UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Remote sensing of Douglas-fir trees newly infested by bark beetles Hall, Peter Michael 1981

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


831-UBC_1981_A6_7 H34.pdf [ 4.99MB ]
JSON: 831-1.0095232.json
JSON-LD: 831-1.0095232-ld.json
RDF/XML (Pretty): 831-1.0095232-rdf.xml
RDF/JSON: 831-1.0095232-rdf.json
Turtle: 831-1.0095232-turtle.txt
N-Triples: 831-1.0095232-rdf-ntriples.txt
Original Record: 831-1.0095232-source.json
Full Text

Full Text

REMOTE SENSING OF DOUGLAS-FIR TREES NEWLY INFESTED BY BARK BEETLES by PETER MICHAEL HALL B. S c , The U n i v e r s i t y of V i c t o r i a , 1972 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE i n THE FACULTY OF GRADUATE STUDIES (Department of Forestry, Forest Entomology) THE UNIVERSITY OF BRITISH COLUMBIA January 1981 0 Peter Michael H a l l , 198I In presenting th i s thes is in pa r t i a l fu l f i lment of the requirements fo r an advanced degree at the Univers i ty of B r i t i s h Columbia, I agree that the L ibrary sha l l make i t f ree ly ava i lab le for reference and study. I further agree that permission for extensive copying of th i s thesis for scho lar ly purposes may be granted by the Head of my Department or by his representat ives. I t i s understood that copying or publ icat ion of th i s thesis for f i nanc ia l gain sha l l not be allowed without my wr i t ten permission. Department of „ The Un ivers i ty of B r i t i s h Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1W5 Date y/^. ft f i i ABSTRACT Two study plots containing Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) newly infested by Douglas-fir beetle (Dendroctonus pseudotsugae Hopk.) were established and photographed with large-scale (1:1000), colour infrared film on July 29. 1979 - approximately three months after possible insect attack. Ground checking confirmed attacked trees and also showed that at the time of photography a l l trees had visually green, healthy-appearing foliage. A l l trees, both attacked and non-attacked i n each plot were matched to their photographic images, and visual photo interpretation for damage types and densitometric analysis of the original transparencies were done. For each tree-crown image included, the yellow, magenta and cyan dye layer density measurements were taken and these values plus three ratios derived from them were tested s t a t i s t i c a l l y using analysis of variance and stepwise discriminant analysis. Significant differences were found between the optical density values of the images of healthy and attacked trees. The ratio values had much smaller variances than did the individual dye layer densities and a l l three ratios showed significant differences between healthy and attacked trees. Stepwise discriminant analysis produced significant separation of damage classes. Two-thirds of the successfully attacked trees were correctly cla s s i f i e d and were confirmed by a second ground check i n January, 1980. It i s concluded that successfully beetle-attacked trees have a unique spectral signature than can be detected on colour infrared a i r photos approxi-mately three months after i n i t i a l attack when the trees s t i l l support visually green, healthy-appearing foliage. i i i TABLE OF CONTENTS TITLE PAGE Page i ABSTRACT i i TABLE OF CONTENTS i i i LIST OF TABLES i v LIST OF FIGURES v ACKNOWLEDGEMENT v i 1. INTRODUCTION 1 1.1 General Impact of Baxk Beetles 1 1.2 Douglas-fir Beetle Life History 3 1.3 Management Considerations 7 l.k Survey and Detection Methods to Date 9 1.5 Early Detection - Methods and Success to Date 11 1.6 Densitometry 17 1.7 Study Objectives 19 2. METHODS 22 2.1 Study Region 22 2.2 Selection and Location of Photo Plots 22 2.3 Photography Z% Z.k Analysis of Photographs 25 2.4.1 Analysis of Densities 25 3. RESULTS AND DISCUSSION 2? 3.1 Data Description 27 3.2 S t a t i s t i c a l Analyses 36 3.2.1 One-May Analysis of Variance 36 3.2.2 Stepwise Discriminant Analysis kO 3.3 Girdled Trees 48 h. CONCLUSIONS 50 5. LITERATURE CITED 54 APPENDIX I. Descriptive s t a t i s t i c s for a l l six variables i n a l l five damage classes 59 APPENDIX II. Histograms of a l l six variables i n 6 2 a l l five damage classes i v LIST OF TABLES TABLE PAGE I Volume Losses (m ) Caused "by Bark Beetles (Dendroctonus sp.) i n B.C. Forest Regions, 2 1971-75 II Structure of Pine Needles (Pinus ponderosae Laws.) Before and 10 Months after Mountain Pine Beetle 13 Attack III F i l t e r s Used to Measure Specific Dye Layers on Colour Infra-red Film on a Densitometer VIII Summary of Stepwise Discriminant Analysis of Three Damage Classes with Green Foliage 20 IV Number of Trees Used for Each Damage Class i n 2 o the Analysis of Dye Layer Densities V Means of Each Variable for Each of the Three Damage Classes with Green-appearing Foliage 37 (healthy, pitched-out and successfully attacked) VI Summary of Stepwise Discriminant Analysis of ^ A l l Five Damage Classes VII Classification Functions Giving Maximum Separation of Successfully Attacked Trees (66.7% correctly classified) when a l l 5 Damage Classes Used i n the Analysis 46 IX Classification Functions Giving Maximum Separation of Successfully Attacked Trees (66.7% correctly classified) when only the ^ Three Green-appearing Classes were used i n the Analysis LIST OF FIGURES FIGURE PAGE 1 Generalized Life Cycle of the Douglas-fir 4 Beetle (Dendroctonus pseudotsugae Hopk.) 2 Portion of the Electromagnetic Spectrum Currently 15 Used i n Attempts at Previsual Detection 3 Schematic X-Section of Colour Infra-red Film 18 4 Reflectance Curves of Various Stages of Insect- 21 attacked Trees and the Resultant Effect on Colour Infra-red Film 5 Topographic Map of the Tranquille River Area 23 with the Two Photo Plots Marked 6 Means and 95% Confidence Intervals of the 30 Individual Dye Layers for the Five Damage Classes 7 Means and 95% Confidence Intervals of the 31 Ratio Variables for the Five Damage Classes 8 Stereo Aerial CIR Photos of Green, Red-topped 33 and Old-dead trees 9 Stereo Aerial CIR Photos of Healthy and 34 Successfully Attacked Trees v i ACKNOWLEDGEMENT I would l i k e to acknowledge the guidance and constructive criticism received from my thesis committee: Dr. J.A. McLean, Dr. P.A. Murtha and Dr. J. Worrall. I also thank Dr. L.H. McMullen of the Canadian Forestry Service (Pacific Forest Research Centre) for showing me and allowing me to use some of his Douglas-fir beetle research plots. Figures were drafted by N. Holm. 1 1. INTRODUCTION 1.1 General Impact of Bark Beetles Bark beetles of the genus Dendroctonus Erichson are extremely important pests. The significance of bark beetle damage i n British Columbia was noted as early as 1913 by J.M. Swaine who carried out his survey i n response to requests made by forestry personnel i n the province (Swaine 1914). These insects, as a group, cause widespread mortality to mature and overmature stands of pine (Pinus L. sp.), spruce (Picea A. Dietr sp.) and Douglas-fir (Pseudotsuga menziesii (Mirb) Franco). Mountain pine beetle (Dendroctonus  ponderosae Hopk.) normally attacks standing, l i v i n g trees and so i s a chronic pest of stands of mature lodgepole and ponderosa pine (Pinus contorta var. l a t i f o l i a Engelm. and P. ponderosa Laws) (Safranyik et a l . 1974). Spruce beetle (D. rufipennis (Kirby)) and Douglas-fir beetle (D. pseudotsugae Hopk.) normally breed i n windthrow and logging slash (Dyer and Taylor 1971. McMullen 1977)' However, when insect population densities are high and tree resistance i s low, these insects may attack and k i l l mature standing trees. Douglas-fir beetle commonly attacks small groups of l i v i n g , although weakened trees (Walters 1956). Of these three insect species, the mountain pine beetle and the Douglas-f i r beetle are chronic pests of British Columbia forests. Each year they i n f l i c t substantial losses on the timber resource (Table i ) . By contrast, spruce beetles rarely k i l l l i v e trees; however, when an epidemic occurs, the results may be catastrophic. The impact of these insects i s great as during epidemics the larger, more commercially valuable trees are k i l l e d f i r s t . Table I. Volume Losses (irr) Caused by Bark Beetles (Dendroctonus sp.) i n B.C. Forest Regions 1971-75 (Cottrell, et a l . 1979) Region Mountain Pine Beetle* Douglas-fir Beetle Spruce Beetle Vancouver 19,647 807 Kamloops 49,282 2,754 175,346 Nelson 98,983 1,152 -Cariboo 90,747 32,477 -Prince George 985 191 -Prince Rupert 52,216 -Total 311,860 37,381 175,346 * includes attack on Lodgepole Pine only 3 The control of these insects, or the minimization of losses caused by these insects, i s dependent on early detection. Promp detection of infested trees allows ample time for management regimes to be implemented, such as logging infested areas or treating single trees to reduce the population of insects i n a valuable stand. 1.2 Douglas-fir Beetle Life History The generalized l i f e cycle of the Douglas-fir beetle i s shown i n Figure 1 (McMullen 197?)• Mature adults emerge from the overwintering sites i n the spring when conditions are conducive to beetle f l i g h t . Under shaded conditions and mid-range relative humidity spontaneous f l i g h t w i l l not occur below 19 to 20°C. However, when exposed to f u l l sunlight, the beetles w i l l f l y at 17°C (Atkins and McMullen I960). Because of this, f l i g h t w i l l occur earliest i n slash areas, i n stand margins and i n relatively open stands. Attacking insects prefer freshly f e l l e d or damaged host material over aged material and also prefer horizontal material over vertical material (McMullen and Atkins 1962). Flying insects select suitable host material by the detection of host volatiles including the oleoresin components alpha-pinene, camphene and limonene (Atkins and McMullen 1958; McMullen and Atkins 1962; Rudinsky 1966; Pitman and Vite 1970). Degradation products such as ethyl alcohol have also been shown to be attractive to flyi n g beetles (Pitman et a l . 1975)* Stressed or damaged standing trees may exude larger quantities of these attractive volatiles, thus explaining why these trees are attacked by f l y i n g beetles while nearby healthy trees are ignored. 4 OVER WINTER MAY JUNE JULY LARVAE \ PUPAE AUG. SEPT. OCT. DEVELOPING ADULTS !nd FLIGHT & REATTACK REPRODUCTIVE ADULTS EGGS LAP VAE PUPAE Figure 1: Generalized L i f e Cycle of the Douglas-fir Beetle (Dendroctonus pseudotsugae Hopk.). 5 Mass attack of the host i s typical of Dendroctonus behaviour (Hopping 1929; Walters 1956; McMullen 1977). The female beetle i n i t i a t e s the egg gallery and produces pheromones for the aggregation of more beetles (Pitman and Vite 1970; Pitman et a l . 1975; Dyer and Lawko 1978). The attacking beetles space themselves over the bark surface area by means of sonic stimuli and repellent pheromones (Hedden and Gara 1976; Rudinsky et a l . 1976; Ryker et a l . 