Use of remote sensing for forest vegetation

Use of remote sensing for forest vegetation management: A problem analysis

The Forestry Chronicle Downloaded from pubs.cif-ifc.org by CARLETON UNIV on 05/04/16 For personal use only.

by Douglas G. pittl, Robert G. wagner2,Ronald J. Hall3, Douglas J. ~ i n gDonald ~, G. ~ e c k i e ~ and Ulf ~ u n e s s o n ~

Forest managers require accurate and timely data that describe vegetation conditions on cutover areas to assess vegetation development and prescribe actions necessary to achieve forest regeneration objectives. Needs for such data are increasing with current emphasis on ecosystem management, escalating silvicultural treatment costs, evolving computer-based decision support tools, and demands for greater accountability. Deficiencies associated with field survey methods of data acquisition (e.g. high costs, subjectivity, and low spatial and temporal coverage) frequently limit decision-making effectiveness. The potential for remotely sensed data to supplement field-collected forest vegetation management data was evaluated in a problem analysis consisting of a comprehensive literature review and consultation with remote sensing and vegetation management experts at a national workshop. Among curently available sensors, aerial photographs appear to offer the most suitable combination of characteristics, including high spatial resolution, stereo coverage, a range of image scales, a variety of film, lens, and camera options, capability for geometric correction, versatility, and moderate cost. A flexible strategy that employs a sequence of 1:10,000-, 15,000-, and 1:500scale aerial photographs is proposed to: 1) accuratelymap cutover areas, 2) facilitate location-specific prescriptions for silvicultura1 treatments, sampling, buffer zones, wildlife areas, etc., and 3) monitor and document conditions and activities at specific points during the regeneration period. Surveys that require very detailed information on smaller plants ( 100 cm on 1:1,000 scale). Goba et al. (1982) suggest scales as small as 1:10,000 for conifer stocking and 1:4,000 for conifer density estimates. In most instances, however, tree counts have not been reliable due to the clumped tendency of natural regeneration and the large contribution of small seedlings (< 50 cm tall) that are not observed on photographs. Regression relationships between photo and ground density are strongly linear and can be used to correct photo estimates if large standard errors can be tolerated (Hall and Aldred 1992).Pilot studies also have illustrated the potential of 1:500-scale photographs for free-to-grow (FTG) surveys (Hunter and Associates 1983). 3~urrent-generationmetric camera lenses may meet or exceed 90 to 100 Ip mm-' AWAR (Area Weighted Average Resolution).

Photo scales of at least 1:10,000 may provide a reasonable compromise between spatial resolution and areal coverage for stratification of regeneration areas prior to detailed sampling (Spencer and Hall 1988). Smaller scales (1:20,000 to 1:60,000) have been found suitable for mapping and updating of cutover boundaries, depending on required accuracy (Gillis and Leckie 1996). Stereo, 1500- to 1:1,000-scalephotographs can be obtained for sampling purposes using 35- or 70-mm camera systems mounted in fixed- or rotary-wing aircraft. Fixed-base systems employ two identical cameras mounted at opposite ends of a rigid boom or at the wing tips, with their lens axes parallel. Booms are mounted parallel or perpendicular to flight. Both cameras are exposed simultaneously to obtain overlapping images of the target vegetation. The advantage of this configuration is that photo scale can be computed, without ground control, by the ratio of photo base to air base. Disadvantages include the fact that the base:height ratio, and thus vertical exaggeration of the stereo model, changes with flying height and most boom systems are designed for helicopters, which are expensive to operate. Timed interval, or sequential, systems employ a single camera that is exposed at specified intervals as the aircraft moves along a flight line (Hall 1984). Stereo photographs are obtained by selecting adjacent photo pairs along the line. In contrast to fixed-base systems, a desired base:height ratio can be maintained through different flying heights by adjusting exposure intervals to maintain photographic overlap. Sequential systems require a radar or laser altimeter for determination of exact camera height at each exposure and, subsequently, photo scale. Some systems are equipped with a second camera that is used to provide 1:2,000- to 1:4,000-scale tracking photographs simultaneous to the larger-scale sample images (Kirby 1980). These are used to facilitate post-flight recovery of sample plots in the field. Spencer and Hall (1988) provide a thorough overview of available Canadian large-scale photo systems to that time. Poole (1994) has recently developed a fixed-base 35-mm photographic system for very light aircraft as well as a gyro-mounted single-camera sequential system for 35-mm or 70-mm photographs. In the latter, the camera is triggered by GPS navigation software instead of timed intervals. Although conventional cameras do not have the capability of capturing images in digital format, various scanning devices can perform analog-to-digital conversions of processed films. Print products can be scanned, but the best results are achieved with first-generation negative or positive transparencies. During the scanning process, the photographic film is recorded as an array of pixels with brightness values translated from the optical density in the film emulsion. The range of density values in the film is typically recorded on an &bit (0-255) range in the digital file. Colour films result in digital files that are three times as large as those scanned from black and white films due to separate scanning of each colour (blue, green, and red in normal colour; green, red, and near IR in CIR). The choice of scanning aperture (spatial resolution) is important, since insufficient resolution will affect post-scanning use of the images and excessive resolution will require large amounts of storage capacity. A 230-mm colour transparency scanned at a resolution of 50 microns will result in a 3-band raster file 4600 by 4600 pixels, approximately 63.5 Mbytes in size. Once a photograph has been scanned, traditional image analysis tools can be applied, such as geometric corrections, geo-referencing, con-

JULYIAUGUST 1997, VOL. 73, NO. 4, THE FORESTRY CHRONICLE

465


trast and brightness adjustments, and statistical pattern recognition. Aerial photography is a mature technology with wellestablished mensurational techniques and a wide variety of available acquisition equipment. With suitable timing, photo scale, and film type, a wide array of FVM data may be acquired from aerial photographs, provided interest is generally in vegetation >0.5 m tall.

Digital Frame Cameras Digital frame cameras (DFCS)are a group of array sensors, of generally higher spatial resolution than video cameras, which are often computer-controlledand have flexible scan/ readout rates. DFCs are poised to displace conventional cameras in aerial photography, offering all the advantages of direct digital images. The use of DFCs permits real-time viewing of results and eliminates film processing and printing charges when hard-copy images are not required. DFCs typically use solidstate imaging technology similar to that in MEIS and CASI (see Airborne line imagers section below), but readout is as an instantaneous 2-D exposure or image frame (King 1993). Current sensor arrays are capable of relatively high spatial resolution and format size is increasing to levels that can match 230-mm photographic coverage (up to 9 K x 7 K elements). However, arrays that have been incorporated into standard imaging cameras are currently limited to about 4 K x 3 K pixels and therefore have limited swath widths. ~aintenanceof an adequate base:height ratio for stereo viewing can also be difficultwith these smaller sensor arrays. Some common 3 K x 2 K sensors, for example, are approximately equivalent to 35-mm format photographs in terms of spatial resolution and coverage. Like photographs, increased coverage is obtained at the expense of decreased spatial resolution for any given array format. Multispectral capabilities can be achieved with multiple cameras, filter wheels, tunable filters, or through internal filtering as in colour and colour IR DFCs. Spectral band widths as small as 5 to 10 nrn can be utilized in custom-designed sensors, while commercial colour~olourIR DFCs use bandwidths of approximately 100 nm. When filter wheels or tunable filters are used, images are slightly offset in time and space and must be geometrically registered afterwards. King (1995) provides a comprehensive summary of current multispectral DFC designs and applications. DFCs meet many advantages of the digital line scanners mentioned below, while being lower in cost, permitting customized sensor design by the user, having levels of technical sophistication and a format similar to photography, and providing easy installation in light aircraft. As two-dimensional arrays become larger, the capability of DFCs for both high resolution and large area coverage will increase. We estimate that commercial DFCs of sufficient size for operational mapping are at least five years distant. In time, conventional aerial cameras will potentially be displaced for many FVM applications as the cost of these higher-resolution solid-state arrays declines. Airborne Videography Video cameras generally have scanning rates, formats, and spatial resolution that comply with television standards. Airborne video offers a low-cost supplement to higherresolution sampling systems such as aerial photographs. Imagery captured continuously (at 30 frames sec-l) along a flight

