The Southern Pine Beetle
Chapter 8: Rating Stand for Susceptibility to SPB
Introduction
Over the past 10 years, more and more emphasis has been focused on the possible prevention of bark beetle outbreaks through forestry practices. In order to prevent attacks, one must know a great deal about insect/host/climatic interactions. This chapter summarizes the state of the art in the application of available knowledge to the prevention problem. Identification of site-stand variables that are associated with SPB attacks has led to a first approximation on many stand rating systems. Development of descriptive and predictive models that would rank forest stands as to SPB susceptibility was a major objective of the Expanded Southern Pine Beetle Research and Applications Program’s coordinated site-stand project. Results of this project can best be understood in light of the historical work done on risk or hazard classification. For the purposes of this chapter, terminology follows that used by researchers in discussing their work.
Early work in the western United States developed tree classification systems applicable to stands under a selection cutting system. Among others, Dunning (1928), Keen (1936, 1943), Keen and Salman (1942), and Salman and Bongberg (1942) developed tree classification systems silvicultural in character and use, but with the primary objective of reducing insect problems. Basically, high-risk trees were logged to reduce potential losses from bark beetles. Johnson (1949) used the percentage of pine volume recently killed by bark beetles, in addition to the percent of present volume of pine in high-risk trees, to develop indices of beetle hazard.
More recent developments in knowledge of bark beetle/site/host relationships have been applied to western forest management problems. Safranyik, Shrimpton, and Whitney (1974) summarized years of work on mountain pine beetle/lodgepole pine biology and ecology and developed management guidelines to reduce losses in Canadian forests. Infestations in stands with an average diameter over 8 inches, that are over 80 years old, and are in the hotter and drier areas of the species range (mild winters) are considered potentially epidemic.
In the Rocky Mountains of the United States, Amman et al. (1977) used tree diameter, tree age, and stand location (elevation and latitude) to predict risk of mountain pine beetle outbreaks in unmanaged lodgepole pine stands. Mahoney (1978) reviewed and tested several such approaches to stand risk classification. Schmid and Frye (1976) developed a system for spruce beetle (Dendroctonus rufipennis [Kirby]) based on psysiographic location, tree diameter, basal area, and percentage of spruce in the canopy. Schenk et al. (1977) and Moore, Schenk, and Hatch (1978) developed and tested a fir engraver stand hazard index for grand fir (Abies grandis) stands based on Crown Competition Factor (Frajicek, Brinkman, and Gingrich 1961) and tree species diversity. Application and evaluation of the utility of these varied forest management tools for the West are progressing, and similar approaches have been developed concurrently in the South.
Classification and Rating of Stands for Risk
The primary task of the coordinated regional project to study site-stand characteristics (see Chapter 4) was to determine what variables (e.g., species composition, stand density, stand age and/or size, site quality, and tree growth rates) were consistently related to outbreaks across a large segment of the beetle’s range (figs. 8-3 and 8-4). Results of this work are summarized in Chapter 4 and in Coster and Searcy (1980). Using these data, researchers have approached stand rating in several ways: (1) Discriminant analysis with primary site, tree, and stand variables, as well as derived variables, from the coordinated regional project; (2) probabilistic models; and (3) qualitative stand risk rating.
Discriminant Analysis
Discriminant analysis is analogous to multiple regression and the same implied statistical assumptions govern (Morrison 1967). In the application of discriminant analysis to the stand rating problem, the range in values of characteristics such as live-crown ratio, radial growth, basal area, and soil depth were examined for both SPB-infested and noninfested plots. Combinations of variates that would discriminate most effectively between infested and noninfested plots were selected for stand hazard models.
By inference, discriminant scores associated with noninfested plots indicate resistance to SPB attack. Those associated with infested plots indicate susceptibility. Noninfested plots consisted of baseline plots established in a line-grid fashion across study areas to characterize general forest conditions. Such plots may or may not have been associated with factors indicative of resistance to bark beetle attack.
Working in the Georgia Piedmont, Belanger, Porterfield, and Rowell (1980) used data from 197 plots (58 SPB-infested, 139 noninfested) to develop a discriminant model with six variables based on data from undisturbed, natural stands. This model correctly classifed 86 percent of 64 SPB-attacked plots from an independent sample.
