Publications by A. Dyk

Compressed hyperspectral imagery for forestry

Mon, 19 Apr 2004

Various compression schemes have been suggested for storage and distribution of hyperspectral remotely sensed data. Hyperspectral forestry applications that rely on the measurement of subtle variations in the spectral signature of the forest canopy can be affected by modification to the spectra induced by compression. As part of an experiment for the Canadian Space Agency (CSA), Hyperion data cubes acquired over the Greater Victoria Watershed District (GVWD) were compressed using the SAMVQ and HSOCVQ algorithms developed by CSA. The data were compressed using compression ratios 10:1 and 20:1 and were returned uncompressed. The data cubes were classified into forest species using the same supervised classification methodology as applied to the original data. The classification accuracies were compared.

For some applications, one can achieve significant reductions in data volume through compression. Of the compression algorithms and ratios tested, SAMVQ 10:1 has the least overall effect but still reduces classification accuracies on difficult to separate classes. While uncompressed data are preferred, SAMVQ 10:1 compression may be suitable for forest inventory.

Hyperspectral remote sensing of conifer chemistry and moisture

Mon, 19 Apr 2004

The chemical and moisture composition of conifer foliage in the Greater Victoria Watershed District (GVWD), Vancouver Island, Canada, was explored using hyperspectral remote sensing data. Imagery acquired from the airborne sensor Advanced Visible/Infrared Imaging Spectrometer (AVIRIS) were evaluated along with sampled foliar chemical and moisture measurements to provide insight into ecological processes occurring within the watershed. Concentrations of nitrogen, total chlorophyll and moisture were used to provide an analysis of the forest canopy, comprised of Coastal Douglas-fir and Western Redcedar.

The AVIRIS data were processed to correct for atmospheric and geometric distortion. The AVIRIS data were used to investigate the relationship between the hyperspectral imagery and the sampled chemical data. A total of 45 plots in the GVWD were sampled from a helicopter. These samples provided both organic and inorganic analysis of the forest canopy. A Partial Least Squares regression was used to analyze the relationship between the data sets in order to extract chemical constituents in the forest canopy. Results indicate that the regression equation explains 81%, 79% and 70% of the variation in nitrogen, total chlorophyll and moisture, respectively. An analysis of the chemical characteristics of the canopy can provide insight into factors controlling growth such as nutrient levels and water deficiencies at the foliar level.

EVEOSD Forest Information Products from AVIRIS and Hyperion

Wed, 03 Mar 2004

Hyperspectral remote sensing can provide forest information products for applications in forest inventory, forest chemistry, and the Kyoto Protocol. One of the forest information products is the high accuracy forest species map produced by the classification of hyperspectral data. As part of the Evaluation and Validation of EO-1 for Sustainable Development (EVEOSD) Project, Hyperion data were acquired in 2001 and 2002. Corresponding AVIRIS data were also acquired. All hyperspectral data acquired were calibrated to reflectance and orthorectified. Experiments were conducted to compare the accuracies of the data sets for mapping forest species. Operational accuracies for forest species recognition were achieved with both AVIRIS and Hyperion. Bioindicators were also developed for mapping chlorophyll, nitrogen, and moisture content. These bioindicators were stratified by forest type. In the Greater Victoria Watershed District and the Hoquiam, Washington State test sites, foliar samples were collected, analyzed, and databases were built, which include foliar chemistry and plot parameters. These sites were used to develop the indicators and to validate their success. The forest information products produced under the EVEOSD project demonstrate some of the benefits to be achieved from an operational hyperspectral satellite.

Monitoring Forests from Space: Hyperspectral and Kyoto Products

Fri, 06 Feb 2004

Remote-sensing data can be used to create products to support national and international agreements on sustainable forest management and the Kyoto Protocol. Examples of products derived from satellite hyperspectral data and from multitemporal Landsat data are presented in this paper. Multitemporal Landsat data from 1985, 1990, 1996 and 2001 were orthorectified and used to create forest classifications and biomass estimates. The multitemporal products were used to create above-ground carbon maps, and reforestation, afforestation and deforestation maps. The above-ground carbon measurements were compared with those derived from a traditional forest inventory for our test site near Hinton, Alberta, Canada. The remote-sensing methods reported twice as much forest area, and half the biomass, as derived from the forest inventory. The total above-ground carbon results for the Hinton test site from the two methods were in general agreement.