466

line with wide swath width may be used simply to idenhfy where large-scale photos were taken, or, in more sophisticated applications, to extrapolate large-scale photo and ground-verification sample information to larger sample (video) strips within a cutover. Typical ground pixel size and coverage ranges for video cameras are shown (Table 2) ranging in format from 4 x 3 mm to 8.8 x 6.6 mm with lens focal length ranges of 6 to 50 mm, flown at altitudes between 300 and 7000 m. Analog data in individual frames can be captured and converted to digital data with a video frame-grabbing board (limiting image resolution to about 792 x 480 pixels). Digital video cam-corders have been developed recently that capture and store 30 digital frames sec-' on 8-mm tape. Multispectral systems, that use similar designs to those described for DFCs, are well established (King 1995). Typically, analog multispectral video signals are digitized in-flight by single or multiple frame-grabbers, thus eliminating analog tape storage. Stereo imagery can be achieved by capturing frames taken a distance apart along a flight line, although low resolution and inferior calibration negate precise photogrammetric applications. Airborne video is not recommended for area-based mapping (Table 2). This is a result of the typically small sensor arrays requiring very short focal length lenses that tend to produce large geometric distortions (unless expensive lenses are used). Airborne video has been used successfully to complement aerial photographs for updating forest maps and assessing regeneration stocking, herbicide coverage, and weed invasion (Hosking et al. 1992; Hosking 1994). Swath-widthadjustments and subsampling were accomplished with a zoom lens, eliminating the need for a second camera. The US Forest Service currently operates several colour video1GPS systems for forest development and damage assessment (Myhre and Silvey 1992). Fraser Inc. of New Brunswick currently use a consumergrade video camera pointing through the chin bubble of a Bell 206 helicopter to image areas of disturbance for some of their map updates (Gillis and Leckie 1996). Several other multispectral video applicationsin forestry have met with varying success. For example, Verreault et al. (1993) were successful in distinguishing and counting conifer regeneration >10 years old with their multiband filter wheel sensor. Lowe et al. (1995) found multispectral video to be more costeffective than either Landsat or CIR aerial photography in classifcation of forest vegetation. Slaymaker et al. (1995)used colour videography to acquire reference data for Landsat mapping of detailed forest ecosystem classes. A simultaneoustwo-scale system was employed, one camera covering a 500-m swath with 1-m pixels, the other covering a 30-m swath with 6-cm pixels. King and Vlcek (1990) found that mixed forest groups, pure conifer stands, and other non-forest types typical of southern Ontario could be classified to 70-90% accuracy using multispectral airborne videography with 2-m pixels. Yuan et al. (1991) demonstrated that quantitative models for individual hardwood tree stress assessment, which incorporated both spectral and textural image measures from large-scale,multispectralvideography, were well correlated with photo-interpreted stress assessment.

Lasers Like radar, lasers are active sensors, emitting and recording their own energy (in the form of light). Unlike the sensors discussed so far, however, lasers do not image the target vegetation but,

JUILLETIAO~T1997, VOL. 73, NO. 4, THE FORESTRY CHRONICLE


rather, they provide precise point information on distance or altitude. The laser system in an aircraft sends a light pulse towards the ground and receives the portion that is reflected back to the laser system. The time between transmitting the pulse and receiving the reflected signal is determined by the distance between the aircraft and the surface. Most laser systems are profilometers, with the laser pointing directly down from the aircraft and providing a series of heights along the flight line. Laser pulse rates are typically 2000 to 4000 pulses sec-l, giving a laser height sample every 1.5 to 5 cm. Typical footprint size of each pulse is 10 to 50 cm in diameter. Vertical height resolution is good, typically being 5 to 10 cm. With a large number of pulses along a track, the laser profile usually has sufficient number of pulses hitting the ground surface through a vegetative canopy that a trace can be made of both a canopy envelope and ground surface. Scanning laser systems that give laser pulse samples at intervals across a swath are also available. There is a 1500m altitude limit to most laser systems. As a laser light pulse travels through a canopy, some of the light is reflected back from different depths within. Some systems, termed lidars, are capable of recording the amount of light reflected from each depth in the canopy. One can see the distance to the top of the canopy and usually the ground surface (represented as the last light reflected). Thus a single pulse gives canopy height and some information on the vertical distribution of vegetation within the canopy. Schreier et al. (1985), Ritchie et al. (1993), and Weltz et al. (1994) provide examples of laser data over regenerating and shrub areas. Aldred and Bonnor (1985) give examples of lidar pulses of forest canopies. The accuracy of typical height estimates can be under 1 m. Information on the distribution of trees and size of gaps between them is also provided. Jacobs et al. (1993) used a laser profiler in conjunction with a solidstate video camera to conduct measurements of tree height (Ga) along with other forest inventory parameters such as species, density, crown and canopy closure. In FVM applications, the main role of the laser would be as a complementary sensor to imaging systems. For example, a laser may be flown with aerial cameras, DFCs, line imagers, or, as in the case of Jacobs et al. (1993), with video. With proper registration of the laser to associated imagery, quantitative samples of vegetation heights and distribution could complement other estimates.

Airborne Line Imagers Line imagers [e.g. Multi-spectral Electro-optical Imaging Scanner (MEIS) (McColl et al. 1983), Wide-angle HighResolution Line imager (Whim) (Neville et al. 1992), and the Compact Airborne Spectrographic Imager (CASI) (Babey and Anger 1993; Anger et al. 1994)l are under evaluation for forest inventory and damage assessment (Ahern et al. 1986;Leckie 1993;Franklin 1994;Paradine et al. 1995). MEIS incorporates linear solid-state sensors in multiple camera housings to generate multispectral imagery. WHiRL is a single channel, highresolution system. CASI is a single video resolution solid-state array sensor, which incorporates diffraction optics for detection of different spectral bands on each line of the array. Both MEIS and CASI offer the advantage of generating multispectral digital data at much higher spatial resolutions than spaceborne optical sensors (> 25 cm). MEIS has demonstrated potential for stratificationin juvenile stands (Kneppeck and Ahern

1987) and conifer regeneration assessment under conditions of minimal competing vegetation (Brand et al. 1991). These sensors, CASI in particular, also offer the greatest number of spectral bands and the smallest band widths among the sensors considered here. However, since each image line is acquired with a unique combination of aircraft position and attitude, special software is required for geometric image correction. The MEIS sensor can produce stereo imagery by using sensor channels oriented to view forward, aft, and nadir along a flight path. CASI and WHiRL do not have a fore-aft stereo viewing capability, but stereo can be achieved by the less-convenient acquisition and viewing of side-lap between adjacent fight lines. Stereo viewing and image manipulation require sophisticated hard- and software, resulting in high data extraction costs. Knowledge of spectral reflectance theory and characteristics of the targetlsensor are required for interpretation of both CASI and MEIS data. In addition, hyper-spectral software is required to make full use of CASI imaging spectrometer capabilities. There are several CASI sensors available for use in Canada and the MEIS sensor is operated commercially. Costs vary, but are generally higher than photographs. Although these sensors are not capable of the very high resolution required for FVM applications, they have potential for density measmments, cutover mapping and stratification, particularly as digital image processing and interpretation techniques evolve.

