Before you begin your inventory

BEFORE YOU BEGIN YOUR INVENTORY1 Jerry Vanclay The most important part of your inventory is the question "Why do you want to do this inventory?". No-one can give specific advice on efficient inventory, until you can answer that question clearly, concisely and completely. To resolve what and how to measure, you first have to decide what information you need. Don't get sidetracked by data collection procedures, and remember that information is more than data. Data are the raw numbers collected in a survey or experiment, and need to be collated in a meaningful way before they can be regarded as information. It is not enough to say that you need a stand table or a count of merchantable trees; to achieve the best outcome, you need to be quite explicit. Why do you need it, and in what form? Is that the final format for these data, or will there be further processing? Are you sure that there is no additional information required? Do you have reliable volume equations for all species? Are there other conversion factors you need? It is much better to ask these questions before the inventory, when deficiencies can still be rectified; it is too late afterwards! Answers to these questions are much easier if we have a clear statement of objectives. Ideally, we should be able to determine information needs from forest policy and management objectives. Unfortunately, this is rarely possible, and inventory planners also have to discuss information needs with management staff. This isn't easy, especially if these staff are not familiar with computers and their capabilities. People who for years have relied on gut feelings and rules of thumb cannot be expected to give a clear statement of what they need from a computer. So don't expect it to be easy, but remember that it is important. Allow plenty of time, talk about the information they need, not about data preparation forms and output proforma, and stimulate discussion by providing mock-ups of the output your propose. Remember that the computer is your slave, not your master; so make it easy for the users, not for the computer. The cost of the computer and the software is small compared with the value of the data it handles, and with the potential implications of poor forest management decisions! What information should you aim to provide? Generally, forest managers and planners need reports detailing: • stocking, basal area, log lengths and/or volume, • by tree species, size (diameter or length) and/or commercial characteristics, and • by individual inventory plots, user selected strata, and/or regional averages. 1 This article is condensed from the paper "Resource Inventory for Land Use Planning" given at the ITTO Seminar on Land Use Planning in Yaoundé, Cameroon, May 1992. They also require forecasts and may ask: • What is the maximum sustainable harvest? • For how long can the present harvest be sustained and what are the implications for the residual forest? • What is the nature (average stem size, species composition, yield per hectare) of future harvests? This information can be compiled from three sources: • area estimates for the existing forest, • stand level inventory of the present forest, and a • growth model to forecast the future forest. These are minimum requirements for management and planning of timber production. You may require analogous information for nontimber products from your forest. So don't simply adopt these ideas, but adapt them to suit your needs and your local situation. One of the most important things you can do is examine existing information, not only to see what you already have, but also to see what you did wrong last time, and to learn from your mistakes. Everybody makes mistakes, but only a fool makes the same mistake twice! Natural changes in most forests are rather slow, so provided that the forest has not been logged or otherwise damaged (e.g. hurricane or wildfire), data may remain valid for many years. Even where the forest has been logged or destroyed, existing data may provide useful information regarding soils, topography, forest type, etc. The amount and quality of this data may influence your choice of sampling design, and may save you a lot of time, money and effort. So be diligent in examining your archives; it pays dividends. Having decided what data we need, and discovered what we already have, we can resolve how to collect the remaining data. There are two issues to be addressed: 1) how many plots and where to place them, and 2) how to measure trees and other attributes on plots. Sampling Design Three simple practical matters may dictate the design: 1) Prior (existing) information may limit your options. Stratified sampling requires sufficient prior information to draw meaningful strata. Without prior information, simple systematic samples such as strip assessments may be the best alternative. Fortunately, prior information is usually available from previous surveys, maps, aerial photographs and satellite data, enabling efficient designs. 2) Area estimates are required for several methods including stratified random sampling. Simple systematic samples avoid the need for independent area estimates, but you may prefer to obtain area estimates from other sources and use more efficient designs. But be careful with area estimates, and remember the big difference between gross and nett area! Table 1. 1 Some Considerations in Sampling Design. Criteria & Consequences Alternatives & Optimal Sampling Method Nature of estimate Forest Characteristics Representative selection Time and resources Critical Unknown/Diverse Unreliable Sufficient Objective Go to 2 Absent Can be estimated Unimportant/Personal Familiar or Uniform Reliable Very limited Subjective Sampling Unavoidable Unknown Possible/Unknown Not required Required Go to 3 Random Correct estimate Unlikely Unlikely or Known Necessary Unimportant Systematic Sampling Probably inflated Possible Clear or Likely Relatively low Strat.random 4 Misjudge pattern Absent or Unlikely High Unrestricted random Sample clustering Obscure/Unknown Geometrical blocks Simple Visible or Well known Statistical blocking Possibly complicated Bias Precision Periodicity Interpolation Estimate of Precision 2 Sampling Error Periodic Bias Pattern in population Sampling intensity 3 Inherent risks Pattern in population 4 Calculations 3) Availability of resources may dictate