1979'). The greatest number of attacks and the greatest proportion of successful attacks on standing trees have been found to be on the upper bole of the tree, half way up the infested height and well above breast height (Furniss 1962). Because of this, estimation of successful attack based on samples taken from breast height may be unreliable as the absence of successful attacks would not necessarily preclude the presence of successful attacks high up the bole. However, evidence of successful attack at breast height indicates a successfully colonized tree. After successful attack, adult beetles begin construction of the egg gallery at the cambium. The gallery i s unbranched and extends vertically. The gallery i s approximately 5mm wide and on the average i s about 27 cm long. Eggs are l a i d i n individual niches excavated i n the side of the gallery. As the eggs hatch, the larvae begin feeding i n the cambial region, producing larvae1 galleries going at right angles to the parent egg gallery, thus girdling the phloem and blocking the translocation of photosynthates down the tree. The larvae pass through several instars and are generally i n the l a s t instar or pupal stage i n the f a l l following the attack. They overwinter i n these stages and complete their development to become adults 6 capable of f l i g h t and attack the next spring. Throughout most of i t s range, the Douglas-fir beetle has a one-year cycle. Associated with the Douglas-fir beetle i s a symbiotic fungus. I t i s introduced into the tree by the attacking insects and colonizes the sapwood. Rapid growth of the fungal mycelia blocks resin canals, thereby protecting the burrowing beetle from pitch flow. Pitch flow i s the main defence mechanism of the tree and a healthy tree can pitch out the attacking beetle (Rudinsky 1966). This pitching-out process leaves pitch pockets i n the wood which can be seen years after the attack (Belluschi et a l . 1965)* Tree water status affects the resin pressure. In trees under water stress, the resin pressure i s reduced and the tree becomes more susceptible to successful insect attack (Rudinsky 1966; Lorio and Hodges 1977). The fungus colonizes the sapwood and may completely block the translocation of water and nutrients up the tree, causing the death of the tree and a f o l i a r colour change. The rate of foliage colour change i s dependent upon weather conditions. The colour progression i s from healthy green to pale green to yellow to red to bare branches. The f i r s t easily visible colour change (from green to yellow) usually occurs i n the spring of the year following the attack. In hot summers, however, the foliage may change colour by the October following attack (Belluschi and Johnson 1969). Detection and appraisal surveys for the Douglas-fir beetle have been dependent upon this colour change. Aerial surveys are usually conducted i n August and numbers of red-topped trees are counted. However, because of the time lag between attack and colour change, the mapping of red-tops represents the mapping of the previous years attack, not the current attack. The time available for treatment and management of the population prior to 7 insect f l i g h t i s reduced. However, comparison of the number of red-tops to the number of older attacks gives a measure of the rate of spread of the infestation. I t 3 Management Considerations Management regimes to reduce the hazards of Douglas-fir beetle attack include single tree treatments, trap tree and trap log programs, logging i n high risk stands and salvage logging of currently infested stands (Hopping 1921; Walters 1956; Lejeune et a l . 1961; McMullen 1977). Single tree treatments are concerned with k i l l i n g the brood i n infested material prior to the insect f l i g h t and attack period. Such treatments include p i l i n g and burning infested material, peeling the bark to expose the insects to dessication and spraying infested material with bark pene-trating insecticides which include ethylene dibromide and lindane (the gamma^isomer of benzene hexachloride) i n fuel o i l . These treatments are expensive and, to be successful, infested material must be located and treated before the insects f l y to attack new hostst The benefits derived from such treatments and the costs involved must be considered before such a program i s carried out. Costs may be amortized over a period of years i f a high-value, high risk stand can be saved u n t i l harvest under long-term logging plans. Trap tree and trap log programs are instituted prior to insect f l i g h t . They are designed to absorb the flyi n g population and thereby decrease the numbers of beetles available to infest desirable timber. In trap log programs, host trees are f e l l e d within the control area. These f e l l e d trees attract and absorb the flying insects. After f l i g h t , these logs are treated as above for single tree treatments or removed for processing. An adequate survey of the stand i s necessary to establish both the amount of infested 8 material (wind-throw and standing trees) and i t s location. After the survey, the number of trees to serve as trap logs sufficient to absorb the attacking insects must be selected and felled. These logs must also be located near the brood producing material to prevent dispersal of the insects. Trees between the trap logs and brood material may become attacked thereby reducing the effectiveness of the trap tree program. Trap logs may be baited with synthetic pheromones to increase their attractiveness and effectiveness and they may be sprayed with a contact insecticide which would k i l l arriving beetles. Whatever the method, insects at the trap log must be k i l l e d to decrease the population and reduce the risk of attack to the remaining stand. Standing trap trees may be used similarly to trap logs. Their effectiveness i s mostly dependent upon the use of an attractant pheromone. Trap trees must be treated after f l i g h t i n the same manner as trap logs unless they had previously been protected with insecticide. An effective trap program i s dependent upon the locating of brood material far enough in advance of insect f l i g h t to establish the trap trees i n optimum numbers and location. Detection of brood material prior to f o l i a r colour change would give greater time available for treatment. Harris et a l . (1978) have proposed a method for detecting recent wind-throw, the other brood source besides standing trees, by detecting gaps in the forest canopy on aerial photographs. Logging of high risk stands i s equivalent to direct competition with the insects for the timber resource. Risk rating systems have been developed for mountain pine beetle (Safranyik et a l . 1974; Amman et a l . 1977; Mahoney 1978) and spruce beetle (Schmid and Frye 1976). Similar risk rating systems 9 could be developed for the Douglas-fir beetle. Once stands have been classified as high, medium or low risk, logging p r i o r i t i e s may be altered so that high risk, high value stands may be u t i l i z e d before they become significantly infested by bark beetles. This method of bark beetle management does not require detection of infested trees to the same extent as the above two methods. It does, however, require information about the biology of the insect and the dynamics of the stands. Harvesting by risk i s a long-term method of insect control necessitating great changes i n current long-range harvesting plans. This type of management may also disrupt other c r i t e r i a for setting harvesting p r i o r i t i e s so that insect damage cannot be the sole reason for harvesting plans. Salvage logging of already 1 infested stands i s a "catch-up" type of management. Its aim i s to u t i l i z e insect k i l l e d timber before significant degradation of wood qualities can occur. When large areas of timber have been k i l l e d , as happens with mountain pine beetle infestations, salvage logging produces large clearcuts without greatly reducing the population of insects. Salvage logging should be carried out i n such a way as to remove insect populations as well as the most recently k i l l e d trees. This requires starting salvage operations at the perimeter of an infestation rather than in the center. Preservation of desirable stands for long periods of time requires the elimination of small, incipient infestations within the stand. Management of stands for control of Douglas-fir beetle must include aspects of a l l of the above methods. Early detection i s a requirement for such treatments. 1.4 Survey and Detection Methods to Date Detection surveys for bark beetle activity have been carried out for several years by the Forest Insect and Disease Survey of the Canadian Forestry 10 Service (McGugan 1956). Detection surveys consist of observations and reports of insect infestations. Their purpose i s to reveal the presence of harmful or potentially harmful infestations of forest insects prior to their development into outbreak proportions, at which time the treatment of the infestation i s d i f f i c u l t (Orr 195*0 • Present Douglas-fir beetle survey methods include aerial sketch-mapping, aerial photography with colour or colour infra-red (CIR) film and some ground checking (Harris and Dawson 1979). In British Columbia these surveys are usually carried out i n the f a l l and numbers of red-tops are counted as they are seen from the a i r . These red-tops represent trees that have been attacked the previous year and which have no brood remaining i n them. Currently infested trees which are s t i l l green are not detected (Belluschi and Johnson 1969; Harris and Dawson 1979). To determine the amount of current infestation, ground surveys are carried out i n areas where red-tops were detected. These ground surveys are used to calculate the ratio of new infested trees to old infested trees. Obviously, this survey for green attacked trees i s expensive. The cost ratio of ground strip cruising to photographic survey i s approximately 100:1 (Heller et a l . 1959; Wear et a l . 1964; Meyer and French 1967; Wert and Roettgering 1968; Klein 1973). Also, only the number of green attacked trees i n known infestations i s assessed. Isolated infested trees or small incipient infestations are not detected. These present surveys, i f conducted annually, yield an historic record of the progress of insect infestations. Also, they indicate areas which have a high likelihood of having current infested trees, based on the presence of an attacking population source. Surveys would be made more effective i f green attacked trees could be detected from the air. Ground surveys would not need to be as extensive 11 and survey f l i g h t s could be carried out ea r l i e r i n the season, allowing a greater amount of time for treatments or formulation of management plans. 