Optical Satellites A current selection of spaceborne imaging spectrometersproduce repetitive, small-scale synoptic images suitable for measuring coarse biophysical parameters and monitoring landscapelevel change [e.g. Landsat MSS and TM, Systeme Probatoire d'observations de la Terre (SPOT) (Table 2), Japanese Earth Resources Satellite (JERS-I), and Indian Resource Satellite (RS1A and'lB)]. Several pilot studies of forest regeneration using Landsat images have met with varying levels of success (Kneppeck and Ahern 1989; OCRS 1989; Leckie et al. 1992; Evans 1994; Franklin et al. 1994).Low spatial resolution and poor reliability caused by cloud cover are important drawbacks. Image costs can also be high (e.g. a full, corrected Landsat scene costs about $5,5004)if data are used for a small number of sites, but inexpensive if applied across large areas. As a source for depletion maps of better than 5 to 10-m accuracy, current satellite images generally do not have sufficient spatial resolution (Gillis and Leckie 1996). A new era of commercial earth observing satellites has begun, however, and new vegetation management applications may arise from future satellite data with spatial resolutions that range from 1 t~ 3 m panchromatic, and 4 to 15 m multispecha1 image bands (Fritz 1996). Within 5 years, high-resolution satellite imagery may offer an alternative to small-scale aerial photographs for large synoptic views, broad stratification, and cutover mapping. Radar Unlike optical sensors that record reflected electro-magnetic energy, radar sensors transmit and receive their own energy in the form of radio signals. For forest land, radar backscatter responds to structure (size, shape, orientation, distribution, and quantity 4All costs shown are in Canadian currency.


467


of vegetation elements), surface orientation and roughness, and dielectric properties, that are mostly governed by moisture content. Radar shows relationships to several of the vegetation parameters of interest in FVM (e.g. ground vegetation density, crown closure, and biomass of regenerating trees). There are numerous radar parameter and processing technique options for extracting information. Radar can be acquired at different wavelengths; the longer the wavelength, the greater the penetration through the vegetation canopy. Radar also can be acquired using different transmit and receive polarizations, each affecting the nature of the response. Polarimetric analysis techniques and sophisticated analysis of multiple wavelengths and polarizations, as well as combination with other remote sensing and ground information, can be applied to a wide range of conditions. Unfortunately, radar backscatter is influenced by a myriad of factors apart from those related to vegetation structure (e.g. slope, aspect, moisture conditions, microrelief, slash, vegetation phenology, etc.). This, in practice, makes the extraction of reliable quantitative information difficult. Although the availability of airborne radar systems for routine forestry applications is limited, there are currently several relevant satellite radars [Radarsat (C band), European Remote Sensing Satellites (C band), and Japanese Earth Resources Satellite (L-band)]. Radarsat pable 2) offers the unique capability of selecting different incidence angles and resolutions. Radar may presently be incapable of providing the detailed informationrequired for FVM applications,but it may well attract future consideration as a possible tool for monitoring general conditions.

Application of Existing Technology Several of the above remote sensing technologies offer the potential to gather data needed for FVM decision making. While no single sensor meets all of the requirements specified above, conventional aerial cameras and film appear to have the widest range of capabilities. This conclusion is supported by the literature and most of the remote sensing experts at the workshop. All focus groups at the workshop recommended a system involving aerial photographs, differing only in the proposed photo scales. The general consensus was to use: 1)photographs of 1:10,000- to 20,000scale for depletion mapping and pn:-establishment planning (depending on cutover size and frequency), 2) photographs of 1:5,000to 10,000-scalefor stratification and establishmentplanning, and 3) combinations of 1:500- to 1,000-scaleand ground sampling for more detailed vegetation assessments, where needed. The following recommendations are intended as a general framework for incorporatingremote sensing (aerial photographs,in particular) into FVM decision-making.

Recommended Approach The typical regeneration period following clearcut harvesting may be viewed as having three main phases where FVM decisions must be made (I 11, ,and III, Fig. 2). In the establishment phase, clearcut boundaries must be identified and mapped and regeneration strategies (a throughf, Fig. 2) defined or confirmed for the various site conditions within each cutover. Site conditions that may influence decisions in this phase include slope, aspect, and soil conditions, in combination with: advanced hardwood or softwood regeneration (a and b, Fig. 2),

468

low organic matter levels, slash, or non-crop vegetation, suitable for planting (c), heavier levels of organic matter, slash, or non-crop vegetation, requiring site preparation prior to artificial regeneration ( d e), and wet areas requiring draining, water courses, human dwellings, main roads, wildlife corridors, recreational areas, etc., requiring special attention and/or modified silvicultural techniques 0. In the early monitoring phase, two to five growing seasons after establishment, vegetation development is examined in terms of management objectives and prescriptions are formulated for areas requiring corrective action. Areas containing excessive non-crop vegetation may require release, over-stocked areas may require pre-commercial thinning, and under-stocked areas may require fill-planting or site preparation and planting. Toward the end of the regeneration period, 5 to 10 years after establishment,vegetation development is again examined to venfy that management objectives have been met (venifiation phase, Fig. 2). The recommended approach is to employ aerial photographs at each of these three regeneration phases to aid the FVM decision-makingprocess. Specifically, decisions in the establishment phase would be enhanced by map-quality, 230-mm format, 1:10,000-scale CIR, stereo photographs taken during the first leaf-off period following harvest (spring or fall, after snow melt or before accumulation) (Fig. 3). This photo scale is small enough to provide complete coverage of one or more cutovers and large enough to provide adequate spatial detail for treatment stratification. CIF film is used to enhance the visibility of conifer advanced regeneration and the presence of wet areas, rock outcrops, excessive slash loads, etc. A newer generation aerial camera with high quality lens (e.g. Wild RC-20, 30) and forward motion compensation is recommended to acquire sharp images and provide greater exposure control for CIF film's relatively low film speed and narrow exposure latitude. A 60% forward overlap would be sufficient for stereo viewing. Photo specifications should be established when contracting photo acquisition, and IR balance should be determined and adjusted with colour compensating filters, if necessary. The I :10,000-scaleCIR photographs would permit accurate depletion mapping and GIs updates that may be used throughout the management cycle. A combination of the analog 1:10,000 photographic information, pre-harvest knowledge and information about specific sites, and management objectives would be used to stratify cutover areas into units requiring specific silviculturaltreatments (minimum treatable size being situation dependent). Strata not readily categorized on the basis of the photographic information and prior knowledge may be visually verified by ground or air (S,, Fig. 2). The transfer of stereoscopically-typed information from analog photos to GIs would depend on end-user sophistication with image analysis. Typed photographs could be scanned, geometrically corrected, geo-referenced, and used as image backdrops in a GIs environment for subsequent data-base updating through on-screen digitizing. Geo-referencing may require GPS field data acquisitionunless existing basemaps provide sufficient detail. Higher levels of sophistication,including auto-interpretation, may assist and, in some cases, displace manual analog photo typing in the future. Hard copies of the 1:10,000photographs