Discriminant score = 2.38664 - 0.02645 (LIVECRWN) + 0.05551 (AVERAD 1)- 0.10111 (SURCLAY) – 0.08016 (SURDEPTH) + 0.00530 (PERLOB) + 0.07842 (SOIL 1),
where
| LIVECRWN | = | Percent live crown |
| AVERAD 1 | = | Radial growth for last 5 years (mm) |
| SURCLAY | = | Percent clay in surface (0-15 cm) horizon |
| SURDEPTH | = | Depth of A horizon (cm) |
| PERLOB | = | Percent loblolly in total pine component |
| SOIL 1 | = | Percent clay per cm A horizon depth |
A land manager’s model was developed using four easy-to-measure variables that might be included in existing inventories. This model was only slightly less effective — 82 percent accurate — in discriminating between infested and noninfested plots from the 64-plot sample.
Discriminant score = 1.24082 - 0.04829 (LIVECRWN) + 0.10006 (AVERAD 1) + 0.00941 (PERLOB) – 0.12903 (SURDEPTH).
The sign and total discriminant score indicate the direction and degree of susceptibility. Average value for infested plots was –0.6312; average score for baseline plots was 0.3343.
Kushmaul et al. (1979) developed three discriminant models based on data collected in the Louisiana, Mississippi, and Texas Gulf Coastal Plain. Stepwise discriminant analysis of data from natural, undisturbed loblolly and shortleaf stands produced the following model.
Discriminant score = 2.33550 – 0.01906 (pine BA) + 0.01484 (average last 10 years’ radial growth) – 0.00829 (understory %) – 0.00613 (surface soil depth) – 1.71662 (bark thickness – fissure).
Discriminant scores less than –0.13514 indicate infested plots; scores above this value indicate uninfested plots. Use of this model is illustrated in table 8-1.
Tests on a subset of the data not used in developing the model (15 infested and 20 noninfested plots) suggested about 74 percent overall accuracy. A model based on commonly measured variables in continuous forest inventory (CFI) plots correctly classified 80 percent of the infested and 70 percent of the noninfested plot subsets.
Discriminant score = 3.06135 – 0.018342 (pine BA) – 0.00705 (age) – 0.00002 (stand density) – 0.00880 (site index) – 0.04085 (total BA/acre).
Scores less than –0.12736 indicate infested plots. Neither of these models was considered practical for users with limited resources, but each gave some insight into variables associated with infested stands.
A simple mode, including only pine BA and last 10 years’ average radial growth, correctly classified 93 percent of the infested plot subset, but only 65 percent of the noninfested subset.
Discriminant score = 0.93080 – 0.02004 (pine BA) + 0.01827 (average last 10 years’ radial growth).
Scores less than –0.12917 indicate infested plots. Lower accuracy for noninfested plots is understandable considering that the noninfested subset did not necessarily represent resistant stands. These stands happened to be uninfested at the time of the data collection, but they could very well possess characteristics commonly associated with infested stands.
Further testing of this and similar simple models, over time and in controlled pilot studies, may be warranted. The need to bore trees for growth measurements might inhibit routine use of the model, but a correlated variable that is easy to estimate, such as live crown ratio, could possibly substitute for growth measurements. Best results probably could be obtained by applying the model to stands that include species and tree size classes that represent favorable SPB habitat.
Table 8-1. — Discriminant scores and susceptibility rankings for five selected stands. (From Kushmaul et al. [1979]).