1.5 Early Detection - Methods and Success to Date Three general methods have been used i n attempting previsual detection: analysis of thermal imagery, analysis of multispectral scanning imagery and analysis of CIR photographs. Early detection of a f o l i a r colour change which i s not detectable through visual means may be termed either extravisual or previsual (Murtha ;I9?8). I t i s extravisual i f a visual change does not occur eventually. This type of damage may be caused by a temporary stress or a general weakening of the tree which does not lead to the death of the tree. The detection i s previsual i f a further colour change does occur. In the case of detecting green bark beetle infested trees, the trees w i l l eventually die and change colour so that the correct term i s previsual. Previsual detection i s based upon the detection of physiologically changed foliage resulting from stress. Stress occurs i n beetle attacked trees when the blue-stain fungus has successfully blocked the sapwood, thereby reducing water transport. The results of this stress are the changes i n the reflective characteristics of the foliage. Gausman (1977)» studying various herbacious plants, determined that near infra-red (NIR) ligh t (700 - 900nm) was reflected by intra* 1 cellular discontinuities including nucleii, crystals and cytoplasm i n addition to the c e l l wall airspace inter-face which accounts for the major portion of the NIR reflectance. Cellular constituents account for approximately 8% of the reflectance at 800nm (Gausman 1977). These components may change i n stressed foliage and produce an increase or a decrease i n the amount of reflected NIR (Knipling 1967). 12 NIR reflectance i n conifers i s generally less than that for broad leaved species, possibly due to their more compact structure and more contained shadows (Kodak 1 9 7 1 ) . It has been postulated that the f i r s t change i n re-flectance of a plant under stress i s a change i n the amount of reflected NIR (Fritz 1967; Murtha 1 9 7 8 ) . Thomas et a l . (1966) showed -that i n cotton, reflectance of NIR increased with water stress possibly due to; an increase i n solute concentration within the c e l l cytoplasm and to changes i n the size and shape of c e l l s and intra-cellular spaces. Olson et a l . (197°) also reported an increase i n NIR reflectance with greater moisture stress (low f o l i a r moisture content) i n yellow birch (Betula alleghaniensis Britton), sugar maple (Acer saccharum Marsh) and white ash (Fraxinus americana L.), a l l of which are broad leaved species. This increase i n reflectance was not found i n beetle attacked conifers. In these conifers, the NIR reflectance was found to decrease by about 10$ (Weber 1965? Heller 1968; Weber and Polcyn 1 9 7 2 ) . These decreases, however, were not visually detectable on CIR film. The decrease of the NIR reflectance found by Weber (1965) and Heller (1968) was found to occur approximately 45 days after i n i t i a l attack. Heller (1968) also found no discernible changes i n needle structure 2 months after attack. Ten months after attack, after the foliage changed colour, significant changes i n morphology occurred (Table II). These changes i n morphology probably reflectance. Preliminary changes occurring within 2 months of attack might account for the 10$ decrease i n NIR reflectance that was found. Other authors (Benson and Sims 1967; Ciesla et a l . 1967? Meyer and French 1967; Ciesla 1977) have also stated that no visual difference between green infested and green healthy trees could be found on colour infra-red photographs although they did state that the colour contrast between old 13 Table I I . Structures of Pine Needles (Pinus ponderosa Laws.) Before and 10 months after Mountain Pine Beetle Attack. (Heller, 1968) Structure Affected Normal Needle Yellow Needle From Infested Pine Resin Canal Open Collapsed or Broken Vascular Bundles Most Cells F i l l e d with Cytoplasm Cytoplasm absent Stomata Intact Broken, Shrunken, Degenerated or Closed Cytoplasm F i l l s Out to Cell Walls Shrunken from Cell Walls, Frequently Absent Cell Walls Normally Thin Thicker by Comparison 14 infested trees and green trees and between deciduous and coniferous trees was greater with colour infra-red film. The haze penetration with infra-red film was also much better. Amberg et a l . (1973)» working with aerial photography of damage caused by Ips typographus (L.), a European bark beetle, found visual differences i n green infested trees. They postulated that bark beetle damage to trees i s manifested by a change i n physiology caused by nutrient deficiency which affects the reflectance properties of the foliage. Previsual detection using thermal imagery makes use of a thermal line scanner which detects emitted radiation greater than about 2600nm (Fig. 2). This energy band can be detected using a multispectral scanner. Results using these devices have been pa r t i a l l y successful. Heller (1968) detected some green infested trees i n the spectral range of 1000 to 2600nm (mid infra-red) but the limit of discernible shades of gray often restricted detection of subtle changes i n moisture stress. Heller (1968) also noted a 6°C increase i n needle temperature i n infested trees. This result agreed with the results of Weber (1965) i n attacked pines and i n Engelmann spruce (Picea engelmannii Parry) infested with spruce beetle (Schmid, 1976). Schmid's results were variable, however, and needle temperatures varied according to wind, shading, needle location (east or west aspect) and orientation to the sun. These factors obscured a consistent difference between healthy and infested trees. A decrease i n reflected NIR and an increase i n f o l i a r temperature i n infested trees form the basis for attempts at previsual detection of beetle infested trees. Later, Heller (1971) detected several green infested pines using a multispectral scanner, but there were also many other objects i n the imagery \ ULTRA VIOLET- -VISIBLE-*, H — h <NEAR*r" 4 ui z 6 o D LU LU J 111 CC m cc o 8 INFRA-RED MID H 1 •THERMAL 10 12 14 16 18 20 22 24 + 26 28 3Q (x100)nm Figure 2: Portion of the Electromagnetic Spectrum Currently Used i n Attempts at Previsual Detection. 16 that had the same temperature range as the target trees. The resolution of such scanners was also not fine enough to distinguish individual trees when the aircraft was flying at over 1500m (Heller, 1971; Weber and Polcyn, 1972). Weber and Polcyn (1972) also had some success with detecting non-faded attacked trees using far infra-red spectral ranges (greater than 900nm). They had d i f f i c u l t i e s with the resolution of the imagery and stated further that greater success could be achieved by covering the entire band-width i n narrow band increments. Other authors (Rohde 1971; Rohde and Olson 1970; Alger et a l . 1978) have also had success using sensors sensitive to this spectral region. The main problems associated with thermal line scanners and multispectral scanners have been poor resolution and poor tree species identification. Analysis of photographs avoids these problems. Most analysis of normal colour and CIR photographs to date has been purely visual interpretation. Most attempts have f a i l e d , with the authors stating that no colour change occurs on infra-red film before i t shows up i n normal colour film (Benson and Sims 1967; Ciesla et a l . 1967; Heller 1968 and 1971; Heller and Wear 1969; Brown 1971; Ciesla 1977). Only Arnberg et a l . (1973) reported success i n visually detecting infested trees on CIR film. More sophisticated analysis of CIR photographs has led to the detection of stress i n plants. Murtha and Hamilton (1969) detected p a r t i a l l y and f u l l y girdled trees prior to f o l i a r colour change. The photographs were analyzed using a microdensitometer to measure the densities of the dye layers i n the film. They were able to show a decrease i n the amount of NIR reflectance i n the damaged trees as indicated by dye-layer density changes. This was also shown by Murtha (1968). Analysis of normal colour film 17 by the densitometric technique did not yield similax results. Microdensitometric analysis was used by Lillesand et a l . (1978) to detect elm trees (Ulmus  americana L.) suffering from Dutch Elm disease. Using discriminant analysis, stressed trees were found to be significantly different from healthy trees. Thus, through the use of densitometric analysis, the 10% decrease i n NIR reflectance reported by Weber (1965) and Heller (1968) could possibly be detected on CIR film. 1.6 Densitometry Microdensitometric analysis of colour infra-red film to detect previsual symptoms of bark beetle attack involves the measurement of the densities of the three dye layers i n the film. Colour infra-red film i s made up of three dye layers, each sensitive to a different region of the visible and near infra-red spectrum (Fritz 1967). These dyes are: yellow dye forming, magenta dye forming and cyan dye forming layers (Fig. 3)« These layers are sensitive to green l i g h t (500 - 600nm), red lig h t (600 - 700nm) and near infra-red li g h t (700 - 9n0nm) respectively. Because a l l three dye layers are sensitive to blue ligh t (400 - 500nm), a Wratten 12 f i l t e r must be used when exposing the film. The Wratten 12 f i l t e r i s a yellow or minus-blue f i l t e r which absorbs blue li g h t . The amount of dye formed i n each layer i s inversely proportional to the amount of incident light i n a specific region of the spectrum (Fritz 1967). Upon development to a positive transparency, the amount of the dye-forming layer not exposed forms a dye i n inverse proportion to the amount of incident light i n the particular range. The density (the amount) of dye formed i s therefore a relative measure of the reflected light i n each of the three spectral regions. These dye layer densities can be measured by using a densitometer SPECTRAL SENSITIVITY GREEN (500-600 nm) RED (600- 700 nm) NIR (700-900 nm) DYE-FORMING LAYER YELLOW MAGENTA CYAN ANTI-HALATION BASE BACKING Figure 3: Schematic X-Section of Colour Infrared Film. 19 which measures the lig h t transmittance at specific spectral regions through a positive transparency. Table III shows the f i l t e r s required to measure the dye layers. On the transparency, a high density of a dye layer indicates a low reflectance i n that specific spectral region. This i s shown schematically i n Figure k (after Murtha, 1978). The diagram shows both the progression of the theoretical reflectance changes i n stressed trees and the resultant effect of the reflectances on the film. 1.7 Study Objectives The objective of this study was to determine i f the change i n NIR reflectance previously detected by Weber (19&5) a n d Heller (1968) relative to bark beetle attack could be detected on CIR photographs before a visual colour change was noticed. A change i n the reflectance alters the amount of dye formed on the transparency. Although no success i n detecting green infested trees has been reported, previous workers relied principally on visual interpretation of CIR photographs. The human eye detects density differences of about 0.1 and notes them as tonal changes. Densitometrie analysis should be able to detect the differences which would escape visual analysis as differences i n density as small as 0.01 are detectable through the use of a microdensitometer. Original transparencies must be used