Sample design

Prior to inventory, a "stratification" needs to be done. This is valid also in a complete inventory (when all stands in the analysis area are included), since some general inventory parameters are set in this part. Considering cost-efficiency, the inventory is commonly based on a two-stage sampling procedure. Here, this starts with selecting the imported stand register of interest, i.e. the analysis area to be in-field inventoried by plotwise sampling in PPS-sampled stands.

NB: The following stand register parameters are, more or less, mandatory to enable stratification: stand area (productive forest land), volume, age, basal area, number of stems and tree species proportions.

• A stratum is defined usually based on volume and age from the stand register

The stands (their productive forest land areas) are related to one of the several user-defined strata. The stratum-matrix should first be arranged according to current forest state, and before that, you might want to separate the analysis area (imported as one stand register) into different domains. A domain is here a certain type of forest (e.g., stands with the dominating tree species Pinus contorta) where a certain inventory-scheme should be applied, e.g. securing a certain sample size.

• When a satisfying stratification of the analysis area has been obtained, the total number of sample stands are decided

First, a stratum will automatically obtain a number of sample stands (depending on the area in the stratum, or perhaps the average volume per hectare?). You should re-distribute the sample stands over the different strata, with a higher sampling intensity in strata with more valuable forests and where forest management decisions are supposed to be more difficult. This part is finished by actually sampling the sample stands in each stratum (done PPS where size is the stand area). Inspect the sampled sample stands regarding, e.g., the total number and each stands corresponding stratum.

NB: There should be at least three sample stands in each stratum. Moreover, aviod extreme differences in the (average) representative area of different strata.

• Next, the presupposed number of plots in a sample stand is set, depending on stand characteristics (chosen by the user)

The plot radius is set in a similar manner. Finally, some additional settings are made, necessary in the inventory (for the sample tree sampling). Inspect the sample stands. When values in corresponding columns (e.g. P1-P3, Stratum, Plots, and Radius) looks ok, check this "stratification" as finished.

NB: The number of plots per stand is set considering costs and assumed losses (caused by bad decisions made from erratic forest descriptions), less than three is only considered stupid. [A corresponding line of reasoning is held for the number of sample trees per plot, where less than 0.5 is stupid, motivated in same manner - do field-measurements when being in field!] The plots are circular, however when, e.g., acting as reference ("ground truth") to remotely sensed data with certain (quadratic) pixel size, you should adjust the plot radius hereof and define radii in meters with two decimals.

Knowing the size of the sample stands (A, the area in m²) and the corresponding number of plots (n), the plot positions are derived externally. Company specific GISs are used to obtain "Theoretical" x- and y-coordinates (in meters, preferrably in RT90) for each plot in each sample stand. The actual number of plots in a stand can differ from the presupposed, since the layout is randomized^*. The actual plot positions, "True" coordinates, are obtained from GPS, preferrably placed at plot centre and averaging positions made during field work. Post-processing of GPS-data will improve the accuracy of the positions even further. Advantages in using external GIS, and perhaps external GPSs too, are numerous; e.g. facilitating navigation in field to stands and plots, using suitable images (e.g. ortophotos) and maps (possible paper print-outs).

^*) Within a sample stand, the location of the first plot is randomly chosen, and from this all other plots are systematically located in a square lattice (with spacing (in m) s = (A/n)**0.5). This simplifying the navigation between plots and, above all, increasing the accuracy in the estimates.