|
|
Sampling frames are constructed from decisions about the elements to be measured, the way the elements are selected, and the sampling design.
|
|
Sample elements
|
Four types of sampling element or unit are common in forest inventory:
In addition, these elements may be relocatable (permanent) or temporary. Permanent sampling units (trees and plots) are established in forests for three main reasons:
- To provide information on growth and yield of forest crops which is then used to construct yield models (yield tables).
- To provide information on the effects of specific experimental treatments, e.g. different thinning regimes, initial plant spacings, fertiliser treatments, etc.
- To provide data which are used in devising new measurement methods and systems for general forest use.
|
|
Sample selection
|
The main approaches to sample selection are:
- Non-random / non-probabilistic. This approach includes subjective (eg selection of the mean tree by personal judgement) and haphazard (eg sticking pins in a map while blindfolded) selection.
- Approaches based on probability where each element has a chance (that can be determined) of being selected. These approaches can be further divided into:
- Unequal probability. The probability of an element being selected is proportional to that element's size, or some other parameter.
- Random. All elements, and all combinations of elements have the same chance of being selected.
- Systemmatic / restricted. All elements, but not all combinations, have the same chance of being selected.
When carried out by experienced professionals, non-probabilistic selection of the mean element in a population can be accurate. However, there is no way of determining the accuracy or reliability of this estimate. Subjective or personal judgements are also liable to bias and therefore are rarely used in important inventories.
|
|
Sample design
|
Different sampling designs can take advantage of pre-existing knowledge of the forest structure (and other information) to improve precision or reduce the cost of an inventory. Examples include:
- Simple Random Sampling. A simple sampling approach that is commonly used when there is no obvious structure or pattern in the population of interest.
- Two-, multi- stage Sampling:
- Stratified Random Sampling. The forest can be divided into two or more groups or strata where the variation between groups is greater than the variation within a group. That is, an element in strata A is likely to be more similar to another element in A than it is to an element in strata B.
- Cluster Sampling. There are no strata or definable patterns in the forest population so that the variation between any two element within an arbitrary group or cluster is at least as great as the variation between clusters. That is, if you calculated the mean of all the clusters in the population, the spread of these cluster means around the population mean will be similar or smaller than the spread of the individual values within a cluster.
- Two-, multi- phase Sampling:
- Regression Sampling. A list of values exists with every element in a forest population included, and these values relate to the parameter of interest for the inventory.
- Double Sampling. Easily measured values can be collected over the forest and an unbiased mean determined. The relationship between these easily collected values and the parameter of interest is determined.
|
|
SubHeading
|
MainTextHereOnwards
|
|
SubHeading
|
MainTextHereOnwards
|
|
SubHeading
|
MainTextHereOnwards
|
|
[temp.htm] Revision: 6/1999
Cris.Brack@anu.edu.au
|