| First Discriminant Model | Values For Each Variable For Stands I Through V | Product (Coefficient Multiplied By Variable Value) | |||||||||
| Coeff. | Variable | I | II | III | IV | V | I | II | III | IV | V |
| A | B | A x B | |||||||||
| -0.01906 | Pine basal area (ft2) | 10 | 70 | 90 | 110 | 220 | -.1906 | -1.3342 | -1.7154 | -2.0966 | -4.1932 |
| +0.01484 | Avg radial growth last 10 yr (mm) |
62 | 58 | 25 | 21 | 30 | +.9201 | +.8607 | +.3710 | +.3116 | +.4452 |
| -0.00829 | Understory (percent) | 50 | 10 | 30 | 70 | 40 | -.4145 | -.0829 | -.2487 | -.5803 | -.3316 |
| -0.00613 | Surface soil depth (cm) | 35 | 60 | 15 | 60 | 60 | -.2146 | -.3678 | -.0920 | -.3678 | -.3678 |
| -1.71662 | Bark thickness - fissure (in) |
.2 | .2 | .5 | .5 | .3 | -.3433 | -.3433 | -.8583 | -.8583 | -.5150 |
| Constant term | 2.3355 | 2.3355 | 2.3355 | 2.3355 | 2.3355 | ||||||
| Discriminant score (constant included) | +2.0926 | +1.0680 | -.2079 | -1.2559 | -2.6269 | ||||||
| Stand ranking based on discriminant score (1 = most susceptible) | 5 | 4 | 3 | 2 | 1 | ||||||
In Arkansas, Ku, Sweeney, and Shelburne (1980a and b), working primarily with shortleaf and loblolly pine, sampled 984 SPB-infested and 509 noninfested stands in a study of site and stand conditions related to a bark beetle outbreak. Their efforts led to an equation that accurately discriminated between infested and noninfested stands 75 percent of the time, both with a small subset and with 240 plots used to develop the equation. (Infested plots used had at least 10 infested trees.) The final equation, based on 268 plots in natural stands on upland flat sites, was
Discriminant score = -1.50 (total BA) + 3.3 (stand age) + 64.3 (last 10 years’ radial growth) + 0.93 (hardwood BA).
Scores greater than 100 indicate low susceptibility, greater than 1 and less than 100 indicate medium susceptibility, and less than 1 indicate high susceptibility.
Ku, Sweeney, and Shelburne believe that this equation applies best to undisturbed natural stands on upland flats — sites from which the data base originated. Its utility on other landforms, in plantations, or on disturbed sites is questionable.
Everyone would like to see a generally applicable stand rating model for the Coastal Plain. Variations in site, stand, vegetation, and climate across the SPB range in the Coastal Plain complicate the task, but Porterfield and Rowell (1980 unpublished) developed a working model that they consider useful.
Using data collected by collaborators from Texas to Virginia, Porterfield and Rowell developed a discriminant analysis to select a set of site-stand variables that, in combination with each other, was most reliable in classifying stands as to infestation status. Individual plots were either SPB-attacked or unattacked, and were in naturally established, undisturbed stands. Infestations had to include five or more SPB-killed trees, an indication that site-stand conditions may have enhanced sustained activity. The data set included 547 infested and 474 uninfested plots and yielded the following discriminant model:
Discriminant score = 1.02559 - 0.00043 (total volume) + 1.33776 (proportion sawtimber) - 2.14726 (average bark thickness) + 0.01878 (10 years’ radial growth) + 0.03205 (slope) - 0.00791 (proportion of total BA in pine)
where
Total volume = ft3 of pine > 4.6 inches d.b.h.
Proportion of sawtimber-sized pine = (ft3 > 9.6 inches)/(ft3 > 4.6 inches)
Average bark thickness = average of fissure and ridge bark thickness (nearest 0.1 inch)
10 years’ radial growth = millimeters (at breast height)
Slope = ground slope in percent
Proportion of total BA in pine = (ft2 of pine BA)/(total BA)
Scores below 0.044185 are best classified as SPB-attacked; those above, unattacked. This model correctly classified 79 percent of the plots used in developing it. With an independent sample of 119 plots (69 SPB-infested, 50 noninfested), it correctly classified 74 percent of the plots. Porterfield and Rowell suggest that the model tended to classify too many infested stands as baseline but note that many misclassifications were borderline. They point out that degrees of susceptibility are continuous, and that a stand’s relative discriminant score in relation to other stands is more important than its classification. Intuitively, we know that many stands exist in a range of degrees of susceptibility to SPB attack. The problem is to find a practical and useful means of identifying the especially susceptible ones.
Porterfield and Rowell’s model uses some variables derived from basic measurements made by collaborators, and would involve considerable effort to apply. It appears to have utility but needs to be tested with sample data from various locations across the Coastal Plain. Species is not an explicit variable in the model; one might consider its application to all southern pine stands, or perhaps to stands stratified by species. Experience indicates that longleaf and slash pine are less suitable hosts for SPB than either loblolly or shortleaf.