JUILLETIAO~~T 1997, VOL. 73, NO. 4, THE FORESTRY CHRONICLE


CIR = colour infrared NC = normal colour S, = subjective survey

Fig. 2. Proposed application of remote sensing approach to forest vegetation management data and decision making.

would be a useful navigation tool for field activities and serve as a subsequent map base for recording actual proceedings. Silviculture contract compliance monitoring also may be facilitated. Decisions in the early monitoring phase (Fig. 2) would be facilitated by supplementary, leaf-on, 230- or 70-rnm, normal colour, 60% forward overlap aerial photographs at a scale of 1:5,000 (Fig. 4). Normal colour film is recommended during the growing season because species differences in CIR radiance are minimal at this time. Photo scale may be increased or decreased to permit cutover coverage with a single stereo

pair. Areas in obvious need of release, thinning, or supplemental planting may be identified on these photographs and mapped. Marginal areas may be visually verified on the ground or by air (S,, Fig. 2), unless circumstances dictate data collection, in which case large-scale photographs andlor ground sampling (So) may be used. This process would allow survey efforts to be focused on the marginal areas and permit the identification and mapping of specific areas requiring treatment. Similar supplementary photographs may be used between age 5 and 10, to aid decision-making during the verification phase (Fig. 2). When the achievement of management objectives cannot


469


Fig. 3. Example of a 1:20,000-scale,230-mm format, leaf-off, colour infrared (CIR) photograph recommended for decision-making in the establishmentphase (Fig. 2). Photo scale is small enough to provide complete coverage of one or more cutovers and large enough to provide adequate spatial detail for accurate depletion mapping and areal stratification for pre-establishment planning. CIR film enhances the visibility of advanced conifer regeneration, as well as the presence of wet areas, water courses, potential frost pockets, rock outcrops, excessive slash loads, and high-use wildlife and recreation areas, that are important for prescription formulation. In the example, the area beneath the lake has been recently cutover. Heavy to moderate slash loadings can be seen as light grey patches; individual pieces of larger debris are visible. Four wet areas show as dark grey or black, and patches of rock, bare soil, and grass show as light grey or white. Residual deciduous (grey) and conifer (pink to red) trees occur in varying densities across the cutover area and in the buffer zone around the lake. Conifer regeneration on an older cutover above and to the right of the lake appears bright red. Note that the original 230-mm format has been cropped to fit the page size. The original photo scale has been maintained. Photo credit: Imagesense Land Information Associates.

be verified from these photographs, verification with an objective survey may be necessary (S,), based on large-scale photographs andlor ground sampling.

470

Whenever large-scale photographs are used as the objective sampling tool, photoscale should be between 1:500 and 1:1,000, with base:height ratios sufficiently large (1:12 or greater) to provide

JUILLETIAO~T1997, VOL. 73, NO. 4, THE FORESTRY CHRONICLE

The Forestry Chronicle Downloaded from pubs.cif-ifc.org by CARLETON UNIV on 05/04/16 For personal use only. Fig. 4. Example of a 1:5,000-scale, 230-mm format, leaf-on, normal colour photograph recommended for decision making in the early monitoring or verification phases (Fig. 2). Complete photographic coverage of individual cutovers between age two and five, in combination with air or field reconnaissance, could be used to identify and map areas requiring thinning, supplemental planting, or release. Similar photographs taken toward the end of the regeneration period (between age five and 10) should serve to verify and document the achivement of management objectives. In the example, portions of a five-year old cutover (A) are in need of release, (B) have been manually released, and (C) have been released with herbicides. Small white dots mark the corners of 10 m x 10 m plots (one of which is shown on LSP in Fig 5). Note that the original 230-mm format has been cropped to fit the page size. The original photo scale has been maintained. Photo credit: Sure-Way Aerial.

adequate vertical exaggeration for precise measurement of tree height (Fig. 5). Photo measurement of variables such as tree height, crown area, percent cover, etc. can be corrected using

a stratified random sample with two-phase correction (Hall 1984; Pitt et al. 1996). Such corrections should be statistically sound (i.e. based on data from at least 30 paired ground and photo plots


471

- 6 = d

s??s 0Q-d

6.S

.S 'G o

,,

k$T a

.Sg.;$

==g : s;,$g 9%.;& 0

~

$i$j

""g .f .Y

-

8 ~3 2 $2, 6


.r(

0

'LE.zs

t;z E.g x % 3

.A

ks

g-g

%

o d M . 3 o 2.G g .+2Z a g s .

z&+.?

,%a 2

% * & d h

> 3 ~ z 0

Ba a E +m aE ng zxog +g

2 2,2O g. t~- 9"M a 8 5 m

mah ggy2

,s.2

0 0 O U S U *

z,

g s $ " 'g G b 0 ~ ,

, a s 2.5 04-0 m a

8.2 e 2

, 2* 33 %rn.a x.2 2.2 g

m

8

G F

0

m

3~

as'", E a g o 3

N

d s g 2 %

3 6:s $5

$ " a % E F, 3 e g, .'3 4 5,. X"6) ,.'.$-&a m "5 243

s

S E > ~ E 2*35 E 3 bb5 3 2

-,a 32

.;c 3 2,% " - 0

8.3%

0 s 0 0

&

03.3s2 -a, 0 E 0 2 2 . 9 ,G

gs

*

g.g:s: - .s s

" g* 3 $$ b)

zg&ss .: m

9

$@

Or, m s 0 0 shod

zgg&$

,'j, g. % 0, %Zt;SE 3

3,-no 0 ~

a, 2 % E 46 % a

~

x 8Y s$ v; - 0 0.5 m tall. In addition, variables such as biomass could be inferred from variables measured with the objective sampling process of the proposed approach. For example, the relationship between biomass and plant height and crown area, or stand height and percent cover, may be used to estimate biomass from large-scale photographs in a two-phase design. The advantage of this effort would be that biomass estimates may be made for large areas much more efficiently than would logistically feasible with a field sample alone. The disadvantage is that twophase equations must be developed for specific cover-types and sites. Information on the early, stand-level horizontal distribution of crop and non-crop plant species could be obtained from the 1:5,000 scale imagery taken between ages 2 to 5, or 5 to 10, depending on the stage of vegetation development. The strata defined for management prescriptions would likely form natural boundaries for further sampling, which could take place on the 1:5,000photographs andlor, depending on the need, largescale photographs and the ground. Manual photo interpretation methods could involve the use of high-density dot grids or comparison to a set of visual standards (Goba et al. 1982).With further research and the anticipated progression towards digital data, textural and pattern-recognition algorithms are possible for this purpose. Forest-level evaluations could be made by integrating information collected for individual strata across a management unit. If juvenile stand conditions are of interest, the 1:10,000 scale photographs taken at age 10+ could be used in a similar fashion. Vegetation changes through time could be monitored and documented at both the stand and forest levels via photographs at the suggested scales. The proposed approach initiates a program of continuous monitoring throughout the regeneration period that may be continued at 5-, lo-, or 15-year cycles. Detailed queries, including diversity; presence of rare and endangeredplants, and evaluations of small (0.5 m tall) are identifiable, as are contiguous patches of annual and perennial herbaceous species >1m2(e.g. raspberry, ferns, grasses, fireweed, etc.). Conifer species of similar form can be difficult to distinguish from one another (e.g. white spruce, black spruce, and balsam fir), particularly at sizes less than 1 m tall. On 1:5,000-scaleimagery, it may only be possible to separate species into broad morphological classes (i.e., tall shrub, low shrub, conifer, herbaceous, graminoid, etc.). Plant physiological status and soils information (type, moisture, and nutrient status) may be inferred from aerial photographs, but detailed ground assessments will be r e q d to quantify these variables, given current remote sensing technology.

