Geordie Hobart

Pixel-Based Image Compositing for Large-Area Dense Time Series Applications and Science

Abstract Free and open access to the more than 40 years of data captured in the Landsat archive, combined with improvements in standardized image products and increasing computer processing and storage capabilities, have enabled the production of large-area, cloud-free, surface reflectance pixel-based image composites. Best-available-pixel (BAP) composites represent a new paradigm in remote sensing that is no longer reliant on scene-based analysis. A time series of these BAP image composites affords novel opportunities to generate information products characterizing land cover, land cover change, and forest structural attributes in a manner that is dynamic, transparent, systematic, repeatable, and spatially exhaustive. Herein, we articulate the information needs associated with forest ecosystem science and monitoring in a Canadian context, and indicate how these new image compositing approaches and subsequent derived products can enable us to address these needs. We highlight some of the issues and opportunities associated with an image compositing approach and demonstrate annual composite products at a national-scale for a single year, with more detailed analyses for two prototype areas using 15 years of Landsat data. Recommendations concerning how to best link compositing decisions to the desired use of the composite (and the information need) are presented, along with future research directions. Résumé L’accès libre et gratuit à plus de 40 ans de données dans l’archive Landsat combiné à l’amélioration des produits d’imagerie standardisés et l’augmentation des capacités de traitement et de stockage informatiques ont permis la production d’images composites basées sur les pixels de réflectance de surface de grande superficie sans nuages. Les composites du « meilleur pixel disponible » (best-available-pixel; BAP) représentent un nouveau paradigme en matière de télédétection qui ne dépend plus de l’analyse par scène. Une série chronologique de ces images composites BAP offre de nouvelles occasions de générer des produits d’information qui caractérisent la couverture terrestre, le changement de la couverture terrestre et les attributs structurels de la forêt d’une manière dynamique, transparente, systématique, répétable et spatialement exhaustive. Ici, nous articulons les besoins d’information liés à la science et à la surveillance des écosystèmes forestiers dans un contexte canadien, et nous indiquons comment ces nouvelles approches de composition d’image et les produits qui en découlent peuvent nous permettre de répondre à ces besoins. Nous soulignons quelques-uns des problèmes et des possibilités associés à une approche de composition d’image et nous démontrons des produits composites annuels à l’échelle nationale pour une année, avec des analyses plus détaillées pour deux zones prototypes utilisant 15 ans de données Landsat. Des recommandations concernant la meilleure façon de lier des décisions de composition d’images à l’utilisation souhaitée du composite (et le besoin d’information) ainsi que les orientations futures de la recherche sont présentées.

Mass data processing of time series Landsat imagery: pixels to data products for forest monitoring

ABSTRACT Free and open access to the Landsat archive has enabled the implementation of national and global terrestrial monitoring projects. Herein, we summarize a project characterizing the change history of Canada’s forested ecosystems with a time series of data representing 1984–2012. Using the Composite2Change approach, we applied spectral trend analysis to annual best-available-pixel (BAP) surface reflectance image composites produced from Landsat TM and ETM+ imagery. A total of 73,544 images were used to produce 29 annual image composites, generating ∼400 TB of interim data products and resulting in ∼25 TB of annual gap-free reflectance composites and change products. On average, 10% of pixels in the annual BAP composites were missing data, with 86% of pixels having data gaps in two consecutive years or fewer. Change detection overall accuracy was 89%. Change attribution overall accuracy was 92%, with higher accuracy for stand-replacing wildfire and harvest. Changes were assigned to the correct year with an accuracy of 89%. Outcomes of this project provide baseline information and nationally consistent data source to quantify and characterize changes in forested ecosystems. The methods applied and lessons learned build confidence in the products generated and empower others to develop or refine similar satellite-based monitoring projects.

Modeling Stand Height, Volume, and Biomass from Very High Spatial Resolution Satellite Imagery and Samples of Airborne LiDAR

Plot-based sampling with ground measurements or photography is typically used to establish and maintain National Forest Inventories (NFI). The re-measurement phase of the Canadian NFI is an opportunity to develop novel methods for the estimation of forest attributes such as stand height, crown closure, volume, and aboveground biomass (AGB) from satellite, rather than, airborne imagery. Based on panchromatic Very High Spatial Resolution (VHSR) images and Light Detection and Ranging (LiDAR) data acquired in the Yukon Territory, Canada, we propose an approach for boreal forest stand attribute characterization. Stand and tree objects are delineated, followed by modeling of stand height, volume, and AGB using metrics derived from the stand and tree crown objects. The calibration and validation of the models are based on co-located LiDAR- derived estimates. A k-nearest neighbor approach provided the best accuracy for stand height estimation (R 2 = 0.76, RMSE = 1.95 m). Linear regression models were the most efficient for estimating stand volume (R 2 = 0.94, RMSE = 9.6 m 3 /ha) and AGB (R 2 = 0.92, RMSE = 22.2 t/ha). This study was implemented for one Canadian ecozone and demonstrated the capacity of a methodology to produce forest inventory attributes with acceptable accuracies offering potential to be applied to other boreal regions.