Probabilistic Models
Hicks et al. (1980) developed a probability of attack model for east Texas forests from site and stand data on 484 SPB-infested and 416 noninfested plots. They first determined the variables most strongly associated with infestations by stepwise discriminant analysis. A discriminant function, including bark thickness in fissures, pine BA, average tree height, and landform category, correctly classified 79 percent of the plots used to develop the function.
Subsequently, Hicks et al. determined that a discriminant function using only pine BA, average height, and landform category correctly classified 72 percent of the sample plots. Then, following an estimate of overall probability of SPB attack in east Texas based on prior incidences of attack per unit area of SPB host types (loblolly-shortleaf pine and oak-pine), they calculated the probability of attack (Pa) according to the general equation:
| Pa | = | 3 å i = 1 |
Pi |
where
| Pi | = | A X D X W |
| Pi | = | probability due to an individual variable |
| A | = | overall probability of attack |
| D | = | ratio of percent of infestation in a variable class to the percent of noninfested plots in that class |
| W | = | variable weight based on the standardized coefficients from discriminant analysis |
For example, Pa for loblolly pine stand averaging 25 m in height, with 28 m2/ha BA, located on a stream terrace is determined as follows:
| A | = | 0.00093 from Hicks’ three-county study area estimation of the area ratio of sampled infestations to SPB host types in the study area (3-year period)(Hicks et al. 1980) |
| D | = | 1.25 for height = 25 m, 2.715 for BA = 28 m2, and 6.857 for landform category = stream terrace |
| W | = | 0.268 for height, 0.498 for BA, and 0.234 for landform category |
| A | D | W | ||||||
| Pi (height) | = | 0.00093 | x | 1.25 | x | 0.268 | = | 0.000311 |
| Pi (basal area) | = | 0.00093 | x | 2.715 | x | 0.498 | = | 0.001257 |
| Pi (landform) | = | 0.00093 | x | 6.857 | x | 0.234 | = | 0.001492 |
Therefore,
Pa = 0.000311 + 0.001257 + 0.001492 = 0.003050,
or about a chance of 1 in 328, if the succeeding 3 years of SPB activity is similar to the preceding 3 years in the area of interest.
Hicks et al. suggest several uses for their probability model, depending on institutional or landowner objectives. The model should be tested with independent data in east Texas. Variables used in the model are meaningful and relatively easy to measure, but the estimate of A for an area of interest and its application to the future years of interest may present serious difficulties.
Another probabilistic model was offered by Daniels et al. (1979). Their model involved use of the logistic function for estimating a continuous measure of SPB incidence or probability of outbreak. This approach provides a more general incidence index than categorical classification methods, and its validity is not dependent upon certain distributional assumptions, as in discriminant analysis.
A continuous probability function is most meaningful, according to Daniels et al., when applied to a specific forest land area and a specific time period (perhaps 1 year). Such a model would be most useful for decision making if probabilities can be estimated as a function of site, stand, and insect population variables that (1) are associated with differences in outbreak probabilities, and (2) can be easily measured in the field or from aerial photographs. Insect population variables may reflect changes in probabilities between endemic and epidemic periods, but useful data are difficult to obtain.
The logistic function,
p = 1 / (1 + e – (b0 + b1x1 + b2x2 +…bkxk) )
cannot be estimated directly (p is not observable), but uninfested (0) or infested (1) may be taken as observed values of p and related to the x variables in the logistic function characterizing stands with and without outbreaks. Direct estimation of the coefficients of the x variables in the logistic function by the maximum likelihood procedure is preferred over discriminant function estimators because the former yields asymptotically unbiased coefficient estimates independently of any distributional assumption about the data (Halperin, Blackwelder, and Verter 1971).
Daniels et al. (1979) developed a logistic regression model based on site and stand variables measured by collaborators in the site-stand regional study (Chapter 4). They used data from 187 natural stands to fit a two-variable model for disturbed and undisturbed categories. Total stand BA (x1) and proportion of the BA in pine (x2) were chosen for the models.
Undisturbed p = -8.599 + 0.044x1 + 3.309x2
Disturbed p = -9.998 + 0.088x1 + 4.801x2
Models with additional independent variables gave similar results.
