Chris Butson

Relating aspen defoliation to changes in leaf area derived from field and satellite remote sensing data

Trembling aspen is the most important deciduous species in the North American boreal forest for which methods are needed to consistently map and monitor insect defoliation patterns to assess impacts on aspen health and productivity. The loss of foliage due to insect defoliation results in changes to leaf area index (ΔLAI) that were studied from both field and satellite image data for a large aspen tortrix (Choristoneura conflictana (Wlk.)) outbreak in boreal forest aspen stands in northern Alberta. Field estimates of defoliation for stands undergoing various levels of defoliation severity and both allometric and optical LAI-2000 measurements of leaf area index were collected and related to Landsat enhanced thematic mapper plus (ETM+) image data representing before- and after-defoliation events. This study was undertaken as part of the Climate Change Impact on the Productivity and Health of Aspen (CIPHA) research initiative that is taking place in western Canada. Field estimates of ΔLAI followed defoliation trends but with substantial scatter that may be attributed to the use of allometric estimates of predefoliation LAI and the subjective nature of visual defoliation assessment. Change estimates derived from both Landsat ETM+ scenes were statistically correlated (correlation r = 0.84, probability P = 0.001) with percent defoliation. Such an approach may be used in the future to calibrate LAI algorithms to other defoliated areas or to map the temporal intensity of these events as they occur.

An Efficient Protocol to Process Landsat Images for Change Detection With Tasselled Cap Transformation

Change detection approaches, such as computing change in spectral indexes through time, are a mature and established science, which is increasingly being applied in operational remote sensing programs. The quality and consistency of the changes detected using these approaches are linked, however, to the processing of the imagery required to address issues related to image radiometry, normalization, and computation of the spectral indexes. These processing steps are typically undertaken independently, providing opportunities for computation errors, increasing disk storage needs, and consuming processing time. In this letter, we present an approach for combining these processing steps to facilitate a more streamlined and computationally efficient approach to change detection using Landsat-5 and -7. The individual elements of the algorithm (raw Landsat-5 or -7, to calibrated Landsat-7, to top-of-atmosphere reflectance, to tasselled cap components) are described, followed by a description and illustration of the protocol to algebraically combine the elements. Rather than producing intermediate outputs, the sequentially integrated data processing protocol operates in memory and produces only the desired outputs. The proposed approach mitigates opportunities for inappropriate scaling between processing steps, the consistency of which is especially important for threshold-based change detection procedures. In addition, savings in both processing time and disk storage are afforded through the combination of processing steps, with processing of the time-1 images reduced from three to two stages and five to two stages for the time-2 images, resulting in savings of 50% and 69% in computing times and disk space requirements, respectively

Landsat ETM+ mosaic of northern Canada

Mapping northern Canada with medium spatial resolution (30 m) Landsat data is important to complement national multiagency activities in forested and agricultural regions, and thus to achieve full Canadian coverage. Northern mapping presents unique challenges due to limited availability of field data for calibration or class labeling. Additional problems are caused by variability between individual Landsat scenes acquired under different atmospheric conditions and at different times. Therefore, the generation of radiometrically consistent coverage is highly desirable to reduce the amount of reference data required for land cover mapping and to increase mapping efficiency and consistency by stabilizing spectra of land cover classes among hundreds of Landsat scenes. The production chain and dataset of a normalized, 90 m resolution Landsat enhanced thematic mapper plus (ETM+) mosaic of northern Canada is presented in this research note. A robust regression technique called Thiel–Sen (TS) is used to normalize Landsat scenes to consistent coarse-resolution VEGETATION (VGT) imagery. The derived dataset is available for any interested user and can be employed in applications aimed at studying processes in the Canadian Arctic regions above the tree line.

An approach for edge matching large-area satellite image classifications

Large-area land cover mapping based on remotely sensed data often requires combining individual or large groups of classified images to produce final map products. Operational and logistical considerations are typically confronted when classifying medium spatial resolution satellite imagery (i.e., Landsat), with the mapping partitioned by spectral, ecological, or political considerations, or combinations thereof. Visual discontinuities can emerge at the locations where logistically based production zones join. Transparent and systematic approaches for addressing discontinuities are desired for the Earth Observation for Sustainable Development of Forests (EOSD) project. This large-area land cover mapping project is producing map products for Canadas forested ecozones. A distributed implementation plan, largely based on grouping provincial and territorial political units, was followed for production. Scene-to-scene discontinuities are rare within each production zone and are primarily related to image acquisition date and phenological state. In contrast, discontinuities can emerge at the production zone boundaries because of differences in support data available or more commonly because of differences in the attribution of density classes. Of the over 475 scenes classified for the EOSD project, it is estimated that fewer than 30 (about 6.3% of total) will require processing to minimize the cross-boundary discontinuity. Options for mitigating the discontinuities are described and demonstrated in the context of different scenarios of overlap found along the EOSD production zone boundaries (complete overlap, partial overlap, and no overlap) using two subsets of a Landsat scene along the shared provincial border between British Columbia and Alberta, Canada. Analysis of image gradients provides a quantitative basis for identification of discontinuities and also relates the results of the likelihood-based relabelling process. Through this process, only density descriptors of cover types are altered, largely maintaining classification integrity. The process as presented is generic and is suitable for addressing edge discontinuities that can emerge when undertaking a large-area land cover classification project.

Landsat ETM+ classification of Fluxnet Canada flux stations

Changes in land use and cover and natural disturbance are thought to be major controls of the dynamic sink/source balance for the immense boreal terrestrial carbon stock. Fluxnet-Canada is a national research network developed to study the influence of climate and disturbance on terrestrial carbon cycling along an east-west transect of Canadian forest and peatlands. The purpose of the present work was to create large-area land cover classifications from satellite imagery to support Fluxnet scaling and modeling studies. The methodology presented in Cihlar et al. (2003) was implemented on seven Fluxnet monitoring sites across Canada. A Landsat ETM+ image covering each site was clustered to 1501 classes using unsupervised K-Means classification prior to 50-cluster merging through classification by generalization (CPG). The resulting clusters were merged to 16-class landcover maps through interactive labeling using cluster bitmaps to create spatial context for the 50 clusters. The bitmaps aided the analyst when ground data was scarce or nonexistent. The products were sent to Fluxnet-Canada site managers for ground validation and the classifications were refined according to the feedback provided. The final landcover maps were used in the calculation of leaf area index (LAI) for the Landsat ETM+ scenes, thus allowing for upscaling of carbon flux measurements based on the correlation of LAI to carbon flux.