for the densitometrie analysis as there i s a reduction i n the amount of information on duplicates. Densitometric analysis w i l l also quantify differences i n dye layer densities allowing s t a t i s t i c a l analysis of reflect-ance changes i n stressed trees. Successful detection using photographs w i l l eliminate some of the problems inherent with thermal line-scanning as well as reduce the costs as the equipment and overall costs of photography are less expensive than 20 Table III. F i l t e r s Used to Measure Specific Dye Layers on Colour Infra-red Film on a Densitometer. Dye Layer To Be Colour of Spectral Reflectance Measured F i l t e r Region being Measured Yellow Blue Green (500 - 600hm) Magenta Green Red (600 - 700nm) Cyan Red near infra-red (700 -900nm) A.SPECTRAL REFLECTANCE PATTERNS 1. HEALTHY GREEN 2. R E D - T O P P E D 80 r 60 40 20 3. DEFOLIATED BLUE GREEN RED NIR BLUE GREEN RED NIR BLUE GREEN RED NIR SPECTRAL REGION B. RELATIVE AMOUNTS OF DYE FORMED AFTER DEVELOPMENT COLOUR ABSORBED BLUE G R E E N RED DYE FORMED VZZZZZZZZZZZZZZZZZZA M7ZZZZZZZZZZZZZZZZZ. WM w/m Y E L L O W WM///, WM/ MAGENTA 'WM W/A CYAN C. RESULTANT COLOUR ON FALSE-COLOUR IMAGE MAGENTA ( B L U E & RED) Y E L L O W ( G R E E N & RED) DARK Figure 4: Reflectance Curves of Various Stages of Insect-attacked Trees and the Resultant Effect on Colour Infrared Film. 22 for thermal imagery. The technology i s also more commonly available. Incorporation of this technique into bark beetle damage surveys would reduce the amount of ground cruising required to estimate the spread of known infestations and thereby reduce the costs involved. High risk, high value stands may also be monitored for the occurrence of small incipient infestations. This would allow more time for treatment of the infested trees leading to the extension of the l i f e of the stand. 2. METHODS 2.1 Study Region The Tranquille Valley i n the Kamloops Forest Region of B r i t i s h Columbia was chosen as the general study area f o r two reasons. F i r s t l y , the Douglas-f i r beetle has been a chronic problem i n this region for several years and, secondly, several other remote sensing studies were being carried out i n this region by other researchers from the University of British Columbia. This led to a more eff i c i e n t u t i l i z a t i o n of the expensive technology used to obtain the CIR photographs. Selection of the actual photo plots was made after ground reconnaissance of the area. The f i n a l locations of the study sites are shown i n Figure 5« 2.2 Selection and Location of Photo Plots The c r i t e r i a for selection of photo plots were as follows: plots must be located i n pure, even-aged stands of mature Douglas-fir; trees infested i n 1978 must be i n the plot or i n close proximity to the plot to provide a source of attacking beetles f o r 1979; there must be some evidence of bark beetle attack (e.g. boring dust on the bole) on some trees with green foliage i n the plot; there must be comparable trees with no symptoms of attack to act as checks. Two such plots were located along the Tranquille River (Fig. 5). The current attacks on trees i n these plots may have been a r t i f i c a l l y induced, 23 Figure 5. Topographic Map of the Tranquille River Area with the Two Photo Plots Marked. 24 however. Dr. L.H. McMullen had halted standing trees with a synthetic attractant Douglas-fir "beetle pheromone prior to beetle f l i g h t which occurred i n the l a s t week of Apri l , 1979. Therefore, some of the attacked trees may not have been attacked under normal circumstances. These areas, though, were judged to have a high probability of containing some successfully attacked trees. Plot markers were put i n position on July 26. Each marker consisted of a white cotton sheet 260 x 165cm. Markers were placed i n an open area at the ends of each plot. White sheets were chosen since they are highly reflective i n a l l wave lengths and would therefore be readily identifiable on aerial photographs. A l l trees i n the plots were ground checked at this time. Five trees were girdled i n plot 2 (Fig. 5) on July 26. These trees were to simulate attacked trees for comparison of their spectral reflectance patterns with healthy trees and with beetle-attacked trees. The would provide further information on reflectance changes following phloem disjunction. The depth of the girdling extended into but did not completely sever the sapwood of each tree. 2.3 Photography Aerial photography of the plots was carried out on July 29, approximately three months after beetle attack. I t was hoped that this time interval would allow sufficient beetle development and reflectance changes to indicate successfully attacked trees on CIR photographs prior to a visual colour change. The photographic f l i g h t was done by Integrated Resource Photography of Vancouver, B.C. Stereo photographs were taken between 18:38 and 18:51 1 Dr. L.H. McMullen. Research Scientist, Canadian Forestry Service, Pacific Forest Research Centre, Victoria, B.C. Personal communication. 25 Greenwich Mean Time or between 11:38 and 11:51 Pacific Daylight Savings Time. Pictures were taken with a Vinten 492 70mm reconnaissance camera with a Leitz f2.0 lens set at f?-5 using a Wratten 12 and a GG20M f i l t e r and Kodak Aerochrome IR film. The resultant false-colour transparencies were at a scale of approximately 1:1100. Forty-four stereo pairs provided photo coverage of the two plots. 2.4 Analysis of Photographs 2 A Macbeth TR-524 Transmission Reflectance Densitometer was used to take density readings on the crowns of trees i n the photographs. The transmission head was used and densities of a l l three dye layers were measured by using the f i l t e r selection of either blue, green or red f i l t e r s to give density measures of the yellow, magenta and cyan dye layers respectively. Only the dominant Douglas-fir were measured i n this study as only mature trees were rated as being suitable for beetle attack. Densities were taken from the images of trees with green foliage, red-tops (attacked and k i l l e d in 1978), and dead defoliated trees (attacked and k i l l e d prior to 1978). Murtha's (1972) damage classification index was also applied i n the visual interpretation of the photographs. 2.4.1 Analysis of Densities A f i n a l ground check was made between January 24 and 27, I98O to establish which trees i n the plots were successfully attacked, attacked but pitched out or unattacked and green. Such trees were identified and matched to their photo images. Successfully attacked trees were those which had evidence of beetle attack and straw-coloured foliage; trees which had pitched-out the 2 Division of Kollmorgen Corporation 26 attacking insects were those with beetle entrance holes on the bole and green foliage; unattacked and healthy trees were those with no evidence of beetle attack on the bole and green foliage. This ground check was necessary to identify trees positively i n each damage class for further s t a t i s t i c a l analysis to determine i f significant differences existed. The variables derived for each tree were the density readings for each of the three dye layers (yellow, magenta and cyan) and three ratios from these readings (magenta/yellow, cyan/magenta and cyan/yellow). The ratios were used because they gave a measure of the relative reflectance between each of the spectral regions. The use of ratios, therefore, should decrease any variance due to film/plot irregularities such as differences i n aspect, slight differences i n scale due to changes i n altitude or slight irregularities i n the film i t s e l f . Histograms and descriptive s t a t i s t i c s (mean, sample size, range and standard error of the mean) were computed for each variable i n a l l the damage classes to check normality of the data and to identify oulier values. One-way analysis of variance (ANOVA) for each variable between the three green-appearing damage classes was carried out to detect significant d i f f e r -ences. Only the green-appearing damage classes were analyzed by this method as the other two classes (red-topped and old dead) were easily distinguished visually. MIDAS (Fox and Guere 1976), a s t a t i s t i c a l computer package, was used for the calculation of ANOVA, histograms and descriptive s t a t i s t i c s . If ANOVA showed a significant difference between the three damage classes for a variable, an a posteriori least significant range tests of means was carried out using the Student-Newman-Keuls test for unequal sample sizes (Sokal and Rohlf 1969). 2? Stepwise discriminant analysis was used to obtain a weighted linear combination of the independent variables that would best separate trees into their proper damage classes. The stepwise discriminant analysis package BMD P?M (Jennrich and Sampson 1977) was used for the analysis. A l l six variables were included so that the best combination could be found although the ratio variables were given 1 p r i o r i t y f o r selection due to their lower va r i a b i l i t y and representation of how one spectral region varied with another. Two sets of discriminant analysis were done, one attempted to discriminate between a l l five damage classes while the other considered only the three green-foliaged classes. 3. RESULTS AND DISCUSSION 3.1 Data Description Densities of the three dye layers (yellow, magenta and cyan) were measured for trees with green foliage, red-tops (attacked and k i l l e d i n 1978) and dead defoliated branches (attacked and k i l l e d prior to 1978). In a l l , the three dye layer densities were measured for each of 636 trees. A ground check made three days prior to the photography showed that both trees with no evidence of attack and trees with fresh boring dust on the bole had green, healthy-appearing foliage. A f i n a l ground check was made i n January, 1980 to determine which trees i n the plots were successfully attacked, attacked but pitched-out, or unattacked and healthy. By this time the foliage of successfully attacked trees had turned a straw colour that was readily apparent from the ground. Unsuccessfully attacked trees were identified as those trees having green foliage and boring dust on the bole. Densitometric measurements of trees i n the same damage class did not vary significantly (p<0.05) between plots so that the data were pooled i n the analysis. A random sample of the healthy trees was selected for comparison with the 28 damaged trees. This was approximately equal i n number to the sizes of the other damage classes. A l l successfully attacked, pitched-out, red-topped and old trees i n the plots were included i n the analysis (Table IV). The variables used for the analysis were the densities of the three dye layers and the three ratios of the densities (magenta/yellow, cyan/ magenta and cyan/yellow). The f i r s t three variables are representative of the reflectance of light i n certain wave lengths (Fig. "}). A high density value indicates a low reflectance and vice versa. The last three variables are more indicative of reflectance