Considerations for Implementation While our recommended approach does not offer a remotesensing alternative for all the data needed for FVM decision making, there are several important features worth noting. First, the approach focuses on stratification of harvested areas into FVM prescription units. A primary advantage of this is the reduction of costs associated with more efficient allocation of silvicultural expenses. While costs of the approach will vary with location, accessibility, project size, equipment availability, etc., typical photograph costs would be about $25 to $50 frame-l for 230-mm format and $10 to $20 frame-' for 70-mm format. At two frames per stereo pair, assuming one stereo pair per 50-ha cutover (230-mm format), total image costs for the regeneration period might be as high as (($50 + $50 + $50) x 2)/50 = $6.00 ha-'. If large-scale photo sampling is conducted, additional photo costs of $15 to $25 ha-l might be expected, depending on sample intensity. Given the high costs of planting ($700-$1100 ha-l) and release ($100 and $300 ha-l), photo costs may be offset largely by the identification of portions of cutovers that do not need either of these treatments. A second important cost-saving feature of this approach is that it employs objective sampling and data collection only where necessary. Such sampling is focused on areas that are difficult to classify on the basis of 1:5,000 scale photographs and prior knowledge of the site. This assumes that supplementary aerial photographs may serve as sufficientjustification and documentation of regeneration conditions under well-defined circumstances, and that actual data need onlv be collected when the situation warrants or when data are ne&led for other purposes (e.g. growth and yield). In this regard, any current legislative requirements for field surveys of all sites should be reviewed. The approach also provides relatively detailed monitoring of the entire forest throughout the regeneration period. The frequency of observation is such that remedial treatments may take place before trees become suppressed and suffer sigmficant growth loss. This contrasts with the status quo, where frequent subjective visits provide limited area coverage and infrequent objective visits are often too late for effective remedial action. Forest managers generally express the desire to purchase a finished data product and have staff free to concentrate on FVM decision making. Photo acquisition, rectification, interpretation, and even mensuration can be transparent to the end-


473


user if these tasks are contracted to one of the many aerial survey5 and consulting firms throughout Canada who have the proper equipment and expertise to do this work. After review of all of the latest and emerging remote sensing technologies, the workshop recommendation that aerial photographs be used for the acquisition of FVM data yields an obvious and perplexing question. Since aerial photography has been in widespread use since the 1950s, why have foresters not yet applied it in this context? Perhaps ths answer lies in the fact that foresters have not yet needed to. Until recently, forestry concerns have focused primarily on fiber extraction and minimizing the costs of reforestation. When combined with the relatively low cost of aerial herbicide application, data collection for FVM decision making has been viewed accordingly as a low priority, only being used when necessary to satisfy regulatory or contractual requirements to place harvested areas back into the inventory. Aerial photogr~phy,and remote sensing in general, has perhaps been thought to be an unnecessary cost and complication for what could be accomplishedby inexpensive seasonal labor and quick visual surveys. As FVM moves into the emerging context of forest ecosystem management (Wagner 1994), however, remote sensing will become an increasingly important tool for gathering information necessary for malung, supporting, and documenting management decisions. For example, the recommended photographic sequence could provide a permanent record of management activities that may be referenced later. Information contained in these photographic records also may be used to address other forest resource concerns as they arise (e.g. wildlife habitat, landscape issues, watershed management, recreation, land-use planning, etc.). Moreover, the same methodology may be applied to monitor and measure mature forests prior to cutting (Lyons 1966,1967) and to quanbfy residual wood volumes after cutting (Kirby and Hall 1979). Aerial photographs, or a combination of visual reconnaissance and a low-cost sensor (i.e. airborne video), also may be used to detect and map insect damage (Croft et al. 1982; Myhre et al. 1987).

Research and Development Directions As remote sensing technology evolves, our recommended approach may be adapted to take advantage of new efficiencies and/or new types of remotely sensed data. Perhaps the most imminent system improvement will be the incorporation of DFCs and the acquisition of digital image data. This advance should bring to the approach many of the advantages of digital data acquisition, including real-time image viewing, wider dynamic range, computerized storage and retrieval of images, and computer-based image classification and interpretation. To realize these advantages, however, several research challenges must be overcome. Factors such as mixed species complexes, vertical structure, topography, time of day and year, vegetation phenology, weather conditions, moisture, stress, etc., complicate computer-based classification procedures by increasing spectral variance and reducing the spectral separability of classes. Factors such as bi-directional reflectance (e.g. sunlit vs. shaded crowns) and other colour variations within and between images (i.e. attributable to the sensor, atmos5 ~ omore r information,contact the InterdepartmentalCommittee on Air Surveys, Ottawa.

474

phere, illumination, view angle, surface reflectance, etc.) necessitate radiometric image correction and make standardized classification schemes difficult (Leckie et al. 1995). The identification and separation of plant species based on "spectral signatures" need to be researched, particularly for immature plants in a FVM setting. Auto-interpretationtechniques based on crown shape are currently being developed for mature forest trees (Franz et al. 1991; Gougeon 1992,1995; Pollock 1994a,b).Although results to date are promising, complete success may not be achieved until additional feature processing algorithms, based on texture and branching structure, are included. Continuation of this work is required for detection and mapping of smaller, immature trees in FVM applications. Colour, tone, texture, and shape, are important characteristics used in the interpretation process. Mimicking human knowledge, experience, and an ability to integrate these characteristics, present major challenges to automation of the interpretation process. Recognizing that the computer can play an important role in managing and presenting data, a logical approach may be to provide tools to assist manual interpretation. These tools may initially include for example: spatially-referenced data bases and historical records that can be quickly and easily retrieved on the computer screen during interpretation, screen-accessed (pop-down) interpretation keys, knowledge-based rules (decision-support systems) that incorporate site factors and silvicultural history to offer "expected" attributes. As satisfactory classification, auto-interpretation and measurement routines are completed, they may best be applied in a supervised fashon, as additions to this suite of tools. Since aerial photographs will likely play an important role in FVM for the foreseeable future, photo specifications should be defined for 70-mm and 230-mm format cameras that meet informationneeds for specific applications (e.g. Fent et al. 1995; Hall and Fent 1996). New and evolving sensor systems also should be researched for FVM, especially for features not easily derived or monitored by existing tools. The low cost and easy access of video, for example, makes this sensor attractive for collection of data supplementary to the main sensor. Electronically shuttered video cameras and gyro-stabilized camera mounts are now available that produce higher quality imagery with reduced image motion and vibration problems. The imminent introduction of High Definition TV (HDTV) also promises to further increase video resolution and small, portable video systems, that produce GPS encoded images, have recently entered the market. As these and other advancements take place, the potential for airborne video to be utilized in FVM should be re-evaluated periodically. Similarly, there is a need to explore the extent to which laser technology can be used to augment the main sensor in FVM applications. While it may not be reasonable to expect a sensor of the present or near future to quanbfy understory vegetation, lidar may be used to detect its presence or absence. Further, the potential for biomass and leaf-area-index (LAI) to be inferred from data generated by this sensor should be determined. The all-weather capability of radar may also be used to advantage for some FVM applications. Both airborne and spaceborne