Differences in sampling intensities of infested and noninfested (baseline) populations required differential weighting of the data. Infested data were estimated to represent 100 times the sampling intensity of noninfested data, so noninfested plots were weighted by 100 in the estimation procedure. Stand size was not part of the basic data, so p’s were for "average" stand sizes and stand size distributions were assumed to be the same for infested and noninfested populations.
The developers of this method believe it has distinct advantages in that (1) it provides a continuous estimate of incidence that may still be partitioned according to user needs or wishes, (2) the probabilities are meaningful since they indicate the chance of an outbreak, and (3) the probabilities may be combined in use with other guideline models for a variety of management objectives.
The approach is attractive and should be explored more extensively. Species or forest type, and tree age or size variations from the 187-plot samples Daniels et al. used would probably limit the utility of these specific models in broad applications.
Qualitative Stand Risk Rating
Efforts to develop qualitative methods of stand risk rating for southern pine beetle include work by Belanger and the Georgia Forestry Commission (R.P. Belanger personal communication). Their system for field evaluation of stand susceptibility to SPB attack includes stand, representative tree, and site characteristics, as follows:
| Ranking the Susceptibility of Stands to SPB Attack | |||
| Stand | |||
| 1. | Shortleaf pine >= 50% total pine | Yes __________ | No __________ |
| 2. | Hardwood component <= 25% total stand | Yes __________ | No __________ |
| 3. | Pine BA >= 130 ft²/acre | Yes __________ | No __________ |
| Representive Tree | |||
| 4. | Radial growth (last 5 years) <= ½ inch | Yes __________ | No __________ |
| 5. | Live crown ration >= 40% | Yes __________ | No __________ |
| Surface Soils 0-6 inches | |||
| 6. | Micaceous red clays | Yes __________ | No __________ |
The yes answers are totaled and a hazard ranking and need for cultural treatment given according to the following diagram.
| Hazard Ranking | |||||||
| Total of "yes" answers | Low | Moderate | High | ||||
| 0 | 1 | 2 | 3 | 4 | 5 | 6 | |
| Cultural treatment | Not Needed | Needed | |||||
This system is currently being tested throughout the Georgia Piedmont by Belanger and Georgia Forestry Commission personnel.
Lorio (1978) proposed the use of available forest resource inventory data such as forest type, tree size and/or age, stand density, and site index for evaluating stand risk to bark beetle attack. Some basic assumptions are made with this approach: (1) All southern pines are susceptible to attack, but loblolly and shortleaf are the primary host species. (2) Lack of knowledge about SPB population dynamics prohibits effective prediction of infestations over time. (3) SPB needs for food and habitat for abundant reproduction are related to recognizable stand characteristics. (4) Stands favorable for SPB food and habitat also constitute potentially large resource losses. (5) Routine forest inventory data, used for a wide variety of forest management planning and decisionmaking purposes, can also be applied effectively to stand risk classification for the SPB.
Continuous Inventory of Stand Conditions (CISC), an automatic data processing system used for National Forests in the South that continuously reflects up-to-date description of timber stands, was used to classify stands on the Kisatchie National Forest in Louisiana. Five criteria in CISC are being used currently in classification: forest type, stand condition class, method of cut, operability, and site index. Forest type is self explanatory. Stand condition class includes consideration for damage, quality, density, and age. There are 15 classes, of which immature poletimber, immature sawtimber, and mature sawtimber are particularly important for risk classification. Method of cut describes the silvicultural treatment needed for the stand, such as clear cutting, thinning, seed tree. Operability indicates the kind and mixture of products to be removed by the method of cut and must be compatible with the method of cut. On the Kisatchie National Forest operable pine poletimber stands must yield at least 3 cords/acre, and sawtimber stands at least 800 fbm/acre (Scribner rule) under a silviculturally acceptable method of cut. Inoperable stands contain very low volume of resource, representing low risk for SPB outbreak even though individual trees might be very susceptible to attack. Site index gives an indication of the quality or capability of the site. The more productive sites can produce larger trees and more of them per acre in less time than poorer sites, and these good sites with moist or wet water regimes constitute potentially greater risk of SPB outbreaks.
