Forest Inventory Maps Research Articles

A model-based performance test for forest classifiers on remote-sensing imagery

Ambiguity between forest types on remote-sensing imagery is a major cause of errors found in accuracy assessments of forest inventory maps. This paper presents a methodology, based on forest plot inventory, ground measurements and simulated imagery, for systematically quantifying these ambiguities in the sense of the minimum distance (MD), maximum likelihood (ML), and frequency-based (FB) classifiers. The method is tested with multi-spectral IKONOS images acquired on areas containing six major communities (oak, pine, fir, primary and secondary high tropical forests, and avocado plantation) of the National Forest Inventory (NFI) map in Mexico. A structural record of the canopy and optical measurements (leaf area index and soil reflectance) were performed on one plot of each class. Intra-class signal variation was modelled using the Discrete Anisotropic Radiative Transfer (DART) simulator of remote-sensing images. Atmospheric conditions were inferred from ground measurements on reference surfaces and leaf optical properties of each forest type were derived from the IKONOS forest signal. Next, all forest types were simulated, using a common environmental configuration, in order to quantify similarity among all forest types, according to MD, ML and FB classifiers. Classes were considered ambiguous when their dissimilarity was smaller than intra-class signal variation. DART proved useful in approximating the pixel value distribution and the ambiguity pattern measured on real forest imagery. In the case study, the oak forest and the secondary tropical forest were both distinguishable from all other classes using an MD classifier in a 25 m window size, whereas pine and primary tropical forests were ambiguous with three other classes using MD. By contrast, only two pairs of classes were found ambiguous for the ML classifier and only one for the FB classifier in that same window size. The avocado plantation was confounded with the primary tropical forest for all classifiers, presumably because the reflectance of both types of forest is governed by a deep canopy and a similar shadow area. We confronted the results of this study with the confusion matrix from the accuracy assessment of the NFI map. An asset of this model-based method is its applicability to a variety of sensor types, eco-zones and class definitions.

Boreal forest mapping at the BOREAS study area using seasonal optical indices sensitive to plant pigment content

In this study, forest vegetation classification was investigated based on seasonal variation of pigments as inferred from visible and near-infrared spectral bands. This analysis was carried out on data collected over the southern study area of the Boreal Ecosystem-Atmosphere Study (BOREAS) with the Compact Airborne Spectrographic Imager (casi) in May and July 1994 and with the medium-resolution imaging spectrometer (MERIS) in May and August 2003. Three modified normalized difference vegetation indices (mNDVIs), which are sensitive to relative proportions among pigments and pigment content, and a red-edge spectral parameter, the wavelength at the reflectance minimum (λ0), were used. Accuracy assessments of the derived land cover maps were performed using a forest inventory map provided by the Saskatchewan Environment and Resource Management Forestry Branch Inventory Unit (SERM–FBIU). The forest vegetation classification using seasonal optical indices (mNDVIs and λ0), as derived from the casi data in May and July, shows an overall accuracy of 84% for all cover types identified, namely dry conifer, wet conifer, mixed stands, aspen, fen, and the disturbed and regenerated area. The classification results also demonstrate that classification using reflectance parameters sensitive to pigment absorption features outperforms that using the reflectance itself. In addition, the classification using seasonal information is better than that using information obtained from a single date, and the spatial patterns were consistent with those achieved using multidate MERIS imagery. The forest vegetation classification using seasonal changes in optical indices (mNDVIs and λ0) derived from the MERIS imagery in May and August revealed a reasonably high overall classification accuracy (72%) for all vegetation cover types identified, namely conifer, mixed stands, and fen.

