Constraints on using AVHRR composite index imagery to study patterns of vegetation cover in boreal forests
Abstract. A wide range of techniques are being developed to map vegetation cover types using multi-date imagery from the Advanced Very High Resolution Radiometer. To date, these techniques do not account for severe constraints which exist for the world's boreal forest. Using composite AVHRR imagery collected over Alaska, we demonstrate how several factors influence the time-series normalized vegetation difference index (NDVI) signatures developed for the boreal forests in this region, including the effects of: (1) clouds and atmospheric haze; (2) climate variations on plant phenology; (3) fire on forest succession; and (4) forest stand patch size with respect to system resolution. Based on the analysis of AVHRR composite data from Alaska, the results of this study show: (1) clouds and haze have distinct effects on the intra-seasonal NDVI signature; (2) there are significant interseasonal variations in NDVI signatures caused by variations in the length of the growing season as well as variations in precipitation and moisture during the growing season; (3) disturbances affect large areas in interior Alaska and forest succession after fire results in significant variations in the inter-seasonal NDVI signatures; and (4) much of the landscape in interior Alaska consists of heterogeneous patches of forest which are much smaller than the resolution cell size of the AVHRR sensor, resulting in significant sub-pixel mixing. Based on these findings, the overall conclusion of this study is scientists using AVHRR to map land cover types in boreal regions must develop approaches which account for these sources of variation.