patterns. The ratios also tend to overcome film and localized reflectance irregularities f a c i l i t a t i n g comparison between different locations and aspects. They may also be more sensitive to differences i n reflectance patterns. Descriptive s t a t i s t i c s including sample size, range, mean and standard error of the mean were derived for each variable i n each damage class and are detailed i n Appendix I and recorded graphically i n Figures 6 and ?. Histograms of each variable i n each damage class are shown i n Appendix II. The reduction of variance attributable to the ratios as well as overlapping of values i s shown i n Figures 6 and ?. The confidence intervals for the ratios are much narrower than for the individual dye layers. Differences between damage classes are also indicated i n Figures 6 and 7 as the more the intervals overlap, the less i s the difference between means. Larger sample sizes would tend to emphasize any differences which exist as the confidence interval decreases with increasing sample sizes. Generally, two distinct groups of observations were obtained (Fig. 6 , 7). One grouping represented the green appearing trees (healthy, pitched-out Table IV. Number of trees used for each damage class i n the analysis of dye layer densities. Damage class Number of trees used Healthy 27 1979 Pitch out 26 1979 Successful attack 15 Red-tops 31 Old dead 27 Total 126 2 o • OLD DEAD o 1978 ATTACK A 1979 SUCCESSFUL ATTACK A 1979 PITCH OUT • HEALTHY (UNATTACKED) Uj < >• t < z - I LU Ul < -O 1 I A 1 o Ul >• J L I A 1 J L .4 .8 DENSITY D = l o 9 % Transm 1.0 1.1 1.2 ission Figure 6: Means and 95% Confidence Intervals of the Individual Dye Laye for the Five Damage Classes. o 2 o o LU > -•—i I—A—I 2 z K- < < >* z LU o < 2 i o 1 h -» - i I—•—I z o < 5 O LU • OLD DEAD o 1978 ATTACK A 1979 SUCCESSFUL ATTACK A 1979 PITCH OUT • HEALTHY (UNATTACKED) h-O—I _l_ _1_ A 1 HA—I I • I J I L .4 .6 .8 1.0 1.1 1.2 R A T I O V A L U E Figure 7: Means and 95% Confidence Intervals of the Ratio Variables for the Five Damage Classes. u> 32 and successfully attacked) and the other group the red-topped and old-dead trees. These two groups are also easily distinguishable from each other visually both from the ground and on the colour infra-red photographs (Figs. 8, 9)- The three green-appearing classes were not greatly different from each other when visually compared on a photograph except for the slight orange hue of the 1979 successfully attacked trees (Fig. 9) ' Because the trees k i l l e d prior to 1979 were visually distinct and because the purpose of this experiment was to distinguish healthy trees from successfully attacked trees, the remainder of the discussion w i l l deal primarily with the three damage classes which had. green foliage. A damage classification system (Murtha 1972) was applied to the trees i n the photographs. In this system trees may be assigned a damage classi-fication index on the basis of morphological or physiological deviations from normal healthy trees. Morphological changes are those which are apparent as changes i n shape or form of the trees such as loss of foliage or broken branches. Physiological deviations from healthy trees are expressed as changes i n the colour or tone of the foliage of affected trees. Classification of damage may be done with either normal colour or CIR film. Morphological damage i s illustrated by old-dead trees (Fig. 8). These trees are given a damage classification of IB, I.e. scattered individuals or small groups of defoliated conifers. Trees that had been dead for one year, or red-tops, are given the classification IIIG, or a conifer with red-brown current and red-brown olderifoliage. Because a l l years of foliage are k i l l e d at the same time, this damage type indicates a tree k i l l e d rapidly i n the present season. The pitched-out trees (Fig. 9) appear as a normal, healthy magenta colour and they would not be included i n a damage appraisal based on visual photo-Figure 8: Stereo a e r i a l CIR photos of green,red-topped and o l d -dead trees. A. Healthy Douglas-fir (magenta tone); B. 1-yr. old dead Douglas-fir with red-brown fo l i a g e (yellow tone); C. Old dead, defoliated tree (grey tone). 34 Figure 9: Stereo a e r i a l CIR photos of healthy and s u c c e s s f u l l y attacked trees. A. Healthy Douglas-fir (magenta tone); B. Beetle infested Douglas-fir with green f o l i a g e (tan-pink tone). 35 interpretation methods. The successfully attacked trees shown i n Figure 9 exhibit a marked colour change indicating a stress situation. These trees appear as intermediate between the healthy magenta and the yellow colouration of the 1978 attacked trees. As such, they were given the classification of 1110b or lighter than normal magenta tone. This, however, i s not completely descriptive and a further class may have to be added to this classification system to more accurately describe this type of tree. Also, some bare branches are evident i n the crown of the successfully attacked tree which i s a morphological change so that the classification under the present system, should be IIIOb/lIA. The IIA describes defoliation concentrated i n the top part of a conifer. Damage classification systems for aerial photographs are useful as they permit quantification of differing types of damage occurring i n close proximity. Also, some damage classes are indicative of specific causal agents or of a limited group of agents, thereby supplying more information than just a notation that some damage i s occurring. The three green appearing classes were not distinguishable from each other when viewed from the ground four days prior to the photographic mission. They are also not greatly different when seen on the colour infra-red photographs (Fig. 9). Notice, however, the slightly orange colour of the successfully attacked trees (Fig. 9). This colouration was consistent i n a l l successfully attacked trees while the other two classes showed the magenta colour of healthy green foliage. This difference i n colour may be explained by comparing the densities of individual dye layers and by comparing their ratio values. 36 3.2 S t a t i s t i c a l Analyses 3.2.1 One-Way Analysis of Variance As red-topped trees and old-dead trees axe readily distinguishable from trees with green foliage on the ground and on CIR photographs, only the success-f u l l y attacked, pitched-out and healthy classes were compared by analyses of variance. Each of the six variables (three dye layer densities and three ratios) was tested separately for the three damage classes. The individual class means were compared using the Student-Newman-Keuls test for unequal sample sizes (a least significant range test of means) i f the F-value from the analysis of variance was significant at the 95% level (Sokal and Rohlf 1969). There were significant differences between damage classes for a l l six variables at the 0.01 level of significance. Results of the tests of means are given i n Table V. For the individual dye layer densities the trees which had pitched-out the attacking insects had consistently .lower densities-which indicated higher reflectance i n the green, red and NIR portions of the spectrum. These data confirm the differences indicated upon examination of the confidence intervals around the means (Fig. 6) . This general higher reflectance may be indicative of greater vigour and therefore possible greater resistance to insect attack. These trees may not have been attacked naturally; rather, the synthetic aggregating pheromone released i n the area (as mentioned previously) may have been responsible for attracting the beetles which i n i t i a t e d these unsuccessful galleries. The trees which pitched-out the attacking beetles may owe their greater vigour to either microsite differences i n nutrients, differences i n infection by pathogens or age differences. At the significance level used i n the test of means (p<0.05) the cyan dye layer density of the images of the pitched-out trees was not significantly different from that of the healthy trees. The same differences were present at the 0.10 level as for the other two dye layers. Table V. Means of each variable for each of the three damage classes with green-appearing foliage (healthy, pitched-out and successfully attacked). Damage Dye layer densities Glass Yellow Magenta Cyan 1 " healthy 1.037a 1 .125 a 0.536a pitched-out 0.807b 0.900 b 0.476 b successfully 1.061a 1.074 a 0.597a attacked Ratios  Magenta/Yellow Cyan/YellowCyan/Magenta 1.099 a 0.526 b 0.480b 1.121 a 0.591 a 0.529 a 1.022 b 0.562 a 0.555 a 1 Values within a column with different superscripts d i f f e r at the O.C-5 level of probability (Student-Newman-Keuls test). 38 The ratio variables best differentiate between successfully attacked trees and healthy trees. As ratios provide a measure of reflectance patterns ( i . e . the reflectance i n one region as compared to another) and because the ratios have a much smaller range within each class, they provide a more sensitive basis of comparison between classes. In the two ratios dealing with the cyan dye layer which measures NIR reflectance, the images of the pitched-out trees and successfully attacked trees had higher values than those of the healthy trees. The pitched-out tree images have higher values because although the density of the cyan layer was equal to the cyan layer density of the healthy tree images, the densities of the yellow and magenta layers were lower i n the pitched-out tree images. These densities and the ratio values are a result of the generally higher reflectance found i n the foliage of pitched-out trees. The significantly higher cyan/magenta and cyan/yellow ratios i n the successfully attacked trees were due to less obvious changes i n the reflectance patterns. While the density values of the individual dye layers for the images of successfully attacked trees did not d i f f e r significantly from the values for the healthy tree images a change i n pattern resulting i n differing ratio values was shown (Figs. 6, 7)- The mean of the cyan dye layer density was higher than that f o r the healthy tree images and the confidence limits did not greatly overlap. There was a considerable overlap of densities of the yellow and magenta dye layers between images of successfully attacked and healthy trees. The generally higher cyan layer densities for the images of successfully attacked trees would yield significantly different ratios. This i s particularly true i f the higher cyan layer densities were from the images of the same trees as the lower values for the yellow and magenta layers. The ratio, therefore, became 39 a measure of the reflectance pattern. The slightly higher cyan layer densities reflect a lowering of the NIR spectral reflectance which i s i n agreement with the underlying hypothesis tested i n this study. The images of successfully attacked trees had a significantly lower magenta/ yellow ratio than either images of healthy trees or pitched-out trees (Table V, Fig. 7). The lowering of the ratio towards a value of 1.0 where