With EO-1 Hyperion data of the Greater Victoria Watershed (on southern Vancouver Island, in British Columbia, Canada), forest species were classified to an accuracy of 90.0% correct. The Hyperion data were orthorectified to a positional accuracy of 10.1 m. Hyperspectral monitoring of forests can be used for forest inventory, forest health and forest chemistry.

Detection and correction of abnormal pixels in hyperion images

Fri, 01 Nov 2002

Hyperion images are currently processed to level 1a (from level 0 or raw data). These level 1a images are files o radiometrically corrected datain units of either watts/(sr x micron x m2) x 40 for VNIR bands or watts/(sr x micron x m2) x 80 for SWIR bands. Each distributed Hyperion level 1a image tape contains a log file, called "(E)-1 identifier).fix.log", that reports the bad or corrupted pixels (called known bad pixels) found during the pre-flight checking, and details how they were fixed. All bad pixels should be corrected in a level 1a image. However, bad pixels are still evident. In addition, there are dark vertical stripes in the image that are not reported in the log file. In this paper, we introduce a method to detect and correct the bad pixels and vertical stripes (we will refer to theses occurrences as abnormal pixels). Images from the Great Victoria Watershed and other EVEOSD test sites are used to determine how stationary the locations of the abnormal pixels are. After abnormal pixel correction a Hyperion image is ready for geometric correction, atmospheric correctiion, and further analysis

Automated methods for atmospheric correction and fusion of multispectral satellite data for national monitoring

Mon, 26 Aug 2002

The Earth Observation for Sustainable Development of Canada's forests (EOSD) project monitors Canada's forests from space. Canada contains ten-percent of the world's forests. Initial EOSD products are land cover, forest change, forest biomass, and automated methods. There are more than 500 LANDSAT TM or ETM+ scenes required for a single coverage of Canada's forests. Multi-temporal analysis using satellite data requires automation for conversion of these data to common units of exoatmospheric radiance or ground reflectance. During the next ten years the EOSD project will use a variety of Landsat optical and Radarsat sensors. A diverse set of ancillary and satellite data formats exist which require the development of adaptable data ingest and processing streams. Legacy LANDSAT TM and ETM+ data are available in a number of different formats from several national and US suppliers. In this paper, we present an automated system for managing processing streams for calibration and atmospheric correction of LANDSAT TM and ETM+ data to create data sets ready to analyze for EOSD products. Using known forest attributes from GIS data and field measurements, we validated our results of studies undertaken to assess spectral signal variability using both at sensor radiance and ground reflectance for LANDSAT TM and ETM+ for a test site on Vancouver Island, BC. We present a strategy for correcting and fusing multi-source and multitemporal satellite data for meeting EOSD requirements.

Monitoring Forests with Hyperion and ALI

Mon, 26 Aug 2002

Hyperion, a hyperspectral sensor, and the Advanced Land Imager (ALI) are carried on NASA’s EO-1 satellite. The Evaluation and Validation of EO-1 for Sustainable Development (EVEOSD) is our project supporting the EO-1 mission. With 10% of the world’s forests and the second largest country by area in the world, Canada has a natural requirement for effective monitoring of its forests. Eight test sites have been selected for EVEOSD, with seven in Canada and one in the US. Extensive fieldwork has been conducted at four of these sites.
A comparison is made of forest classification results from Hyperion, ALI, and the ETM+ of Landsat-7 for the Greater Victoria Watershed. The data have been radiometrically corrected and ortho-rectified. Feature selection and statistical transforms are used to reduce the Hyperion feature space from 220 channels to 12 features. Classes chosen for discrimination included Douglas Fir, Hemlock, Western Red Cedar, Lodgepole Pine and Red Alder. Overall classification accuracies obtained for each sensor were: Hyperion 92.9%, ALI 84.8%, and ETM+ 75.0%. Hyperspectral remote sensing provides significant advantages and greater accuracies over ETM+ for forest discrimination. The EO-1 sensors, Hyperion and ALI, provide data with excellent discrimination for Pacific Northwest forests in comparison to Landsat-7.