HABITAT CLASSIFICATION MODELING WITH INCOMPLETE DATA: PUSHING THE HABITAT ENVELOPE

Habitat classification models (HCMs) are invaluable tools for species conservation, land-use planning, reserve design, and metapopulation assessments, particularly at broad spatial scales. However, species occurrence data are often lacking and typically limited to presence points at broad scales. This lack of absence data precludes the use of many statistical techniques for HCMs. One option is to generate pseudo-absence points so that the many available statistical modeling tools can bb used. Traditional techniques generate pseudo-absence points at random across broadly defined species ranges, often failing to include biological knowledge concerning the species-habitat relationship. We incorporated biological knowledge of the species-habitat relationship into pseudo-absence points by creating habitat envelopes that constrain the region from which points were randomly selected. We define a habitat envelope as an ecological representation of a species, or species feature's (e.g., nest) observed distribution (i.e., realized niche) based on a single attribute, or the spatial intersection of multiple attributes. We created HCMs for Northern Goshawk (Accipiter gentilis atricapillus) nest habitat during the breeding season across Utah forests with extant nest presence points and ecologically based pseudo-absence points using logistic regression. Predictor variables were derived from 30-m USDA Landfire and 250-m Forest Inventory and Analysis (FIA) map products. These habitat-envelope-based models were then compared to null envelope models which use traditional practices for generating pseudo-absences. Models were assessed for fit and predictive capability using metrics such as kappa, threshold-independent receiver operating characteristic (ROC) plots, adjusted deviance (D(adj)2), and cross-validation, and were also assessed for ecological relevance. For all cases, habitat envelope-based models outperformed null envelope models and were more ecologically relevant, suggesting that incorporating biological knowledge into pseudo-absence point generation is a powerful tool for species habitat assessments. Furthermore, given some a priori knowledge of the species-habitat relationship, ecologically based pseudo-absence points can be applied to any species, ecosystem, data resolution, and spatial extent.

Accuracy of forest inventory mapping: Some implications for boreal forest management

We tested the accuracy of forest resource inventory (FRI) mapping by measuring tree species composition and density on 129 forest stands in two boreal forest regions in Ontario, Canada. For each stand we used chi-square analysis to determine significant differences between interpreted and ground data and then tallied differences to estimate percent error by stand type. To determine whether or not errors were compensatory over large areas, we combined stands and ground results within larger areas of 1, 2, and 5 km radii. We used our corrected information in the contemporary Ontario forest modelling process to compare long-term, large-scale effect of errors for model results from the original FRI. For this model, we compared available harvest volumes for important species and habitat availability for six wildlife indicator species. We observed that 83 of 129 stands were incorrectly classified by species composition. Approximately 30% were also misclassified by broad forest categories of conifer, mixed or deciduous. Common boreal species, including jack pine, black spruce and trembling aspen were incorrectly classified in about half of the cases. Rate of misclassification of species among forest types was inconsistent. Errors were not compensatory across larger areas when stands were combined. Our model indicated that resulting errors in available harvest volumes were compensatory in the jack pine-black spruce-dominated forests, but not in the black spruce-mixedwoods types, where 10–20% less softwood fibre was available for harvest than predicted from the original FRI. Similarly, less poplar was predicted over 30–50 years in all management units. These errors could potentially have financial implications for forest companies and for forest management. Errors in tree species classification altered preferred habitat availability for some wildlife species but not sufficiently to suggest lack of forest sustainability. The implications for researchers seeking specific stand types for sampling is that they should expect a 30–60% error rate, depending on classification, and select additional stands to ground-truth accordingly. The study provides clear implications for any jurisdiction using aerial photograph interpretation in forest management.

Estimating areal means and variances of forest attributes using the k-Nearest Neighbors technique and satellite imagery

The k-Nearest Neighbor ( k-NN) technique has become extremely popular for a variety of forest inventory mapping and estimation applications. Much of this popularity may be attributed to the non-parametric, multivariate features of the technique, its intuitiveness, and its ease of use. When used with satellite imagery and forest inventory plot data, the technique has been shown to produce useful estimates of many forest attributes including forest/non-forest, volume, and basal area. However, variance estimators for quantifying the uncertainty of means or sums of k-NN pixel-level predictions for areas of interest (AOI) consisting of multiple pixels have not been reported. The primary objectives of the study were to derive variance estimators for AOI estimates obtained from k-NN predictions and to compare precision estimates resulting from different approaches to k-NN prediction and different interpretations of those predictions. The approaches were illustrated by estimating proportion forest area, tree volume per unit area, tree basal area per unit area, and tree density per unit area for 10-km AOIs. Estimates obtained using k-NN approaches and traditional inventory approaches were compared and found to be similar. Further, variance estimates based on different interpretations of k-NN predictions were similar. The results facilitate small area estimation and simultaneous and consistent mapping and estimation of multiple forest attributes.

