akar.earthSampling error: Definition
What is sampling error in carbon inventories? How plot sampling creates uncertainty, how it is calculated and reduced, and why it cuts credit volumes.
Sampling error is the difference between an estimate based on a sample of plots and the true value for the whole area that would be found by measuring everything. In carbon inventories, it comes from measuring only some plots and reflects how variable carbon stocks are and how many plots were measured.
Why it matters
Sampling error is usually the largest quantified source of uncertainty in a forest carbon project. Most methodologies set a target precision, often expressed as the half-width of a 90% or 95% confidence interval relative to the mean. If precision is worse than the target, projects must apply uncertainty deductions, which reduce credits. So sampling error has a direct price.
How it is calculated and reduced
- Standard error: standard deviation of plot values divided by the square root of the number of plots.
- Precision: standard error × t-value for the confidence level, divided by the mean, expressed as a percentage.
- Stratification: sampling uniform strata separately lowers overall error; see stratification.
- More plots: error falls roughly with the square root of plot numbers, so halving error needs about four times as many plots.
- Model-assisted methods: using remote sensing maps as auxiliary data can improve precision when they correlate well with plot data.
Sampling error does not capture measurement errors, allometric model error or bias. Some methodologies require those to be addressed separately. Our plot sampling design guide covers practical design.
Indonesia angle
Indonesian projects often show high variability: patchy degraded forest, mixed agroforestry and many small smallholder plots. That pushes sampling error up. Cost-effective answers include good stratification, consistent plot protocols, mobile data capture with photos and GPS, and linking plots to remote sensing. Logistics in remote islands make extra plots expensive, so design before fieldwork pays off.
Common mistakes
- Placing plots by convenience, near roads, which biases results.
- Reporting precision without stating the confidence level.
- Assuming low sampling error means an unbiased estimate.
Related terms
See confidence interval, stratification and uncertainty deduction, or explore the dMRV hub.
Frequently asked questions
How is sampling error calculated?
From the variation among plot values and the number of plots: the standard error of the mean, multiplied by a t-value for the chosen confidence level, often expressed as a percentage of the mean.
How can a project reduce sampling error?
Add more plots, stratify into more uniform groups, use larger plots or better plot layout, or use remote sensing as auxiliary data in model-assisted estimation.