industrial, and 10% parkland, they would ensure their sampling locations maintain these proportions. However, instead of randomly choosing points within each zone, the consultant selects accessible or convenient locations until each quota is filled. As noted by Researcher.Life, quota sampling ensures that important subgroups are proportionally represented while remaining flexible about exact sampling locations.
Environmental consultants frequently use quota sampling when regulatory requirements specify that different land use categories or property types must be assessed, but don’t mandate how specific locations within each category are chosen. It is a practical compromise that balances representation with logistical flexibility.
The trade-off is clear: quota sampling guarantees that all important categories are covered, but the non-random selection within each category introduces potential bias. The findings may reflect the characteristics of conveniently selected spots rather than the full range of conditions within each zone.
Other non-probability approaches
Beyond convenience and quota sampling, environmental researchers sometimes use purposive (judgmental) sampling, where locations are deliberately chosen based on expert knowledge. For instance, a scientist investigating contamination near an industrial outfall might concentrate sample collection around the discharge point because professional experience suggests that’s where pollutant levels will be highest. According to NJIT’s environmental sampling guide, judgmental sampling concentrates resources on areas of greatest concern but limits the ability to draw conclusions about the broader environment.
Snowball sampling is less common in environmental fieldwork but can be used in community-based environmental health studies, where one participant refers the researcher to others with relevant exposure or information.
How to choose the right sampling method
Selecting an appropriate sampling method requires balancing several factors, and experienced environmental scientists weigh these carefully before heading into the field.
Study objectives and required precision: If the goal is to produce statistically defensible results for regulatory compliance, scientific publication, or legal proceedings, probability sampling methods are essential. For preliminary assessments or screening studies, non-probability methods may suffice.
Environmental heterogeneity: Uniform environments work well with simple random or systematic sampling. Complex ecosystems with distinct zones – such as a coastal area with industrial, residential, and agricultural stretches – benefit from stratified approaches that ensure each zone is properly represented.
Resource constraints: Budget, time, and personnel limitations often dictate what is practical. Research published in Ecological Indicators emphasises that environmental monitoring programs must be designed to use limited resources efficiently while still producing reliable data. Systematic sampling may be more cost-effective than purely random sampling for large areas, while convenience sampling may be the only option for dangerous or inaccessible sites.
Regulatory requirements: Many environmental monitoring programs operate under specific sampling protocols set by agencies like the EPA. Understanding these requirements upfront prevents costly mistakes and ensures that collected data will be accepted by regulators.
Combining methods: In practice, environmental studies frequently combine multiple sampling approaches. A researcher might use stratified sampling to divide a study area into zones and then apply systematic sampling within each stratum. Or a preliminary convenience sample might inform the design of a more rigorous probability-based study. The U.S. Forest Service has noted that in complex environmental situations, both probabilistic and non-probabilistic procedures often need to be used together due to practical constraints of cost and incomplete knowledge.
Common pitfalls to avoid
Even experienced researchers can make sampling mistakes that compromise data quality. Spatial clustering – unconsciously placing sampling points too close together – reduces independence between samples and can misrepresent the study area. Temporal bias occurs when samples are always collected at the same time of day or season, missing important variations. And access bias subtly affects many studies: when researchers can only reach certain parts of a study area, the resulting data may reflect conditions at accessible sites rather than the environment as a whole.
The key to avoiding these pitfalls is careful planning. Document the sampling strategy in advance, justify why a particular method was chosen, and acknowledge any limitations in the final analysis. Transparent reporting of sampling methodology helps other scientists evaluate the reliability of the findings and builds credibility in the results.
What do you think? Given that environmental monitoring budgets are often tight and field conditions unpredictable, how should researchers decide when the statistical rigour of probability sampling is worth the extra cost – and when practical non-probability methods provide sufficient insight for the decisions at hand?
References
- https://www.epa.gov/quality/selecting-sampling-design
- https://web.njit.edu/~kebbekus/analysis/SAMPLING.htm
- https://www.epa.gov/sites/default/files/2015-06/documents/g5s-final.pdf
- https://www.cambridge.org/core/journals/prehospital-and-disaster-medicine/article/population-research-convenience-sampling-strategies/B0D519269C76DB5BFFBFB84ED7031267
- https://researcher.life/blog/article/what-is-quota-sampling-definition-advantages-disadvantages-and-examples/
- https://www.sciencedirect.com/science/article/pii/S1470160X20310359
- https://link.springer.com/article/10.1023/A:1006316418865
Leave a Reply