the reflectance of the two spectral regions would be equal indicates a change i n the normal reflectance of the green and red spectral regions. This may be caused by a decrease i n the green reflectance and an increase i n the red reflectance. There was no significant difference between healthy and successfully attacked trees i n either of the individual dye layers concerned i n this ratio (Table V) and their confidence intervals overlap extensively (Fig. 6). However, there must be a consistent relationship to produce significantly differing ratios. This indicates that the higher values of the yellow dye layer (lower green reflectance) were from the same successfully attacked tree images that had a lower magenta dye layer density (higher red reflectance). The ratios represent a measure of reflectance patterns showing how the change i n reflectance of one spectral region i s related to the change i n the reflectance of another. The shifting of green and red reflectance and the above mentioned slight decreasing of NIR reflectance explain the orangish hue of the images of success-f u l l y attacked trees (Fig. 9). The red reflectance increase may have been a result of the chlorophyll beginning to break down, unmasking the reflectance from carotenoids and anthocyanins. This i s a preliminary step to the all-red appearing one-year-old attacked trees, producing a colour on CIR film intermediate between the magenta colour of healthy foliage and the yellow colour of red dead foliage. This slight change i n green colour was not noticed just prior to the 40 photographic mission, so the -visual change must have "been slight. Although the sample sizes i n this study were relatively small (Table IV), significant differences i n reflectance as measured by film dye layer density were found to exist between healthy and attacked tree images. Finding specific spectral signatures unique to each damage class would allow computer-assisted discrimination between classes based on multispectral data. To determine i f this i s possible, stepwise discriminant analysis was employed using the generalized spectral data available from film dye layer analysis. Greater discrimination between classes could probably be achieved i f a greater number of smaller discrete spectral bands were considered, as i s possible with a multispectral scanner. The dye layer densities offer spectral reflectance data i n only three broad spectral ranges. 3.2.2 Stepwise Discriminant Analysis A stepwise discriminant analysis program, BMD P?M, was used to derive the best weighted linear combination of reflectance variables and ratios that would distinguish between the damage classes. Two analyses were done: one discriminating between a l l five classes and the other discriminating between only the three classes with green foliage. The second analysis was done as the red-topped and old-dead classes are readily separated from each other and other damage classes from the ground and from photographs. Before any variables were selected i n discriminant analysis, F values were calculated that were equivalent to those obtained from one-way analysis of variance for each variable over a l l the classes concerned. The f i r s t variable selected was that which had the highest F value and was therefore the variable which best discriminated between classes. After a variable had been selected, the variance associated with i t was removed from the variance/covariance matrix, 41 and the next best discriminating variable as indicated by the highest F value was then included i n the discriminant function. This process was repeated u n t i l no further variation could be removed. A classification function using the variables entered with coefficients calculated from known individuals i s derived for each class. Each case i s tested and allocated to a class by the classification functions. The values of the individual variables for each case are substituted i n each of the functions (one classification function for each class). The function which yields the highest value determines which class that case i s assigned to. In jack-khife classification, which was used i n this study, the functions are re-evaluated excluding the case being considered. Thus each case i s considered as an unknown giving a better measure of the power of the discriminant functions. As the class of each case was known, a percentage of correct classification was determined. A U-statistic i s calculated for the analysis. A low U value indicates a high degree of successful discrimination. That i s , the lower the U-statistic the more powerful the discriminant functions. Table VI i s a summary of the discriminant analysis with the 5 classes: old-dead, red-top, successful attack, pitched-out and unattacked healthy. The f i r s t variable entered was the cyan/magenta ratio. This represents the ratio of NIR to red reflectance. As seen from Figure 7 this serves to divide the five classes into two groups: trees without green foliage and trees with green appearing foliage. The low total correct discrimination i s due to old-dead and red-topped trees being mutually misclassified and the three green-appearing classes being misclassified among themselves. The next variable, the cyan/yellow dye layer ratio, separated the old-dead from the red-topped Table VI. Summary of Stepwise Discriminant Analysis of a l l Five Damage Classes Variable Percent Classified Correct^ Step No. Entered Removed Included U-Statistic Approximate F-Statistic Old Dead Red Top Successfully Pitched-Attacked out Healthy Total 1 C / M 2 1 0 . 1 0 2 4 265.276 74.1 4 1 . 9 53.3 19.2 70.4 51.6 2 C / Y 2 0.0358 128 .428 96.3 ' 8 0 . 6 53-3 73.1 63.O 75.4 3 M/Y 3 0.0298 72.763 92.6 83.9 66.7 53.8 44.4 69.0 4 M 4 0 . 0 2 4 5 53 . 468 92.6 83.9 6 0 . 0 69.2 4 8 . 1 72.2 5 C 5 0.0202 43.627 92.6 83.9 6 0 . 0 73.1 51.9 73-8 1 Jack-kinfe classification 2 M = Magenta dye layer (red sensitive) C = Cyan dye layer (NIR sensitive) C/M = Cyan to Magenta ratio C/Y = Cyan to Yellow ratio M/Y = Magenta to Yellow ratio 43 trees while s t i l l leaving some confusion among the green-appearing classes. The remaining three variables included i n the f i n a l discriminant function serve primarily to discriminate between the three green-appearing classes. The magenta to yellow dye layer ratio (green to red spectral reflectance ratio) was entered at the third step. As mentioned i n the section dealing with one-way analysis of variance, this ratio separated the successfully attacked trees from the other two green categories. With this variable included, 66.? percent of the successfully attacked trees were assigned to the correct damage class. Discrimination success of the old-dead and red-topped trees remained high and the overall correct classification dropped slightly as there was confusion between the pitched-out and healthy trees. The addition of the f i n a l two variables (magenta and cyan dye layer densities) served to increase the discrimination between pitched-out and healthy trees while decreasing the correct classification of successfully attacked trees to 60 percent. This decrease represents the further misclassification of one tree. As the purpose of the analysis was to discriminate between successfully attacked trees and other green-appearing classes the inclusion of the l a s t two variables serves no purpose even though the overall successful discrimin-ations was improved. The classification functions for a l l five damage classes which gave the best separation of successfully attacked trees are shown i n Table VII. Significant differences based on the discriminant functions were found between a l l five groups. The greatest differences exist between the discoloured classes and the green-appearing classes. This i s consistent with the appearance of the two groups on the photographs (Figs. 8, 9). Significant differences 44 Table VII Classification functions giving maximum separation of successfully attacked trees (66.7% correctly classified) when a l l 5 damage classes used i n the analysis. coefficients for group Variable old-dead red-tops successful attack pitch-out healthy M / Y 1 .1193-240 1218 .685 1219.694 1265.199 1262.817 C / M 1369.481 1442 .803 1361.470 1392.511 1394.387 C / Y -1256.544 -1368.426 -1327.085 -1356.570 -1368.716 Constant -666.425 -663.982 -629.865 -678.460 -669.976 1 M/Y = Magenta/Yellow dye layer ratio C/M = Cyan/Magenta dye layer ratio C/Y = Cyan/Yellow dye layer ratio 45 also exist between a l l three classes with green foliage. The discriminant analysis considers and combines a l l the information from a l l included vari-ables between a l l classes. A summary of the discriminant analysis for the three green-appearing classes (successfully attacked, pitched-out and healthy) i s : .shown i n Table VIII. In this case the f i r s t variable selected was the magenta to yellow dye layer ratio. This variable served primarily to correctly classify 66.7 percent of the successfully attacked trees. This was the same maximum separation of successfully attacked trees as was found i n the analysis of a l l five groups. Table IX shows the classification functions using this variable which yielded the 66.7 percent correct classification of successfully attacked trees. The remainder of the trees i n the other classes were poorly identified producing an overall successful classification percentage of 45.6. The addition of further variables to the function improved the classification of the other classes while again reducing the correct classification of successfully attacked trees to 60.0 percent. The f i n a l overall successful classification was 66.2 percent. F-test comparisons between the three green-foliaged damage classes based on the f i n a l three variables included i n the discriminant functions showed significant differences between a l l three groups. The greatest difference was found to be between the successfully attacked trees and the remaining two classes. Successful discriminant analysis depends upon each class having a unique reflectance pattern. Variation exists within each class. In the case of successfully attacked trees the degree of change i n reflectance may depend upon the degree of infection by the beetle/fungus complex. Trees with more attacks Table VIII. Summary of Stepwise Discriminant Analysis of Three Damage Glasses with Green Foliage Variable Percent Classified Correct No. Approximate Successfully Pitched-Step Entered Removed Included U-Statistic F-Statistic Attacked Out Healthy Total 1 M / Y 2 1 0.7522 10.706 66.7 53-8 25 .9 4 5 . 6 2 C / Y 2 0.5868 9 . 7 7 5 60.0 73.1 48.1 6O.3 3 M 3 0.5430 7 . 9 4 8 60.0 76 .9 51.9 63.2 4 C 4 0.4879 6.691 60.0 80.8 51.9 6 4 . 7 5 M/Y 3 0 . 4 9 4 0 8 . 8 7 9 60.0 80.8 5 5 . 