JUILLETIAO~JT

1997, VOL. 73, NO. 4, THE FORESTRY CHRONICLE


radar should be evaluated,particularly in terms of fulfilling landscape-level data inquiries and monitoring changes through time. For regenerating forest sites, radar backscatter is influenced by the density and crown closure of ground vegetation; stands with higher density vegetation generally having higher backscatter. There have been numerous studies showing a relationship between radar response and biomass (Wu 1987; LeToan et al. 1992; Ranson and Sum 1994). These relationships are particularly strong for stands under 15 years old and are better for the longer wavelength radars (i.e. L and P band) and cross polarization. Most tests, however, have been conducted over uniform pine plantations. Although further research is needed, it may be possible to determine broad regeneration levels, and multiple-year sequences of imagery may provide information on regeneration levels for monitoring purposes. New high-resolution optical satellite images (1- to 3-m resolution) may offer similar possibilities. Finally, this problem analysis has dealt entirely with FVM problems as they relate to even-aged management practices on clearcuts > 10 ha. Forest managers are under increasing pressure to reduce clearcut size (and in some cases, avoid the practice entirely). Partial cuts and uneven-aged silvicultural practices present new FVM challenges that are just beginning to be addressed by forest researchers. Along with practices, the role that remote sensing may play in decision making should be part of this work.

Implementation and Research Coordination

We recommended that a network of remote sensing experts, FVM specialists, and forest managers be assembled to implement the proposed approach and c&rdinate futureresearch a&ed at its improvement. The first task of this network would be to design and execute a pilot study for the proposed approach on a typical provincial forest license area. Efficacy and full costbenefit evaluation could be completed within a 5-year period, with design changes and final recommendations being available to the general forestry community by year 5. At the same time, this network might coordinate research efforts focusing on direct system improvements such as the following:

1) Preparation for the imminent digital era, including: a) discovery and definition of spectral signatures for young crop and non-crop forest plants, b) adaptation of auto-int-tation techniques based on crown shape and spectral reflectance to young crop and non-crop plants, c) further development of auto-interpretive techniques based on texture and pattern-recognition, d) continuation of the development of techniques for dealing with radiometric distortion (within and between images), and e) integration of these techniques in a computer-based suite of tools that will enhance interpreter consistency, objectivity and efficiency. 2) Streamlining the integration of image capture and GPS and promotion of strategies for better and faster photographGIs integration. 3) Refinement of filmselection, exposure, processing parameten, timing, and photo scales for specific FVM applications.

4) Investigation of relationships between image measures

and physical measures, including LA1 and biomass. 5) Investigation of remote sensing applicationsin partial cutting systems. Additionally, this network should participate in the development and evaluation of new sensors with potential for FVM applications. It is recommended that immediate focus be on: 1) DFCs as replacements for conventional cameras within the framework of the adopted system.

2) h r n e video as a low-cost replacement for aaial photographs when high levels of spatial resolution, detailed measurements, and hard-copy images are not required. 3) Laser technology for complementing aerial photographs, DFCs, or low-cost video with detailed vegetation height information or for understory detection and quantification (biomass, LAI, etc.). 4) Multitemporal air- and spaceborne radar and its possible integration with optical data for monitoring vegetation patterns and changes over time at the landscape level, depletion mapping, broad stratification,and its interferometriccapabilities (e.g. inferences on vegetation height and biomass). 5 ) New high-resolution satellite imagery for depletion mapping and broad stratification.

In light of new forest-management legislation and decision support tools that will place increased data-demands on forest managers, these recommendations hopefully offer a proactive approach to streamlining the data acquisition process in FVM.

Acknowledgments We wish to thank the Ontario Ministry of Natural Resources (VegetationManagement Alternatives Program), Canadian Forest Service (Sault Ste. Marie), Abitibi Price Inc. (Thunder Bay), and the Canadian Institute of Forestry for the financial resources necessary to complete this problem analysis. The time and effort of the many workshop participants, many of whom provided thoughtful and constructive suggestions for improving this manuscript, are also gratefully acknowledged.We are especially indebted to Heather McLeod for her assistance with the organization of the workshop, Mike Irvine for his support (both of Ontario Ministry of Natural Resources), and Andrij Obarymskyj (Canadian Forest Service) for his assistance with graphics.

References Ahern, F.J., W.J. Bennett and E.G. Kettela. 1986. An initial evaluation of two digital airborne imagers for surveying spruce budworm defoliation. Photograrnm. Eng. Remote Sensing 52: 1647-1654. Aldred, A.H. and G.M. Bonner. 1985. Applications of airborne lasers to forest surveys. Petawawa National Forestry Institute, Chalk River, Ontario. Inf. Rep. PI-X-51.62~. Anger, C.D., S. Mah and S.K. Babey. 1994. Technological enhancements to the compact airborne spectographic Imager (CASI). Proc. First International Airborne Remote Sensing Conf. and Exhibition. Sept. 11-15, Strasbourg, France. Vol. 2, pp. 205-213. Babey, S.K. and C.D. Anger. 1993. Compact airborne spectographic imager (CASI): a progress review. Proc. SPIE, Vol. 1937. pp. 152-163. Ball, W.J. and V.S. Kolabinski. 1979. An aerial reconnaissance of softwood regeneration on mixedwood sites in Saskatchewan.Northern Forest Research Centre, Edmonton, Alberta. Inf. Rep. NOR-X-216. 14 p.