The spatial and temporal distribution of fires on Sakhalin Island, Russia

In the 1990s, catastrophic fires affected ~8 million ha of forest lands in the Russian Far East, including forests of Sakhalin Island. A study that correlated the spatial distribution of burned area and topographic features (elevation, slope, aspect) was carried out for Sakhalin Island. Burned area information derived from forest inventory maps (1935 to 1990) and satellite imagery (1998) was digitised and entered into a Geographic Information System. The burned area locations were correlated with topographic information; the normalisation procedure allows for analysis of the dependence of the fire scars on landscape features. The analyses show that fires occur primarily on the eastern, south- and north-eastern facing areas; >90% of fires occur at elevations lower than 300 m, and >95% occur on slopes <10 degrees. For the period 1935 to 1998, ~54% of the Sakhalin Island forest land territory was burned. From the total area of fire scars, formed from 1935 to 1998, 90.5% occurred owing to single fires, 8.6% of fire scars were the result of burning by two fires, 0.9% of fire scars were from three fires, and 0.03% from four fires. A fire return interval for the study region is ~120 years.

Physically based inversion modeling for unsupervised cluster labeling, independent forest classification, and LAI estimation using MFM-5-Scale

Unsupervised clustering is important for regional- to national-scale forest inventories where supervised training data are impractical or unavailable. However, labeling clusters in terms of land-cover classes can be labour intensive and problematic, and clustering and related methods do not provide biophysical-structural information (BSI). Canopy reflectance models such as 5-Scale are powerful forest remote sensing tools; however, 5-Scale can only be run in forward mode and is not invertible to obtain the required forest information. This problem was solved using multiple-forward-mode (MFM) coupled with 5-Scale to enable MFM-5-Scale inversion of land cover and BSI using a look-up table (MFM-LUT) approach that matches satellite image reflectance values with modeled reflectance values that have associated land cover and BSI, such as density, leaf area index (LAI), and crown dimensions, as well as subpixel-scale component fractions. MFM requires no training data or a priori BSI and can optionally be stratified (generalized) by species, structural, hierarchical, mixed forest, and other class definitions. In this paper, MFM-5-Scale was used with Landsat thematic mapper (TM) imagery at the Boreal Ecosystem-Atmosphere Study (BOREAS) southern study area (SSA) modeling subarea (MSA) in Saskatchewan, Canada. MFM-5-Scale was used to label unsupervised cluster sets (n = 17 and 97) from a previous land-cover classification by progressive generalization (CPG), with the best results obtained from independent, stand-alone MFM classification (87%, 76%, and 71% for the three hierarchies of 16 forest type, species, and density classes) validated against the provincial (SERM) forest inventory map and also compared with a standard maximum likelihood (ML) classification. Further, MFM-5-Scale estimated LAI at 24 BOREAS plots within ±0.57 LAI compared with ground-based tracing radiation and architecture of canopies (TRAC) LAI validation data. BSI is not provided by CPG clustering or ML. Based on this and other studies, we conclude that MFM provides an inversion modeling context for sophisticated forest radiative transfer models to retrieve a higher level of land cover and BSI, with detailed LUTs providing a rich set of forest information suitable for query, analysis, and follow-on simulation studies. These methods can augment existing regional- to national-scale remote sensing based inventories by providing a robust cluster labeling and BSI capability or can provide stand-alone capabilities over a variety of applications and scales.

Factors Responsible for the Co-occurrence of Forested and Unforested Rock Outcrops in the Boreal Forest

Rock outcrops in the boreal forest of Quebec can show either of two different states: a forested state with >25% tree cover, and an unforested state (<25% tree cover). We tested three different hypotheses that might explain the co-occurrence of forested and unforested rock outcrops: (1) differences in bedrock geology, with unforested outcrops associated to bedrock types inimical to tree growth; (2) unforested outcrops as recently disturbed sites undergoing secondary succession towards a forested state; (3) unforested outcrops as an alternative stable state to forested outcrops, induced by post-fire regeneration failure. Digitized forest inventory maps were used along with bedrock geology maps and time-since-fire maps to compare forested and unforested outcrops for bedrock geology type and date of the last fire. Field surveys were conducted on 28 outcrops (14 forested, 14 unforested) to gather information regarding tree species composition and site characteristics (thickness of the organic matter layer, percent cover of lichens, mosses and ericaceous shrubs). None of the three hypotheses explain the co-occurrence of forested and unforested rock outcrops in the boreal forest of Quebec. Both outcrop types occur on the same bedrock geology types. Unforested outcrops are not recently disturbed sites in early-successional states, as no clear distinction could be made in tree species composition and date of the last fire between the two outcrop types. Forested and unforested outcrops are not alternative stable states, as unforested outcrops are unstable and cannot maintain themselves through time in the prolonged absence of fire. Hence, unforested rock outcrops could be viewed as degraded, diverging post-fire types maintained by the late Holocene disturbance regime, characterized by high fire frequencies.