6 66.2 1 Jack-knife classification 2 M = Magenta dye layer (red sensitive) C = Cyan dye layer (NIR sensitive) C/Y = Cyan to Yellow ratio M/Y = Magenta to Yellow ratio 47 Table IX. Classification functions giving maximum separation of successfully attacked trees (66.7% correctly classified) when only the three green-appearing classes were used i n the analysis. Coefficients for Group  Variable Successful Attack Pitch-Out Healthy M/Y1 226.029 247.984 249.993 Constant -116.599 -140.127 -134.587 1 M/Y = Magenta/Yellow dye layer ratio 48 may degenerate faster than those trees with few attacks. This may be related to the rate of spread of the blue stain fungus through the sapwood. The more innoculations by attacking beetles the faster the sapwood becomes completely colonized and the faster the reflectance pattern degenerates from the normal pattern. Also, site differences may affect the speed of colonization. Time of attack i s probably not a factor as the population of Douglas-fir beetles emerges from the hosts and attacks new trees over a short period of time. Even with this additional possible source of variation, significant differences were found between the groups. Discriminant analysis of these tree classes has shown differences i n reflectance patterns i n each of the classes. Discriminant analysis on dye-layer density data shows promise as an aid to damage assessment from aerial photographs or other imagery. I t i s speculated that measurements from a greater number of narrow spectral regions, as i s possible with a multispectral scanner, may provide better separation of damage classes. 3.3 Girdled Trees The five trees girdled i n late July, 1979 prior to the photographic f l i g h t were intended to serve as simulated successfully attacked trees although the effect of the fungus was not properly reproduced. Densitometric readings on the f o l i a r reflectance of these a r t i f i c a l l y stressed trees were to have served as comparisons with readings taken from attacked trees. It was hoped that these trees would provide further infomration on the reflectance changes occurring i n stressed trees. Density data were not obtained for these trees since the plot was not covered by the a i r photos. 49 These trees were observed during the f i n a l ground check i n January, 1980. They s t i l l had healthy appearing green foliage which was i n contrast to the successfully beetle attacked trees whose foliage had changed to a yellow colour. Although the phloem had been severed and the sapwood pa r t i a l l y severed, required nutrients were s t i l l available to the foliage through the remaining sapwood and these trees may continue to l i v e indefinitely due to root grafting with surrounding healthy Douglas-fir. The change i n foliage colour i n the successfully attacked trees indicates that the sapwood had been completely colonized by the blue-staining fungi associated with the Douglas-fir beetle. Death of trees attacked by the Douglas-fir beetle can therefore be attributed primarily to the successful colonization of the sapwood by the fungus. If this i s the case, the main role of the insect i s that of a vector for the fungus which not only causes the death of the tree but also protects i t s vector by blocking resin canals, thereby ensuring successful beetle broods to act as vectors i n the next year. Because of the importance of the fungus, the rate of f o l i a r colour change may be related to the rate of development of the fungus rather than to the development rate of the beetle larvae. To predict when f o l i a r colour change wi l l occur i t w i l l be necessary to determine the biological requirements of the fungus. Further research should be conducted to determine when reflectance changes occur i n relation to the degree of colonization of the sapwood by the fungus. In this study, a slight foliage reflectance change of unknown value occurred approximately three months after i n i t i a l attack by the beetle and introduction 50 of the fungus. Some changes i n reflectance may occur prior to this visual change. Trees could be girdled (completely severing the sapwood) at weekly intervals and a single photographic f l i g h t using both colour and CIR films could be made after several such treatments which would include a l l such treated trees as well as untreated trees. The time progression of reflectance changes could then be determined which would aid i n determining when beetle survey f l i g h t s should be made to detect the f i r s t changes i n spectral reflectance patterns. Additionally, sets of trees could be girdled to varying depths of the sapwood and then photographed. Analysis of the dye layer densities of these images may give some indication as to the progression of reflectance changes i n relation to the amount of active sapwood available for nutrient transport. 4. CONCLUSIONS The present study demonstrated that Douglas-fir trees successfully attacked by the Douglas-fir beetle could be detected on colour infra-red photographs prior to a visual foliage colour change. Such detection depended upon a change i n the amount of reflected NIR light prior to any change i n the visual spectral regions. This change i n the NIR reflectance produced a change i n the density of the cyan dye layer of the film. Slight changes, after the f i r s t stages of insect attack, were not apparent visually from the ground, but densitometric analysis of film dye layer densities showed the changes occurring. Departures from the dye layer densities of normal trees signify physiological stress i n the trees, possibly caused by the activity of the Douglas-fir beetle/blue stain fungus complex. Significant differences were found i n the dye layer densities between healthy and attacked trees (Table V). The most significant differences were 51 found when ratios of dye layer densities were compared. Ratios have much smaller variances than the single dye layer densities and provide a measure of the reflectance pattern i n trees showing how one spectral region interacts with another. Discriminant analysis determined the optimum combination of variables that correctly classified individual cases to their proper class. This analysis made use of the unique spectral qualities associated with each damage class. Highly successful discrimination was achieved when the classes are vastly different. Such was the case when old-dead and red-topped trees are compared to the green-appearing damage classes. Poorer discrimination occurred when the differences between classes were not so extreme. In such instances a larger number of samples may help to define the classification functions for each class. However, even with the limited number of samples used i n this study nearly 67 percent of the successfully attacked trees were classified correctly, indicating that these trees do have a unique spectral signature. Discriminant analysis shows promise i n distinguishing damage types from spectral reflectance data. In this study the significantly different magenta dye layer to yellow dye layer density ratio (ratio of red to green reflectance) found i n the successfully attacked trees indicated that some slight visual colour change had already occurred at the time the photographs were taken. Although the difference was not noted from the ground immediately prior to the photographic mission, this change was possibly an intermediate step between the i n i t i a l change i n NIR reflectance and the visual change apparent i n straw-coloured trees. Further experiments should be carried out to determine the earliest possible time after 52 insect attack that differences can be found i n the dye layer densities, especially the cyan dye layer which i s responsive to NIR reflectance. This time period w i l l probably be variable from year to year depending on the seasonal temperature as i t affects the growth and spread of the blue stain fungus. The rate of colonization of the sapwood by the fungus should also be studied under a variety of temperature regimes. I n i t i a l photographic detection f l i g h t s could then be carried out after a pre-determined number of day-degrees since i n i t i a l beetle attack. Chemical analysis of foliage from healthy and success-f u l l y attacked trees should also be carried out to determine the physiological changes occurring which influence the changes i n reflectance patterns. Determination of the smallest photograph scale necessary for the detection of the f i r s t reflectance change i s also necessary. A scale of 1:1100, as was used i n this study, i s too large to be practical for insect surveys except as one of the f i n a l stages i n a multi-stage sampling regime. A small scale i s necessary i f this technique i s to be used i n the future for cost effective insect-damage surveys. Larger scales may be used i n high-value, high-risk stands where expense of survey i s less of a factor. This technique may never identify a l l the infested trees i n a stand. However, because Douglas-fir beetle infested trees tend to occur i n small clumps, the correct identification of one tree w i l l lead to the identification of others when the area i s ground checked. This technique should provide starting points for the ground cruises allowing more optimal allocation of resources. Although this study concerned i t s e l f with Douglas-fir beetle detection, a l l bark beetles of the genus Dendroctonus behave in.a similar fashion. Also, a l l members of this genus which are economically important have associated 53 blue stain fungi. Because of this, this technique may be used i n the detection of other bark beetle caused damage. In the case of the mountain pine beetle, detection of newly attacked trees from the a i r months prior to a f o l i a r colour change would greatly improve the possibility of controlling this insect and avoiding widespread epidemics such as are now i n progress i n British Columbia. 54 5. LITERATURE CITED Alger, L.A., P.J. Egan and H.J. Heikkenen. 1978. Previsual Detection of Stressed Loblolly Pine (Pinus taeda L.). Symp. Rem. Sens. Veg. Dam. Assess.:65-72. Amman, G.D., M.D. McGregor, D.B. Cahill and W.H. Klein. 1977* Guidelines for Reducing Losses of Lodgepole Pine to the Mountain Pine Beetle in Unmanaged Stands in the Rocky Mountains. U.S.D.A. For. Serv. Gen. Tech. Rept. INT-36. 19pp. Arnberg, W., L. Wastenson and B. Lekander. 1973• Use of Aerial Photographs for Early Detection of Bark Beetle Infestations of Spruce. Ambio 2: 77-83. Atkins, M.D. and L.H. McMullen. 1958. Selection of Host Material by the Douglas-fir Beetle. Bi-Mon. Prog. Rept. 14:3. Atkins, M.D. and L.H. McMullen. I960. On Certain Factors Influencing Douglas-f i r Beetle Populations. Special Paper. Proc. 5th World For. Cong.: 857-859. Bellusehi, P.G., N.E. Johnson and H.J. Heikkenen. 1965* Douglas-fir Defects Caused by the Douglas-fir Beetle. Jour. For. 63:252-256. Bellusehi, P.G. and N.E. Johnson. 1969. The Rate of Crown Fade of Trees Killed by the Douglas-fir Beetle in Southwestern Oregon. Jour. For. 6700 -32. Benson, M.L. and W.G. Sims. 1967* False-Color Film Fails in Practice. Jour. For. 65:904. Brown, H.D. 1971 Declining Loblolly Pine Not Reliably Detected with Aerial Color Film. 3rd Bien. Wrkshp. Col. Aer. Photog. in the Pint. Sci. and Related Fields:246-254. Ciesla, W.M. 1977* Color vs. Color IR Photos for Forest Insect Surveys. 