contributed by Genevieve Tucker
Sampling sufficiency provides monitoring program personnel and land managers with an estimate of sample sizes required to achieve monitoring objectives. Before determining if change is occurring across a landscape, it is necessary to determine if the changes in monitoring data are due to natural variability within the ecosystem of if these in fact are true changes in ecosystem processes. Analysis of sampling sufficiency asks “is the variability within a plot greater than the variability between plots?” If the answer is yes, the sample size is sufficient to answer the monitoring question at the stated home#power, assuming a given level of correlation (home#rho), and a given false change rate (home#alpha).
1. Management objectives, the desired ecological state.
2. Sampling objectives, the desired detection ability of the monitoring program.
- Example: The ability to detect a 20% difference ([[glossary:home#MDD]]=0.2) in foliar cover after 10 years
3. Assumptions, the statistical assumptions in detecting change.
- Example: The assumption that there is a correlation of 0.6 (rho=0.6) between sampling events. The desired certainty of detecting a difference is 80% (power = 0.8). There is a willingness to accept a 20% chance (alpha=0.2) that a conclusion will be reached that a difference exists when it really does not.
4. Results from sampling and/or pilot data, summarized as indicators (e.g. annual grass herbaceous cover, bare ground cover). The indicators selected are dependent on the management and monitoring objectives.
- Example: Foliar Cover in Stratum X
|Line||Plot A||Plot B||Plot C||Plot D|
The Multi-Scale Sample Requirements Evaluation Tool (MSSRET) is a tool for assisting managers and researchers in determining appropriate sample requirements. If there are less than 10 plots per stratum, the data may be formatted in the format list above and use the MSSRET found at https://landscapetoolbox.org/mssret/MSSRET.html. If there are more than ten plots per stratum, it is recommended to use a combination of the Excel spreadsheet MSSRET developed by the NRCS and the Landscape Toolbox page.
This list of questions could be asked of the following data sample:
– Which strata meet the minimum number of plots required to detect a MDD of 10%? 20%?
– Which strata do not?
– Are there factors besides the number of plots which might be affecting the variance (e.g. multiple states within a stratum?)
– Which strata are insufficiently sampled to run a sampling sufficiency?
|Stratum||n||abs MDD||n_10||n_20||abs MDD||n_10||n_20||abs MDD||n_10||n_20||abs MDD||n_10||n_20|
For the example data above, the data suggest that CPBWM, CPID and AB need additional plots in order to determine sampling sufficiency to detect a 20% MDD. The next sampling priorities are FWMM, AFPD, TMCW and AFS as they are not sufficiently sampled for answering monitoring questions relative to two indicators. If additional monitoring resources exist, field crews can then focus on sampling the rest of the strata. Fortunately, in this system shrub encroachment is a key ecosystem indicator and the sample data collected thus far indicate that would be able to detect a 20% change in shrub cover, and for some strata, a 10% change in cover.
The process of determining if data has been sufficiently sampled is intended to provide guidelines for future sample design. The boundary of “sufficiently sampled” may change depending on management objectives, sampling unit scale, temporal scale and other disturbances. As with all monitoring data, these data must be interpreted within the context of ecosystem dynamics and land potential.