Geometric Correction and Validation of Hyperion and ALI Data for EVEOSD

Mon, 26 Aug 2002

Precise geometric correction of EO-1’s Hyperion data is essential to link ground spectral data and satellite hyperspectral data. Two scenes have been selected from sites of the EVEOSD (Evaluation and Validation of EO-1 for Sustainable Development of Forests) project. One site is the Greater Victoria Watershed District (GVWD) located on south Vancouver Island, BC and the other is Hoquiam located in southwestern Washington State. Ground Control Point (GCP) collection has been performed using a feature fitting method in which high accuracy, orthorectified photo-derived polygons of features are used for tie-down. For example lakes are adjusted to match the same feature obvious in the hyperspectral imagery. This technique allows for easier estimation of a GCP’s precise fit to the imagery. A third (11) of the GCPs were identified as check points to validate the geometric models. GCPs were collected independently from both the VNIR and SWIR arrays of the Hyperion sensor to determine the adjustment factor required to remove the displacement and skew between these arrays. The adjustment can then be applied to GCPs collected from one array to make a compatible geometric correction model for both arrays. The polynomial and rational function correction methods have been applied to both scenes with various orders applied to each function. The effect of terrain distortion removal is evaluated in using the rational function method. Hyperion data can be geocorrected with surprising accuracy. For example, we obtained 10 m RMS on check points with the rational function. With a second order polynomial we achieved 13 m RMS without terrain correction. The accuracy of this latter result is due to the small swath width of the sensor. Applying terrain correction does improve the accuracy of geometric correction in areas with high relief. A similar procedure was applied to EO-1’s ALI sensor and this paper compares the results for Hyperion and ALI geometric fidelity.

AVIRIS imagery for forest attribute information: anisotropic effects and limitations in multi-temporal data

Fri, 18 Nov 2005

Hyperspectral data can provide valuable forest information, such as forest species, stand density, biochemistry, and forest structure. It is also well known that optical radiometric properties of forest objects vary with the angles of illumination and view angle. The anisotropy of the forest canopy can restrict the determination of the forest parameters of interest. In high relief areas such as Vancouver Island, Canada the impact of illumination effects presents numerous additional complexities. The authors present the results of a study undertaken to assess forest attribute determination from AVIRIS data acquired over the Greater Victoria Watershed District Test Site (GVWD) on Vancouver Island B.C., Canada on two dates. A comparison of data from a number of test plots is carried out using AVIRIS imagery acquired in 1993 and 1994. Inventory information (such as stem density, species distribution, biomass, etc.) for these plots is known as a result of field sampling and data fusion of the AVIRIS Hyperspectral data with high spatial resolution (1 m) MEIS data and AirSAR data For GVWD, the dominant forest species is Douglas fir. Similarly aged stands on different slopes and at various aspects provide a sampling of view angles. Acquisitions at different times of the day sample the variation in illumination angles. AVIRIS reflectances from 1993 and 1994 are used to determine the limitations imposed by a range of off-nadir angles and BRDF effects.

Determination of above ground carbon in Canada’s forests

Thu, 31 May 2001

Determination of above ground carbon in Canada's forests - A multi-source approach

Tue, 19 Sep 2000

Canada is a signatory to the Kyoto Protocol and must report on reforestation, afforestation and deforestation activities since 1990. Reporting commitments also include a baseline estimate of forest carbon stocks in 1990 and the monitoring of changes in carbon stocks leading up to the reporting period 2008 to 2012.

Canada has 10% of the world's forests (418 million hectares), which account for a significant amount of stored carbon. The determination of above-ground carbon stocks in the forest can be based on several sources: remote sensing, models of vegetation growth, book-keeping carbon models, and traditional forest inventories. Estimating above-ground carbon with remote sensing requires the fusion and integration of remote sensing data with topographic, forest cover and other geospatial information. Multi-temporal LANDSAT TM imagery was used in conjunction with GIS data to compute above-ground biomass from which the carbon content is determined. In addition to biomass, other key factors, which play a role in the determination of carbon stocks, include species and age distribution, forest structure, and climate variables. The paper reports on remote sensing experiments to determine the above-ground carbon stocks for a test site near Hinton, AB. It is expected that this approach will be useful in supporting Canada's reporting commitments on the sustainability its forest resources.

Earth observation for sustainable development of forests (EOSD) - A National Project.

Fri, 09 Nov 2001