Initial retrospective tests of the classification, based on 25 months of infestation data (June 1975 - June 1977), showed that stands classed as high risk had 10.5, medium 6.3, and low 3.2 infestations per 1,000 acres. A second predictive test was extended another 30 months on the Catahoula Ranger District with the following results: high 12.1, medium 5.8, and low 3.1 infestations per 1,000 acres (fig. 8-5).
These results on over 100,000 acres over 4½ years strongly indicate some practical utility of the system. Stands classed as high and medium risk comprised only about 23 percent of the study area (fig. 8-6). Identification of such stands significantly reduces the area of primary concern relative to potential SPB outbreak and provides criteria that can be used in deciding which stands to thin or regenerate, what sequence of treatment to follow during a cutting cycle, and what stands to monitor for potential problems.
Tests and revisions of the approach are continuing on the Kisatchie National Forest. CISC does not include BA, so method of cut and operability are used as broad indicators of density in risk classification of stands. Currently, provisions are being made on the KNF so that foresters may include BA in CISC as well as enter an SPB risk classification code if desired. Criteria have been prepared for prescriptionists to use on a trial basis. These include basal area criteria based on thinning guides currently in use on the Kisatchie, and on stand data from 318 SPB infestations collected over a 25-month period (Lorio and Sommers 1980).
Aerial Approaches
Sader and Miller (1976 unpublished) developed a risk-rating system based on a study conducted in Copiah County, Mississippi. The study revealed that a trained interpreter could estimate four forest stand and topographic variables from 1:24,000 color infrared imagery. These primary variables were species composition, stand size (sawtimber, poletimber, etc.), stand density, and topographic position. Sader and Miller assigned a numerical weight to each variable according to their estimate of the degree of influence each has in creating conditions favorable for SPB attack. Secondary variables within each primary variable were assigned a numerical rank based on relative susceptibility to initial attack and potential for infestation spread.
Sader and Miller’s weight and risk values for stand and topographic variables are given in table 8-2. In their study, dense pure pine stands with sawtimber-size trees on ridges (broad inter-stream divides) had the greatest probability of initial attack and infestation spread. Such stands also constituted the greatest potential resource loss.
Examples of use of their system are:
Stand 1: Pine, sawtimber, dense, ridge position
(PsDR) = 154 (high risk)
Stand 2: Pine-hardwood, pole, sparse, lower 1/3 slope position
(PHpSL 1/3) = 53.5 (low risk)
No explicit differentiation of pine species was included in the system, but loblolly and shortleaf were the predominant species in Copiah County. Sawtimber was defined as trees over 11.5 inches d.b.h.; large poletimber to sawtimber as 9.6 to 11.5 inches d.b.h., and poletimber as 5.0 to 9.5 inches d.b.h. Stand density criteria were, dense > 120 ft2 BA/acre, normal 80-119 ft2 BA/acres, and sparse < 80 ft2/acre.
Sader and Miller used 235 infestations from 1974 in developing their model and 42 infestations from the decreased activity in 1975 in attempting to evaluate it. Their evaluation did not include consideration of the relative area occupied by each of their five risk classes (high, moderately high to high, moderately high, moderate, and low). They concluded that the approach was encouraging but in need of further refinement through data input over a longer timespan.
Table 8-2. — Weight and rank values for stand and topographic variables (from Sader and Miller 1979 unpublished)
| Weight | Primary Variable | Secondary Variable | Rank | Total |
| 15 | Species composition | Pine (P) | 3 | 45 |
| Pine-hardwood (P-H) | 1.5 | 22.5 | ||
| 7 | Stand size | Sawtimber (s) | 3 | 21 |
| Large pole to small sawtimber (pl) | 2 | 14 | ||
| Pole (p) | 1 | 7 | ||
| 12 | Density | Dense (D) | 4 | 48 |
| Normal (N) | 3 | 36 | ||
| Sparse(S) | 1 | 12 | ||
| 8 | Topographic Position | Ridge (R) | 5 | 40 |
| Upper slope (U1/3) | 4 | 32 | ||
| Middle slope (M1/3) | 2 | 16 | ||
| Lower slope (L1/3) | 1.5 | 12 | ||
| Minor bottom (Mb) | 1.5 | 12 |
Mason (1979) has developed a similar approach to risk rating in east Texas, based on photo-interpretable variables. He first examined the work of Hicks et al. (1979) and determined which of the stand and site variables that they reported to be closely associated with SPB infestations could be assessed with reasonable accuracy by photo interpretation methods. These were total BA/acre, tree species, average tree diameter, average height, and landform.