475


Brand, D.G., D.G. Leckie, and E.E. Cloney. 1991. Forest regeneration surveys: Design, data collection, and analysis. For. Chron. 67: 649-657. Burton, PJ. 1993. Some limitations inherent to static indices of plant competition. Can. J. For. Res. 23: 2141-2152. Canadian Council of Forest Ministers. 1996. Compendium of Canadian Forestry Statistics 1995, National Forest Database. Natural Resources Canada, Communications Branch, Ottawa, Ontario, KIA OE4.152 p. Caylor, J.A. 1989. Film, camera, and mission considerations to reduce image motion effects on photos. Proc. 12th Biennial Workshop on Color Aerial Photography and Videography in the Plant Sciences. Reno, Nevada. American Society of Photogramrnetry and Remote Sensing. pp. 46-63. Croft, F.C., R.C. Heller and D.A. Hamilton, Jr. 1982. How to interpret tree mortality on large-scale color aerial photographs.USDA Forest Service, General Technical Report INT-24. Intermountain Forest and Range Experiment Station, Ogden, UT. 13 p. Evans, D.L. 1994. Forest cover from Landsat thematic mapper data for use in the Catahoula Ranger District GIs. USDA, Forest Service, South For. Exp. Sta. Gen. Tech. Rep. SO-99. 14 p. Fent, L., RJ. Hall and R.K. Nesby. 1995. Aerial films for forest inventory: optimizing film parameters. Photograrnm.Eng. Remote Sensing 61(3): 281-289. Fleming, J. 1978. Exploiting the variability of aerochrome infrared film. Photogramm. Eng. Remote Sensing 44(5): 601-605. Flowerday, A.D. 1982. Low-altitude infrared photography as a crop management tool. pp. 141-158. In Remote Sensing for Resource Management. C.J. Johannsen and J.L. Sanders, (eds.). Soil Conservation Society of America, Ankeny, Iowa. Franklin, S.E. 1994. Discrimination of subalpine forest species and canopy density using digital CASI, SPOT, PLA and Landsat TM data. Photogramm. Eng. Remote Sensing. 60(10): 1233-1241. Franklin, S.E., RT. Gillespie, B.D. Titus and D.B. Pike. 1994.Aerial and satellite sensor detection of Kalmia angustifolia at forest regeneration sites in central Newfoundland. Int. J. Remote Sensing 15(13): 2553-2557. Franz, E., M.R. Gebhardt and K.B. Unklesbay. 1991. Shape description of completely visible and partially occluded leaves for identifying plants in digital images. Trans. ASAE 34(2): 673-681. Fritz, L.W. 1996. The era of commercial Earth observation satellites. Photogramm. Eng. Remote Sensing 62(1): 39-45. Gi,M.D. and D.G. Leckie. 19%. Forest inventory update in Canada. For. Chron. 72: 138-156. Goba, N., S. Pala and J. Narraway. 1982. An instruction manual on the assessment of regeneration success by aerial survey. Ontario Centre for Remote Sensing, Ontario Ministry of Natural Resources, Toronto, ON. 5 1 p. Gougeon, F.A. 1992. Individual tree identification from high resolution MEIS images. pp. 117-128. In Proceedings of the International Forum on Airborne Multispectral Scanning for Forestry and Mapping (with emphasis on MEIS). D.G. Leckie and M.D. Gillis. Val-Morin, Quebec, April 13-16,1992. Petawawa National Forestry Institute, Chalk River, Ontario. Inf. Rep. PI-X-113. Gougeon, F.A. 1995. A crown-following approach to the automatic delineation of individual tree crowns in high spatial resolution aerial images. Can. J. Remote Sensing 21(3): 274-284. Graham, R. and R.E. Read. 1986. Manual of aerial photography. Focal Press, London. Hall, R.J. 1984. Use of large-scale aerial photographs in regeneration assessments. Can. For. Serv. North. For. Res. Cent., Edmonton, Alberta. Inf. Rep. NOR-X-264.31 p. Hall, RJ. and A.H. Aldred. 1992. Forest regeneration appraisal with large-scale aerial photographs. For. Chron. 68(1): 142-150. Hall, R.J. and L. Fent. 1996. A comparison of four black-andwhite aerial filmsand their texture on interpreting images of forest species composition. Proc. 18th Can. Symp. on Remote Sensing, Vancouver,

476

BC. pp. 118-122. Hosking, G. 1994. Airborne video for resource monitoring - a new technology takes off. What's new in Forest Research, No. 233. New Zealand Forest Research Institute, Rotorua, NZ. Hosking, G.P., J.G. Firth, R.K. Brownlie and W.B. Shaw. 1992. Airborne video -new toy or new tool? A New Zealand perspective. Proc. Resource Technology 92 - Information Technology for Environmental Management. November 1&18,1992, Taipei, Taiwan. Hunter and Associates. 1983. Free-to-grow evaluation compiled from large scale photo sampling. Prepared for the Ontario Ministry of Natural Resources, Cochrane District. 2695 North Sheridan Way, Ste. 120, Mississauga, ON L5K 2N6. 11 p. Jacobs, D.M., D.L. Evans and J.C. Ritchie. 1993. Laser profiler and aerial video data for forest assessments. Proc. ACSMIASPRS Annual Convention, New Orleans, LA, Feb. 15-18, American Society of Photogramrnetry and Remote Sensing, Vol. 11, pp. 135-142. King, DJ. 1993. Digital frame cameras: Characteristicsrelated to remote sensing. pp. 31-35. In Rweedings of the International F o m on Airborne Multispectral Scanning for Forestry and Mapping (with emphasis on MEIS). D.G. Leckie and M.D. Gillis, (eds.). Petawawa National Forestry Institute, Chalk River, Ontario. Inf. Rep. PI-X-113. King, D.J. 1995. Airborne multispectral digital camera and vidw sensors: a critical review of system designs and applications. Can. J. Remote Sensing, Special Issue on Aerial Optical Remote Sensing 21(3): 245-273. King, D.J. and J. Vlcek. 1990. Development of a multispectral video system and its application in forestry. Can. J. Remote Sensing 16(1): 15-22. Kirby, C.L. 1980. A camera and interpretationsystem for assessment of forest regeneration. Can. For. Serv. North. For. Res. Cent., Edmonton, Alberta. Inf. Rep. NOR-X-221. 8 p. Kirby, C.L. and R.J. Hall. 1979. The estimation of logging residues on large-scale aerial photographs. Proc. Int. Fire Management Workshop, Environ. Can., Can. For. Sew., North For. Res. Cent., Edmonton, Alberta. Inf. Rep. NOR-X-215. pp. 57-62. Kodak Aerial Systems. 1995. Kodak Aerochrome HS film SO359. Aerial data. AS-207. Kodak Aerial Systems. 1996. High resolution mapping film. Aerial advantage. 9(1): 1. Kneppeck, I.D. and F.J. Ahern. 1987. Evaluation of a multispectral linear array sensor for assessing juvenile stand conditions. Proc. Twenty-First International Symposium on Remote Sensing of Environment, Ann Arbor, Michigan, October 26-30, 1987. pp. 955-969. Kneppeck, I.D. and F.J. Ahern. 1989. Stratification of a regenerating burned forest in Alberta using thematic mapper and C-SAR images. Proc. 12thCanadian Symposiumon Remote Sensing, Vancouver, 1989. pp. 1391-1396. Leckie, D.G. 1993. Application of airborne multispectrd scanning to forest inventory mapping. pp. 86-93. In Proc. International Forum on Airborne Multispectral Scanning for Forestry Mapping (with emphasis on MEIS). D.G. Leckie and M.D. Gillis (eds.). Petawawa National Forestry Institute, Chalk River, Ontario. Inf. Rep. PI-X-113. Leckie, D.G., M.D. Gillis and S.P. Joyce. 1992. A forest monitoring system based on satellite imagery. In Proc. 15th Canadian Symposium on Remote Sensing. J.K. Hornsby, D.J. King and N.A. Prout (eds.). June 1 4 , 1992, Toronto, ON. Leckie, D.G., J. Beaubien, J.R Gibson, N.T. O'Niell, T. Piekutowski and S.P. Joyce. 1995. Data processing and analysis for MIFUCAM: a trial of MEIS imagery for forest inventory mapping. Can. J. Remote Sensing 21(3): 337-356. LeToan, T., A. Beaudoin, J. Riom and D. Guyon. 1992. Relating forest biomass to SAR data. IEEEi Trans. Geoscienceand Remote Sensing 30(2): 4 0 3 4 1 1. Light, D.L. 19%. Film cameras or digital sensors? The challenge ahead for digital imaging. Photogramm. Eng. Remote Sensing 62(3): 285-291.