Comparison of Aviris and AISA Airborne Hyperspectral Sensing for Above-Ground Forest Carbon Mapping

Monitoring of the 418 million ha of forests in Canada is needed to ensure the sustainable development of these forests. Hyperspectral sensing can provide mapping of forest species, forest health, and above-ground biomass. Airborne two-meter AISA hyperspectral and LIDAR data were acquired by the University of Victoria (UVic) over the Greater Victoria Watershed District (GVWD) test site and compared to NASAs AVIRIS data that had been acquired in the summer of 2002 at 4 m spatial resolution. Tree heights derived from LIDAR data, and allometric equations were used to provide independent ground estimates of biomass. Between-sensor calibration calibrated the AISA data to the same basis as the AVIRIS data. The calibrated reflectance data were used to generate forest species classifications, and biomass estimates for the test site. Average classification accuracies exceeded 89% in mapping major forest species. These products were used to create a map of above-ground carbon for the forested portion of the GVWD test site.

Forest inventory stand height estimates from very high spatial resolution satellite imagery calibrated with lidar plots

Many areas of forest across northern Canada are challenging to monitor on a regular basis as a result of their large extent and remoteness. Although no forest inventory data typically exist for these northern areas, detailed and timely forest information for these areas is required to support national and international reporting obligations. We developed and tested a sample-based approach that could be used to estimate forest stand height in these remote forests using panchromatic Very High Spatial Resolution (VHSR, < 1 m) optical imagery and light detection and ranging (lidar) data. Using a study area in central British Columbia, Canada, to test our approach, we compared four different methods for estimating stand height using stand-level and crown-level metrics generated from the VHSR imagery. ‘Lidar plots’ (voxel-based samples of lidar data) are used for calibration and validation of the VHSR-based stand height estimates, similar to the way that field plots are used to calibrate photogrammetric estimates of stand height in a conventional forest inventory or to make empirical attribute estimates from multispectral digital remotely sensed data. A k-nearest neighbours (k-NN) method provided the best estimate of mean stand height (R 2 = 0.69; RMSE = 2.3 m, RMSE normalized by the mean value of the estimates (RMSE-%) = 21) compared with linear regression, random forests, and regression tree methods. The approach presented herein demonstrates the potential of VHSR panchromatic imagery and lidar to provide robust and representative estimates of stand height in remote forest areas where conventional forest inventory approaches are either too costly or are not logistically feasible. While further evaluation of the methods is required to generalize these results over Canada to provide robust and representative estimation, VHSR and lidar data provide an opportunity for monitoring in areas for which there is no detailed forest inventory information available.

Grid-enabled OGC environment for EO data and services in support of Canada’s forest applications

System of agents for forest observation research with advanced hierarchies (SAFORAH) is a grid-enabled environment that makes optimum use of distributed data storage facilities. It supports EO data collaboration among various Canadian research organizations and provides ready access to the most current, consistent and reliable forest resource information through various network environments. SAFORAH was built with an extensible data grid network based on the Globus Toolkit and transparently presents the distributed data storage facilities across multiple organizations and government agencies as one single data archive to users to facilitate easy EO data archiving and distribution. Currently, four Canadian forestry centres located in Victoria British Columbia,Cornerbrook Newfoundland,Edmonton,Alberta and Laurentian Quebec, and environment Canadas Canadian Wildlife service are operationally connected via Canarie to the SAFORAH data grid. The connection to a large petabyte storage facility under development at the University of Victoria (UVic) was also established. SAFORAH has been modified to provide data services based on the OGC standard using NWGISS to support the distribution of comprehensive, integrated and thematic EO data for forest applications. SAFORAH has been used in the EOSD project to support the national monitoring of Canadas forests. This use has demonstrated that SAFORAH is a practical tool for collaborative research to facilitate the sharing of EO data. The extension of SAFORAH to handle OGC requests has enabled the distribution of landcover product tiles from multiple storage sites seamlessly to national and international users. The web address for SAFORAH is www.saforah.org.

