Step

Why

How

1.Define boundary of project area.

Samples have to represent the area you are interested in, hence you need to define it exactly.

On any map, draw the boundary of the area of interest (AOI). This is best done digitally. Google Earth is a suitable available tool also accessible through the CBP tools.

2.Identify distinct regions of your project area that you anticipate will have different SOC levels (if interested in stocks) and/or have different levels of change (if interested in SOC change). These define Strata

Sampling is most efficient if distinct areas are deliberately targeted, rather than leaving it to chance. If interested in specific interventions it makes sense to put more sampling effort into measuring them compared with the rest of the landscape.

Based on literature and local knowledge, select and map the boundaries of strata. Typically 2-10 strata will be appropriate.

3.Define a hierarchical grid to sample from

Sampling from a grid simplifies all data processing and links to other data sources. Cost and logistics usually require sampling, so that sample locations appear clusters in the landscape.

Choose top-level grid with a size that is appropriate for the dimensions of your strata example. Choose the lower level grids using the standard dimensions recommended. Display your selected grids on a map. Here we need a simple map as an example than the complex hierarchical modeling

4.Choose a sample size

Taking too few samples leads to estimates with inadequate precision and is a waste of money. Taking too many samples generates more data than you need and is also a waste of money.

Decide how precise you want estimates to be. Use standard formulae to estimates of expected variation SOC to come up with an initial sample size determination. Online tool to determine sample size spreadsheet with examples

5.Choose a sampling scheme

You need to decide how to allocate the samples size to difference strata and different levels of the hierarchy

Use the principles to come up with a proposed design.

6.Estimate the cost of your sampling

You need to be sure the planned sampling is within budget before starting field work.

Estimate the cost as a combination of the fixed costs (e.g equipment) plus variable costs (e.g. labour and travel), based on either your own information or experiences of others.

7.Reconcile cost and precision

If your initial design is too expensive you need to change something. Simply taking fewer samples is not an answer – you need to objectively change what you are aiming to get from the measurement.

Return to steps 1, 2, 4 or 5 and decided what to adjust, before recalculating the cost and precision

8.Generate the sample locations

You need exactly locations to visit for sampling. Random sampling (within strata, and based on grids) is used to ensure that there are no biases.

Using standard code to generate random sample locations that conform to the strata and grids proposed. If samples end up in inappropriate locations (e.g, on a road, in water), generate a replacement. Output the list of sampling locations to your GPS.

9.Sampling for change in SOC

If you are selecting samples in order to measure change from a baseline or previous measurement, then you can usually improve on simply taking a second sample using these steps.

Choose the second or change detection sample to balance the strategies of (a) sampling the same locations to get precise estimate of change and (b) sampling new locations to 'fill gaps' in the first sample and hence get a good estimate of the stock at Time 2.