designs that are feasible. It is important that the data are reliable and that the inventory is completed. If you attempt something too big or too complex, it may never get finished. If your staff find it too confusing, they may make too many mistakes. So stick with something you can do well, and can carry it through to completion. Grand designs don't look so good when they can't be completed! Table 1 highlights some considerations which influence the selection of a sampling design. It takes the form of a binary key which may help select an appropriate method. Start at step 1 in the first row of the table and consider the questions in the left column; if you think that the central column best describes your situation, proceed to the next row (step 2). Alternatively, the right column suggests one possible design that may fit your circumstances and highlights some possible consequences. For example, Table 1 indicates that if we seek a reliable estimate and require interpolation to prepare a site quality or forest type map, we should employ systematic sampling. Alternatively, if our estimate is critical, is to have known precision, and is to be obtained using a small sample, it may be better to use stratified random sampling. Most forest services started out with no prior information and no remote sensing (air photos, satellite imagery, etc), so strip assessment was a good approach to use. It is still good, but it is expensive, and is better for locating a resource than for quantifying it. Other methods allow prior information to be used, reduce the cost and increase the precision of the inventory. So think carefully about the alternatives summarized in Table 1. For many applications, some form of stratified random sampling may be optimal. Either statistical (i.e. strata drawn from prior information such as forest type maps) or geometric blocking (i.e. regular blocks ignoring variation in the forest) may be used, depending on prior information. With either method, three principles offer the greatest precision for a fixed outlay: • The precision of the final estimate is influenced most by the initial stratification; • Precision is gained by dividing the population into many strata, provided strata contain at least two plots; • Further improvement can be achieved by sampling proportional to the variance within the stratum. One important rule remains: Quality. It is better to have a few reliable plots than a lot of unreliable ones. It costs a lot of time and effort to inventory forests, so make sure that you do it right: aim for quality, not quantity. Type of Plot (or Point Sample) The sampling design dictates the placement of plots, but we still have to resolve the type of samples to be taken. They may comprise long narrow strip samples, square plots, variable radius plots, or plotless point samples, and all offer advantages in some circumstances. Three factors dictate the type of plot you should use: 1) Distribution (of parameter of interest): How does the main parameter vary with the measured objects? Is it is uniform (e.g. number) or does it vary with tree size (e.g. volume)? How do attributes vary within the stratum: are trees all the same size (e.g. plantation) or is there a big range of tree sizes (e.g. natural forest)? 2) Variation: Do you wish to capture the variation within plots (e.g. to minimize standard error associated with a resource survey) or between plots (e.g. to get homogeneous plots for growth studies)? 3) Edge effects: Are edge effects critical in your application? Plots providing data on stand dynamics and for growth models should be homogeneous (i.e. the plot should be relatively uniform), edge effects should be minimized (e.g. trees outside the plot affect growth within the plot, so minimize perimeter relative to area), and plot boundaries should be easy to mark and to relocate (i.e. straight edges). These factors dictate square, fixed-area, permanent plots. To check survival in a new plantation, we would use a different approach. Permanent plots are unnecessary and edge effects are irrelevant. The variance should be minimized within strata, and maximized within plots. Thus strip samples, oriented across the topography, would be optimal. In contrast, when estimating standing volume in a natural (uneven-aged) forest, we shouldn't waste time measuring many small trees (most of the volume is in the big trees), so variable radius plots are appropriate. This could involve smaller subplots for the smaller trees, or sampling with probability proportional to size with an angle gauge. Point sampling is fast and efficient, and is widely used in many forests. Plot Permanence Permanent plots cost more than temporary plots, so there is no point using them without good reason. If you want to measure changes, you must use permanent plots (otherwise you cannot tell if differences are due to time or location), but temporary plots suffice for estimates of present status. Inventory systems may comprise a combination of many temporary and few permanent plots. With such a system, what proportion of plots should be permanent? Like everything else in inventory, that depends on what you want to determine. Theoretical guidelines draw on the relative costs of each and may indicate about twenty percent permanent plots in a continuous forest inventory. In contrast, fewer permanent plots may suffice for growth model development, especially if optimally located. Number of Plots Required Statistical formulae often indicate more plots than the forest manager can afford, and this raises several questions. Are the appropriate formulae being used, is the specified precision really required, and if so, is there a danger that the system will cost more than the resulting data are worth? Numbers of temporary plots can be varied over time to suit changing resources, but permanent plots require an on-going commitment to standards and to remeasurement. Permanent plots provide useful data only when regularly remeasured and when standards and records are maintained, so the number of such plots may be dictated by resources available (funds, manpower and skills) rather than by theoretical considerations. But don't overestimate your capability, as a few reliable plots are better than many incomplete or inaccurate plot records. What to Measure Much has been written on this topic, so a brief summary will suffice. You should measure and record attributes from each of the following categories: 1) Plot establishment details including descriptive location and numeric co-ordinates, plot dimensions and orientation. 