Satellite-based detection of Canadian boreal forest fires: Development and application of the algorithm

This study presents a comprehensive investigation of fires across the Canadian boreal forest zone by means of satellite-based remote sensing. A firedetection algorithm was designed to monitor fires using daily Advanced Very High Resolution Radiometer (AVHRR) images. It exploits information from multichannel AVHRR measurements to determine the locations of fires on satellite pixels of about 1 km2 under clear sky or thin smoke cloud conditions. Daily fire maps were obtained showing most of the active fires across Canada (except those obscured by thick clouds). This was achieved by first compositing AVHRR scenes acquired over Canada on a given day and then applying the fire-detection algorithm. For the fire seasons of 1994-1998, about 800 NOAA/AVHRR daily mosaics were processed. The results provide valuable nation-wide information on fire activities in terms of their locations, burned area, starting and ending dates, as well as development. The total burned area as detected by satellite across Canada is estimated to be approximately 3.9, 4.9, 1.3, 0.4 and 2.4 million hectares in 1994, 1995, 1996, 1997 and 1998, respectively. The peak month of burning varies considerably from one year to another between June and August, as does the spatial distribution of fires. In general, conifer forests appear to be more vulnerable to burning and fires tend to grow larger than in deciduous forests.     

Detection and removal of cloud contamination from AVHRR images

Describes the characteristics and performance of an algorithm designed to remove the effects of cloud contamination from AVHRR composite images over land. Using information in the temporal NDVI profile, the algorithm detects cloud-contaminated pixels, and optionally replaces these with interpolated values of individual AVHRR channels or channel transformations. As tested, the algorithm detects only those effects of residual clouds that result in a temporary decrease of the NDVI. It was found that the algorithm is capable of identifying clouds of varying opacity as well as cloud shadows present in the composites, and that its use appears preferable to compositing over longer periods. The algorithm can be applied without the need for ancillary information, e.g., on expected surface radiance conditions. Although developed for use in biospheric studies in Canada, it could also be applied at other latitudes where the seasonal trajectories of satellite-derived variables exhibit a single maximum or, by extension, a single minimum     

Approaches to fractional land cover and continuous field mapping: A comparative assessment over the BOREAS study region

Subpixel land cover mapping involves the estimation of surface properties using sensors whose spatial sampling is coarse enough to produce mixtures of the properties within each pixel. This study evaluates five algorithms for mapping subpixel land cover fractions and continuous fields of vegetation properties within the BOREAS study area. The algorithms include a conventional “hard”, per-pixel classifier, a neural network, a clustering/look-up-table approach, multivariate regression, and linear least squares inversion. A land cover map prepared using a Landsat TM mosaic was adopted as the source of fine scale calibration and validation data. Coarse scale mixtures of five basic land cover classes and continuous vegetation fields, both corresponding to the field of view of SPOT-VEGETATION imagery (1.15-km pixel size), were synthesised from the TM mosaic using a modelled point spread function. Two measures of land cover distribution were used, fractions of fine scale land cover categories and continuous fields of vegetation structural characteristics. The subpixel algorithms were applied using both proximate (<100 km) and distant (>400 km) separation between training and validation regions. “Hard” classification performed poorly in estimating proportions or continuous fields. The neural network, look-up-table and multivariate regression algorithms produced good matches of spatial patterns and regional land cover composition for the proximate treatment. However, all three methods exhibited substantial biases with the distant treatment due to the characteristics of the training data. Linear least squares inversion offers a relatively unbiased but less precise alternative for subpixel proportion and fraction mapping as it avoids calibration to the a priori distribution of land cover in the training data. In general, a combination of multivariate regression for proximate training data and linear least squares inversion for distant training data resulted in woody fraction estimates within 20% of the Landsat TM classification-based estimates.