An automated object-based approach for the multiscale image segmentation of forest scenes

Over the last decade the analysis of Earth observation data has evolved from what were predominantly per-pixel multispectral-based approaches, to the development and application of multiscale object-based methods. To empower users with these emerging object-based approaches, methods need to be intuitive, easy to use, require little user intervention, and provide results closely matching those generated by human interpreters. In an attempt to facilitate this, we present multiscale object-specific segmentation (MOSS) as an integrative object-based approach for automatically delineating image-objects (i.e., segments) at multiple scales from a high-spatial resolution remotely sensed forest scene. We further illustrate that these segments cognitively correspond to individual tree crowns, ranging up to forest stands, and describe how such a tool may be used in computer-assisted forest inventory mapping. MOSS is composed of three primary components: object-specific analysis (OSA), object-specific upscaling (OSU), and a new segmentation algorithm referred to as size constrained region merging (SCRM). The rationale for integrating these methods is that the first two provide the third with object-size parameters that otherwise would need to be specified by a user. Analysis is performed on an IKONOS-2 panchromatic image that represents a highly fragmented forested landscape in the Sooke Watershed on southern Vancouver Island, BC, Canada.

Boreal mixedwood species composition in relationship to topography and white spruce seed dispersal constraint

In this study we integrated digital terrain models, forest inventory maps, optical remote sensing and field data to analyze the spatial structure of a 4850 km 2 boreal mixedwood forest landscape in northeastern British Columbia. We built independent maps of forest cover and landform using a Bayesian classification algorithm and quantitative surface analysis. These data were used to test the strength of the association between topographic position and forest cover using a modified electivity index. We then used logistic regression to test whether the probability of a site being occupied by either mixedwood or hardwood is correlated to its distance from white spruce ( Picea glauca) seed sources. The relationship between forest cover and topography showed significant departures from randomness, with white spruce preferentially associated with channels and concave slopes, and hardwoods preferentially associated with ridges and convex slopes. The analysis of mixedwood and hardwood stand distribution showed a positive correlation between hardwood occurrence and distance from spruce stands, suggesting that the dispersal limitations of white spruce is a significant influence on landscape vegetation dynamics. Overall, the results support the hypothesis that mixedwood dynamics are the product of ecological processes at multiple scales. Furthermore, these dynamics are only revealed by taking a varied approach to both data gathering and analyses.

Stand delineation and composition estimation using semi-automated individual tree crown analysis

Stand delineation and species composition estimation are cornerstones of forest inventory mapping and key elements to forest management decision making. Improved mapping techniques are constantly being sought in terms of speed, consistency, accuracy, level of detail, and overall effectiveness. Semi-automated analysis of high-resolution imagery at the individual tree crown level may offer such benefits. Methods, however, need to be developed and tested under a variety of forest conditions. High-resolution (60 cm) multispectral airborne imagery was acquired over a predominantly young conifer forest and plantation test area on the west coast of Canada. Automated tree isolation algorithms were applied to the data in order to delineate tree crowns or clusters of crowns. An object-oriented single tree classification was conducted using a maximum likelihood classifier. Stands of similar species composition, closure, and stem density were defined through a sequence that first generated images of these parameters from the automated delineation and classification, used these as input to an unsupervised classification, and then filtered and smoothed the resulting classification clusters. Because of the dense nature of the stands and small crowns on the site, the isolation process often delineated clusters of several trees. Species classification accuracy was determined by comparing the average stand composition from the automated technique to that derived from ground transects or plots. Species classification was good, with average composition error (difference between field measured and automated composition) over all 16 test stands being 7.25%. Most errors for individual species in stands were below 20%, but a few were up to 30%. The automatically generated stand boundaries mimicked well those of known plantation and interpreted inventory boundaries. The automated technique created a few larger stands and some additional small stands in areas of complex forest structure. Overall, for the young fairly uniform stands of the site, both stand delineation and species composition estimation were of a quality suitable for operational use in inventory and forest management. Further development and testing is needed to extend results to situations covering large areas, multiple flight lines, varied topography, and different forest conditions.