2) Site variables including full descriptive and numerical characterization of the plot, forest type and site quality. 3) Trees species, size, vigour and characteristics. On permanent plots, all trees should identified, numbered, tagged and mapped, and co-ordinates should be recorded. 4) Other species present (shrubs, herbs and other species) and their abundance should also be documented. 5) Temporal data such as droughts, floods and heavy seed crops should be recorded, especially for permanent plots. Resource Forecasts Growth models are beyond the scope of this paper, but you should recognize that you will eventually need them to provide information for management and planning. You should also realize that growth models cannot be created overnight, but require several years of data from permanent plots. The sooner you establish these plots, the sooner you can build growth models and prepare yield forecasts. Don't rush out and put plots anywhere, but give some careful thought to the type of plots you need, where to put them, and how to manage them. Permanent plots are a long term commitment, and a little extra care and effort at the outset will pay dividends. Growth models coupled with area and inventory data provide the best way to estimate sustainable harvests and to investigate the impacts of alternative harvesting strategies. Without a growth model, you have to make more assumptions, but you can still estimate the sustainable harvest. One way is to estimate the volume production (per unit area) and multiply by the nett area. Avoid the temptation to overestimate; unless you have very good evidence to the contrary, you should not assume that production will exceed one cubic metre per hectare per annum, on average. Although small areas may exhibit much higher yields, most natural forest estates seem to produce between half and one m³/year per nett hectare. Yields estimates for gross areas will be lower! Another method is to quantify the potential harvest from some typical forest areas, and to estimate the time before a second harvest would be silviculturally and economically viable. For an estimate of the allowable cut, divide the potential harvest by the cycle length and multiply by the nett productive area, but remember it is subjective. Be objective when selecting stands for these estimates. The harvest volume depends on the nature of the initial stand, the size and number of trees removed, the skills of the individuals involved, and damage to harvested stems during felling and handling. These may be subjectively estimated or determined through field studies. The time until the next viable harvest depends on the nature of residual stand (influenced by initial stand, removals and damage), its growth rate, harvesting damage and other losses during the cycle. It may be estimated from permanent plot data, or in some cases, ring counts. In the short term, harvesting practices and the state of the residual stand are more important then theoretical yield estimates. If the forest is well managed and harvesting leaves the residual stand in a productive condition, then a future harvest is assured. However, the long term continuity of the harvest depends on the reliability of such yield estimates. Implications Resource managers have to be pragmatic. We can't wait until we know everything about the whole resource; we have to make do with the data at hand. But we still need inventory. Information is a prerequisite for good management, and informed management requires reliable data. We cannot ignore our information needs; sooner or later we'll need the data, and the sooner we have it the sooner we can use it. It is preferable to identify these needs, plan how to satisfy them, and implement these plans when feasible. This requires a careful study of short and long term needs, examining the alternatives, choosing an appropriate solution, and timetabling its implementation. We need the best data and the best management possible, for ourselves, for our forests, and for our critics. We have to do the best we can with the resources at hand, and that means working co-operatively and efficiently, without wasting resources or data. It means careful gathering of data, and skillful collating and reporting. In short, • Clear, concise and explicit objectives. • Making full use of all existing data. • Planning for the future; starting small, building on success, and working towards an integrated system for long-term needs. • Planning carefully, documenting and reviewing proposals before commencing inventory. • Implementing only those inventory proposals which are feasible within existing resource constraints (time, funds, staff numbers and skills). Further Reading Beetson, T., Nester, M. and Vanclay, J.K., 1991. Enhancing a permanent plot system in natural forests. Proceedings of IUFRO Conference on "Optimal Design of Forest Experiments and Forest Surveys", 10-13 September 1991, London. Hamilton, D.A., 1979. Setting precision for resource inventories: The manager and the mensurationist. Journal of Forestry 77(10):667670. LaBau, V.J. and Cunia, T. (eds), 1990. State-of-the-art methodology of forest inventory: a symposium proceedings, 30 July-5 August 1989, Syracuse, NY. USDA Forest Service General Technical Report PNW-GTR-263, Pacific Northwest Research Station, Portland OR, USA. 592 p. Vanclay, J.K., 1990. Integrated resource monitoring: current trends and future needs. In: Global Natural Resource Monitoring and Assessments: Preparing for the 21st century. Proceedings of conference, Sept 24-30, 1989, Venice, Italy. American Society for Photogrammetry and Remote Sensing, Bethesda, Md, USA. P. 650-658. Vanclay, J.K., 1991. Data requirements for developing growth models for tropical moist forests. Commonwealth Forestry Review 70(4):248-271. Vanclay, J.K. and Preston, R., 1989. Sustainable timber harvesting in the rainforests of north Queensland. In: Forest Planning for People, Proceedings of 13th biennial conference of the Institute of Foresters of Australia, Leura, NSW, 18-22 September 1989. IFA, Sydney, p.181-191. Whitmore, T.C., 1989. Guidelines to avoid remeasurement problems in permanent sample plots in tropical rainforests. Biotropica 21(3):282-283.