Sampling design
The documented plan for where, when and how often measurements are taken, chosen to make estimates defensible at a stated precision.
A sampling design answers four questions in advance: what is sampled, where samples are located, how many there are, and how often they are repeated. Each answer carries statistical consequences that cannot be repaired after the fact.
Good designs account for spatial autocorrelation, access constraints and cost. Purely random locations may be statistically clean but operationally impossible; systematic grids are efficient but can alias against periodic landscape features; stratified and model-based designs trade assumptions for efficiency.
Verifiers scrutinise the design because it determines whether an estimate can be trusted at all. A large but biased sample is worse than a modest random one, and an undocumented deviation from the design is treated as an unquantified error.
Frequently asked questions
What is the difference between sampling design and monitoring plan?
Sampling design specifies the statistical structure of measurement; the monitoring plan wraps that into the full schedule, responsibilities and data management of a project.
Does remote sensing replace field sampling?
No. Remote sensing extends coverage but still needs field plots to calibrate and validate the relationship between signal and carbon.