Diurnal and spatial variation of xylem dielectric constant in Norway Spruce (Picea abies [L.] Karst.) as related to microclimate, xylem sap flow, and xylem chemistry

Spatial and temporal variations in vegetation dielectric properties strongly influence the microwave backscatter characteristics of forested landscapes. This paper examines the relationship between xylem tissue dielectric constant, xylem sap flux density, and xylem sap chemical composition as measured in the stems of two Norway Spruce (Picea abies [L.] Karst.) trees in the Fichtelgebirge region of Northern Bavaria, Germany. Dielectric constant and xylem sap flux were monitored continuously from June through October 1995, at several heights along the tree trunks. At the end of the measurement series, each tree was harvested, and its xylem sap extracted and analyzed to determine the concentrations of amino acids and cations. Results show that the sap flux density was correlated with vapor pressure deficit (VPD) at all heights in the stem. In contrast, the xylem tissue dielectric constant is influenced by VPD but can exhibit a significant temporal lag relative to changes in VPD. This lag varies with position along the tree trunk. The temporal variability of the dielectric constant is compared with both trees at several positions along the tree trunks. Results of xylem sap chemical analysis are presented. We show that spatial and temporal variability in the xylem tissue dielectric constant is influenced not only by water content, but by variations in xylem sap chemistry as well. This has important implications for microwave remote sensing of forested landscapes, as useful information may be acquired regarding stand physiology and water relations and where variations in dielectric properties within individual trees and across geographic areas can be significant error sources for forest inventory mapping.

Prediction of the spatial distribution and relative abundance of ground-dwelling mammals using remote sensing imagery and simulation models

We present an approach that allows current, retrospective and future relative abundances of mammal species to be predicted across landscapes. A spatial generalized regression model of species relative abundance based on habitat quality and time since disturbance was combined with coverages of the spatial distribution of habitat quality derived from a simulation model which predicts the historical and future spatial arrangement of forest habitat. The strength of this approach is that the input habitat data can be derived as part of a standard forest inventory mapping program with the addition of high spatial resolution remote sensing imagery. Furthermore, it operates at the scale used for wildlife management in Australia, which makes it widely applicable. To demonstrate the approach we use data collected over 20 years on the long-nosed potoroo (Potorous tridactylus) and the large wallabies (red-necked wallaby, Macropus rufogriseus, and swamp wallaby, Wallabia bicolor) and their habitats following wildfire. Results indicate the relative abundance of the potoroo has increased, from initially sparse numbers of less than 0.5 % of plot-night occurrences to close to 3% approximately twenty years after a major fire event. The large wallabies by contrast decreased in relative abundance from about 20% since the major fire event. Presently the relative abundance of large wallabies was modelled at 2% of plot-nights with tracks which was very low. Predictions of future relative abundance without additional disturbance were low, with the region likely to be unsuitable for the species in the next 5 years. These models offer tools for investigating the current and historical abundances of key species which can provide data to forest managers for wildlife management thereby translating current scientific understanding into tools suitable for every-day use by forest managers.

Post-fire development of canopy structure and composition in black spruce forests of Abitibi, Québec: a landscape scale study

Fire reconstruction and forest inventory maps provided an opportunity to study changes in stand-level characteristics following fire using a data set comprised of all forest stands of fire origin in an area of over 10Â 000 km2. We assigned the date of the most recent fire occurrence to over 31Â 000 forest stands in an ecoforestry database. We categorized stands on different substrates into age classes to investigate differences in canopy composition, cover and height, and incidence of secondary disturbance. Stands with over 75% Picea mariana (Mill.) BSP dominated all age classes on organic sites. On other substrates, there was a change in canopy composition from deciduous stands and stands dominated by Pinus banksiana Lamb. to Picea mariana stands after about 100 yr. This transition was later for xeric sites. After a peak in canopy cover and height at about 100 yr, there was a decrease in the area occupied by stands with dense, tall canopies. Structural development was slower on less productive sites. There was little incidence of spruce budworm outbreaks. Partial disturbance by windthrow coincided with canopy break-up at 100 yr, but appeared to have little effect on overall canopy structure in later stages. Structural diversity was independent of compositional diversity; on organic sites, stands with similar composition had different canopy structure. Diversity of stands with different composition and structure was greatest in the first 150 yr following fire. Maintaining stands in different stages of structural development on the landscape would serve to maintain regional biodiversity.