Data Fusion Study Between Polarimetric SAR, Hyperspectral and Lidar Data for Forest Information

ALOS PALSAR L-band quad-pol data were acquired over our study area on Vancouver Island in British Columbia in the summer of 2007. The site has significant topographic relief and high biomass in this temperate coastal rainforest. Our emphasis was on integration and fusion techniques of polarimetric SAR, hyperspectral and LIDAR data for useful forest information extraction. The polarimetric SAR techniques and analysis methods studied in this project drew on the work of other researchers. The Jong-Sen Lee algorithm for polarization compensation for terrain azimuth slope variations was implemented and tested. The Shane Cloude decomposition method, with basic types of scattering analysis for reducing sensitivity to topography effects was examined and applied. In this study, the hyperspectral data was used for providing high spectral resolution information, such as major forest species and land-cover characterization, and the LIDAR data were utilized to generate information related to vertical structure of both the underlying topography and the forest structure. The combination of these data sources and techniques provided an opportunity to examine the potential capabilities of polarimetric SAR and the synergy of the fused data for forest classification.

Forest information products from hyperspectral data — Victoria and Hoquiam test sites

Canada contains 10% of the worlds forests, covering an area of 418 million ha. Remote sensing can provide rapid coverage of these forests. Forest information needs are moving from broad forest typing to mapping of major forest species. Hyperspectral sensing can be used to create products for forest inventory, forest health, biomass, and aboveground carbon. The challenge is to develop a process to extract from hyperspectral data the relevant forest information in a timely and reproducible manner with a high degree of confidence in the quality of the resultant products. To answer this challenge, experiments have been conducted on two test sites: the Greater Victoria Watershed District (GVWD) on Vancouver Island, BC, and the Hoquiam test site in Washington State. Three types of data were collected to perform the analysis. First, Hyperion and AVIRIS hyperspectral images were collected for both test sites. Secondly, ground truth reflectance data were acquired from locations within the GVWD test sites simultaneously with these image collections. Finally, foliar samples were taken and chemically analyzed from 30 Douglas fir plots in the GVWD test site and five plots in the Hoquiam test site to provide a sample set of known values to compare against the spectral imagery. This paper reports on the results for chlorophyll and nitrogen mapping for these sites. Difficulties associated with chemistry signature extension are discussed. For nitrogen, an r2 of 0.915 was obtained between ground measured nitrogen and nitrogen predicted from AVIRIS data.

Disturbance-Informed Annual Land Cover Classification Maps of Canada's Forested Ecosystems for a 29-Year Landsat Time Series

ABSTRACT Land cover classification of large geographic areas over multiple decades at an annual time step is now possible based upon free and open access to the Landsat data archive. Annual gap-free, best-available-pixel, surface reflectance, image composites and annual forest change maps have been generated for Canada for the years 1984 to 2012. Using these data, we demonstrate the Virtual Land Cover Engine (VLCE), a framework for change-informed annual land cover mapping, over the 650 million ha forested ecosystems of Canada, to produce a 29-year data cube of land cover. Post-processing aimed to reduce spurious class transitions is undertaken integrating change information, land cover transition likelihoods, and year-on-year class membership likelihoods. Validation was assessed for a single year (2005) using independent data for an overall accuracy of 70.3% (± 2.5%). Key results are the detailed capture of trends in land cover, illustration of land cover links to disturbance processes, and insights related to the general stability of land cover over time with stand replacing disturbance followed by regeneration of forests. The portable mapping framework and resultant data products offer an integrated, long baseline, disturbance-informed and detailed depiction of land cover to meet science and program related information needs.

Unsupervised nonparametric classification of polarimetric SAR data using the K-nearest neighbor graph

Polarimetric SAR classifications are often based on assumptions about the shape of clusters in the data space. Such a scheme will fail for nonlinear structures in the feature space, unless the classification algorithm has the capacity to describe cluster shapes in sufficient generality. Existing polarimetric SAR classification methods are faced by this exact problem: typically they initialize clusters in the Cloude-Pottier parameter space [1], further optimizing them in the coherency matrix space [2, 3]. Methods using K-means [2] or agglomeration [3] require clusters that are spherical, or compact and well separated, respectively. In the Cloude-Pottier space, these requirements are not met, so initialization in the Cloude-Pottier space cannot be consistent with optimization by K-means or agglomeration. This paper sets out to address this problem, by implementing a new data-driven clustering approach, for arbitrarily shaped clusters. It is applied to quad-polarisation data, demonstrating the new methodologys potential for forest land-cover type discrimination.