AN APPLICATION OF POINT-IN-TIME EXTENSIVE FOREST INVENTORY DATA ANALYSIS TO ASSESSMENT OF SPRUCE BARK BEETLE INFESTATION TRENDS AND DYNAMICS Willem W. S. van Hees U.S. Forest Service, Pacific Northwest Research Station Forestry Sciences Laboratory, Suite 200, 3301 C Street Anchorage, AK 99503-3954 ABSTRACT: Estimations of areas of forestland and numbers of trees affected by a spruce bark beetle infestation on Alaska's Kenai Peninsula are presented. Selected forest stand conditions pertinent to infestation dynamics are examined. Application to improving understanding of infestation trends and dynamics while providing logistical efficiency is demonstrated. INTRODUCTION The purpose of this paper is to demonstrate that it is possible to use data from an extensive forest inventory to assess some dynamics of a widespread insect infestation. Extensive inventory, in this context, means two things -- large area coverage and low sampling intensity. The point of this demonstration is that logistical and financial saving might be realized when the needs driving traditional forest inventories are combined with needs of regional land use planners to understand the current state, and likely progress, of insect infestations. Often, multiple observations in time must be procured in order to estimate conditions or trends in dynamic systems such as insect infestations. The forest inventory used here as an example was conducted on Alaska's Kenai Peninsula, ( Figure 1) by the Anchorage Forestry Sciences Laboratory of the US Forest Service's Pacific Northwest Research Station. The inventory unit area is about 2.3 million acres (van Hees and Larson, 1992). Nearly 40 percent of this, almost 2 million acres, is forested. Not all of this is productive forest land. Only 480 thousand acres are classed as timberland. That is, forest land capable of producing 20 cubic feet of wood, or more, per acre per year. The white spruce component of the forest land on the Kenai Peninsula has been subjected to an ongoing infestation by spruce bark beetles for a number of decades. Rapid expansion of the infestation has recently occurred for reasons including favorable weather conditions and an aging resource. Most forest land on the Peninsula is not affected by the infestation; only timberland, the more productive segment is so favored. The particular infestation dynamics addressed in this paper are those relating to changes in area of infestation, numbers and sizes of trees affected, and the relation between tree stocking and recent radial growth in the face of the infestation. METHODS AND DATA The inventory design used on the Kenai Peninsula was a fairly standard stratified two-phase design. In the first phase, about 5,600 photo points were systematically distributed over 1:60,000-scale aerial photos and then interpreted. Each photo point was classified into one of four land classes; timberland, other forest land, nonforest, and water. From the photo points, a random sample of about 1,200 points was selected for ground visitation. Because the focus of the inventory was on timber, only timberland locations and other forest land locations photo-interpreted as being marginally capable of producing 20 cubic feet per acre per year were actually visited on the ground. 129 plots were selected for ground visitation. The ground plots were 5-point, variable radius clusters which covered about 5 acres. Inventory design error standards were plus or minus 3% per million acres of timberland and plus or minus 10% per billion cubic feet of gross tree volume given 68% confidence limits. In order to develop an understanding of a dynamic system like an insect infestation from a point-in-time observation such as a single forest inventory, it is necessary that the observations of the system be separated into classes or groups respresenting various stages of the system. The analyses referred to in this paper require knowledge of the current stage of infestation of the stands in which the ground plots were placed. To obtain this, forested ground plots were separated into four infestation categories (van Hees, 1992). The separation was based on whether or not there were beetle attacked trees on the plot, whether or not those trees were living or dead, and if dead, how long. The stage-of-infestation categories were: uninfested, potential, ongoing, and past. Uninfested -- Plots where no live or dead tree showed any sign of beetle attack. Potential infestation -- Plots where only live trees showed signs of beetle attack and plots where live trees and trees dead more the 5 years showed signs of attack. This latter group represents areas of potential re-infestation. Ongoing infestation -- Plots where live trees and trees dead less than 5 years showed signs of beetle attack plus plots where live trees, trees dead less than 5 years, and those dead more than 5 years showed indication of attack. Past infestation -- Plots where only dead trees showed signs of beetle attack. Estimates of the area of productive forestland affected by each infestation category were developed by summing the area expansion factors for the plots that fell in each stage-of-infestation category. As a way of checking whether or not the inventory estimates were at all reflective of reality, an effort to validate these estimates was make by comparing them with estimates derived from successive annual aerial survey maps. By "subtracting" the current annual aerial survey map from the prior one, a "difference" map was produced that allowed development of estimates of the area of new, or potential infestation. Maps of successive prior years were overlaid to derive difference maps for the other infestation categories. Comparing the inventory