Stratified sampling
A sampling design that divides an area into homogeneous groups and samples each separately to reduce estimate variance.
Stratified sampling splits a population into strata that are internally similar and externally different, then draws samples within each stratum in proportion to its size, its variability, or its importance.
In carbon accounting, strata might be forest types, soil classes, age since disturbance or management regimes. Because variance is averaged within groups rather than across the whole area, a stratified design usually achieves the same precision with far fewer plots than simple random sampling — the single largest cost driver in a field campaign.
The design must be fixed before fieldwork and documented. Post-hoc stratification chosen after seeing results inflates apparent precision and is one of the more common methodological weaknesses found in verification.
Frequently asked questions
How many plots are enough?
Enough that the confidence interval around the estimate meets the programme's threshold. The required number follows from observed variance, the accepted error and the stratification used.
Can strata be defined from satellite data?
Yes, and it is common practice: land-cover and biomass maps provide objective, pre-existing strata that can be frozen before sampling begins.