Comparison of forestry-based remote sensing methodologies to evaluate woodland caribou habitat in non-forested areas of Newfoundland

Forest inventory maps and a manual interpretation of forestry-enhanced Landsat imagery are compared to the results of a detailed aerial photograph interpretation used to map habitat for caribou (Rangifer tarandus terra novae) in a relatively unforested region of Newfoundland. This comparison serves as an illustration of the pitfalls inherent in using readily available remote sensing technologies in applications for which they were not intended. The non-forest classes in the Newfoundland Forest Inventory are too broad to describe single vegetation communities, and only rarely are vegetation communities found entirely within a single forest inventory class. For example, "bog" is relatively well associated with wetland vegetation classes and "barren" with upland classes, but "scrub" is a misleading term used to describe both forest and non-forest communities. An earlier (global) forest classification for Newfoundland has a more reliable association of scrub with forest, but a less reliable identification of bog than later updates to the forest inventory in the study area. Landsat imagery applications for forest inventory updates do not appear useful in identifying non-forest vegetation communities. Caution should be taken in using forest inventory maps in wildlife habitat applications when the habitat includes important non-forest components. Key words: forest inventory, habitat classification, Landsat imagery, mapping, remote sensing

FOREST TYPE AND WILDFIRE IN THE ALBERTA BOREAL MIXEDWOOD: WHAT DO FIRES BURN?

Two determinants of fire behavior are fire weather and spatial variation in fuels. Their relative importance in boreal forests has been unclear. I evaluated the effect of fuels on a ∼74 000-km2 landscape in the boreal mixedwood region of western Canada. My data were the compositions, or the proportional areas of different forest types, of 48 mapped lightning fires and of their immediate surroundings. I measured areal compositions from forest inventory maps, using a five-way classification representing deciduous forest, three types of coniferous forest, and wetlands. The fires burned between 1980 and 1993. Fire sizes ranged from 70 ha to 70 000 ha. By multivariate linear regression, fire surroundings explain 57% of the variation in forest types within mapped fires. Fire compositions are not representative of the study area as a whole, or of a fire's surroundings, and are unrelated to fire size and location within the study area. Using the model, I predicted the areas of the five types burned within all other lightning fires >200 ha in the study area during 1961–1996 and estimated type-specific mean annual burn rates. These rates vary by an order of magnitude. Deciduous stands burn at the lowest rate, and black spruce stands burn at the highest rate. Fires exhibit significant preferences between forest types at both local and regional scales. Preference orderings are similar at both scales and are generally consistent with the rank order of estimated burn rates. Preferential burning may result from between-class differences in vertical canopy structure and foliage characteristics. The statistical model and the postulated variations in fire behavior between classes indicate that landscape-scale fuels management may be feasible in this system. The rank ordering of burning frequencies and preferences is the inverse of the planned disturbance rates under forest management.

Forest Characterization in the Central Plateau of Yellowstone National Park Using Multispectral High Resolution Digital Photography

Detailed knowledge of forest structure is an important component in research focusing on forest biodiversity monitoring, carbon budgeting studies, fire modeling and forest inventory estimation. In forest inventory mapping, high spatial resolution multispectral imagery are becoming a valuable and often a critical tool to effective resource management planning. In this report, we describe and illustrate the forestry applications of a multispectral digital imaging camera system (DuncanTech MS3100) that was experimentally flown over several study areas in the Central Plateau of Yellowstone National Park, in July, 2001. The purpose of this demonstration is three fold; first we will apply geostatistical methods, co-kriging specifically, to model forest canopy components such as height, canopy thiclmess, species, density and structure; second, we will examine the spectral and spatial characteristics of seedling regeneration in the 1988 burn sites; finally, forest stand characteristics such as stem counts, crown species and amount of deadfall (potential fire fuel) will be estimated for selected sites. Our results will be compared to information acquired from coarser resolution, multispectral Landsat TM data obtained on the same date.