After preliminary geographical and host type characterization with 1:250,000 LANDSAT false color composites, Mason selected 10 U.S. Geological Survey 15-minute quadrangles representative of the east Texas piney woods. Within each of these he chose an 18,000-acre test block for detailed habitat mapping with the use of large- and small-scale color infrared photography (1:5,000, 1:10,000, and 1:60,000) (fig. 8-7 A-C). Using high-intensity dot grids (10,000 dots/in2) and specially developed equations, he estimated percent pine stocking (composition), BA/acre, average d.b.h., and percent crown closure. Stand height was visually estimated within broad classes by stereo observation and field reconnaissance.
Data were extrapolated from large-scale strips to areas covered by small-scale photos, to produce habitat type maps and landform overlays at a scale of 1:24,000. Ground checks verified that photo estimates of mapping variables were 92 percent accurate, and stepwise discriminant analysis yielded two photo-applicable models:
Infested stand characteristics = 7.76 (HTC) + 4.42 (BAC) + 4.49 (LDC) – 26.05
Baseline characteristics = 7.06 (HTC) + 3.40 (BAC) + 5.02 (LDC) – 22.92
where
HTC = Height class
BAC = Basal area class
LDC = Landform class.
These equations were about as accurate (71 percent) as models based on more variables or on additional variables not suitable for photo interpretation (Hicks et al. 1979).
Hazard ratings were assigned to stands within the 18,000-acre test blocks. Results with one such block based on 1973-1979 SPB infestation data in the Texas Forest Service’s Operational Control System were encouraging.
| Hazard Class | ||||
| Very High | High | Moderate | Low | |
| Infestation/1,000 acres | 10 | 10 | 6 | 2 |
| Trees killed/1,000 acres | 92 | 399 | 107 | 2 |
Evaluation is continuing on the nine additional 18,000-acre test blocks, and plans are being developed to test the system outside the are where it was developed.
Mason (personal communication) recognizes the need to differentiate somehow between the terms "hazard" and "risk," which are often used interchangeably. In one sense hazard may be high for an individual tree or small group of pines within an essentially hardwood stand, but the risk in terms of potential loss of resource may be quite low for the stand as a whole. Aerial photo interpretation techniques provide a useful means of recognizing and sorting out such differences.
Aerial approaches like Sader and Miller’s and Mason’s should be especially useful in situations where resource inventory data are lacking or are not available in a useful form.
Current Status and Future Needs
Great effort has been marshaled by the ESPBRAP to focus on the development of useful tools for forest managers to avoid and deal with southern pine beetle problems. A variety of approaches has yielded valuable insight into some basic interrelationships between the SPB and its host and site environment. Some potentially useful tools have been developed. Conscious effort should be given to developing guidelines for specific potential users in the context of the entire forest management problem. The SPB is only one small aspect of the complexity involved in management of the southern pine forests. So the closer research can come to developing useful tools that are easily understood, and that can be easily integrated into forest management planning systems, the more likely managers are to use them. Getting consideration for potential SPB outbreak incorporated into overall planning systems would be a major accomplishment in forestry practice in the South.
The essential basic knowledge needed for developing practical guidelines to forest managers is at hand. A large vacuum exists, however, with regard to knowledge of the basic relationships between host stand characteristics and SPB population dynamics. Future work toward refinement of SPB population dynamics models should involve closely integrated studies of population dynamics and tree, stand, and site characteristics. Hope for methods to predict continued growth or collapse of individual infestations and for predicting the start, continuation, or collapse of outbreaks hinges on such integrated research.
Developed by the University of Georgia Bugwood Network in cooperation with USDA Forest Service - Forest Health Protection, USDA APHIS PPQ, Georgia Forestry Commission, Texas Forest Service
and the Pests and Diseases Image Library - Australia
Home | Image Usage | Accessibility Policy | Privacy Policy | Disclaimers | Contact Us
Last updated on Monday, August 14, 2006 at 10:52 AM
www.barkbeetles.org version 2.0