estimates with the mapped estimates ( Figure 2 ) shows that the inventory estimates are likely reasonable. Given the vagaries of aerial mapping and sampling errors, the two sets of estimates agree well. From these area estimates it is apparent that the infestation has expanded and there are indications it will continue to expand. The evidence for recent expansion is that the area of ongoing infestation is larger than the area where the infestation has passed through. Indication that the infestation could continue to expand is seen in the magnitude of area with potential infestation in context with the area of past infestation. Trees unsuccessfully attacked in stands once attacked are more likely to succumb in subsequent attacks because of probable lack of vigor that led to the first attack and stress induced by the first attack. The existence of a sizable, vulnerable resource (the area where infestation has passed through already) along with an area of potential infestation of nearly equal size points at an infestation that can at least maintain its current magnitude. For the resource manager interested in quantities of timber resource currently and potentially available for salvage, plot level information can easily be used to estimate numbers of trees and volumes. The pieces of plot level information referred to here are trees-per-acre and volumes-per-acre by diameter class figures. Segregating this information by infestation category can give the manager an idea of the resource already available and that which will likely become available in the near future. So, how much of the resource is involved? How much is currently alive, how much is dead, and what part is likely to die in the near future? This analysis started with a look at the overall situation. Nearly half of the current total white spruce resource (including both living and dead trees) is in beetle attacked areas ( Figure 3 ). And most of that half is in areas where the infestation is ongoing. The next largest component is in areas where the infestation has passed through. This supports the notion that the infestation is expanding. Also, this picture provides graphic evidence that the infestation has been around for some time. A clearer picture of the impact of the infestation is formed if percent of total numbers of white spruce is examined ( Figure 4 ). In those areas where the infestation has passed through and in those areas where it is ongoing, the trees in almost every diameter class that have been attacked or are at risk of attack represent between 15 and 30 percent of the total resource in that diameter class. An observation of real interest here is that for those areas where the infestation is ongoing or potential, most of the diameter classes are suffering equally within the category. Definite indication that all segments of the white spruce resource are currently in trouble or will be so shortly. The next questions then, are -- how much of the live resource is actually under attack and how much of the dead resource is so because it was attacked? Figure 5 shows numbers of just the live tree component separated by attack status - that is under attack or not. The 'past' category has no live trees under attack -- by definition. Because of the numbers of trees in the uninfested category, the scale of this graphic tends to downplay the importance of the number of live trees under attack. By examining the percent of live trees in each category showing signs of attack, the impact of the infestation becomes apparent ( Figure 6 ). Of the live trees in areas where the infestation is ongoing, between 5 and 25 percent of the trees in the affected diameter classes are under attack. In the areas where there are potential infestations, between 2 and 12 percent of the trees in the affected diameter classes are under attack. Looking at the dead component ( Figure 7 ), it is noticeable that, except for areas where there are potential infestations, most of the dead trees in each diameter class are dead due to beetle attack. Again, looking at the same information but in terms of percentages provides a dramatic graphic of the impact of the infestation ( Figure 8 ). For most diameter classes, 60 percent or more of the trees died because of beetle attack. These snapshots of the past, current, and likely future condition of the white spruce resource are only a sample. However, at this point a few conclusions can be drawn. ONE -- It is apparent that the infestation has increased substantially in recent years, and TWO -- it is likely the infestation will maintain, if not increase, its magnitude for the near future. The assertion that the infestation has increased derives from three observations of the data: 1. When estimates of the percent of all dead trees affected by beetle attack are compared for areas of ongoing versus areas of past infestation it is noticeable that except for the 4- and 6-inch diameter classes, most succeeding diameter classes show higher relative numbers died due to beetle attack in the areas of ongoing infestation than in areas where the infestation passed through. In other words, more of the resource is dead because of ongoing infestations than because of past infestations. 2. As noted earlier, the estimated area of timberland affected by ongoing infestations is high, 121 thousand areas. This is greater than the estimated area where infestations have run their course; 85 thousand acres -- a definite indication of expansion. 