Land cover mapping at BOREAS using red edge spectral parameters from CASI imagery

Scientific and technical challenges remain significant to accurate classification of land cover and forest species as a result of the many spectral and spatial variables influencing surface reflectance, coupled with the constraints imposed by the spectral and spatial characteristics of the remote sensing instrumentation. The use of systematic differences in canopy pigment or chemistry by cover type or by species as a basis for land cover classification has very recently emerged as a potentially new approach. In this study, classification of land cover is investigated, based on chlorophyll content variations as inferred from spectral bands in the red edge reflectance region. This analysis was carried out on data collected with the Compact Airborne Spectrographic Imager (CASI) for a 16 km × 12 km image mosaic over the submodeling grid of the southern study area at the Boreal Ecosystem‐Atmosphere Study (BOREAS). The analysis demonstrates that land cover mapping, based solely on red edge spectral parameters, appears to be feasible, robust, and for some cover classes outperforms other current classification methods. Classification accuracy assessments of the derived land cover maps were performed using a forest inventory map provided by the Saskatchewan Environment and Resource Management Forestry Branch‐Inventory Unit (SERM‐FBIU). The red edge parameter‐based land cover classification showed producer's accuracies which exceeded 68.6% for all classes identified: conifers (however, without an ability to separate wet from dry conifers), mixed stands, fen, and disturbed and regeneration features. The corresponding user's accuracies for these classes ranged between 58 and 66%, with the overall classification accuracies of 61.15% and Kappa coefficient (K) of 0.52. In comparison, the corresponding Kappa coefficients for the cover classification using 16 channel CASI data and for a TM‐based classification, were 0.36 and 0.29, respectively. Results of this study suggest that whereas land cover classification accuracy improvements for the important but illusive fen cover type in the boreal ecosystem are possible using classifications based on red edge parameters, significant uncertainties remain in the estimated aerial extent.

La cartographie écoforestière ou le défi d’intégrer le milieu physique et le couvert forestier

Abstract:In Quebec, for the past thirty years, ecological land classifications have been proposed to support the use of ecological information by forest managers. Recently, it has been suggested to replace the classical forest inventory maps, which use only dendrometric variables, by integrated forest/site maps that would use both ecological and forest variables. Several methods were tested at the Foret Montmorency, the experimental forest of Universite Laval located in a hilly boreal area influenced by minor natural and human disturbances (1-50 ha). The production of a large scale (1:20 000 – 1:10 000) integrated forest/site map turned out to be much more difficult than anticipated. In a fragmented forest such as the one in the pilot area, the concept of a homogenous “ecoforest stand” can be considered as utopic. Because of soil heterogeneity and also of the large number of discrepancies between vegetation and soil limits, the delimitation of integrated vegetation/site units was operationally difficult. ...

Data Processing and Analysis for MIFUCAM: A Trial of MEIS Imagery for Forest Inventory Mapping

RÉSUMÉLa méthode d'inventaire forestier utilisant le capteur aéroporté MEIS (MIFUCAM) est un procédé expérimental visant à dresser l'inventaire des forêts à partir d'imagerie aérienne multispectrale numérique à haute résolution. Des données ont été acquises sur deux aires d'essai au Québec, Canada, à l'aide du capteur à réseau rectiligne MEIS (balayeur imageur électro-optique multispectral). L'aire principale, une surface de 10 km sur 14 km, a été survolée en vingt-six segments d'une résolution de soixante centimètres. Les principales étapes du traitement sont : le calibrage des capteurs, la correction radiométrique pour tenir compte de l'effet de l'atmosphère et des angles de vue, la correction géométrique, l'amélioration de l'image ainsi que la délimitation et l'interprétation des peuplements forestiers. Ces étapes sont décrites et démontrées. D'autres travaux sont en cours, mais des progrès ont été réalisés en vue de déterminer la possibilité d'utiliser l'imagerie numérique à haute résolution pour dresser l'inventaire des forêts ainsi que les mises au point requises pour améliorer les systèmes afin que la technologie puisse devenir opérationnelle.