Discussion of the Merits of the two Designs

Both the grid based and the sorted polygon list designs have feature which make them attractive under different condition. There are no well established criteria for assessing conclusively the advantages of one over the other. The choice of one design over the other depends more on subjective judgments of how it meets the objectives of the proposed inventory, than on the quantitative analysis of the statistical advantages.

There are several criteria which could assist a forest biometrician in choosing a sam­

pling design for a specific situation. Some of these criteria are discussed in the following sections.

Simplicity

A polygon list can be sorted by several criteria. For the forest timber participants, the list could be sorted by species, height, age, volume, etc .. The criteria by which the list is sorted would influence the polygon selection. The question then is: what combination of the criteria would be used, and what influence would the other participants in the inventory have in the choice of combinations?

The sorted polygon list design would be simple and straight forward if the sorted list consisted of a small number of polygons, and if the variation among the vegetation cover types were small. In addition, the sample obtained would be representative of the popu­

lation if the proportions of the different cover types within the population were nearly equal ^(WARWICK and LININGER 1975).

The vegetation cover types in British Columbia are quite diverse, and there are some cover types which form only small proportions of the vegetation in most management units.

Consequently, using the sorted lists without stratification would not produce desirable results.

With the introduction of stratification, the determination of sample size, and the choice an allocation rule to ensure adequate representation of each stratum, complicate the planning process for the sampling. For this reason, the sorted polygon design is not easy to implement.

The grid based design on the other hand, only requires a base map for a sampling frame.

The grids are established on the base map and ground sampling follows without the complex sorting and stratification processes.

Flexibility

The design based on sample selection from a sorted list would be based on management units. One or more management units would be sampled at a time. Statistics for the province as a whole would be obtained by rolling up the individual management unit values after all the units are sampled. There are several types of management units, i.e., Timber Supply Areas (TSAs), Tree Farm Licenses (TFLs), Parks, and other privately and publicly owned lands. Overall, the number of management-unit-type areas may be as large as 70.

Conducting inventories in each of these management units through the two stage process of photo classification followed by ground sampling, would be cumbersome and slow. It would take 15 to 20 years to obtain the final provincial statistics.

The grid based design, however, would provide provincial statistics within 3 to 5 years, because sampling at the grid points would not require the time consuming photo classification component.

For the second component, which is management unit based, the sampling would target specific vegetation cover types that are not sufficiently represented by the grid points within the management unit. This feature would offer the flexibility of allowing more sampling depending on local planning needs. The criteria for increasing sample size may include the economic value of the vegetation cover type, under-representation of a cover type, or special studies on one or more vegetation cover types.

The grid based design also offers the flexibility of the use of a combination of recent and new photography to obtain photo based at�ribute estimates. This could lead to substantial cost savings since this component is the most costly in the inventory.

Forming Common Locations for Multi-Resource Sampling

The primary objective of the proposed new inventory is to design an inventory which will allow for integrated resource management. This means that the ground samples should be selected without bias toward one particular cover type, and that no prior assumptions of value or importance should be assigned to any cover type. A sampling design based on a provincial grid concept would be the most suitable design to meet this objective.

The sorted polygon list design on the other hand would require stratification to ensure representativeness. The stratification process assumes that: a) the cover type identification, and estimation of polygon areas based on photo interpretation are accurate, and that, b) some strata are more important than others.

In reality, the accuracy of cover type identification is questionable. For the forested polygons, past experience shows that identification of the leading tree species may be wrong on up to 40% of the polygons in some instances. The magnitude of the error of the polygon area estimation is not known.

Furthermore, the forest classification polygons, which would most likely be the basis for forming a sorted list may not coincide with the ecology, soil type, range, and other participants' classification polygons. This situation would lead to conflict and bitterness among the multi-resource inventory participants in the long-run.

Change Detection, updating, and Projection

The provincial grid would form permanent sampling point which would be revisited at regular intervals. In the long-run, this would provide a wealth of information on growth rates, depletion rates, and stand dynamics in general. Consequently, the inventory information would be projectable into the future with better accuracy. In addition, the grid based inventory would facilitate easier adjustment of the inventory information after logging, blow-down, fire, etc. In other words, more current inventory information would be available at all times.

Because of the nature of the plot layout within the sample polygons, it would be difficult to establish a re measurement system to monitor change, for the sampling design based on a sorted polygon list. Furthermore, it would be difficult to account for the effects of changes such as fire logging, etc., particularly when only parts of a sample polygon were affected.

This, in fact, is one of the major weaknesses of the last (1961-1977) inventory. Without a well coordinated update program, the relevance of an inventory deteriorates with time.

Old Growth and Second Growth Information

One of the major forest management issues in British Columbia, is the need for information on the quantity, and geographic distribution of old growth forests in the province. The sampling design based on a sorted polygon list is not amenable to providing information on this aspect of forestry.

The provincial grid would cover the entire land base more or less uniformly. This would allow for the determination of the geographic distribution, and estimation of the pro -portions of old growth and second growth forest

Summary Statistics

The proposed new inventory should provide broad summary statistics for provincial general and strategic planning, as well as management unit summary statistics for regional and district planning.

The sorted polygon list design would probably provide inventory information more effectively at the management unit level. The two component, grid based design on the other hand, would be more efficient at the provincial level.