JUILLETIAO~TT1997, VOL. 73, NO. 4, THE FORESTRY CHRONICLE

_


Lillesand, T.M. and R.W. Kiefer. 1994. Remote sensing and image interpretation. 3rd edition. John Wiley & Sons, Inc., New York. 750 p. Lowe, J.L., B.P. Oswald, T.L. Coleman, W. Tadesse, J.H. Everitt, D.E. Escobar and M.R. Davis. 1995. Comparison of conventional ground sampling and remote sensing techniques for mapping of forest vegetation. Proc. 15th Biennial Workshop on Color Photography and Videography in Resource Assessment. Am. Soc. Photogramm. and Rem. Sens., Terre Haute, IN. May 2-3. pp. 31-52. Lyons, E.H. 1966. Fixed-airbase 70-mm photography, a new tool for forest sampling. For. Chron. 42: 420-43 1. Lyons, E.H. 1967. Forest sampling with 70-mm fixed-base photography from helicopters. Photogrammetria 22: 213-23 1. McCoU, W.D., R.A. Neville and S.M. T i 1983. Multidetector Electrooptical Imaging Scanner MEIS 11. Proc. 8th Canadian Symp. Remote Sensing, Montreal, Quebec, May 3-6. pp. 7 1-79. Mussio, L. and D. L. Light. 1995. ISPRS Commission I: sensors, platforms, and imagery symposium. Photogramm. Eng. Remote Sensing 61(11): 1339-1334. Myhre, R.J., A.S. Munson, D.E. Meisner and S. Dewhurst. 1987. Assessment of a color infrared aerial video system for forest insect detection and evaluation. Proc. 1lth Biennial Workshop on Color Aerial Photography and Videography in the Plant Sciences. Am. Soc. Photogramm. and Remote Sensing, Bethesda, Maryland. pp. 244251. Myhre, R.J. and B. Silvey. 1992. An airborne video system developed within forest pest management - status and activities. pp. 291-300. In Proc. of Fourth Forest Service Remote Sensing Applications Conference, Remote Sensing and Natural Resource Management, Orlando, FL, April 1992. Nelson, H.A. 1977. Assessment of forest plantations from low altitude aerial photography. In Proc, of the I lth International Symposium on Remote Sensing of the Environment, 25-29 Apr. 1977, Ann Arbor, MI. Environmental Research Institute of Michigan (ed.). Center for Remote Sensing Information and Analysis, Ann Arbor, MI. 2: 1515-1522. Neville, R.A., R. Marois, J.W. Schwartz and S.M. Till. 1992. Wide-angle high-resolution line-imager prototype flight test results. Applied Optics 3 l(18): 3463-3472. OCRS. 1989. Forest regeneration assessment from high-resolution satellite data. COFRDA Project No. 02SE.OIK35-6-0076, OCRS. Paradine, D., P. Howarth and P. Treib. 1995. Empirical identification of forest ecological parameters from remote sensing imagery of northwestern Ontario. Proc. 17th Canadian Symp. Remote Sensing. June 13-1 5, Saskatoon, Saskatchewan. pp. 61 1-616. Pitt, D.G. 1994. Use of large-scale 35-mm aerial photographs for assessment of vegetation management research plots. Ph.D. dissertation, School of Forestry, Auburn University, AL. 247 p. Pitt, D.G. and G.R. Glover. 1993. Large-scale 35-mm aerial photographs for assessment of vegetation-managementresearch plots in eastern Canada. Can. J. For. Res. 23: 2159-2169. Pitt, D.G. and G.R Glover. 1996. Measurements of woody plant attributes from large-scale aerial photographs. N.Z. J. For. Sci. 26('/2): 5S73. Pitt, D.G., G.R Glover and R.A. Jones. 1996. Two-phase sampling of woody and herbaceous plant communities using large-scale aerial photographs. Can. J. For. Res. 26(4): 509-524.

Pollock, R.J. 1994a. A model-based approach to automatically locating tree crowns in high spatial resolution images. Proc. Conference on Image and Signal Processing for Remote Sensing, Sept. 1994,Rome. EOSISPIE, Vol. 23 15. Pollock, R.J. 199413. A model-based approach to automatically locating individual tree crowns in high-resolution images of forest canopies. Proc. First International Airborne Remote Sensing Conference and Exhibition, Strasbourg, France, Sept. 11-1 5, 1994. Poole, P. 1994. Appropriate geomatics technology for local earth observation. Yale School of Forestry and Environmental Studies Bulletin Series, No. 98. pp. 156-163. Ranson, K.J. and G. Sun. 1994. Mapping biomass of a northern forest ecosystem using multifrequency SAR data. IEEE Trans. Geoscience and Remote Sensing 32(2): 38S396. Ritchie, J.C., D.L. Evans, D. Jacobs, J.H. Everitt and M.A. Weltz. 1993. Measuring canopy structure with an airborne laser altimeter. Trans. ASAE 36(4): 123S1238. Schreier, H., J. Lougheed, C. Tucker and D. Leckie. 1985. Automated measurements of terrain reflection and height variations using an airborne infrared laser system. Int. J. Remote Sensing 6(1): 101-1 13. Slaymaker,D.M., K.M.L. Jones, C.R Griffin, and J.T. Finn. 1995. Mapping deciduous forests in New England using aerial videography and multi-temporal Landsat TM imagery. Dept. of Forestry and Wildlife Management, University of Massachusetts, Amherst, MA, 15 p. Spencer, R.D. and R.J. Hall. 1988. Canadian large-scale aerial photographic systems. Photogramm. Eng. Remote Sensing 54(4): 475482. Tappeiner, J.C. I1 and R.G. Wagner. 1987. Principles of silvicultural prescriptions for vegetation management. pp. 399-429. In Vegetation management for conifer production. J.D. Walstad and P.J. Kuch (eds.). John Wiley and Sons Inc., New York. Verreault, R , G.H. Lemieux and S. McLaughlin. 1993. La videographie aerienne a multispectrale (VAM) appliquee au monitoring de la regeneration forestiere. Proc. 16th Canadian Symposium on Remote Sensing, Sherbrooke, Quebec, June 7-1 0, Can. Rem. Sens. Soc., Ottawa, ON. p. 6 4 7 6 5 1. Wagner, R.G. 1994. Toward integrated forest vegetation management. J. For. 92(11): 26-30. Wagner, R.G. and S.R. Radosevich. 1991. Neighborhood predictors of interspecific competition in young Douglas-fir plantations. Can. J. For. Res. 21: 821-828. Warner, W.S., R.W. Graham and R.E. Read. 1996. Small format aerial photography. American Society for Photogrammetry and Remote Sensing, Bethesda, MD. 348 p. Weltz, M., J.C. Ritchie and H.D. Fox. 1994. Caparison of laser and field measurements of vegetation heights and canopy cover. Water Resourc. Res. 30(5): 131 1-1319. Wu, S.T. 1987. Potential application of multipolarization SAR for pineplantation biomass estimation. IEEE TRans. Geoscience and Remote Sensing. Ge-25(3): 403-409. Yuan, X., D.J. King and J. Vlcek. 1991. Sugar maple decline assessment based on spectral and textural analysis of multispectral aerial videography. Remote Sensing of Environment 37(1): 47-54.


477