3. Half the current white spruce resource is within beetle attack areas. The possibility that the infestation would likely maintain and perhaps increase in magnitude is based on two observations: 1. The residual, live-tree resource on areas where the infestation has passed is larger than the live-tree resource in either areas of ongoing infestation or areas of potential infestation. As mentioned above, this is a vulnrable resource and will succumb easily if any subsequent attack occurs. 2. The estimated area of potential infestation, 81 thousand acres, is already 2/3 the area in currently ongoing infestations (121 thousand acres). This suggests that at least for the near term (next 5 years), the infestation can maintain its current magnitude. Although most of the analyses presented so far refer to numbers of trees, these could be easily converted to volume based analyses. An understanding of how many acres, trees, or cubic feet of volume are being subjected to infestation is useful for regional awareness of the extent of the problem and for assessing the possible flow of expected forest products. However, for the manager interested in improving the health of the forest in order to help it better ward off insect attacks in the future, it is important to understand some of the stand dynamics involved; such as the relation between radial growth and stand stocking as affected by the infestation. Table 1. shows least-squares regression equations for 10-year mean plot radial growth as a function of the natural log of the number of trees per hectare hreater than 9 cm in diameter, by state of infestation (van Hees, in press). Note that the values of the regression intercept decrease steadily from potential infestation to past infestation. This pattern among the estimated relations indicates that as the infestation progresses mean plot radial growth decreases. Table 1. Regressions of plot mean spruce radial growth on plot stocking of all species > 9 cm dbh, Kenai Peninsula, Alaska, 1987. Category Regression relation R-squared n -------- ------------------- --------- --- Potential y = 35.955 - 4.269 ln(tph9) 0.514 11 Ongoing y = 24.483 - 2.362 ln(tph9) 0.265 22 Past y = 14.323 - 0.922 ln(tph9) 0.112 25 Hard et al. y = 12.961 - 1.360 ln(tph9) 0.671 25 The possibility that stand density differences might account for radial growth differences seen here was addressed by examining plot mean number of live trees per hectare by stage of infestation. This was live-tree stocking by all species, not just the white spruce. T-tests indicated the differences between the means were not different from zero at the 0.05 level of significance (van Hees, in press). Also, except for one plot in the ongoing infestation group, all plots used for the three stages of infestation were on sites capable of producing between 20 and 50 cubic feet per acre per year. So, the plot groupings used for each of the three infestation stages are similar in that plot mean stand density levels are likely the same and the site qualities are essentially the same. Stand stocking level differences then, are likely not the only important explanatory variable for changes in the estimated regression relations. Effects of the spruce beetle infestation are also consistent with patterns observed in the estimated regression relations. For verification, the regression relations developed here were compared with an estimate made by Hard, who conducted small-scale studies of the same infestation (Hard et al., 1983). Hard examined the relation between mean cumulative 5-year radial growth of spruce on each plot and live-tree stocking in an ongoing infestation on the Peninsula. Hard's result, in magnitude, is nearly one-half that found for the ongoing infestation category in my study. This is because Hard's estimate of radial growth was for a 5-year period as opposed to the 10-year period used in this study. Although the R-square of the estimated regression relation in Hard's study is higher, the similarity of estimated regression coefficients indicates the inventory data may provide managers with estimated relations that would not be wildly divergent from true relations. So, the inventory data verify what is intuitively reasonable. As stocking and stress from attack go up, radial growth goes down. But now the manager would have a numeric model to work with. SUMMARY What can a resource manager do with this kind of information? Some questions that can be addressed using point-in-time forest inventory data include: "Do we need to be concerned that the infestation needs attention? How much area are we going to be dealing with? What are the product sizes and quantities that could be available?" At this point, the purpose of this paper has been demonstrated. One can conduct an extensive forest inventory and from it, garner information that will provide a manager with material regarding the current state and the likely near-term future state of an infestation, some idea of expected forest products flow, and an ability to model various stand dynamics in the face of an infestation. METRIC EQUIVALENTS 1 inch = 2.54 centimeters 1 acre = 0.4047 hectare 1 cubic foot per acre = 0.07 cubic meters per hectare LITERATURE CITED Hard, J. S., R. A. Werner and E. H. Holsten. 1983. Susceptibility of white spruce to attack by spruce beetles during the early years of an outbreak in Alaska. Can. J. For. Res. 13:678-684. van Hees, Willem W. S. 1992. An anlytical method to assess spruce beetle impacts on white spruce resources, Kenai Peninsula, Alaska. USDA For. Serv. Res. Paper PNW-RP-446. 15 p. van Hees, Willem W. S. and Edward H. Holsten. [In press] An evaluation of selected spruce bark beetle infestation dynamics using point-in-time extensive forest inventory data, Kenai Peninsula, Alaska. Can. J. For. Res. van Hees, Willem W. S. and Frederick R. Larson. 1991. Timberland resources of the Kenai Peninsula, Alaska, 1987. USDA For. Serv. Resource Bull. PNW-RB-180. 56 p.