6th Bien. Wrkshp. Aer. Col. Photog. in the Pint. Sci. and Related Fields:31-42. Ciesla, W.M., J.C. Bell Jr. and J.W. Curlin. 1967. Color Photos and the Southern Pine Beetle. Photogram. Eng. 33:883-888. Cottrell, C.B., L.S. Unger and R.L. Fiddick. 1979. Timber Killed by Insects in British Columbia, 1971-1975* Environ. Can., Can. For. Serv. Rept. BC-X-189. 31pp. Dyer, E.D.A. and D.W. Taylor. 1971. Spruce Beetle Brood Production i n Logging Slash and Windthrown Trees in British Columbia. Environ. Can., Can. For. Serv. Info. Rept. BC-X-62. 18pp. 55 Dyer, E.D.A. and CM. Lawko. 1978. Effect of Seudenol on Spruce Beetle and Douglas-fir Beetle Aggregation. Bi-Mon. Res. Notes 34:30-32. Fox, D.J. and K.E. Guire. 1976. Documentation for MIDAS. 3rd Ed. S t a t i s t i c a l Research Laboratory, Univ. of Michigan. 203pp. F r i t z , N.L. 1967. Optimum Methods for Using Infrared-Sensitive Color Films. Photogram. Eng. 33:1128-1138. Furniss, M.M. 1962. Infestation Patterns of Douglas-fir Beetle i n Standing and Windthrown Trees i n Southern Idaho. Jour. Econ. Ent. 55:486-491. Gausman, H.W. 1977. Reflectance of Leaf Components. Rem. Sens, of Envir. 6:1-9. Harris, J.W.E., A.F. Dawson and R.G. Brown. 1978. Detecting Windthrow, Potential Foci for Bark Beetle Infestation, by Simple Aerial Photographic Techniques. Bi-Mon. Res. Notes 34:29. Harris, J.W.E. and A.F. Dawson. 1979. Evaluation of Aerial Forest Pest Damage Survey Techniques i n British Columbia. Envir. Can., Can. For. Serv. Info. Rept. BC-X-198. 22pp. Hedden, R.L. and R.I. Gara. 1976. Spatial Attack Pattern of a Western Washington Douglas-fir Beetle Population. For. Sci. 22:100-102. Heller, R.C. 1968. Previsual Detection of Ponderosa Pine Trees Dying From Bark Beetle Attack. Proc. 5th Symp. Rem. Sens, of Envir.«387-434. Heller, R.C. 1971. Detection and Characterization of Stress Symptoms i n Forest Vegetation. Proc. Int. Wrkshp. Earth Res. Surv. Syst. Vol I I : 108-122. Heller, R.C, R.C Aldrich and F.W. Bailey. 1959. An Evaluation of Aerial Photography for Detecting Southern Pine Beetle Damage. Photogram. Eng. 25:595-606. Heller, R.C. and J.F. Wear. 1969. Sampling Forest Insect Epidemics with Color Films. 6th Symp. on Rem. Sens, of the Envir.:1157-ll67. Hopping, R. 1929. The Control of Bark Beetle Outbreaks i n B r i t i s h Columbia. Can. Dept. Agric. Ent. Br. Circ. No. 15, 15pp. Jennrich, R. and P. Sampson. 1977. ?7M Stepwise Discriminant Analysis. In BMDP Biomedical Computer Programs, M.B. Brown (Ed.), pp. 711-36. Klein, W.H. 1973- Beetle-Killed Pine Estimates. Photogram Eng. 39:385-388. Knipling, E.B. 1967. Physical and Physiological Basis for Differences i n Reflectance of Healthy and Diseased Plants. Proc. Wrkshp. on Infrared Col. Photog. i n Pint. Sci. Flo. Dept. A g r i c , Div. Plant Indus., Winterhaven, Fla. 24pp. 56 Kodak. 1971' Kodak Data for Aerial Photography. Eastman Kodak Co., Kodak Publ. No. M-29. 80pp. Lejeune, R.R., L.H. McMullen and M.D. Atkins. I96I. The Influence of Logging on Douglas-fir Beetle Populations. For. Ghron. 3?:308-314. Lillesand, T.M., P.D. Manion and B.B. Eav. 1978. Quantification of Urban Tree Stress Through Microdensitometrie Analysis of Aerial Photography. St. Univ. New York, Col. Envir. Sci. and For., Syracuse. 57PP« Lorio, P.L. and J.D. Hodges. 1977« Tree Water Status Affects Induced Southern Pine Beetle Attack and Brood Production. U.S.D.A. For. Serv. Res. Pap. SO-135. 7PP. Mahoney, R.L. 1978. Lodgepole Pine/Mountain Pine Beetle Risk Classification Methods and Their Application. Proc. Symp. oh Theory and Practice of Mountain Pine Beetle Management in Lodgepole Pine Forests:106-113. Meyer, M.P. and D.W. French. 1967« Detection of Diseased Trees. Photogram. Eng. 37:1035-1040. McGugan, B.M. 1956. The Canadian Forest Insect Survey. Proc. Tenth Int. Cong. Entomol. 4:219-231. McMullen, L.H. and M.D. Atkins. 1962. On the Flight and Host Selection of the Douglas-fir Beetle, Dendroctonus pseudotsugae Hopk. (Coleoptera; Scolytidae). Can. Ent. 94:1309-1325. McMullen, L.H. 1977. Douglas-fir Beetle in British Columbia. Environ. Can., Can. For. Serv. FPL. 14. 6pp. Murtha, P.A. 1968. Near-Infrared Detection of Simulated Animal Damage on Conifers. Ph.D. Thesis, Cornell Univ. 85pp. Murtha, P.A. 1972. A Guide to Air Photo Interpretation of Forest Damage in Canada. Envir. Can., Can. For. Serv. Publ. 1292. 62pp. Murtha, P.A. 1978. Remote Sensing and Vegetation Damage: A Theory for Detection and Assessment. Photogram. Eng. 44:1147-1158. Murtha, P.A. and L.S. Hamilton. 1969. Detection of Simulated Damage on Conifers Using Near Infrared Film. Jour. For., Nov., 1969:827-829. Olson, C.E. Jr., W.G. Rohde and J.M. Ward. 1970. Remote Sensing of Changes in Morphology and Physiology of Trees Under Stress. Ann. Prog. Rept. for Earth Resources Survey Prog., Off. of Space Sciences and Applications, NASA. - Remote Sensing Applications in Forestry. 26pp. Orr, L.W. 1954. The Role of Surveys in Forest Insect Control. Jour. For. 52:250-252. 57 Pitman, G.B. and J.P. Vite. 1970. Field Response of Dendroctonus pseudotsugae (Coleoptera: Scolytidae) to Synthetic Frontalin, Ann. Ent. Soc. Amer. Pitman, G.B., R.H. Hedden and R.I. Gara. 1 9 7 5 ' Synergistic Effects of Ethyl Alcohol on the Aggregation of Dendroctonus pseudotsugae (Col., Scolytidae) i n Response to Pheromones. Z. ang. Ent. 78s203-208. Rohde, W.G. 1971. Multispectral Enhancement of Disease i n Forest Stands 3rd. Bien. Wrkshp. Col. Amer. Photog. i n the Pint. Sci. : 1 3 1-l43. Rohde, W.G. and C E . Olson, J r . 1 9 7 0 . Detecting Tree Moisture Stress. Photogram. Eng. 36:561-566. Rudinsky, J.A. 1966. Host Selection and Invasion by the Douglas-fir Beetle, Dendroctonus pseudotsugae Hopkins, i n Coastal Douglas-fir Forests. Can. Ent. 98:98-111. Rudinsky, J.A., L.C Ryker, R.R. Michael, L.M. Libbey and M.E. Morgan. 1976. Sound Production i n Scolytidae: Female Sonic Stimulus of Male Pheromone Release i n Two Dendroctonus Beetles. J. Insect Physiol. 22(12):l675-l68l. Ryker, L.C, L.M. Libbey and J.A. Rudinsky. 1979* Comparison of Volatile Compounds and Stridulation Emitted by the Douglas-fir Beetle from Idaho and Western Oregon Populations. Envir. Ent. 8:789-798. Safranyik. L., D.M. Shrimpton, H.W. Whitney. 1974. Management of Lodgepole Pine to Reduce Losses from the Mountain Pine Beetle. Environ. Can., Can. For. Serv. For. Tech. Rept. 1. 24pp. Schmid, J.M. 1976. Temperatures, Growth, and F a l l of Needles on Engelmann Spruce Infested by Spruce Beetles. U.S.D.A. For. Serv. Res. Note RM-331. 4pp. Schmid, J.M. and R.H. Frye. 1976. Stand Ratings for Spruce Beetles, U.S.D.A. For. Serv. Res. Note RM-309. 4pp. Sokal, R.R. and F.J. Rohlf. 1 9 6 9 . Biometry. W.H. Freeman and Co., San Francisco. 776pp . Swaine, J.M. 1914. Forest Insect Conditions i n British Columbia. A Preliminary Survey. Can. Dept. Agric. Div. of Entomol. Entomol. Bull. No. 7. 41pp. Thomas, J.R., V.I. Myers, M.D. Heilman and C L . Wiegand. 1966. Factors Affecting Light Reflectance of Cotton. Proc. 4th Symp. Rem. Sens, of Environ. O O 5 - 3 1 2 . Walters, J. 1956. Biology and Control of the Douglas-fir beetle i n the Interior of Br i t i s h Columbia. Can. Dept. Agric. For. Bi o l . Div. Publ. 975. l l p p . 58 Wear, J.F., R.B. Pope and P.G. Lauterbach. 1964. Estimating Beetle-Killed Douglas-fir by Aerial Photo and Field Plots. Jour. For. 62:309-315-Weber, F.P. 1965. Exploration of Changes i n Reflected and Emitted Radiation Properties for Early Remote Detection of Tree Vigour Decline. M.Sc. Thesis. Univ. of Michigan, Sch. of Nat. Res. 101pp. Weber, F.P. and F.C. Polcyn. 1972. Remote Sensing to Detect Stress i n Forests. Photogram. Eng. 28:163-175. Wert, S.L. and B. Roettgering. 1968. Douglas-fir Beetle Survey With Colour Photos. Photogram. Eng. 34:1243-1248. 59 Appendix I. Descriptive Statistics for a l l six variables i n a l l five damage classes. Yellow Dye Layer Density (Green Sensitive - 500 - 600nm) Damage class n range X Sx Healthy green 27 0.640 - 1.650 1.037 O.O532 1979 pitch-out 26 0.540 - 1.260 0.807 0.0324 1979 attack 15 0.680 - I . 5 8 O 1.061 O.O659 1978 attack 31 0.140 - I .38O O.713 O.O374 old attack 27 0.100 - 1.220 O.696 0.0429 Total 126 0.100 - 1.650 0.840 0.0240 Magenta Dye Layer Density (Red Sensitive - 600 - 700nm) Damage class n range X Sx Healthy green 27 O.76O -- I.63O 1.125 0.0473 1979 pitch-out 26 O.58O -- 1-335 0.900 O.O309 1979 attack 15 0.755 -• 1.500 1.074 0.0570 1978 attack 31 0.370 -- 1.120 0.592 0.0301 old attack 27 0.390 -- 1.100 0.687 0.0382 Total 126 0.370 -- I.63O 0.848 0.0256 Cyan Dye Layer Density (NIR sensitive - ?00 - 900nm) Damage class n range Sx Healthy green 27 0.370 - 0.740 0.536 0.0216 1979 pitch-out 26 O.36O - 0.705 0.476 0.0199 1979 attack 15 O.38O - 0 . 8 9 5 0.597 0.0394 1978 attack 31 0.370 - 1.120 0.575 0.0272 old attack 27 0.480 - 1.140 0.742 0.0362 Total 126 O.36O - 1.140 0.582 0.0150 Magenta/Yellow Density Ratio (Red reflectance/Green reflectance) Damage class n range X Sx Healthy green 27 0.988 -- 1.209 1.099 O.OI32 1979 pitch-out 26 I.034 - 1.206 1.121 0.0100 1979 attack 15 0.873 -- 1.206 1.022 0.0224 1978 attack 31 0.662 -- 1.000 0.792 O.OI33 old attack 27 0.848 -- 1.022 0.935 0.00908 Total 126 0.662 -• 1.209 0.984 0.0128 61 Cyan/Magenta Density Ratio (NIR reflectance/Red. reflectance) Damage class n range X Sx Healthy green 27 0.389 - 0.627 0.480 0.0110 1979 pitch-out 26 0.430 - 0.690 0.529 0.0127 1979 attack 15 0.462 - 0.670 0.555 0.0195 1978 attack 31 0.677 - 1.137 0.988 0.0246 old attack 27 0.946 - 1.286 1.093 0.0152 Total 126 0.389 - 1.286 0.756 0.0248 Cyan/Yellow Density Ratio (NIR reflectance/Green reflectance) Damage class n range x Sx Healthy green 27 0.416 -- 0.673 0.526 0.0116 1979 pitch-out 26 0.512 -- 0.741 0.591 0.0109 1979 attack 15 0.485 -- 0.639 0.562 0.0124 1978 attack 31 0.538 -- 0.964 0.779 0.0200 old attack 27 0.894 -• 1.182 1.021 0.0150 Total 126 0.416 -• 1.182 0.712 0.0180 62 APPENDIX II. Histograms of a l l six variables i n a l l five damage classes. Page 63: Blue F i l t e r ; Yellow Dye Layer; Green Sensitive. Page 6k'. Green F i l t e r ; Magenta Dye Layer; Red Sensitive. Page 65s Red F i l t e r ; Cyan Dye Layer; NIR Sensitive. Page 66: Green/Blue Ratio i . e . Red/Green Reflectance Ratio. Page 67: Red/Green Ratio i . e . NIR/Red Reflectance Ratio. Page 68: Red/Blue Ratio i.e. NIR/Green Reflectance Ratio. BLUE FILTER ; YELLOW DYE LAYER ; GREEN SENSITIVE GREEN HEALTHY 1979 PITCH OUT * 1979 SUCCESSFUL ATTACK 1978 ATTACK, RED TOP • OLD DEAD TOTAL 0.1 .5 1.0 1.5 DENSITY 64 GREEN FILTER ; MAGENTA DYE LAYER ; RED SENSITIVE 1 0 r 1 GREEN HEALTHY 10 -5 -(/) Z o I C C L U (/) 0 3 O LL o 15 10 5 10 5 25 20 15 10 5 1979 PITCH OUT 1979 S U C C E S S F U L ATTACK RED TOP OLD DEAD TOTAL 0.1 .5 1.0 1.5 DENSITY 65 RED FILTER; CYAN DYE LAYER; NIR SENSITIVE to § DC UJ CO CQ o o d 1 0 5 15 10 5 10 5 15 10 5 10 5 40 30 20 10 0.1 .5 1.0 DENSITY GREEN HEALTHY 1979 PITCH OUT 1979 SUCCESSFUL ATTACK RED TOP OLD DEAD TOTAL 1.5 66 G R E E N / B L U E R A T I O ie R E D / G R E E N R E F L E C T A N C E R A T I O R A T I O RED/GREEN RATIO ie MIR/RED REFLECTANCE RATIO RATIO 68 RED/BLUE RATIO ie NIR/GREEN REFLECTANCE RATIO RATIO 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items